diff options
Diffstat (limited to 'src')
31 files changed, 21557 insertions, 18837 deletions
diff --git a/src/libtracker-data/tracker-data-manager.c b/src/libtracker-data/tracker-data-manager.c index b7d2ffea2..5a811bd79 100644 --- a/src/libtracker-data/tracker-data-manager.c +++ b/src/libtracker-data/tracker-data-manager.c @@ -3642,14 +3642,11 @@ ontology_get_fts_properties (gboolean only_new, } static void -rebuild_fts_tokens (TrackerDBInterface *iface, - gboolean creating_db) +rebuild_fts_tokens (TrackerDBInterface *iface) { - if (!creating_db) { - g_debug ("Rebuilding FTS tokens, this may take a moment..."); - tracker_db_interface_sqlite_fts_rebuild_tokens (iface); - g_debug ("FTS tokens rebuilt"); - } + g_debug ("Rebuilding FTS tokens, this may take a moment..."); + tracker_db_interface_sqlite_fts_rebuild_tokens (iface); + g_debug ("FTS tokens rebuilt"); /* Update the stamp file */ tracker_db_manager_tokenizer_update (); @@ -4636,9 +4633,9 @@ tracker_data_manager_init (TrackerDBManagerFlags flags, tracker_db_manager_set_current_locale (); #if HAVE_TRACKER_FTS - rebuild_fts_tokens (iface, is_first_time_index); + rebuild_fts_tokens (iface); } else if (!read_only && tracker_db_manager_get_tokenizer_changed ()) { - rebuild_fts_tokens (iface, is_first_time_index); + rebuild_fts_tokens (iface); #endif } diff --git a/src/libtracker-data/tracker-data-update.c b/src/libtracker-data/tracker-data-update.c index fcf0cc594..9cdfcaf46 100644 --- a/src/libtracker-data/tracker-data-update.c +++ b/src/libtracker-data/tracker-data-update.c @@ -949,7 +949,6 @@ tracker_data_resource_buffer_flush (GError **error) if (resource_buffer->fts_updated) { TrackerProperty *prop; GArray *values; - gboolean create = resource_buffer->create; GPtrArray *properties, *text; properties = text = NULL; @@ -982,8 +981,7 @@ tracker_data_resource_buffer_flush (GError **error) tracker_db_interface_sqlite_fts_update_text (iface, resource_buffer->id, (const gchar **) properties->pdata, - (const gchar **) text->pdata, - create); + (const gchar **) text->pdata); update_buffer.fts_ever_updated = TRUE; g_ptr_array_free (properties, TRUE); g_ptr_array_free (text, TRUE); @@ -1465,17 +1463,26 @@ get_old_property_values (TrackerProperty *property, if (tracker_property_get_fulltext_indexed (prop) && check_property_domain (prop)) { const gchar *property_name; + GString *str; gint i; old_values = get_property_values (prop); property_name = tracker_property_get_name (prop); + str = g_string_new (NULL); /* delete old fts entries */ for (i = 0; i < old_values->len; i++) { - tracker_db_interface_sqlite_fts_delete_text (iface, - resource_buffer->id, - property_name); + GValue *value = &g_array_index (old_values, GValue, i); + if (i != 0) + g_string_append_c (str, ','); + g_string_append (str, g_value_get_string (value)); } + + tracker_db_interface_sqlite_fts_delete_text (iface, + resource_buffer->id, + property_name, + str->str); + g_string_free (str, TRUE); } } diff --git a/src/libtracker-data/tracker-db-interface-sqlite.c b/src/libtracker-data/tracker-db-interface-sqlite.c index c4f011905..1bd455548 100644 --- a/src/libtracker-data/tracker-db-interface-sqlite.c +++ b/src/libtracker-data/tracker-db-interface-sqlite.c @@ -90,8 +90,7 @@ struct TrackerDBInterface { gpointer busy_user_data; gchar *busy_status; - gchar *fts_insert_str; - gchar *fts_delete_str; + gchar *fts_properties; }; struct TrackerDBInterfaceClass { @@ -1352,7 +1351,7 @@ tracker_db_interface_sqlite_fts_init (TrackerDBInterface *db_interface, tracker_fts_init_db (db_interface->db, properties); if (create && - !tracker_fts_create_table (db_interface->db, "fts", + !tracker_fts_create_table (db_interface->db, "fts5", properties, multivalued)) { g_warning ("FTS tables creation failed"); } @@ -1360,33 +1359,18 @@ tracker_db_interface_sqlite_fts_init (TrackerDBInterface *db_interface, fts_columns = _fts_create_properties (properties); if (fts_columns) { - GString *insert, *select, *delete, cols; - gint i = 0; - - insert = g_string_new ("INSERT INTO fts (docid"); - select = g_string_new ("SELECT rowid"); - delete = g_string_new ("UPDATE fts SET docid=?"); - - while (fts_columns[i]) { - g_string_append_printf (insert, ", \"%s\"", - fts_columns[i]); - g_string_append_printf (select, ", \"%s\"", - fts_columns[i]); - g_string_append_printf (delete, ", \"%s\"=\"\"", - fts_columns[i]); - i++; - } - - g_string_append (select, " FROM fts_view WHERE rowid=?"); - g_string_append (insert, ") "); - g_string_append (insert, select->str); + GString *fts_properties; + gint i; - g_string_free (select, TRUE); - db_interface->fts_insert_str = g_string_free (insert, FALSE); + fts_properties = g_string_new (NULL); - g_string_append (delete, " WHERE docid=?"); - db_interface->fts_delete_str = g_string_free (delete, FALSE); + for (i = 0; fts_columns[i] != NULL; i++) { + g_string_append_printf (fts_properties, ", \"%s\"", + fts_columns[i]); + } + db_interface->fts_properties = g_string_free (fts_properties, + FALSE); g_strfreev (fts_columns); } #endif @@ -1399,104 +1383,134 @@ tracker_db_interface_sqlite_fts_alter_table (TrackerDBInterface *db_interface, GHashTable *properties, GHashTable *multivalued) { - if (!tracker_fts_alter_table (db_interface->db, "fts", properties, multivalued)) { + if (!tracker_fts_alter_table (db_interface->db, "fts5", properties, multivalued)) { g_critical ("Failed to update FTS columns"); } } +static gchar * +tracker_db_interface_sqlite_fts_create_query (TrackerDBInterface *db_interface, + gboolean delete, + const gchar **properties) +{ + GString *insert_str, *values_str; + gint i; + + insert_str = g_string_new ("INSERT INTO fts5 ("); + values_str = g_string_new (NULL); + + if (delete) { + g_string_append (insert_str, "fts5,"); + g_string_append (values_str, "'delete',"); + } + + g_string_append (insert_str, "rowid"); + g_string_append (values_str, "?"); + + for (i = 0; properties[i] != NULL; i++) { + g_string_append_printf (insert_str, ",\"%s\"", properties[i]); + g_string_append (values_str, ",?"); + } + + g_string_append_printf (insert_str, ") VALUES (%s)", values_str->str); + g_string_free (values_str, TRUE); + + return g_string_free (insert_str, FALSE); +} + +static gchar * +tracker_db_interface_sqlite_fts_create_delete_all_query (TrackerDBInterface *db_interface) +{ + GString *insert_str; + + insert_str = g_string_new (NULL); + g_string_append_printf (insert_str, + "INSERT INTO fts5 (fts5, rowid %s) " + "SELECT 'delete', rowid %s FROM fts_view " + "WHERE rowid = ?", + db_interface->fts_properties, + db_interface->fts_properties); + return g_string_free (insert_str, FALSE); +} + gboolean tracker_db_interface_sqlite_fts_update_text (TrackerDBInterface *db_interface, int id, const gchar **properties, - const gchar **text, - gboolean create) + const gchar **text) { TrackerDBStatement *stmt; GError *error = NULL; + gchar *query; + gint i; - if (!create) { - stmt = tracker_db_interface_create_statement (db_interface, - TRACKER_DB_STATEMENT_CACHE_TYPE_UPDATE, - &error, - "DELETE FROM fts WHERE docid=?"); - - if (!stmt || error) { - if (error) { - g_warning ("Could not create FTS update statement: %s", - error->message); - g_error_free (error); - } - return FALSE; - } - - tracker_db_statement_bind_int (stmt, 0, id); - tracker_db_statement_execute (stmt, &error); - g_object_unref (stmt); - - if (error) { - g_warning ("Could not update FTS text: %s", error->message); - g_error_free (error); - return FALSE; - } - } - + query = tracker_db_interface_sqlite_fts_create_query (db_interface, + FALSE, properties); stmt = tracker_db_interface_create_statement (db_interface, TRACKER_DB_STATEMENT_CACHE_TYPE_UPDATE, &error, - "%s", - db_interface->fts_insert_str); + "%s", query); + g_free (query); - if (!stmt || error) { - if (error) { - g_warning ("Could not create FTS insert statement: %s\n", - error->message); - g_error_free (error); - } - return FALSE; - } + if (!stmt || error) { + if (error) { + g_warning ("Could not create FTS insert statement: %s\n", + error->message); + g_error_free (error); + } + return FALSE; + } - tracker_db_statement_bind_int (stmt, 0, id); - tracker_db_statement_execute (stmt, &error); - g_object_unref (stmt); + tracker_db_statement_bind_int (stmt, 0, id); + for (i = 0; text[i] != NULL; i++) { + tracker_db_statement_bind_text (stmt, i + 1, text[i]); + } - if (error) { - g_warning ("Could not insert FTS text: %s", error->message); - g_error_free (error); - return FALSE; - } + tracker_db_statement_execute (stmt, &error); + g_object_unref (stmt); - return TRUE; + if (error) { + g_warning ("Could not insert FTS text: %s", error->message); + g_error_free (error); + return FALSE; + } + + return TRUE; } gboolean -tracker_db_interface_sqlite_fts_delete_text (TrackerDBInterface *db_interface, - int id, - const gchar *property) +tracker_db_interface_sqlite_fts_delete_text (TrackerDBInterface *db_interface, + int rowid, + const gchar *property, + const gchar *old_text) { TrackerDBStatement *stmt; GError *error = NULL; + const gchar *properties[] = { property, NULL }; + gchar *query; + query = tracker_db_interface_sqlite_fts_create_query (db_interface, + TRUE, properties); stmt = tracker_db_interface_create_statement (db_interface, TRACKER_DB_STATEMENT_CACHE_TYPE_UPDATE, &error, - "UPDATE fts SET \"%s\" = '' WHERE docid = ?", - property); + "%s", query); + g_free (query); if (!stmt || error) { - if (error) { - g_warning ("Could not create FTS update statement: %s\n", - error->message); - g_error_free (error); - } + g_warning ("Could not create FTS delete statement: %s", + error ? error->message : "No error given"); + g_clear_error (&error); return FALSE; } - tracker_db_statement_bind_int (stmt, 0, id); + tracker_db_statement_bind_int (stmt, 0, rowid); + tracker_db_statement_bind_text (stmt, 1, old_text); tracker_db_statement_execute (stmt, &error); g_object_unref (stmt); if (error) { - g_warning ("Could not execute FTS update: %s", error->message); + g_warning ("Could not delete FTS text: %s", error->message); g_error_free (error); return FALSE; } @@ -1510,12 +1524,15 @@ tracker_db_interface_sqlite_fts_delete_id (TrackerDBInterface *db_interface, { TrackerDBStatement *stmt; GError *error = NULL; + gchar *query; + query = tracker_db_interface_sqlite_fts_create_delete_all_query (db_interface); stmt = tracker_db_interface_create_statement (db_interface, TRACKER_DB_STATEMENT_CACHE_TYPE_UPDATE, &error, - "%s", - db_interface->fts_delete_str); + "%s", query); + g_free (query); + if (!stmt || error) { if (error) { g_warning ("Could not create FTS delete statement: %s", @@ -1526,7 +1543,6 @@ tracker_db_interface_sqlite_fts_delete_id (TrackerDBInterface *db_interface, } tracker_db_statement_bind_int (stmt, 0, id); - tracker_db_statement_bind_int (stmt, 1, id); tracker_db_statement_execute (stmt, &error); g_object_unref (stmt); @@ -1542,7 +1558,7 @@ tracker_db_interface_sqlite_fts_delete_id (TrackerDBInterface *db_interface, void tracker_db_interface_sqlite_fts_rebuild_tokens (TrackerDBInterface *interface) { - tracker_fts_rebuild_tokens (interface->db, "fts"); + tracker_fts_rebuild_tokens (interface->db, "fts5"); } #endif @@ -1592,8 +1608,7 @@ tracker_db_interface_sqlite_finalize (GObject *object) db_interface = TRACKER_DB_INTERFACE (object); close_database (db_interface); - g_free (db_interface->fts_insert_str); - g_free (db_interface->fts_delete_str); + g_free (db_interface->fts_properties); g_message ("Closed sqlite3 database:'%s'", db_interface->filename); diff --git a/src/libtracker-data/tracker-db-interface-sqlite.h b/src/libtracker-data/tracker-db-interface-sqlite.h index 18c020450..a7e265ef5 100644 --- a/src/libtracker-data/tracker-db-interface-sqlite.h +++ b/src/libtracker-data/tracker-db-interface-sqlite.h @@ -52,17 +52,17 @@ void tracker_db_interface_sqlite_wal_hook (TrackerD void tracker_db_interface_sqlite_fts_alter_table (TrackerDBInterface *interface, GHashTable *properties, GHashTable *multivalued); -int tracker_db_interface_sqlite_fts_update_text (TrackerDBInterface *interface, - int id, +gboolean tracker_db_interface_sqlite_fts_update_text (TrackerDBInterface *db_interface, + int id, const gchar **properties, - const gchar **text, - gboolean create); + const gchar **text); -gboolean tracker_db_interface_sqlite_fts_delete_text (TrackerDBInterface *db_interface, - int id, - const gchar *property); +gboolean tracker_db_interface_sqlite_fts_delete_text (TrackerDBInterface *interface, + int rowid, + const gchar *property, + const gchar *old_text); gboolean tracker_db_interface_sqlite_fts_delete_id (TrackerDBInterface *interface, - int id); + int rowid); void tracker_db_interface_sqlite_fts_update_commit (TrackerDBInterface *interface); void tracker_db_interface_sqlite_fts_update_rollback (TrackerDBInterface *interface); diff --git a/src/libtracker-data/tracker-sparql-expression.vala b/src/libtracker-data/tracker-sparql-expression.vala index d69b908f7..91c6c10a0 100644 --- a/src/libtracker-data/tracker-sparql-expression.vala +++ b/src/libtracker-data/tracker-sparql-expression.vala @@ -709,7 +709,7 @@ class Tracker.Sparql.Expression : Object { var variable = context.get_variable (v); sql.append (variable.sql_expression); - fts_sql = "tracker_offsets(offsets(\"fts\"),fts_property_names())"; + fts_sql = "tracker_offsets(\"fts5\")"; return PropertyType.STRING; } else if (uri == FTS_NS + "snippet") { bool is_var; @@ -718,7 +718,10 @@ class Tracker.Sparql.Expression : Object { var variable = context.get_variable (v); var fts = new StringBuilder (); - fts.append_printf ("snippet(\"fts\""); + fts.append_printf ("snippet(\"fts5\""); + + // lookup column + fts.append (", -1"); // "start match" text if (accept (SparqlTokenType.COMMA)) { @@ -741,9 +744,6 @@ class Tracker.Sparql.Expression : Object { fts.append (", '...'"); } - // lookup column - fts.append (", -1"); - // Approximate number of words in context if (accept (SparqlTokenType.COMMA)) { fts.append (", "); diff --git a/src/libtracker-data/tracker-sparql-pattern.vala b/src/libtracker-data/tracker-sparql-pattern.vala index ec9d5471e..73740b7c2 100644 --- a/src/libtracker-data/tracker-sparql-pattern.vala +++ b/src/libtracker-data/tracker-sparql-pattern.vala @@ -341,14 +341,14 @@ class Tracker.Sparql.Pattern : Object { } if (queries_fts_data && fts_subject != null) { - // Ensure there's a docid to match on in FTS queries + // Ensure there's a rowid to match on in FTS queries if (!first) { sql.append (", "); } else { first = false; } - sql.append ("%s AS docid ".printf (fts_subject.sql_expression)); + sql.append ("%s AS rowid ".printf (fts_subject.sql_expression)); } // literals in select expressions need to be bound before literals in the where clause @@ -459,9 +459,9 @@ class Tracker.Sparql.Pattern : Object { str.append (fts_var); } - str.append (" FROM fts JOIN ("); + str.append (" FROM fts5 JOIN ("); sql.prepend (str.str); - sql.append_printf (") AS ranks USING (docid) WHERE fts %s".printf (match_str.str)); + sql.append_printf (") AS ranks ON fts5.rowid=rowid WHERE fts5 %s".printf (match_str.str)); } context = context.parent_context; @@ -1349,7 +1349,7 @@ class Tracker.Sparql.Pattern : Object { } else if (prop == null) { if (current_predicate == "http://www.tracker-project.org/ontologies/fts#match") { // fts:match - db_table = "fts"; + db_table = "fts5"; share_table = false; is_fts_match = true; fts_subject = context.get_variable (current_subject); @@ -1477,7 +1477,7 @@ class Tracker.Sparql.Pattern : Object { binding.table = table; binding.type = subject_type; if (is_fts_match) { - binding.sql_db_column_name = "docid"; + binding.sql_db_column_name = "rowid"; } else { binding.sql_db_column_name = "ID"; } @@ -1532,13 +1532,12 @@ class Tracker.Sparql.Pattern : Object { binding.literal = object; // binding.data_type = triple.object.type; binding.table = table; - binding.sql_db_column_name = "fts"; + binding.sql_db_column_name = "fts5"; triple_context.bindings.append (binding); - sql.append_printf ("\"%s\".\"docid\" AS \"ID\", ", + sql.append_printf ("\"%s\".\"rowid\" AS \"ID\", ", binding.table.sql_query_tablename); - sql.append_printf ("tracker_rank(matchinfo(\"%s\".\"fts\", 'cl'),fts_column_weights()) " + - "AS \"%s_u_rank\", ", + sql.append_printf ("\"%s\".\"rank\" AS \"%s_u_rank\", ", binding.table.sql_query_tablename, context.get_variable (current_subject).name); } else { diff --git a/src/libtracker-fts/Makefile.am b/src/libtracker-fts/Makefile.am index 8021c78b7..327dd03af 100644 --- a/src/libtracker-fts/Makefile.am +++ b/src/libtracker-fts/Makefile.am @@ -6,24 +6,9 @@ AM_CPPFLAGS = \ noinst_LTLIBRARIES = libtracker-fts.la -fts4_sources = \ - fts3_aux.c \ - fts3.c \ - fts3_expr.c \ - fts3.h \ - fts3_hash.c \ - fts3_hash.h \ - fts3_icu.c \ - fts3Int.h \ - fts3_porter.c \ - fts3_snippet.c \ - fts3_term.c \ - fts3_tokenizer1.c \ - fts3_tokenizer.c \ - fts3_tokenizer.h \ - fts3_unicode.c \ - fts3_unicode2.c \ - fts3_write.c +fts5_sources = \ + fts5.c \ + fts5.h libtracker_fts_la_SOURCES = \ tracker-fts.c \ @@ -34,7 +19,7 @@ libtracker_fts_la_SOURCES = \ tracker-fts-tokenizer.h if !HAVE_BUILTIN_FTS - libtracker_fts_la_SOURCES += $(fts4_sources) + libtracker_fts_la_SOURCES += $(fts5_sources) endif libtracker_fts_la_LIBADD = \ @@ -42,7 +27,7 @@ libtracker_fts_la_LIBADD = \ $(BUILD_LIBS) \ $(LIBTRACKER_FTS_LIBS) -EXTRA_DIST = $(fts4_sources) +EXTRA_DIST = $(fts5_sources) # Configuration / GSettings gsettings_ENUM_NAMESPACE = org.freedesktop.Tracker diff --git a/src/libtracker-fts/fts3.c b/src/libtracker-fts/fts3.c deleted file mode 100644 index 3dc62ba8d..000000000 --- a/src/libtracker-fts/fts3.c +++ /dev/null @@ -1,5350 +0,0 @@ -/* -** 2006 Oct 10 -** -** The author disclaims copyright to this source code. In place of -** a legal notice, here is a blessing: -** -** May you do good and not evil. -** May you find forgiveness for yourself and forgive others. -** May you share freely, never taking more than you give. -** -****************************************************************************** -** -** This is an SQLite module implementing full-text search. -*/ - -/* -** The code in this file is only compiled if: -** -** * The FTS3 module is being built as an extension -** (in which case SQLITE_CORE is not defined), or -** -** * The FTS3 module is being built into the core of -** SQLite (in which case SQLITE_ENABLE_FTS3 is defined). -*/ - -/* The full-text index is stored in a series of b+tree (-like) -** structures called segments which map terms to doclists. The -** structures are like b+trees in layout, but are constructed from the -** bottom up in optimal fashion and are not updatable. Since trees -** are built from the bottom up, things will be described from the -** bottom up. -** -** -**** Varints **** -** The basic unit of encoding is a variable-length integer called a -** varint. We encode variable-length integers in little-endian order -** using seven bits * per byte as follows: -** -** KEY: -** A = 0xxxxxxx 7 bits of data and one flag bit -** B = 1xxxxxxx 7 bits of data and one flag bit -** -** 7 bits - A -** 14 bits - BA -** 21 bits - BBA -** and so on. -** -** This is similar in concept to how sqlite encodes "varints" but -** the encoding is not the same. SQLite varints are big-endian -** are are limited to 9 bytes in length whereas FTS3 varints are -** little-endian and can be up to 10 bytes in length (in theory). -** -** Example encodings: -** -** 1: 0x01 -** 127: 0x7f -** 128: 0x81 0x00 -** -** -**** Document lists **** -** A doclist (document list) holds a docid-sorted list of hits for a -** given term. Doclists hold docids and associated token positions. -** A docid is the unique integer identifier for a single document. -** A position is the index of a word within the document. The first -** word of the document has a position of 0. -** -** FTS3 used to optionally store character offsets using a compile-time -** option. But that functionality is no longer supported. -** -** A doclist is stored like this: -** -** array { -** varint docid; (delta from previous doclist) -** array { (position list for column 0) -** varint position; (2 more than the delta from previous position) -** } -** array { -** varint POS_COLUMN; (marks start of position list for new column) -** varint column; (index of new column) -** array { -** varint position; (2 more than the delta from previous position) -** } -** } -** varint POS_END; (marks end of positions for this document. -** } -** -** Here, array { X } means zero or more occurrences of X, adjacent in -** memory. A "position" is an index of a token in the token stream -** generated by the tokenizer. Note that POS_END and POS_COLUMN occur -** in the same logical place as the position element, and act as sentinals -** ending a position list array. POS_END is 0. POS_COLUMN is 1. -** The positions numbers are not stored literally but rather as two more -** than the difference from the prior position, or the just the position plus -** 2 for the first position. Example: -** -** label: A B C D E F G H I J K -** value: 123 5 9 1 1 14 35 0 234 72 0 -** -** The 123 value is the first docid. For column zero in this document -** there are two matches at positions 3 and 10 (5-2 and 9-2+3). The 1 -** at D signals the start of a new column; the 1 at E indicates that the -** new column is column number 1. There are two positions at 12 and 45 -** (14-2 and 35-2+12). The 0 at H indicate the end-of-document. The -** 234 at I is the delta to next docid (357). It has one position 70 -** (72-2) and then terminates with the 0 at K. -** -** A "position-list" is the list of positions for multiple columns for -** a single docid. A "column-list" is the set of positions for a single -** column. Hence, a position-list consists of one or more column-lists, -** a document record consists of a docid followed by a position-list and -** a doclist consists of one or more document records. -** -** A bare doclist omits the position information, becoming an -** array of varint-encoded docids. -** -**** Segment leaf nodes **** -** Segment leaf nodes store terms and doclists, ordered by term. Leaf -** nodes are written using LeafWriter, and read using LeafReader (to -** iterate through a single leaf node's data) and LeavesReader (to -** iterate through a segment's entire leaf layer). Leaf nodes have -** the format: -** -** varint iHeight; (height from leaf level, always 0) -** varint nTerm; (length of first term) -** char pTerm[nTerm]; (content of first term) -** varint nDoclist; (length of term's associated doclist) -** char pDoclist[nDoclist]; (content of doclist) -** array { -** (further terms are delta-encoded) -** varint nPrefix; (length of prefix shared with previous term) -** varint nSuffix; (length of unshared suffix) -** char pTermSuffix[nSuffix];(unshared suffix of next term) -** varint nDoclist; (length of term's associated doclist) -** char pDoclist[nDoclist]; (content of doclist) -** } -** -** Here, array { X } means zero or more occurrences of X, adjacent in -** memory. -** -** Leaf nodes are broken into blocks which are stored contiguously in -** the %_segments table in sorted order. This means that when the end -** of a node is reached, the next term is in the node with the next -** greater node id. -** -** New data is spilled to a new leaf node when the current node -** exceeds LEAF_MAX bytes (default 2048). New data which itself is -** larger than STANDALONE_MIN (default 1024) is placed in a standalone -** node (a leaf node with a single term and doclist). The goal of -** these settings is to pack together groups of small doclists while -** making it efficient to directly access large doclists. The -** assumption is that large doclists represent terms which are more -** likely to be query targets. -** -** TODO(shess) It may be useful for blocking decisions to be more -** dynamic. For instance, it may make more sense to have a 2.5k leaf -** node rather than splitting into 2k and .5k nodes. My intuition is -** that this might extend through 2x or 4x the pagesize. -** -** -**** Segment interior nodes **** -** Segment interior nodes store blockids for subtree nodes and terms -** to describe what data is stored by the each subtree. Interior -** nodes are written using InteriorWriter, and read using -** InteriorReader. InteriorWriters are created as needed when -** SegmentWriter creates new leaf nodes, or when an interior node -** itself grows too big and must be split. The format of interior -** nodes: -** -** varint iHeight; (height from leaf level, always >0) -** varint iBlockid; (block id of node's leftmost subtree) -** optional { -** varint nTerm; (length of first term) -** char pTerm[nTerm]; (content of first term) -** array { -** (further terms are delta-encoded) -** varint nPrefix; (length of shared prefix with previous term) -** varint nSuffix; (length of unshared suffix) -** char pTermSuffix[nSuffix]; (unshared suffix of next term) -** } -** } -** -** Here, optional { X } means an optional element, while array { X } -** means zero or more occurrences of X, adjacent in memory. -** -** An interior node encodes n terms separating n+1 subtrees. The -** subtree blocks are contiguous, so only the first subtree's blockid -** is encoded. The subtree at iBlockid will contain all terms less -** than the first term encoded (or all terms if no term is encoded). -** Otherwise, for terms greater than or equal to pTerm[i] but less -** than pTerm[i+1], the subtree for that term will be rooted at -** iBlockid+i. Interior nodes only store enough term data to -** distinguish adjacent children (if the rightmost term of the left -** child is "something", and the leftmost term of the right child is -** "wicked", only "w" is stored). -** -** New data is spilled to a new interior node at the same height when -** the current node exceeds INTERIOR_MAX bytes (default 2048). -** INTERIOR_MIN_TERMS (default 7) keeps large terms from monopolizing -** interior nodes and making the tree too skinny. The interior nodes -** at a given height are naturally tracked by interior nodes at -** height+1, and so on. -** -** -**** Segment directory **** -** The segment directory in table %_segdir stores meta-information for -** merging and deleting segments, and also the root node of the -** segment's tree. -** -** The root node is the top node of the segment's tree after encoding -** the entire segment, restricted to ROOT_MAX bytes (default 1024). -** This could be either a leaf node or an interior node. If the top -** node requires more than ROOT_MAX bytes, it is flushed to %_segments -** and a new root interior node is generated (which should always fit -** within ROOT_MAX because it only needs space for 2 varints, the -** height and the blockid of the previous root). -** -** The meta-information in the segment directory is: -** level - segment level (see below) -** idx - index within level -** - (level,idx uniquely identify a segment) -** start_block - first leaf node -** leaves_end_block - last leaf node -** end_block - last block (including interior nodes) -** root - contents of root node -** -** If the root node is a leaf node, then start_block, -** leaves_end_block, and end_block are all 0. -** -** -**** Segment merging **** -** To amortize update costs, segments are grouped into levels and -** merged in batches. Each increase in level represents exponentially -** more documents. -** -** New documents (actually, document updates) are tokenized and -** written individually (using LeafWriter) to a level 0 segment, with -** incrementing idx. When idx reaches MERGE_COUNT (default 16), all -** level 0 segments are merged into a single level 1 segment. Level 1 -** is populated like level 0, and eventually MERGE_COUNT level 1 -** segments are merged to a single level 2 segment (representing -** MERGE_COUNT^2 updates), and so on. -** -** A segment merge traverses all segments at a given level in -** parallel, performing a straightforward sorted merge. Since segment -** leaf nodes are written in to the %_segments table in order, this -** merge traverses the underlying sqlite disk structures efficiently. -** After the merge, all segment blocks from the merged level are -** deleted. -** -** MERGE_COUNT controls how often we merge segments. 16 seems to be -** somewhat of a sweet spot for insertion performance. 32 and 64 show -** very similar performance numbers to 16 on insertion, though they're -** a tiny bit slower (perhaps due to more overhead in merge-time -** sorting). 8 is about 20% slower than 16, 4 about 50% slower than -** 16, 2 about 66% slower than 16. -** -** At query time, high MERGE_COUNT increases the number of segments -** which need to be scanned and merged. For instance, with 100k docs -** inserted: -** -** MERGE_COUNT segments -** 16 25 -** 8 12 -** 4 10 -** 2 6 -** -** This appears to have only a moderate impact on queries for very -** frequent terms (which are somewhat dominated by segment merge -** costs), and infrequent and non-existent terms still seem to be fast -** even with many segments. -** -** TODO(shess) That said, it would be nice to have a better query-side -** argument for MERGE_COUNT of 16. Also, it is possible/likely that -** optimizations to things like doclist merging will swing the sweet -** spot around. -** -** -** -**** Handling of deletions and updates **** -** Since we're using a segmented structure, with no docid-oriented -** index into the term index, we clearly cannot simply update the term -** index when a document is deleted or updated. For deletions, we -** write an empty doclist (varint(docid) varint(POS_END)), for updates -** we simply write the new doclist. Segment merges overwrite older -** data for a particular docid with newer data, so deletes or updates -** will eventually overtake the earlier data and knock it out. The -** query logic likewise merges doclists so that newer data knocks out -** older data. -*/ - -#include "fts3Int.h" -#if !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3) - -#if defined(SQLITE_ENABLE_FTS3) && !defined(SQLITE_CORE) -# define SQLITE_CORE 1 -#endif - -#include <assert.h> -#include <stdlib.h> -#include <stddef.h> -#include <stdio.h> -#include <string.h> -#include <stdarg.h> - -#include "fts3.h" -#ifndef SQLITE_CORE -# include "sqlite3ext.h" - SQLITE_EXTENSION_INIT1 -#endif - -static int fts3EvalNext(Fts3Cursor *pCsr); -static int fts3EvalStart(Fts3Cursor *pCsr); -static int fts3TermSegReaderCursor( - Fts3Cursor *, const char *, int, int, Fts3MultiSegReader **); - -/* -** Write a 64-bit variable-length integer to memory starting at p[0]. -** The length of data written will be between 1 and FTS3_VARINT_MAX bytes. -** The number of bytes written is returned. -*/ -int sqlite3Fts3PutVarint(char *p, sqlite_int64 v){ - unsigned char *q = (unsigned char *) p; - sqlite_uint64 vu = v; - do{ - *q++ = (unsigned char) ((vu & 0x7f) | 0x80); - vu >>= 7; - }while( vu!=0 ); - q[-1] &= 0x7f; /* turn off high bit in final byte */ - assert( q - (unsigned char *)p <= FTS3_VARINT_MAX ); - return (int) (q - (unsigned char *)p); -} - -/* -** Read a 64-bit variable-length integer from memory starting at p[0]. -** Return the number of bytes read, or 0 on error. -** The value is stored in *v. -*/ -int sqlite3Fts3GetVarint(const char *p, sqlite_int64 *v){ - const unsigned char *q = (const unsigned char *) p; - sqlite_uint64 x = 0, y = 1; - while( (*q&0x80)==0x80 && q-(unsigned char *)p<FTS3_VARINT_MAX ){ - x += y * (*q++ & 0x7f); - y <<= 7; - } - x += y * (*q++); - *v = (sqlite_int64) x; - return (int) (q - (unsigned char *)p); -} - -/* -** Similar to sqlite3Fts3GetVarint(), except that the output is truncated to a -** 32-bit integer before it is returned. -*/ -int sqlite3Fts3GetVarint32(const char *p, int *pi){ - sqlite_int64 i; - int ret = sqlite3Fts3GetVarint(p, &i); - *pi = (int) i; - return ret; -} - -/* -** Return the number of bytes required to encode v as a varint -*/ -int sqlite3Fts3VarintLen(sqlite3_uint64 v){ - int i = 0; - do{ - i++; - v >>= 7; - }while( v!=0 ); - return i; -} - -/* -** Convert an SQL-style quoted string into a normal string by removing -** the quote characters. The conversion is done in-place. If the -** input does not begin with a quote character, then this routine -** is a no-op. -** -** Examples: -** -** "abc" becomes abc -** 'xyz' becomes xyz -** [pqr] becomes pqr -** `mno` becomes mno -** -*/ -void sqlite3Fts3Dequote(char *z){ - char quote; /* Quote character (if any ) */ - - quote = z[0]; - if( quote=='[' || quote=='\'' || quote=='"' || quote=='`' ){ - int iIn = 1; /* Index of next byte to read from input */ - int iOut = 0; /* Index of next byte to write to output */ - - /* If the first byte was a '[', then the close-quote character is a ']' */ - if( quote=='[' ) quote = ']'; - - while( ALWAYS(z[iIn]) ){ - if( z[iIn]==quote ){ - if( z[iIn+1]!=quote ) break; - z[iOut++] = quote; - iIn += 2; - }else{ - z[iOut++] = z[iIn++]; - } - } - z[iOut] = '\0'; - } -} - -/* -** Read a single varint from the doclist at *pp and advance *pp to point -** to the first byte past the end of the varint. Add the value of the varint -** to *pVal. -*/ -static void fts3GetDeltaVarint(char **pp, sqlite3_int64 *pVal){ - sqlite3_int64 iVal; - *pp += sqlite3Fts3GetVarint(*pp, &iVal); - *pVal += iVal; -} - -/* -** When this function is called, *pp points to the first byte following a -** varint that is part of a doclist (or position-list, or any other list -** of varints). This function moves *pp to point to the start of that varint, -** and sets *pVal by the varint value. -** -** Argument pStart points to the first byte of the doclist that the -** varint is part of. -*/ -static void fts3GetReverseVarint( - char **pp, - char *pStart, - sqlite3_int64 *pVal -){ - sqlite3_int64 iVal; - char *p; - - /* Pointer p now points at the first byte past the varint we are - ** interested in. So, unless the doclist is corrupt, the 0x80 bit is - ** clear on character p[-1]. */ - for(p = (*pp)-2; p>=pStart && *p&0x80; p--); - p++; - *pp = p; - - sqlite3Fts3GetVarint(p, &iVal); - *pVal = iVal; -} - -/* -** The xDisconnect() virtual table method. -*/ -static int fts3DisconnectMethod(sqlite3_vtab *pVtab){ - Fts3Table *p = (Fts3Table *)pVtab; - int i; - - assert( p->nPendingData==0 ); - assert( p->pSegments==0 ); - - /* Free any prepared statements held */ - for(i=0; i<SizeofArray(p->aStmt); i++){ - sqlite3_finalize(p->aStmt[i]); - } - sqlite3_free(p->zSegmentsTbl); - sqlite3_free(p->zReadExprlist); - sqlite3_free(p->zWriteExprlist); - sqlite3_free(p->zContentTbl); - sqlite3_free(p->zLanguageid); - - /* Invoke the tokenizer destructor to free the tokenizer. */ - p->pTokenizer->pModule->xDestroy(p->pTokenizer); - - sqlite3_free(p); - return SQLITE_OK; -} - -/* -** Construct one or more SQL statements from the format string given -** and then evaluate those statements. The success code is written -** into *pRc. -** -** If *pRc is initially non-zero then this routine is a no-op. -*/ -static void fts3DbExec( - int *pRc, /* Success code */ - sqlite3 *db, /* Database in which to run SQL */ - const char *zFormat, /* Format string for SQL */ - ... /* Arguments to the format string */ -){ - va_list ap; - char *zSql; - if( *pRc ) return; - va_start(ap, zFormat); - zSql = sqlite3_vmprintf(zFormat, ap); - va_end(ap); - if( zSql==0 ){ - *pRc = SQLITE_NOMEM; - }else{ - *pRc = sqlite3_exec(db, zSql, 0, 0, 0); - sqlite3_free(zSql); - } -} - -/* -** The xDestroy() virtual table method. -*/ -static int fts3DestroyMethod(sqlite3_vtab *pVtab){ - Fts3Table *p = (Fts3Table *)pVtab; - int rc = SQLITE_OK; /* Return code */ - const char *zDb = p->zDb; /* Name of database (e.g. "main", "temp") */ - sqlite3 *db = p->db; /* Database handle */ - - /* Drop the shadow tables */ - if( p->zContentTbl==0 ){ - fts3DbExec(&rc, db, "DROP TABLE IF EXISTS %Q.'%q_content'", zDb, p->zName); - } - fts3DbExec(&rc, db, "DROP TABLE IF EXISTS %Q.'%q_segments'", zDb,p->zName); - fts3DbExec(&rc, db, "DROP TABLE IF EXISTS %Q.'%q_segdir'", zDb, p->zName); - fts3DbExec(&rc, db, "DROP TABLE IF EXISTS %Q.'%q_docsize'", zDb, p->zName); - fts3DbExec(&rc, db, "DROP TABLE IF EXISTS %Q.'%q_stat'", zDb, p->zName); - - /* If everything has worked, invoke fts3DisconnectMethod() to free the - ** memory associated with the Fts3Table structure and return SQLITE_OK. - ** Otherwise, return an SQLite error code. - */ - return (rc==SQLITE_OK ? fts3DisconnectMethod(pVtab) : rc); -} - - -/* -** Invoke sqlite3_declare_vtab() to declare the schema for the FTS3 table -** passed as the first argument. This is done as part of the xConnect() -** and xCreate() methods. -** -** If *pRc is non-zero when this function is called, it is a no-op. -** Otherwise, if an error occurs, an SQLite error code is stored in *pRc -** before returning. -*/ -static void fts3DeclareVtab(int *pRc, Fts3Table *p){ - if( *pRc==SQLITE_OK ){ - int i; /* Iterator variable */ - int rc; /* Return code */ - char *zSql; /* SQL statement passed to declare_vtab() */ - char *zCols; /* List of user defined columns */ - const char *zLanguageid; - - zLanguageid = (p->zLanguageid ? p->zLanguageid : "__langid"); - sqlite3_vtab_config(p->db, SQLITE_VTAB_CONSTRAINT_SUPPORT, 1); - - /* Create a list of user columns for the virtual table */ - zCols = sqlite3_mprintf("%Q, ", p->azColumn[0]); - for(i=1; zCols && i<p->nColumn; i++){ - zCols = sqlite3_mprintf("%z%Q, ", zCols, p->azColumn[i]); - } - - /* Create the whole "CREATE TABLE" statement to pass to SQLite */ - zSql = sqlite3_mprintf( - "CREATE TABLE x(%s %Q HIDDEN, docid HIDDEN, %Q HIDDEN)", - zCols, p->zName, zLanguageid - ); - if( !zCols || !zSql ){ - rc = SQLITE_NOMEM; - }else{ - rc = sqlite3_declare_vtab(p->db, zSql); - } - - sqlite3_free(zSql); - sqlite3_free(zCols); - *pRc = rc; - } -} - -/* -** Create the %_stat table if it does not already exist. -*/ -void sqlite3Fts3CreateStatTable(int *pRc, Fts3Table *p){ - fts3DbExec(pRc, p->db, - "CREATE TABLE IF NOT EXISTS %Q.'%q_stat'" - "(id INTEGER PRIMARY KEY, value BLOB);", - p->zDb, p->zName - ); - if( (*pRc)==SQLITE_OK ) p->bHasStat = 1; -} - -/* -** Create the backing store tables (%_content, %_segments and %_segdir) -** required by the FTS3 table passed as the only argument. This is done -** as part of the vtab xCreate() method. -** -** If the p->bHasDocsize boolean is true (indicating that this is an -** FTS4 table, not an FTS3 table) then also create the %_docsize and -** %_stat tables required by FTS4. -*/ -static int fts3CreateTables(Fts3Table *p){ - int rc = SQLITE_OK; /* Return code */ - int i; /* Iterator variable */ - sqlite3 *db = p->db; /* The database connection */ - - if( p->zContentTbl==0 ){ - const char *zLanguageid = p->zLanguageid; - char *zContentCols; /* Columns of %_content table */ - - /* Create a list of user columns for the content table */ - zContentCols = sqlite3_mprintf("docid INTEGER PRIMARY KEY"); - for(i=0; zContentCols && i<p->nColumn; i++){ - char *z = p->azColumn[i]; - zContentCols = sqlite3_mprintf("%z, 'c%d%q'", zContentCols, i, z); - } - if( zLanguageid && zContentCols ){ - zContentCols = sqlite3_mprintf("%z, langid", zContentCols, zLanguageid); - } - if( zContentCols==0 ) rc = SQLITE_NOMEM; - - /* Create the content table */ - fts3DbExec(&rc, db, - "CREATE TABLE %Q.'%q_content'(%s)", - p->zDb, p->zName, zContentCols - ); - sqlite3_free(zContentCols); - } - - /* Create other tables */ - fts3DbExec(&rc, db, - "CREATE TABLE %Q.'%q_segments'(blockid INTEGER PRIMARY KEY, block BLOB);", - p->zDb, p->zName - ); - fts3DbExec(&rc, db, - "CREATE TABLE %Q.'%q_segdir'(" - "level INTEGER," - "idx INTEGER," - "start_block INTEGER," - "leaves_end_block INTEGER," - "end_block INTEGER," - "root BLOB," - "PRIMARY KEY(level, idx)" - ");", - p->zDb, p->zName - ); - if( p->bHasDocsize ){ - fts3DbExec(&rc, db, - "CREATE TABLE %Q.'%q_docsize'(docid INTEGER PRIMARY KEY, size BLOB);", - p->zDb, p->zName - ); - } - assert( p->bHasStat==p->bFts4 ); - if( p->bHasStat ){ - sqlite3Fts3CreateStatTable(&rc, p); - } - return rc; -} - -/* -** Store the current database page-size in bytes in p->nPgsz. -** -** If *pRc is non-zero when this function is called, it is a no-op. -** Otherwise, if an error occurs, an SQLite error code is stored in *pRc -** before returning. -*/ -static void fts3DatabasePageSize(int *pRc, Fts3Table *p){ - if( *pRc==SQLITE_OK ){ - int rc; /* Return code */ - char *zSql; /* SQL text "PRAGMA %Q.page_size" */ - sqlite3_stmt *pStmt; /* Compiled "PRAGMA %Q.page_size" statement */ - - zSql = sqlite3_mprintf("PRAGMA %Q.page_size", p->zDb); - if( !zSql ){ - rc = SQLITE_NOMEM; - }else{ - rc = sqlite3_prepare(p->db, zSql, -1, &pStmt, 0); - if( rc==SQLITE_OK ){ - sqlite3_step(pStmt); - p->nPgsz = sqlite3_column_int(pStmt, 0); - rc = sqlite3_finalize(pStmt); - }else if( rc==SQLITE_AUTH ){ - p->nPgsz = 1024; - rc = SQLITE_OK; - } - } - assert( p->nPgsz>0 || rc!=SQLITE_OK ); - sqlite3_free(zSql); - *pRc = rc; - } -} - -/* -** "Special" FTS4 arguments are column specifications of the following form: -** -** <key> = <value> -** -** There may not be whitespace surrounding the "=" character. The <value> -** term may be quoted, but the <key> may not. -*/ -static int fts3IsSpecialColumn( - const char *z, - int *pnKey, - char **pzValue -){ - char *zValue; - const char *zCsr = z; - - while( *zCsr!='=' ){ - if( *zCsr=='\0' ) return 0; - zCsr++; - } - - *pnKey = (int)(zCsr-z); - zValue = sqlite3_mprintf("%s", &zCsr[1]); - if( zValue ){ - sqlite3Fts3Dequote(zValue); - } - *pzValue = zValue; - return 1; -} - -/* -** Append the output of a printf() style formatting to an existing string. -*/ -static void fts3Appendf( - int *pRc, /* IN/OUT: Error code */ - char **pz, /* IN/OUT: Pointer to string buffer */ - const char *zFormat, /* Printf format string to append */ - ... /* Arguments for printf format string */ -){ - if( *pRc==SQLITE_OK ){ - va_list ap; - char *z; - va_start(ap, zFormat); - z = sqlite3_vmprintf(zFormat, ap); - va_end(ap); - if( z && *pz ){ - char *z2 = sqlite3_mprintf("%s%s", *pz, z); - sqlite3_free(z); - z = z2; - } - if( z==0 ) *pRc = SQLITE_NOMEM; - sqlite3_free(*pz); - *pz = z; - } -} - -/* -** Return a copy of input string zInput enclosed in double-quotes (") and -** with all double quote characters escaped. For example: -** -** fts3QuoteId("un \"zip\"") -> "un \"\"zip\"\"" -** -** The pointer returned points to memory obtained from sqlite3_malloc(). It -** is the callers responsibility to call sqlite3_free() to release this -** memory. -*/ -static char *fts3QuoteId(char const *zInput){ - int nRet; - char *zRet; - nRet = 2 + (int)strlen(zInput)*2 + 1; - zRet = sqlite3_malloc(nRet); - if( zRet ){ - int i; - char *z = zRet; - *(z++) = '"'; - for(i=0; zInput[i]; i++){ - if( zInput[i]=='"' ) *(z++) = '"'; - *(z++) = zInput[i]; - } - *(z++) = '"'; - *(z++) = '\0'; - } - return zRet; -} - -/* -** Return a list of comma separated SQL expressions and a FROM clause that -** could be used in a SELECT statement such as the following: -** -** SELECT <list of expressions> FROM %_content AS x ... -** -** to return the docid, followed by each column of text data in order -** from left to write. If parameter zFunc is not NULL, then instead of -** being returned directly each column of text data is passed to an SQL -** function named zFunc first. For example, if zFunc is "unzip" and the -** table has the three user-defined columns "a", "b", and "c", the following -** string is returned: -** -** "docid, unzip(x.'a'), unzip(x.'b'), unzip(x.'c') FROM %_content AS x" -** -** The pointer returned points to a buffer allocated by sqlite3_malloc(). It -** is the responsibility of the caller to eventually free it. -** -** If *pRc is not SQLITE_OK when this function is called, it is a no-op (and -** a NULL pointer is returned). Otherwise, if an OOM error is encountered -** by this function, NULL is returned and *pRc is set to SQLITE_NOMEM. If -** no error occurs, *pRc is left unmodified. -*/ -static char *fts3ReadExprList(Fts3Table *p, const char *zFunc, int *pRc){ - char *zRet = 0; - char *zFree = 0; - char *zFunction; - int i; - - if( p->zContentTbl==0 ){ - if( !zFunc ){ - zFunction = ""; - }else{ - zFree = zFunction = fts3QuoteId(zFunc); - } - fts3Appendf(pRc, &zRet, "docid"); - for(i=0; i<p->nColumn; i++){ - fts3Appendf(pRc, &zRet, ",%s(x.'c%d%q')", zFunction, i, p->azColumn[i]); - } - if( p->zLanguageid ){ - fts3Appendf(pRc, &zRet, ", x.%Q", "langid"); - } - sqlite3_free(zFree); - }else{ - fts3Appendf(pRc, &zRet, "rowid"); - for(i=0; i<p->nColumn; i++){ - fts3Appendf(pRc, &zRet, ", x.'%q'", p->azColumn[i]); - } - if( p->zLanguageid ){ - fts3Appendf(pRc, &zRet, ", x.%Q", p->zLanguageid); - } - } - fts3Appendf(pRc, &zRet, " FROM '%q'.'%q%s' AS x", - p->zDb, - (p->zContentTbl ? p->zContentTbl : p->zName), - (p->zContentTbl ? "" : "_content") - ); - return zRet; -} - -/* -** Return a list of N comma separated question marks, where N is the number -** of columns in the %_content table (one for the docid plus one for each -** user-defined text column). -** -** If argument zFunc is not NULL, then all but the first question mark -** is preceded by zFunc and an open bracket, and followed by a closed -** bracket. For example, if zFunc is "zip" and the FTS3 table has three -** user-defined text columns, the following string is returned: -** -** "?, zip(?), zip(?), zip(?)" -** -** The pointer returned points to a buffer allocated by sqlite3_malloc(). It -** is the responsibility of the caller to eventually free it. -** -** If *pRc is not SQLITE_OK when this function is called, it is a no-op (and -** a NULL pointer is returned). Otherwise, if an OOM error is encountered -** by this function, NULL is returned and *pRc is set to SQLITE_NOMEM. If -** no error occurs, *pRc is left unmodified. -*/ -static char *fts3WriteExprList(Fts3Table *p, const char *zFunc, int *pRc){ - char *zRet = 0; - char *zFree = 0; - char *zFunction; - int i; - - if( !zFunc ){ - zFunction = ""; - }else{ - zFree = zFunction = fts3QuoteId(zFunc); - } - fts3Appendf(pRc, &zRet, "?"); - for(i=0; i<p->nColumn; i++){ - fts3Appendf(pRc, &zRet, ",%s(?)", zFunction); - } - if( p->zLanguageid ){ - fts3Appendf(pRc, &zRet, ", ?"); - } - sqlite3_free(zFree); - return zRet; -} - -/* -** This function interprets the string at (*pp) as a non-negative integer -** value. It reads the integer and sets *pnOut to the value read, then -** sets *pp to point to the byte immediately following the last byte of -** the integer value. -** -** Only decimal digits ('0'..'9') may be part of an integer value. -** -** If *pp does not being with a decimal digit SQLITE_ERROR is returned and -** the output value undefined. Otherwise SQLITE_OK is returned. -** -** This function is used when parsing the "prefix=" FTS4 parameter. -*/ -static int fts3GobbleInt(const char **pp, int *pnOut){ - const char *p; /* Iterator pointer */ - int nInt = 0; /* Output value */ - - for(p=*pp; p[0]>='0' && p[0]<='9'; p++){ - nInt = nInt * 10 + (p[0] - '0'); - } - if( p==*pp ) return SQLITE_ERROR; - *pnOut = nInt; - *pp = p; - return SQLITE_OK; -} - -/* -** This function is called to allocate an array of Fts3Index structures -** representing the indexes maintained by the current FTS table. FTS tables -** always maintain the main "terms" index, but may also maintain one or -** more "prefix" indexes, depending on the value of the "prefix=" parameter -** (if any) specified as part of the CREATE VIRTUAL TABLE statement. -** -** Argument zParam is passed the value of the "prefix=" option if one was -** specified, or NULL otherwise. -** -** If no error occurs, SQLITE_OK is returned and *apIndex set to point to -** the allocated array. *pnIndex is set to the number of elements in the -** array. If an error does occur, an SQLite error code is returned. -** -** Regardless of whether or not an error is returned, it is the responsibility -** of the caller to call sqlite3_free() on the output array to free it. -*/ -static int fts3PrefixParameter( - const char *zParam, /* ABC in prefix=ABC parameter to parse */ - int *pnIndex, /* OUT: size of *apIndex[] array */ - struct Fts3Index **apIndex /* OUT: Array of indexes for this table */ -){ - struct Fts3Index *aIndex; /* Allocated array */ - int nIndex = 1; /* Number of entries in array */ - - if( zParam && zParam[0] ){ - const char *p; - nIndex++; - for(p=zParam; *p; p++){ - if( *p==',' ) nIndex++; - } - } - - aIndex = sqlite3_malloc(sizeof(struct Fts3Index) * nIndex); - *apIndex = aIndex; - *pnIndex = nIndex; - if( !aIndex ){ - return SQLITE_NOMEM; - } - - memset(aIndex, 0, sizeof(struct Fts3Index) * nIndex); - if( zParam ){ - const char *p = zParam; - int i; - for(i=1; i<nIndex; i++){ - int nPrefix; - if( fts3GobbleInt(&p, &nPrefix) ) return SQLITE_ERROR; - aIndex[i].nPrefix = nPrefix; - p++; - } - } - - return SQLITE_OK; -} - -/* -** This function is called when initializing an FTS4 table that uses the -** content=xxx option. It determines the number of and names of the columns -** of the new FTS4 table. -** -** The third argument passed to this function is the value passed to the -** config=xxx option (i.e. "xxx"). This function queries the database for -** a table of that name. If found, the output variables are populated -** as follows: -** -** *pnCol: Set to the number of columns table xxx has, -** -** *pnStr: Set to the total amount of space required to store a copy -** of each columns name, including the nul-terminator. -** -** *pazCol: Set to point to an array of *pnCol strings. Each string is -** the name of the corresponding column in table xxx. The array -** and its contents are allocated using a single allocation. It -** is the responsibility of the caller to free this allocation -** by eventually passing the *pazCol value to sqlite3_free(). -** -** If the table cannot be found, an error code is returned and the output -** variables are undefined. Or, if an OOM is encountered, SQLITE_NOMEM is -** returned (and the output variables are undefined). -*/ -static int fts3ContentColumns( - sqlite3 *db, /* Database handle */ - const char *zDb, /* Name of db (i.e. "main", "temp" etc.) */ - const char *zTbl, /* Name of content table */ - const char ***pazCol, /* OUT: Malloc'd array of column names */ - int *pnCol, /* OUT: Size of array *pazCol */ - int *pnStr /* OUT: Bytes of string content */ -){ - int rc = SQLITE_OK; /* Return code */ - char *zSql; /* "SELECT *" statement on zTbl */ - sqlite3_stmt *pStmt = 0; /* Compiled version of zSql */ - - zSql = sqlite3_mprintf("SELECT * FROM %Q.%Q", zDb, zTbl); - if( !zSql ){ - rc = SQLITE_NOMEM; - }else{ - rc = sqlite3_prepare(db, zSql, -1, &pStmt, 0); - } - sqlite3_free(zSql); - - if( rc==SQLITE_OK ){ - const char **azCol; /* Output array */ - int nStr = 0; /* Size of all column names (incl. 0x00) */ - int nCol; /* Number of table columns */ - int i; /* Used to iterate through columns */ - - /* Loop through the returned columns. Set nStr to the number of bytes of - ** space required to store a copy of each column name, including the - ** nul-terminator byte. */ - nCol = sqlite3_column_count(pStmt); - for(i=0; i<nCol; i++){ - const char *zCol = sqlite3_column_name(pStmt, i); - nStr += (int)strlen(zCol) + 1; - } - - /* Allocate and populate the array to return. */ - azCol = (const char **)sqlite3_malloc(sizeof(char *) * nCol + nStr); - if( azCol==0 ){ - rc = SQLITE_NOMEM; - }else{ - char *p = (char *)&azCol[nCol]; - for(i=0; i<nCol; i++){ - const char *zCol = sqlite3_column_name(pStmt, i); - int n = (int)strlen(zCol)+1; - memcpy(p, zCol, n); - azCol[i] = p; - p += n; - } - } - sqlite3_finalize(pStmt); - - /* Set the output variables. */ - *pnCol = nCol; - *pnStr = nStr; - *pazCol = azCol; - } - - return rc; -} - -/* -** This function is the implementation of both the xConnect and xCreate -** methods of the FTS3 virtual table. -** -** The argv[] array contains the following: -** -** argv[0] -> module name ("fts3" or "fts4") -** argv[1] -> database name -** argv[2] -> table name -** argv[...] -> "column name" and other module argument fields. -*/ -static int fts3InitVtab( - int isCreate, /* True for xCreate, false for xConnect */ - sqlite3 *db, /* The SQLite database connection */ - void *pAux, /* Hash table containing tokenizers */ - int argc, /* Number of elements in argv array */ - const char * const *argv, /* xCreate/xConnect argument array */ - sqlite3_vtab **ppVTab, /* Write the resulting vtab structure here */ - char **pzErr /* Write any error message here */ -){ - Fts3Hash *pHash = (Fts3Hash *)pAux; - Fts3Table *p = 0; /* Pointer to allocated vtab */ - int rc = SQLITE_OK; /* Return code */ - int i; /* Iterator variable */ - int nByte; /* Size of allocation used for *p */ - int iCol; /* Column index */ - int nString = 0; /* Bytes required to hold all column names */ - int nCol = 0; /* Number of columns in the FTS table */ - char *zCsr; /* Space for holding column names */ - int nDb; /* Bytes required to hold database name */ - int nName; /* Bytes required to hold table name */ - int isFts4 = (argv[0][3]=='4'); /* True for FTS4, false for FTS3 */ - const char **aCol; /* Array of column names */ - sqlite3_tokenizer *pTokenizer = 0; /* Tokenizer for this table */ - - int nIndex; /* Size of aIndex[] array */ - struct Fts3Index *aIndex = 0; /* Array of indexes for this table */ - - /* The results of parsing supported FTS4 key=value options: */ - int bNoDocsize = 0; /* True to omit %_docsize table */ - int bDescIdx = 0; /* True to store descending indexes */ - char *zPrefix = 0; /* Prefix parameter value (or NULL) */ - char *zCompress = 0; /* compress=? parameter (or NULL) */ - char *zUncompress = 0; /* uncompress=? parameter (or NULL) */ - char *zContent = 0; /* content=? parameter (or NULL) */ - char *zLanguageid = 0; /* languageid=? parameter (or NULL) */ - - assert( strlen(argv[0])==4 ); - assert( (sqlite3_strnicmp(argv[0], "fts4", 4)==0 && isFts4) - || (sqlite3_strnicmp(argv[0], "fts3", 4)==0 && !isFts4) - ); - - nDb = (int)strlen(argv[1]) + 1; - nName = (int)strlen(argv[2]) + 1; - - aCol = (const char **)sqlite3_malloc(sizeof(const char *) * (argc-2) ); - if( !aCol ) return SQLITE_NOMEM; - memset((void *)aCol, 0, sizeof(const char *) * (argc-2)); - - /* Loop through all of the arguments passed by the user to the FTS3/4 - ** module (i.e. all the column names and special arguments). This loop - ** does the following: - ** - ** + Figures out the number of columns the FTSX table will have, and - ** the number of bytes of space that must be allocated to store copies - ** of the column names. - ** - ** + If there is a tokenizer specification included in the arguments, - ** initializes the tokenizer pTokenizer. - */ - for(i=3; rc==SQLITE_OK && i<argc; i++){ - char const *z = argv[i]; - int nKey; - char *zVal; - - /* Check if this is a tokenizer specification */ - if( !pTokenizer - && strlen(z)>8 - && 0==sqlite3_strnicmp(z, "tokenize", 8) - && 0==sqlite3Fts3IsIdChar(z[8]) - ){ - rc = sqlite3Fts3InitTokenizer(pHash, &z[9], &pTokenizer, pzErr); - } - - /* Check if it is an FTS4 special argument. */ - else if( isFts4 && fts3IsSpecialColumn(z, &nKey, &zVal) ){ - struct Fts4Option { - const char *zOpt; - int nOpt; - } aFts4Opt[] = { - { "matchinfo", 9 }, /* 0 -> MATCHINFO */ - { "prefix", 6 }, /* 1 -> PREFIX */ - { "compress", 8 }, /* 2 -> COMPRESS */ - { "uncompress", 10 }, /* 3 -> UNCOMPRESS */ - { "order", 5 }, /* 4 -> ORDER */ - { "content", 7 }, /* 5 -> CONTENT */ - { "languageid", 10 } /* 6 -> LANGUAGEID */ - }; - - int iOpt; - if( !zVal ){ - rc = SQLITE_NOMEM; - }else{ - for(iOpt=0; iOpt<SizeofArray(aFts4Opt); iOpt++){ - struct Fts4Option *pOp = &aFts4Opt[iOpt]; - if( nKey==pOp->nOpt && !sqlite3_strnicmp(z, pOp->zOpt, pOp->nOpt) ){ - break; - } - } - if( iOpt==SizeofArray(aFts4Opt) ){ - *pzErr = sqlite3_mprintf("unrecognized parameter: %s", z); - rc = SQLITE_ERROR; - }else{ - switch( iOpt ){ - case 0: /* MATCHINFO */ - if( strlen(zVal)!=4 || sqlite3_strnicmp(zVal, "fts3", 4) ){ - *pzErr = sqlite3_mprintf("unrecognized matchinfo: %s", zVal); - rc = SQLITE_ERROR; - } - bNoDocsize = 1; - break; - - case 1: /* PREFIX */ - sqlite3_free(zPrefix); - zPrefix = zVal; - zVal = 0; - break; - - case 2: /* COMPRESS */ - sqlite3_free(zCompress); - zCompress = zVal; - zVal = 0; - break; - - case 3: /* UNCOMPRESS */ - sqlite3_free(zUncompress); - zUncompress = zVal; - zVal = 0; - break; - - case 4: /* ORDER */ - if( (strlen(zVal)!=3 || sqlite3_strnicmp(zVal, "asc", 3)) - && (strlen(zVal)!=4 || sqlite3_strnicmp(zVal, "desc", 4)) - ){ - *pzErr = sqlite3_mprintf("unrecognized order: %s", zVal); - rc = SQLITE_ERROR; - } - bDescIdx = (zVal[0]=='d' || zVal[0]=='D'); - break; - - case 5: /* CONTENT */ - sqlite3_free(zContent); - zContent = zVal; - zVal = 0; - break; - - case 6: /* LANGUAGEID */ - assert( iOpt==6 ); - sqlite3_free(zLanguageid); - zLanguageid = zVal; - zVal = 0; - break; - } - } - sqlite3_free(zVal); - } - } - - /* Otherwise, the argument is a column name. */ - else { - nString += (int)(strlen(z) + 1); - aCol[nCol++] = z; - } - } - - /* If a content=xxx option was specified, the following: - ** - ** 1. Ignore any compress= and uncompress= options. - ** - ** 2. If no column names were specified as part of the CREATE VIRTUAL - ** TABLE statement, use all columns from the content table. - */ - if( rc==SQLITE_OK && zContent ){ - sqlite3_free(zCompress); - sqlite3_free(zUncompress); - zCompress = 0; - zUncompress = 0; - if( nCol==0 ){ - sqlite3_free((void*)aCol); - aCol = 0; - rc = fts3ContentColumns(db, argv[1], zContent, &aCol, &nCol, &nString); - - /* If a languageid= option was specified, remove the language id - ** column from the aCol[] array. */ - if( rc==SQLITE_OK && zLanguageid ){ - int j; - for(j=0; j<nCol; j++){ - if( sqlite3_stricmp(zLanguageid, aCol[j])==0 ){ - int k; - for(k=j; k<nCol; k++) aCol[k] = aCol[k+1]; - nCol--; - break; - } - } - } - } - } - if( rc!=SQLITE_OK ) goto fts3_init_out; - - if( nCol==0 ){ - assert( nString==0 ); - aCol[0] = "content"; - nString = 8; - nCol = 1; - } - - if( pTokenizer==0 ){ - rc = sqlite3Fts3InitTokenizer(pHash, "simple", &pTokenizer, pzErr); - if( rc!=SQLITE_OK ) goto fts3_init_out; - } - assert( pTokenizer ); - - rc = fts3PrefixParameter(zPrefix, &nIndex, &aIndex); - if( rc==SQLITE_ERROR ){ - assert( zPrefix ); - *pzErr = sqlite3_mprintf("error parsing prefix parameter: %s", zPrefix); - } - if( rc!=SQLITE_OK ) goto fts3_init_out; - - /* Allocate and populate the Fts3Table structure. */ - nByte = sizeof(Fts3Table) + /* Fts3Table */ - nCol * sizeof(char *) + /* azColumn */ - nIndex * sizeof(struct Fts3Index) + /* aIndex */ - nName + /* zName */ - nDb + /* zDb */ - nString; /* Space for azColumn strings */ - p = (Fts3Table*)sqlite3_malloc(nByte); - if( p==0 ){ - rc = SQLITE_NOMEM; - goto fts3_init_out; - } - memset(p, 0, nByte); - p->db = db; - p->nColumn = nCol; - p->nPendingData = 0; - p->azColumn = (char **)&p[1]; - p->pTokenizer = pTokenizer; - p->nMaxPendingData = FTS3_MAX_PENDING_DATA; - p->bHasDocsize = (isFts4 && bNoDocsize==0); - p->bHasStat = isFts4; - p->bFts4 = isFts4; - p->bDescIdx = bDescIdx; - p->bAutoincrmerge = 0xff; /* 0xff means setting unknown */ - p->zContentTbl = zContent; - p->zLanguageid = zLanguageid; - zContent = 0; - zLanguageid = 0; - TESTONLY( p->inTransaction = -1 ); - TESTONLY( p->mxSavepoint = -1 ); - - p->aIndex = (struct Fts3Index *)&p->azColumn[nCol]; - memcpy(p->aIndex, aIndex, sizeof(struct Fts3Index) * nIndex); - p->nIndex = nIndex; - for(i=0; i<nIndex; i++){ - fts3HashInit(&p->aIndex[i].hPending, FTS3_HASH_STRING, 1); - } - - /* Fill in the zName and zDb fields of the vtab structure. */ - zCsr = (char *)&p->aIndex[nIndex]; - p->zName = zCsr; - memcpy(zCsr, argv[2], nName); - zCsr += nName; - p->zDb = zCsr; - memcpy(zCsr, argv[1], nDb); - zCsr += nDb; - - /* Fill in the azColumn array */ - for(iCol=0; iCol<nCol; iCol++){ - char *z; - int n = 0; - z = (char *)sqlite3Fts3NextToken(aCol[iCol], &n); - memcpy(zCsr, z, n); - zCsr[n] = '\0'; - sqlite3Fts3Dequote(zCsr); - p->azColumn[iCol] = zCsr; - zCsr += n+1; - assert( zCsr <= &((char *)p)[nByte] ); - } - - if( (zCompress==0)!=(zUncompress==0) ){ - char const *zMiss = (zCompress==0 ? "compress" : "uncompress"); - rc = SQLITE_ERROR; - *pzErr = sqlite3_mprintf("missing %s parameter in fts4 constructor", zMiss); - } - p->zReadExprlist = fts3ReadExprList(p, zUncompress, &rc); - p->zWriteExprlist = fts3WriteExprList(p, zCompress, &rc); - if( rc!=SQLITE_OK ) goto fts3_init_out; - - /* If this is an xCreate call, create the underlying tables in the - ** database. TODO: For xConnect(), it could verify that said tables exist. - */ - if( isCreate ){ - rc = fts3CreateTables(p); - } - - /* Check to see if a legacy fts3 table has been "upgraded" by the - ** addition of a %_stat table so that it can use incremental merge. - */ - if( !isFts4 && !isCreate ){ - int rc2 = SQLITE_OK; - fts3DbExec(&rc2, db, "SELECT 1 FROM %Q.'%q_stat' WHERE id=2", - p->zDb, p->zName); - if( rc2==SQLITE_OK ) p->bHasStat = 1; - } - - /* Figure out the page-size for the database. This is required in order to - ** estimate the cost of loading large doclists from the database. */ - fts3DatabasePageSize(&rc, p); - p->nNodeSize = p->nPgsz-35; - - /* Declare the table schema to SQLite. */ - fts3DeclareVtab(&rc, p); - -fts3_init_out: - sqlite3_free(zPrefix); - sqlite3_free(aIndex); - sqlite3_free(zCompress); - sqlite3_free(zUncompress); - sqlite3_free(zContent); - sqlite3_free(zLanguageid); - sqlite3_free((void *)aCol); - if( rc!=SQLITE_OK ){ - if( p ){ - fts3DisconnectMethod((sqlite3_vtab *)p); - }else if( pTokenizer ){ - pTokenizer->pModule->xDestroy(pTokenizer); - } - }else{ - assert( p->pSegments==0 ); - *ppVTab = &p->base; - } - return rc; -} - -/* -** The xConnect() and xCreate() methods for the virtual table. All the -** work is done in function fts3InitVtab(). -*/ -static int fts3ConnectMethod( - sqlite3 *db, /* Database connection */ - void *pAux, /* Pointer to tokenizer hash table */ - int argc, /* Number of elements in argv array */ - const char * const *argv, /* xCreate/xConnect argument array */ - sqlite3_vtab **ppVtab, /* OUT: New sqlite3_vtab object */ - char **pzErr /* OUT: sqlite3_malloc'd error message */ -){ - return fts3InitVtab(0, db, pAux, argc, argv, ppVtab, pzErr); -} -static int fts3CreateMethod( - sqlite3 *db, /* Database connection */ - void *pAux, /* Pointer to tokenizer hash table */ - int argc, /* Number of elements in argv array */ - const char * const *argv, /* xCreate/xConnect argument array */ - sqlite3_vtab **ppVtab, /* OUT: New sqlite3_vtab object */ - char **pzErr /* OUT: sqlite3_malloc'd error message */ -){ - return fts3InitVtab(1, db, pAux, argc, argv, ppVtab, pzErr); -} - -/* -** Implementation of the xBestIndex method for FTS3 tables. There -** are three possible strategies, in order of preference: -** -** 1. Direct lookup by rowid or docid. -** 2. Full-text search using a MATCH operator on a non-docid column. -** 3. Linear scan of %_content table. -*/ -static int fts3BestIndexMethod(sqlite3_vtab *pVTab, sqlite3_index_info *pInfo){ - Fts3Table *p = (Fts3Table *)pVTab; - int i; /* Iterator variable */ - int iCons = -1; /* Index of constraint to use */ - int iLangidCons = -1; /* Index of langid=x constraint, if present */ - - /* By default use a full table scan. This is an expensive option, - ** so search through the constraints to see if a more efficient - ** strategy is possible. - */ - pInfo->idxNum = FTS3_FULLSCAN_SEARCH; - pInfo->estimatedCost = 500000; - for(i=0; i<pInfo->nConstraint; i++){ - struct sqlite3_index_constraint *pCons = &pInfo->aConstraint[i]; - if( pCons->usable==0 ) continue; - - /* A direct lookup on the rowid or docid column. Assign a cost of 1.0. */ - if( iCons<0 - && pCons->op==SQLITE_INDEX_CONSTRAINT_EQ - && (pCons->iColumn<0 || pCons->iColumn==p->nColumn+1 ) - ){ - pInfo->idxNum = FTS3_DOCID_SEARCH; - pInfo->estimatedCost = 1.0; - iCons = i; - } - - /* A MATCH constraint. Use a full-text search. - ** - ** If there is more than one MATCH constraint available, use the first - ** one encountered. If there is both a MATCH constraint and a direct - ** rowid/docid lookup, prefer the MATCH strategy. This is done even - ** though the rowid/docid lookup is faster than a MATCH query, selecting - ** it would lead to an "unable to use function MATCH in the requested - ** context" error. - */ - if( pCons->op==SQLITE_INDEX_CONSTRAINT_MATCH - && pCons->iColumn>=0 && pCons->iColumn<=p->nColumn - ){ - pInfo->idxNum = FTS3_FULLTEXT_SEARCH + pCons->iColumn; - pInfo->estimatedCost = 2.0; - iCons = i; - } - - /* Equality constraint on the langid column */ - if( pCons->op==SQLITE_INDEX_CONSTRAINT_EQ - && pCons->iColumn==p->nColumn + 2 - ){ - iLangidCons = i; - } - } - - if( iCons>=0 ){ - pInfo->aConstraintUsage[iCons].argvIndex = 1; - pInfo->aConstraintUsage[iCons].omit = 1; - } - if( iLangidCons>=0 ){ - pInfo->aConstraintUsage[iLangidCons].argvIndex = 2; - } - - /* Regardless of the strategy selected, FTS can deliver rows in rowid (or - ** docid) order. Both ascending and descending are possible. - */ - if( pInfo->nOrderBy==1 ){ - struct sqlite3_index_orderby *pOrder = &pInfo->aOrderBy[0]; - if( pOrder->iColumn<0 || pOrder->iColumn==p->nColumn+1 ){ - if( pOrder->desc ){ - pInfo->idxStr = "DESC"; - }else{ - pInfo->idxStr = "ASC"; - } - pInfo->orderByConsumed = 1; - } - } - - assert( p->pSegments==0 ); - return SQLITE_OK; -} - -/* -** Implementation of xOpen method. -*/ -static int fts3OpenMethod(sqlite3_vtab *pVTab, sqlite3_vtab_cursor **ppCsr){ - sqlite3_vtab_cursor *pCsr; /* Allocated cursor */ - - UNUSED_PARAMETER(pVTab); - - /* Allocate a buffer large enough for an Fts3Cursor structure. If the - ** allocation succeeds, zero it and return SQLITE_OK. Otherwise, - ** if the allocation fails, return SQLITE_NOMEM. - */ - *ppCsr = pCsr = (sqlite3_vtab_cursor *)sqlite3_malloc(sizeof(Fts3Cursor)); - if( !pCsr ){ - return SQLITE_NOMEM; - } - memset(pCsr, 0, sizeof(Fts3Cursor)); - return SQLITE_OK; -} - -/* -** Close the cursor. For additional information see the documentation -** on the xClose method of the virtual table interface. -*/ -static int fts3CloseMethod(sqlite3_vtab_cursor *pCursor){ - Fts3Cursor *pCsr = (Fts3Cursor *)pCursor; - assert( ((Fts3Table *)pCsr->base.pVtab)->pSegments==0 ); - sqlite3_finalize(pCsr->pStmt); - sqlite3Fts3ExprFree(pCsr->pExpr); - sqlite3Fts3FreeDeferredTokens(pCsr); - sqlite3_free(pCsr->aDoclist); - sqlite3_free(pCsr->aMatchinfo); - assert( ((Fts3Table *)pCsr->base.pVtab)->pSegments==0 ); - sqlite3_free(pCsr); - return SQLITE_OK; -} - -/* -** If pCsr->pStmt has not been prepared (i.e. if pCsr->pStmt==0), then -** compose and prepare an SQL statement of the form: -** -** "SELECT <columns> FROM %_content WHERE rowid = ?" -** -** (or the equivalent for a content=xxx table) and set pCsr->pStmt to -** it. If an error occurs, return an SQLite error code. -** -** Otherwise, set *ppStmt to point to pCsr->pStmt and return SQLITE_OK. -*/ -static int fts3CursorSeekStmt(Fts3Cursor *pCsr, sqlite3_stmt **ppStmt){ - int rc = SQLITE_OK; - if( pCsr->pStmt==0 ){ - Fts3Table *p = (Fts3Table *)pCsr->base.pVtab; - char *zSql; - zSql = sqlite3_mprintf("SELECT %s WHERE rowid = ?", p->zReadExprlist); - if( !zSql ) return SQLITE_NOMEM; - rc = sqlite3_prepare_v2(p->db, zSql, -1, &pCsr->pStmt, 0); - sqlite3_free(zSql); - } - *ppStmt = pCsr->pStmt; - return rc; -} - -/* -** Position the pCsr->pStmt statement so that it is on the row -** of the %_content table that contains the last match. Return -** SQLITE_OK on success. -*/ -static int fts3CursorSeek(sqlite3_context *pContext, Fts3Cursor *pCsr){ - int rc = SQLITE_OK; - if( pCsr->isRequireSeek ){ - sqlite3_stmt *pStmt = 0; - - rc = fts3CursorSeekStmt(pCsr, &pStmt); - if( rc==SQLITE_OK ){ - sqlite3_bind_int64(pCsr->pStmt, 1, pCsr->iPrevId); - pCsr->isRequireSeek = 0; - if( SQLITE_ROW==sqlite3_step(pCsr->pStmt) ){ - return SQLITE_OK; - }else{ - rc = sqlite3_reset(pCsr->pStmt); - if( rc==SQLITE_OK && ((Fts3Table *)pCsr->base.pVtab)->zContentTbl==0 ){ - /* If no row was found and no error has occured, then the %_content - ** table is missing a row that is present in the full-text index. - ** The data structures are corrupt. */ - rc = FTS_CORRUPT_VTAB; - pCsr->isEof = 1; - } - } - } - } - - if( rc!=SQLITE_OK && pContext ){ - sqlite3_result_error_code(pContext, rc); - } - return rc; -} - -/* -** This function is used to process a single interior node when searching -** a b-tree for a term or term prefix. The node data is passed to this -** function via the zNode/nNode parameters. The term to search for is -** passed in zTerm/nTerm. -** -** If piFirst is not NULL, then this function sets *piFirst to the blockid -** of the child node that heads the sub-tree that may contain the term. -** -** If piLast is not NULL, then *piLast is set to the right-most child node -** that heads a sub-tree that may contain a term for which zTerm/nTerm is -** a prefix. -** -** If an OOM error occurs, SQLITE_NOMEM is returned. Otherwise, SQLITE_OK. -*/ -static int fts3ScanInteriorNode( - const char *zTerm, /* Term to select leaves for */ - int nTerm, /* Size of term zTerm in bytes */ - const char *zNode, /* Buffer containing segment interior node */ - int nNode, /* Size of buffer at zNode */ - sqlite3_int64 *piFirst, /* OUT: Selected child node */ - sqlite3_int64 *piLast /* OUT: Selected child node */ -){ - int rc = SQLITE_OK; /* Return code */ - const char *zCsr = zNode; /* Cursor to iterate through node */ - const char *zEnd = &zCsr[nNode];/* End of interior node buffer */ - char *zBuffer = 0; /* Buffer to load terms into */ - int nAlloc = 0; /* Size of allocated buffer */ - int isFirstTerm = 1; /* True when processing first term on page */ - sqlite3_int64 iChild; /* Block id of child node to descend to */ - - /* Skip over the 'height' varint that occurs at the start of every - ** interior node. Then load the blockid of the left-child of the b-tree - ** node into variable iChild. - ** - ** Even if the data structure on disk is corrupted, this (reading two - ** varints from the buffer) does not risk an overread. If zNode is a - ** root node, then the buffer comes from a SELECT statement. SQLite does - ** not make this guarantee explicitly, but in practice there are always - ** either more than 20 bytes of allocated space following the nNode bytes of - ** contents, or two zero bytes. Or, if the node is read from the %_segments - ** table, then there are always 20 bytes of zeroed padding following the - ** nNode bytes of content (see sqlite3Fts3ReadBlock() for details). - */ - zCsr += sqlite3Fts3GetVarint(zCsr, &iChild); - zCsr += sqlite3Fts3GetVarint(zCsr, &iChild); - if( zCsr>zEnd ){ - return FTS_CORRUPT_VTAB; - } - - while( zCsr<zEnd && (piFirst || piLast) ){ - int cmp; /* memcmp() result */ - int nSuffix; /* Size of term suffix */ - int nPrefix = 0; /* Size of term prefix */ - int nBuffer; /* Total term size */ - - /* Load the next term on the node into zBuffer. Use realloc() to expand - ** the size of zBuffer if required. */ - if( !isFirstTerm ){ - zCsr += sqlite3Fts3GetVarint32(zCsr, &nPrefix); - } - isFirstTerm = 0; - zCsr += sqlite3Fts3GetVarint32(zCsr, &nSuffix); - - if( nPrefix<0 || nSuffix<0 || &zCsr[nSuffix]>zEnd ){ - rc = FTS_CORRUPT_VTAB; - goto finish_scan; - } - if( nPrefix+nSuffix>nAlloc ){ - char *zNew; - nAlloc = (nPrefix+nSuffix) * 2; - zNew = (char *)sqlite3_realloc(zBuffer, nAlloc); - if( !zNew ){ - rc = SQLITE_NOMEM; - goto finish_scan; - } - zBuffer = zNew; - } - assert( zBuffer ); - memcpy(&zBuffer[nPrefix], zCsr, nSuffix); - nBuffer = nPrefix + nSuffix; - zCsr += nSuffix; - - /* Compare the term we are searching for with the term just loaded from - ** the interior node. If the specified term is greater than or equal - ** to the term from the interior node, then all terms on the sub-tree - ** headed by node iChild are smaller than zTerm. No need to search - ** iChild. - ** - ** If the interior node term is larger than the specified term, then - ** the tree headed by iChild may contain the specified term. - */ - cmp = memcmp(zTerm, zBuffer, (nBuffer>nTerm ? nTerm : nBuffer)); - if( piFirst && (cmp<0 || (cmp==0 && nBuffer>nTerm)) ){ - *piFirst = iChild; - piFirst = 0; - } - - if( piLast && cmp<0 ){ - *piLast = iChild; - piLast = 0; - } - - iChild++; - }; - - if( piFirst ) *piFirst = iChild; - if( piLast ) *piLast = iChild; - - finish_scan: - sqlite3_free(zBuffer); - return rc; -} - - -/* -** The buffer pointed to by argument zNode (size nNode bytes) contains an -** interior node of a b-tree segment. The zTerm buffer (size nTerm bytes) -** contains a term. This function searches the sub-tree headed by the zNode -** node for the range of leaf nodes that may contain the specified term -** or terms for which the specified term is a prefix. -** -** If piLeaf is not NULL, then *piLeaf is set to the blockid of the -** left-most leaf node in the tree that may contain the specified term. -** If piLeaf2 is not NULL, then *piLeaf2 is set to the blockid of the -** right-most leaf node that may contain a term for which the specified -** term is a prefix. -** -** It is possible that the range of returned leaf nodes does not contain -** the specified term or any terms for which it is a prefix. However, if the -** segment does contain any such terms, they are stored within the identified -** range. Because this function only inspects interior segment nodes (and -** never loads leaf nodes into memory), it is not possible to be sure. -** -** If an error occurs, an error code other than SQLITE_OK is returned. -*/ -static int fts3SelectLeaf( - Fts3Table *p, /* Virtual table handle */ - const char *zTerm, /* Term to select leaves for */ - int nTerm, /* Size of term zTerm in bytes */ - const char *zNode, /* Buffer containing segment interior node */ - int nNode, /* Size of buffer at zNode */ - sqlite3_int64 *piLeaf, /* Selected leaf node */ - sqlite3_int64 *piLeaf2 /* Selected leaf node */ -){ - int rc; /* Return code */ - int iHeight; /* Height of this node in tree */ - - assert( piLeaf || piLeaf2 ); - - sqlite3Fts3GetVarint32(zNode, &iHeight); - rc = fts3ScanInteriorNode(zTerm, nTerm, zNode, nNode, piLeaf, piLeaf2); - assert( !piLeaf2 || !piLeaf || rc!=SQLITE_OK || (*piLeaf<=*piLeaf2) ); - - if( rc==SQLITE_OK && iHeight>1 ){ - char *zBlob = 0; /* Blob read from %_segments table */ - int nBlob; /* Size of zBlob in bytes */ - - if( piLeaf && piLeaf2 && (*piLeaf!=*piLeaf2) ){ - rc = sqlite3Fts3ReadBlock(p, *piLeaf, &zBlob, &nBlob, 0); - if( rc==SQLITE_OK ){ - rc = fts3SelectLeaf(p, zTerm, nTerm, zBlob, nBlob, piLeaf, 0); - } - sqlite3_free(zBlob); - piLeaf = 0; - zBlob = 0; - } - - if( rc==SQLITE_OK ){ - rc = sqlite3Fts3ReadBlock(p, piLeaf?*piLeaf:*piLeaf2, &zBlob, &nBlob, 0); - } - if( rc==SQLITE_OK ){ - rc = fts3SelectLeaf(p, zTerm, nTerm, zBlob, nBlob, piLeaf, piLeaf2); - } - sqlite3_free(zBlob); - } - - return rc; -} - -/* -** This function is used to create delta-encoded serialized lists of FTS3 -** varints. Each call to this function appends a single varint to a list. -*/ -static void fts3PutDeltaVarint( - char **pp, /* IN/OUT: Output pointer */ - sqlite3_int64 *piPrev, /* IN/OUT: Previous value written to list */ - sqlite3_int64 iVal /* Write this value to the list */ -){ - assert( iVal-*piPrev > 0 || (*piPrev==0 && iVal==0) ); - *pp += sqlite3Fts3PutVarint(*pp, iVal-*piPrev); - *piPrev = iVal; -} - -/* -** When this function is called, *ppPoslist is assumed to point to the -** start of a position-list. After it returns, *ppPoslist points to the -** first byte after the position-list. -** -** A position list is list of positions (delta encoded) and columns for -** a single document record of a doclist. So, in other words, this -** routine advances *ppPoslist so that it points to the next docid in -** the doclist, or to the first byte past the end of the doclist. -** -** If pp is not NULL, then the contents of the position list are copied -** to *pp. *pp is set to point to the first byte past the last byte copied -** before this function returns. -*/ -static void fts3PoslistCopy(char **pp, char **ppPoslist){ - char *pEnd = *ppPoslist; - char c = 0; - - /* The end of a position list is marked by a zero encoded as an FTS3 - ** varint. A single POS_END (0) byte. Except, if the 0 byte is preceded by - ** a byte with the 0x80 bit set, then it is not a varint 0, but the tail - ** of some other, multi-byte, value. - ** - ** The following while-loop moves pEnd to point to the first byte that is not - ** immediately preceded by a byte with the 0x80 bit set. Then increments - ** pEnd once more so that it points to the byte immediately following the - ** last byte in the position-list. - */ - while( *pEnd | c ){ - c = *pEnd++ & 0x80; - testcase( c!=0 && (*pEnd)==0 ); - } - pEnd++; /* Advance past the POS_END terminator byte */ - - if( pp ){ - int n = (int)(pEnd - *ppPoslist); - char *p = *pp; - memcpy(p, *ppPoslist, n); - p += n; - *pp = p; - } - *ppPoslist = pEnd; -} - -/* -** When this function is called, *ppPoslist is assumed to point to the -** start of a column-list. After it returns, *ppPoslist points to the -** to the terminator (POS_COLUMN or POS_END) byte of the column-list. -** -** A column-list is list of delta-encoded positions for a single column -** within a single document within a doclist. -** -** The column-list is terminated either by a POS_COLUMN varint (1) or -** a POS_END varint (0). This routine leaves *ppPoslist pointing to -** the POS_COLUMN or POS_END that terminates the column-list. -** -** If pp is not NULL, then the contents of the column-list are copied -** to *pp. *pp is set to point to the first byte past the last byte copied -** before this function returns. The POS_COLUMN or POS_END terminator -** is not copied into *pp. -*/ -static void fts3ColumnlistCopy(char **pp, char **ppPoslist){ - char *pEnd = *ppPoslist; - char c = 0; - - /* A column-list is terminated by either a 0x01 or 0x00 byte that is - ** not part of a multi-byte varint. - */ - while( 0xFE & (*pEnd | c) ){ - c = *pEnd++ & 0x80; - testcase( c!=0 && ((*pEnd)&0xfe)==0 ); - } - if( pp ){ - int n = (int)(pEnd - *ppPoslist); - char *p = *pp; - memcpy(p, *ppPoslist, n); - p += n; - *pp = p; - } - *ppPoslist = pEnd; -} - -/* -** Value used to signify the end of an position-list. This is safe because -** it is not possible to have a document with 2^31 terms. -*/ -#define POSITION_LIST_END 0x7fffffff - -/* -** This function is used to help parse position-lists. When this function is -** called, *pp may point to the start of the next varint in the position-list -** being parsed, or it may point to 1 byte past the end of the position-list -** (in which case **pp will be a terminator bytes POS_END (0) or -** (1)). -** -** If *pp points past the end of the current position-list, set *pi to -** POSITION_LIST_END and return. Otherwise, read the next varint from *pp, -** increment the current value of *pi by the value read, and set *pp to -** point to the next value before returning. -** -** Before calling this routine *pi must be initialized to the value of -** the previous position, or zero if we are reading the first position -** in the position-list. Because positions are delta-encoded, the value -** of the previous position is needed in order to compute the value of -** the next position. -*/ -static void fts3ReadNextPos( - char **pp, /* IN/OUT: Pointer into position-list buffer */ - sqlite3_int64 *pi /* IN/OUT: Value read from position-list */ -){ - if( (**pp)&0xFE ){ - fts3GetDeltaVarint(pp, pi); - *pi -= 2; - }else{ - *pi = POSITION_LIST_END; - } -} - -/* -** If parameter iCol is not 0, write an POS_COLUMN (1) byte followed by -** the value of iCol encoded as a varint to *pp. This will start a new -** column list. -** -** Set *pp to point to the byte just after the last byte written before -** returning (do not modify it if iCol==0). Return the total number of bytes -** written (0 if iCol==0). -*/ -static int fts3PutColNumber(char **pp, int iCol){ - int n = 0; /* Number of bytes written */ - if( iCol ){ - char *p = *pp; /* Output pointer */ - n = 1 + sqlite3Fts3PutVarint(&p[1], iCol); - *p = 0x01; - *pp = &p[n]; - } - return n; -} - -/* -** Compute the union of two position lists. The output written -** into *pp contains all positions of both *pp1 and *pp2 in sorted -** order and with any duplicates removed. All pointers are -** updated appropriately. The caller is responsible for insuring -** that there is enough space in *pp to hold the complete output. -*/ -static void fts3PoslistMerge( - char **pp, /* Output buffer */ - char **pp1, /* Left input list */ - char **pp2 /* Right input list */ -){ - char *p = *pp; - char *p1 = *pp1; - char *p2 = *pp2; - - while( *p1 || *p2 ){ - int iCol1; /* The current column index in pp1 */ - int iCol2; /* The current column index in pp2 */ - - if( *p1==POS_COLUMN ) sqlite3Fts3GetVarint32(&p1[1], &iCol1); - else if( *p1==POS_END ) iCol1 = POSITION_LIST_END; - else iCol1 = 0; - - if( *p2==POS_COLUMN ) sqlite3Fts3GetVarint32(&p2[1], &iCol2); - else if( *p2==POS_END ) iCol2 = POSITION_LIST_END; - else iCol2 = 0; - - if( iCol1==iCol2 ){ - sqlite3_int64 i1 = 0; /* Last position from pp1 */ - sqlite3_int64 i2 = 0; /* Last position from pp2 */ - sqlite3_int64 iPrev = 0; - int n = fts3PutColNumber(&p, iCol1); - p1 += n; - p2 += n; - - /* At this point, both p1 and p2 point to the start of column-lists - ** for the same column (the column with index iCol1 and iCol2). - ** A column-list is a list of non-negative delta-encoded varints, each - ** incremented by 2 before being stored. Each list is terminated by a - ** POS_END (0) or POS_COLUMN (1). The following block merges the two lists - ** and writes the results to buffer p. p is left pointing to the byte - ** after the list written. No terminator (POS_END or POS_COLUMN) is - ** written to the output. - */ - fts3GetDeltaVarint(&p1, &i1); - fts3GetDeltaVarint(&p2, &i2); - do { - fts3PutDeltaVarint(&p, &iPrev, (i1<i2) ? i1 : i2); - iPrev -= 2; - if( i1==i2 ){ - fts3ReadNextPos(&p1, &i1); - fts3ReadNextPos(&p2, &i2); - }else if( i1<i2 ){ - fts3ReadNextPos(&p1, &i1); - }else{ - fts3ReadNextPos(&p2, &i2); - } - }while( i1!=POSITION_LIST_END || i2!=POSITION_LIST_END ); - }else if( iCol1<iCol2 ){ - p1 += fts3PutColNumber(&p, iCol1); - fts3ColumnlistCopy(&p, &p1); - }else{ - p2 += fts3PutColNumber(&p, iCol2); - fts3ColumnlistCopy(&p, &p2); - } - } - - *p++ = POS_END; - *pp = p; - *pp1 = p1 + 1; - *pp2 = p2 + 1; -} - -/* -** This function is used to merge two position lists into one. When it is -** called, *pp1 and *pp2 must both point to position lists. A position-list is -** the part of a doclist that follows each document id. For example, if a row -** contains: -** -** 'a b c'|'x y z'|'a b b a' -** -** Then the position list for this row for token 'b' would consist of: -** -** 0x02 0x01 0x02 0x03 0x03 0x00 -** -** When this function returns, both *pp1 and *pp2 are left pointing to the -** byte following the 0x00 terminator of their respective position lists. -** -** If isSaveLeft is 0, an entry is added to the output position list for -** each position in *pp2 for which there exists one or more positions in -** *pp1 so that (pos(*pp2)>pos(*pp1) && pos(*pp2)-pos(*pp1)<=nToken). i.e. -** when the *pp1 token appears before the *pp2 token, but not more than nToken -** slots before it. -** -** e.g. nToken==1 searches for adjacent positions. -*/ -static int fts3PoslistPhraseMerge( - char **pp, /* IN/OUT: Preallocated output buffer */ - int nToken, /* Maximum difference in token positions */ - int isSaveLeft, /* Save the left position */ - int isExact, /* If *pp1 is exactly nTokens before *pp2 */ - char **pp1, /* IN/OUT: Left input list */ - char **pp2 /* IN/OUT: Right input list */ -){ - char *p = *pp; - char *p1 = *pp1; - char *p2 = *pp2; - int iCol1 = 0; - int iCol2 = 0; - - /* Never set both isSaveLeft and isExact for the same invocation. */ - assert( isSaveLeft==0 || isExact==0 ); - - assert( p!=0 && *p1!=0 && *p2!=0 ); - if( *p1==POS_COLUMN ){ - p1++; - p1 += sqlite3Fts3GetVarint32(p1, &iCol1); - } - if( *p2==POS_COLUMN ){ - p2++; - p2 += sqlite3Fts3GetVarint32(p2, &iCol2); - } - - while( 1 ){ - if( iCol1==iCol2 ){ - char *pSave = p; - sqlite3_int64 iPrev = 0; - sqlite3_int64 iPos1 = 0; - sqlite3_int64 iPos2 = 0; - - if( iCol1 ){ - *p++ = POS_COLUMN; - p += sqlite3Fts3PutVarint(p, iCol1); - } - - assert( *p1!=POS_END && *p1!=POS_COLUMN ); - assert( *p2!=POS_END && *p2!=POS_COLUMN ); - fts3GetDeltaVarint(&p1, &iPos1); iPos1 -= 2; - fts3GetDeltaVarint(&p2, &iPos2); iPos2 -= 2; - - while( 1 ){ - if( iPos2==iPos1+nToken - || (isExact==0 && iPos2>iPos1 && iPos2<=iPos1+nToken) - ){ - sqlite3_int64 iSave; - iSave = isSaveLeft ? iPos1 : iPos2; - fts3PutDeltaVarint(&p, &iPrev, iSave+2); iPrev -= 2; - pSave = 0; - assert( p ); - } - if( (!isSaveLeft && iPos2<=(iPos1+nToken)) || iPos2<=iPos1 ){ - if( (*p2&0xFE)==0 ) break; - fts3GetDeltaVarint(&p2, &iPos2); iPos2 -= 2; - }else{ - if( (*p1&0xFE)==0 ) break; - fts3GetDeltaVarint(&p1, &iPos1); iPos1 -= 2; - } - } - - if( pSave ){ - assert( pp && p ); - p = pSave; - } - - fts3ColumnlistCopy(0, &p1); - fts3ColumnlistCopy(0, &p2); - assert( (*p1&0xFE)==0 && (*p2&0xFE)==0 ); - if( 0==*p1 || 0==*p2 ) break; - - p1++; - p1 += sqlite3Fts3GetVarint32(p1, &iCol1); - p2++; - p2 += sqlite3Fts3GetVarint32(p2, &iCol2); - } - - /* Advance pointer p1 or p2 (whichever corresponds to the smaller of - ** iCol1 and iCol2) so that it points to either the 0x00 that marks the - ** end of the position list, or the 0x01 that precedes the next - ** column-number in the position list. - */ - else if( iCol1<iCol2 ){ - fts3ColumnlistCopy(0, &p1); - if( 0==*p1 ) break; - p1++; - p1 += sqlite3Fts3GetVarint32(p1, &iCol1); - }else{ - fts3ColumnlistCopy(0, &p2); - if( 0==*p2 ) break; - p2++; - p2 += sqlite3Fts3GetVarint32(p2, &iCol2); - } - } - - fts3PoslistCopy(0, &p2); - fts3PoslistCopy(0, &p1); - *pp1 = p1; - *pp2 = p2; - if( *pp==p ){ - return 0; - } - *p++ = 0x00; - *pp = p; - return 1; -} - -/* -** Merge two position-lists as required by the NEAR operator. The argument -** position lists correspond to the left and right phrases of an expression -** like: -** -** "phrase 1" NEAR "phrase number 2" -** -** Position list *pp1 corresponds to the left-hand side of the NEAR -** expression and *pp2 to the right. As usual, the indexes in the position -** lists are the offsets of the last token in each phrase (tokens "1" and "2" -** in the example above). -** -** The output position list - written to *pp - is a copy of *pp2 with those -** entries that are not sufficiently NEAR entries in *pp1 removed. -*/ -static int fts3PoslistNearMerge( - char **pp, /* Output buffer */ - char *aTmp, /* Temporary buffer space */ - int nRight, /* Maximum difference in token positions */ - int nLeft, /* Maximum difference in token positions */ - char **pp1, /* IN/OUT: Left input list */ - char **pp2 /* IN/OUT: Right input list */ -){ - char *p1 = *pp1; - char *p2 = *pp2; - - char *pTmp1 = aTmp; - char *pTmp2; - char *aTmp2; - int res = 1; - - fts3PoslistPhraseMerge(&pTmp1, nRight, 0, 0, pp1, pp2); - aTmp2 = pTmp2 = pTmp1; - *pp1 = p1; - *pp2 = p2; - fts3PoslistPhraseMerge(&pTmp2, nLeft, 1, 0, pp2, pp1); - if( pTmp1!=aTmp && pTmp2!=aTmp2 ){ - fts3PoslistMerge(pp, &aTmp, &aTmp2); - }else if( pTmp1!=aTmp ){ - fts3PoslistCopy(pp, &aTmp); - }else if( pTmp2!=aTmp2 ){ - fts3PoslistCopy(pp, &aTmp2); - }else{ - res = 0; - } - - return res; -} - -/* -** An instance of this function is used to merge together the (potentially -** large number of) doclists for each term that matches a prefix query. -** See function fts3TermSelectMerge() for details. -*/ -typedef struct TermSelect TermSelect; -struct TermSelect { - char *aaOutput[16]; /* Malloc'd output buffers */ - int anOutput[16]; /* Size each output buffer in bytes */ -}; - -/* -** This function is used to read a single varint from a buffer. Parameter -** pEnd points 1 byte past the end of the buffer. When this function is -** called, if *pp points to pEnd or greater, then the end of the buffer -** has been reached. In this case *pp is set to 0 and the function returns. -** -** If *pp does not point to or past pEnd, then a single varint is read -** from *pp. *pp is then set to point 1 byte past the end of the read varint. -** -** If bDescIdx is false, the value read is added to *pVal before returning. -** If it is true, the value read is subtracted from *pVal before this -** function returns. -*/ -static void fts3GetDeltaVarint3( - char **pp, /* IN/OUT: Point to read varint from */ - char *pEnd, /* End of buffer */ - int bDescIdx, /* True if docids are descending */ - sqlite3_int64 *pVal /* IN/OUT: Integer value */ -){ - if( *pp>=pEnd ){ - *pp = 0; - }else{ - sqlite3_int64 iVal; - *pp += sqlite3Fts3GetVarint(*pp, &iVal); - if( bDescIdx ){ - *pVal -= iVal; - }else{ - *pVal += iVal; - } - } -} - -/* -** This function is used to write a single varint to a buffer. The varint -** is written to *pp. Before returning, *pp is set to point 1 byte past the -** end of the value written. -** -** If *pbFirst is zero when this function is called, the value written to -** the buffer is that of parameter iVal. -** -** If *pbFirst is non-zero when this function is called, then the value -** written is either (iVal-*piPrev) (if bDescIdx is zero) or (*piPrev-iVal) -** (if bDescIdx is non-zero). -** -** Before returning, this function always sets *pbFirst to 1 and *piPrev -** to the value of parameter iVal. -*/ -static void fts3PutDeltaVarint3( - char **pp, /* IN/OUT: Output pointer */ - int bDescIdx, /* True for descending docids */ - sqlite3_int64 *piPrev, /* IN/OUT: Previous value written to list */ - int *pbFirst, /* IN/OUT: True after first int written */ - sqlite3_int64 iVal /* Write this value to the list */ -){ - sqlite3_int64 iWrite; - if( bDescIdx==0 || *pbFirst==0 ){ - iWrite = iVal - *piPrev; - }else{ - iWrite = *piPrev - iVal; - } - assert( *pbFirst || *piPrev==0 ); - assert( *pbFirst==0 || iWrite>0 ); - *pp += sqlite3Fts3PutVarint(*pp, iWrite); - *piPrev = iVal; - *pbFirst = 1; -} - - -/* -** This macro is used by various functions that merge doclists. The two -** arguments are 64-bit docid values. If the value of the stack variable -** bDescDoclist is 0 when this macro is invoked, then it returns (i1-i2). -** Otherwise, (i2-i1). -** -** Using this makes it easier to write code that can merge doclists that are -** sorted in either ascending or descending order. -*/ -#define DOCID_CMP(i1, i2) ((bDescDoclist?-1:1) * (i1-i2)) - -/* -** This function does an "OR" merge of two doclists (output contains all -** positions contained in either argument doclist). If the docids in the -** input doclists are sorted in ascending order, parameter bDescDoclist -** should be false. If they are sorted in ascending order, it should be -** passed a non-zero value. -** -** If no error occurs, *paOut is set to point at an sqlite3_malloc'd buffer -** containing the output doclist and SQLITE_OK is returned. In this case -** *pnOut is set to the number of bytes in the output doclist. -** -** If an error occurs, an SQLite error code is returned. The output values -** are undefined in this case. -*/ -static int fts3DoclistOrMerge( - int bDescDoclist, /* True if arguments are desc */ - char *a1, int n1, /* First doclist */ - char *a2, int n2, /* Second doclist */ - char **paOut, int *pnOut /* OUT: Malloc'd doclist */ -){ - sqlite3_int64 i1 = 0; - sqlite3_int64 i2 = 0; - sqlite3_int64 iPrev = 0; - char *pEnd1 = &a1[n1]; - char *pEnd2 = &a2[n2]; - char *p1 = a1; - char *p2 = a2; - char *p; - char *aOut; - int bFirstOut = 0; - - *paOut = 0; - *pnOut = 0; - - /* Allocate space for the output. Both the input and output doclists - ** are delta encoded. If they are in ascending order (bDescDoclist==0), - ** then the first docid in each list is simply encoded as a varint. For - ** each subsequent docid, the varint stored is the difference between the - ** current and previous docid (a positive number - since the list is in - ** ascending order). - ** - ** The first docid written to the output is therefore encoded using the - ** same number of bytes as it is in whichever of the input lists it is - ** read from. And each subsequent docid read from the same input list - ** consumes either the same or less bytes as it did in the input (since - ** the difference between it and the previous value in the output must - ** be a positive value less than or equal to the delta value read from - ** the input list). The same argument applies to all but the first docid - ** read from the 'other' list. And to the contents of all position lists - ** that will be copied and merged from the input to the output. - ** - ** However, if the first docid copied to the output is a negative number, - ** then the encoding of the first docid from the 'other' input list may - ** be larger in the output than it was in the input (since the delta value - ** may be a larger positive integer than the actual docid). - ** - ** The space required to store the output is therefore the sum of the - ** sizes of the two inputs, plus enough space for exactly one of the input - ** docids to grow. - ** - ** A symetric argument may be made if the doclists are in descending - ** order. - */ - aOut = sqlite3_malloc(n1+n2+FTS3_VARINT_MAX-1); - if( !aOut ) return SQLITE_NOMEM; - - p = aOut; - fts3GetDeltaVarint3(&p1, pEnd1, 0, &i1); - fts3GetDeltaVarint3(&p2, pEnd2, 0, &i2); - while( p1 || p2 ){ - sqlite3_int64 iDiff = DOCID_CMP(i1, i2); - - if( p2 && p1 && iDiff==0 ){ - fts3PutDeltaVarint3(&p, bDescDoclist, &iPrev, &bFirstOut, i1); - fts3PoslistMerge(&p, &p1, &p2); - fts3GetDeltaVarint3(&p1, pEnd1, bDescDoclist, &i1); - fts3GetDeltaVarint3(&p2, pEnd2, bDescDoclist, &i2); - }else if( !p2 || (p1 && iDiff<0) ){ - fts3PutDeltaVarint3(&p, bDescDoclist, &iPrev, &bFirstOut, i1); - fts3PoslistCopy(&p, &p1); - fts3GetDeltaVarint3(&p1, pEnd1, bDescDoclist, &i1); - }else{ - fts3PutDeltaVarint3(&p, bDescDoclist, &iPrev, &bFirstOut, i2); - fts3PoslistCopy(&p, &p2); - fts3GetDeltaVarint3(&p2, pEnd2, bDescDoclist, &i2); - } - } - - *paOut = aOut; - *pnOut = (int)(p-aOut); - assert( *pnOut<=n1+n2+FTS3_VARINT_MAX-1 ); - return SQLITE_OK; -} - -/* -** This function does a "phrase" merge of two doclists. In a phrase merge, -** the output contains a copy of each position from the right-hand input -** doclist for which there is a position in the left-hand input doclist -** exactly nDist tokens before it. -** -** If the docids in the input doclists are sorted in ascending order, -** parameter bDescDoclist should be false. If they are sorted in ascending -** order, it should be passed a non-zero value. -** -** The right-hand input doclist is overwritten by this function. -*/ -static void fts3DoclistPhraseMerge( - int bDescDoclist, /* True if arguments are desc */ - int nDist, /* Distance from left to right (1=adjacent) */ - char *aLeft, int nLeft, /* Left doclist */ - char *aRight, int *pnRight /* IN/OUT: Right/output doclist */ -){ - sqlite3_int64 i1 = 0; - sqlite3_int64 i2 = 0; - sqlite3_int64 iPrev = 0; - char *pEnd1 = &aLeft[nLeft]; - char *pEnd2 = &aRight[*pnRight]; - char *p1 = aLeft; - char *p2 = aRight; - char *p; - int bFirstOut = 0; - char *aOut = aRight; - - assert( nDist>0 ); - - p = aOut; - fts3GetDeltaVarint3(&p1, pEnd1, 0, &i1); - fts3GetDeltaVarint3(&p2, pEnd2, 0, &i2); - - while( p1 && p2 ){ - sqlite3_int64 iDiff = DOCID_CMP(i1, i2); - if( iDiff==0 ){ - char *pSave = p; - sqlite3_int64 iPrevSave = iPrev; - int bFirstOutSave = bFirstOut; - - fts3PutDeltaVarint3(&p, bDescDoclist, &iPrev, &bFirstOut, i1); - if( 0==fts3PoslistPhraseMerge(&p, nDist, 0, 1, &p1, &p2) ){ - p = pSave; - iPrev = iPrevSave; - bFirstOut = bFirstOutSave; - } - fts3GetDeltaVarint3(&p1, pEnd1, bDescDoclist, &i1); - fts3GetDeltaVarint3(&p2, pEnd2, bDescDoclist, &i2); - }else if( iDiff<0 ){ - fts3PoslistCopy(0, &p1); - fts3GetDeltaVarint3(&p1, pEnd1, bDescDoclist, &i1); - }else{ - fts3PoslistCopy(0, &p2); - fts3GetDeltaVarint3(&p2, pEnd2, bDescDoclist, &i2); - } - } - - *pnRight = (int)(p - aOut); -} - -/* -** Argument pList points to a position list nList bytes in size. This -** function checks to see if the position list contains any entries for -** a token in position 0 (of any column). If so, it writes argument iDelta -** to the output buffer pOut, followed by a position list consisting only -** of the entries from pList at position 0, and terminated by an 0x00 byte. -** The value returned is the number of bytes written to pOut (if any). -*/ -int sqlite3Fts3FirstFilter( - sqlite3_int64 iDelta, /* Varint that may be written to pOut */ - char *pList, /* Position list (no 0x00 term) */ - int nList, /* Size of pList in bytes */ - char *pOut /* Write output here */ -){ - int nOut = 0; - int bWritten = 0; /* True once iDelta has been written */ - char *p = pList; - char *pEnd = &pList[nList]; - - if( *p!=0x01 ){ - if( *p==0x02 ){ - nOut += sqlite3Fts3PutVarint(&pOut[nOut], iDelta); - pOut[nOut++] = 0x02; - bWritten = 1; - } - fts3ColumnlistCopy(0, &p); - } - - while( p<pEnd && *p==0x01 ){ - sqlite3_int64 iCol; - p++; - p += sqlite3Fts3GetVarint(p, &iCol); - if( *p==0x02 ){ - if( bWritten==0 ){ - nOut += sqlite3Fts3PutVarint(&pOut[nOut], iDelta); - bWritten = 1; - } - pOut[nOut++] = 0x01; - nOut += sqlite3Fts3PutVarint(&pOut[nOut], iCol); - pOut[nOut++] = 0x02; - } - fts3ColumnlistCopy(0, &p); - } - if( bWritten ){ - pOut[nOut++] = 0x00; - } - - return nOut; -} - - -/* -** Merge all doclists in the TermSelect.aaOutput[] array into a single -** doclist stored in TermSelect.aaOutput[0]. If successful, delete all -** other doclists (except the aaOutput[0] one) and return SQLITE_OK. -** -** If an OOM error occurs, return SQLITE_NOMEM. In this case it is -** the responsibility of the caller to free any doclists left in the -** TermSelect.aaOutput[] array. -*/ -static int fts3TermSelectFinishMerge(Fts3Table *p, TermSelect *pTS){ - char *aOut = 0; - int nOut = 0; - int i; - - /* Loop through the doclists in the aaOutput[] array. Merge them all - ** into a single doclist. - */ - for(i=0; i<SizeofArray(pTS->aaOutput); i++){ - if( pTS->aaOutput[i] ){ - if( !aOut ){ - aOut = pTS->aaOutput[i]; - nOut = pTS->anOutput[i]; - pTS->aaOutput[i] = 0; - }else{ - int nNew; - char *aNew; - - int rc = fts3DoclistOrMerge(p->bDescIdx, - pTS->aaOutput[i], pTS->anOutput[i], aOut, nOut, &aNew, &nNew - ); - if( rc!=SQLITE_OK ){ - sqlite3_free(aOut); - return rc; - } - - sqlite3_free(pTS->aaOutput[i]); - sqlite3_free(aOut); - pTS->aaOutput[i] = 0; - aOut = aNew; - nOut = nNew; - } - } - } - - pTS->aaOutput[0] = aOut; - pTS->anOutput[0] = nOut; - return SQLITE_OK; -} - -/* -** Merge the doclist aDoclist/nDoclist into the TermSelect object passed -** as the first argument. The merge is an "OR" merge (see function -** fts3DoclistOrMerge() for details). -** -** This function is called with the doclist for each term that matches -** a queried prefix. It merges all these doclists into one, the doclist -** for the specified prefix. Since there can be a very large number of -** doclists to merge, the merging is done pair-wise using the TermSelect -** object. -** -** This function returns SQLITE_OK if the merge is successful, or an -** SQLite error code (SQLITE_NOMEM) if an error occurs. -*/ -static int fts3TermSelectMerge( - Fts3Table *p, /* FTS table handle */ - TermSelect *pTS, /* TermSelect object to merge into */ - char *aDoclist, /* Pointer to doclist */ - int nDoclist /* Size of aDoclist in bytes */ -){ - if( pTS->aaOutput[0]==0 ){ - /* If this is the first term selected, copy the doclist to the output - ** buffer using memcpy(). */ - pTS->aaOutput[0] = sqlite3_malloc(nDoclist); - pTS->anOutput[0] = nDoclist; - if( pTS->aaOutput[0] ){ - memcpy(pTS->aaOutput[0], aDoclist, nDoclist); - }else{ - return SQLITE_NOMEM; - } - }else{ - char *aMerge = aDoclist; - int nMerge = nDoclist; - int iOut; - - for(iOut=0; iOut<SizeofArray(pTS->aaOutput); iOut++){ - if( pTS->aaOutput[iOut]==0 ){ - assert( iOut>0 ); - pTS->aaOutput[iOut] = aMerge; - pTS->anOutput[iOut] = nMerge; - break; - }else{ - char *aNew; - int nNew; - - int rc = fts3DoclistOrMerge(p->bDescIdx, aMerge, nMerge, - pTS->aaOutput[iOut], pTS->anOutput[iOut], &aNew, &nNew - ); - if( rc!=SQLITE_OK ){ - if( aMerge!=aDoclist ) sqlite3_free(aMerge); - return rc; - } - - if( aMerge!=aDoclist ) sqlite3_free(aMerge); - sqlite3_free(pTS->aaOutput[iOut]); - pTS->aaOutput[iOut] = 0; - - aMerge = aNew; - nMerge = nNew; - if( (iOut+1)==SizeofArray(pTS->aaOutput) ){ - pTS->aaOutput[iOut] = aMerge; - pTS->anOutput[iOut] = nMerge; - } - } - } - } - return SQLITE_OK; -} - -/* -** Append SegReader object pNew to the end of the pCsr->apSegment[] array. -*/ -static int fts3SegReaderCursorAppend( - Fts3MultiSegReader *pCsr, - Fts3SegReader *pNew -){ - if( (pCsr->nSegment%16)==0 ){ - Fts3SegReader **apNew; - int nByte = (pCsr->nSegment + 16)*sizeof(Fts3SegReader*); - apNew = (Fts3SegReader **)sqlite3_realloc(pCsr->apSegment, nByte); - if( !apNew ){ - sqlite3Fts3SegReaderFree(pNew); - return SQLITE_NOMEM; - } - pCsr->apSegment = apNew; - } - pCsr->apSegment[pCsr->nSegment++] = pNew; - return SQLITE_OK; -} - -/* -** Add seg-reader objects to the Fts3MultiSegReader object passed as the -** 8th argument. -** -** This function returns SQLITE_OK if successful, or an SQLite error code -** otherwise. -*/ -static int fts3SegReaderCursor( - Fts3Table *p, /* FTS3 table handle */ - int iLangid, /* Language id */ - int iIndex, /* Index to search (from 0 to p->nIndex-1) */ - int iLevel, /* Level of segments to scan */ - const char *zTerm, /* Term to query for */ - int nTerm, /* Size of zTerm in bytes */ - int isPrefix, /* True for a prefix search */ - int isScan, /* True to scan from zTerm to EOF */ - Fts3MultiSegReader *pCsr /* Cursor object to populate */ -){ - int rc = SQLITE_OK; /* Error code */ - sqlite3_stmt *pStmt = 0; /* Statement to iterate through segments */ - int rc2; /* Result of sqlite3_reset() */ - - /* If iLevel is less than 0 and this is not a scan, include a seg-reader - ** for the pending-terms. If this is a scan, then this call must be being - ** made by an fts4aux module, not an FTS table. In this case calling - ** Fts3SegReaderPending might segfault, as the data structures used by - ** fts4aux are not completely populated. So it's easiest to filter these - ** calls out here. */ - if( iLevel<0 && p->aIndex ){ - Fts3SegReader *pSeg = 0; - rc = sqlite3Fts3SegReaderPending(p, iIndex, zTerm, nTerm, isPrefix, &pSeg); - if( rc==SQLITE_OK && pSeg ){ - rc = fts3SegReaderCursorAppend(pCsr, pSeg); - } - } - - if( iLevel!=FTS3_SEGCURSOR_PENDING ){ - if( rc==SQLITE_OK ){ - rc = sqlite3Fts3AllSegdirs(p, iLangid, iIndex, iLevel, &pStmt); - } - - while( rc==SQLITE_OK && SQLITE_ROW==(rc = sqlite3_step(pStmt)) ){ - Fts3SegReader *pSeg = 0; - - /* Read the values returned by the SELECT into local variables. */ - sqlite3_int64 iStartBlock = sqlite3_column_int64(pStmt, 1); - sqlite3_int64 iLeavesEndBlock = sqlite3_column_int64(pStmt, 2); - sqlite3_int64 iEndBlock = sqlite3_column_int64(pStmt, 3); - int nRoot = sqlite3_column_bytes(pStmt, 4); - char const *zRoot = sqlite3_column_blob(pStmt, 4); - - /* If zTerm is not NULL, and this segment is not stored entirely on its - ** root node, the range of leaves scanned can be reduced. Do this. */ - if( iStartBlock && zTerm ){ - sqlite3_int64 *pi = (isPrefix ? &iLeavesEndBlock : 0); - rc = fts3SelectLeaf(p, zTerm, nTerm, zRoot, nRoot, &iStartBlock, pi); - if( rc!=SQLITE_OK ) goto finished; - if( isPrefix==0 && isScan==0 ) iLeavesEndBlock = iStartBlock; - } - - rc = sqlite3Fts3SegReaderNew(pCsr->nSegment+1, - (isPrefix==0 && isScan==0), - iStartBlock, iLeavesEndBlock, - iEndBlock, zRoot, nRoot, &pSeg - ); - if( rc!=SQLITE_OK ) goto finished; - rc = fts3SegReaderCursorAppend(pCsr, pSeg); - } - } - - finished: - rc2 = sqlite3_reset(pStmt); - if( rc==SQLITE_DONE ) rc = rc2; - - return rc; -} - -/* -** Set up a cursor object for iterating through a full-text index or a -** single level therein. -*/ -int sqlite3Fts3SegReaderCursor( - Fts3Table *p, /* FTS3 table handle */ - int iLangid, /* Language-id to search */ - int iIndex, /* Index to search (from 0 to p->nIndex-1) */ - int iLevel, /* Level of segments to scan */ - const char *zTerm, /* Term to query for */ - int nTerm, /* Size of zTerm in bytes */ - int isPrefix, /* True for a prefix search */ - int isScan, /* True to scan from zTerm to EOF */ - Fts3MultiSegReader *pCsr /* Cursor object to populate */ -){ - assert( iIndex>=0 && iIndex<p->nIndex ); - assert( iLevel==FTS3_SEGCURSOR_ALL - || iLevel==FTS3_SEGCURSOR_PENDING - || iLevel>=0 - ); - assert( iLevel<FTS3_SEGDIR_MAXLEVEL ); - assert( FTS3_SEGCURSOR_ALL<0 && FTS3_SEGCURSOR_PENDING<0 ); - assert( isPrefix==0 || isScan==0 ); - - memset(pCsr, 0, sizeof(Fts3MultiSegReader)); - return fts3SegReaderCursor( - p, iLangid, iIndex, iLevel, zTerm, nTerm, isPrefix, isScan, pCsr - ); -} - -/* -** In addition to its current configuration, have the Fts3MultiSegReader -** passed as the 4th argument also scan the doclist for term zTerm/nTerm. -** -** SQLITE_OK is returned if no error occurs, otherwise an SQLite error code. -*/ -static int fts3SegReaderCursorAddZero( - Fts3Table *p, /* FTS virtual table handle */ - int iLangid, - const char *zTerm, /* Term to scan doclist of */ - int nTerm, /* Number of bytes in zTerm */ - Fts3MultiSegReader *pCsr /* Fts3MultiSegReader to modify */ -){ - return fts3SegReaderCursor(p, - iLangid, 0, FTS3_SEGCURSOR_ALL, zTerm, nTerm, 0, 0,pCsr - ); -} - -/* -** Open an Fts3MultiSegReader to scan the doclist for term zTerm/nTerm. Or, -** if isPrefix is true, to scan the doclist for all terms for which -** zTerm/nTerm is a prefix. If successful, return SQLITE_OK and write -** a pointer to the new Fts3MultiSegReader to *ppSegcsr. Otherwise, return -** an SQLite error code. -** -** It is the responsibility of the caller to free this object by eventually -** passing it to fts3SegReaderCursorFree() -** -** SQLITE_OK is returned if no error occurs, otherwise an SQLite error code. -** Output parameter *ppSegcsr is set to 0 if an error occurs. -*/ -static int fts3TermSegReaderCursor( - Fts3Cursor *pCsr, /* Virtual table cursor handle */ - const char *zTerm, /* Term to query for */ - int nTerm, /* Size of zTerm in bytes */ - int isPrefix, /* True for a prefix search */ - Fts3MultiSegReader **ppSegcsr /* OUT: Allocated seg-reader cursor */ -){ - Fts3MultiSegReader *pSegcsr; /* Object to allocate and return */ - int rc = SQLITE_NOMEM; /* Return code */ - - pSegcsr = sqlite3_malloc(sizeof(Fts3MultiSegReader)); - if( pSegcsr ){ - int i; - int bFound = 0; /* True once an index has been found */ - Fts3Table *p = (Fts3Table *)pCsr->base.pVtab; - - if( isPrefix ){ - for(i=1; bFound==0 && i<p->nIndex; i++){ - if( p->aIndex[i].nPrefix==nTerm ){ - bFound = 1; - rc = sqlite3Fts3SegReaderCursor(p, pCsr->iLangid, - i, FTS3_SEGCURSOR_ALL, zTerm, nTerm, 0, 0, pSegcsr - ); - pSegcsr->bLookup = 1; - } - } - - for(i=1; bFound==0 && i<p->nIndex; i++){ - if( p->aIndex[i].nPrefix==nTerm+1 ){ - bFound = 1; - rc = sqlite3Fts3SegReaderCursor(p, pCsr->iLangid, - i, FTS3_SEGCURSOR_ALL, zTerm, nTerm, 1, 0, pSegcsr - ); - if( rc==SQLITE_OK ){ - rc = fts3SegReaderCursorAddZero( - p, pCsr->iLangid, zTerm, nTerm, pSegcsr - ); - } - } - } - } - - if( bFound==0 ){ - rc = sqlite3Fts3SegReaderCursor(p, pCsr->iLangid, - 0, FTS3_SEGCURSOR_ALL, zTerm, nTerm, isPrefix, 0, pSegcsr - ); - pSegcsr->bLookup = !isPrefix; - } - } - - *ppSegcsr = pSegcsr; - return rc; -} - -/* -** Free an Fts3MultiSegReader allocated by fts3TermSegReaderCursor(). -*/ -static void fts3SegReaderCursorFree(Fts3MultiSegReader *pSegcsr){ - sqlite3Fts3SegReaderFinish(pSegcsr); - sqlite3_free(pSegcsr); -} - -/* -** This function retreives the doclist for the specified term (or term -** prefix) from the database. -*/ -static int fts3TermSelect( - Fts3Table *p, /* Virtual table handle */ - Fts3PhraseToken *pTok, /* Token to query for */ - int iColumn, /* Column to query (or -ve for all columns) */ - int *pnOut, /* OUT: Size of buffer at *ppOut */ - char **ppOut /* OUT: Malloced result buffer */ -){ - int rc; /* Return code */ - Fts3MultiSegReader *pSegcsr; /* Seg-reader cursor for this term */ - TermSelect tsc; /* Object for pair-wise doclist merging */ - Fts3SegFilter filter; /* Segment term filter configuration */ - - pSegcsr = pTok->pSegcsr; - memset(&tsc, 0, sizeof(TermSelect)); - - filter.flags = FTS3_SEGMENT_IGNORE_EMPTY | FTS3_SEGMENT_REQUIRE_POS - | (pTok->isPrefix ? FTS3_SEGMENT_PREFIX : 0) - | (pTok->bFirst ? FTS3_SEGMENT_FIRST : 0) - | (iColumn<p->nColumn ? FTS3_SEGMENT_COLUMN_FILTER : 0); - filter.iCol = iColumn; - filter.zTerm = pTok->z; - filter.nTerm = pTok->n; - - rc = sqlite3Fts3SegReaderStart(p, pSegcsr, &filter); - while( SQLITE_OK==rc - && SQLITE_ROW==(rc = sqlite3Fts3SegReaderStep(p, pSegcsr)) - ){ - rc = fts3TermSelectMerge(p, &tsc, pSegcsr->aDoclist, pSegcsr->nDoclist); - } - - if( rc==SQLITE_OK ){ - rc = fts3TermSelectFinishMerge(p, &tsc); - } - if( rc==SQLITE_OK ){ - *ppOut = tsc.aaOutput[0]; - *pnOut = tsc.anOutput[0]; - }else{ - int i; - for(i=0; i<SizeofArray(tsc.aaOutput); i++){ - sqlite3_free(tsc.aaOutput[i]); - } - } - - fts3SegReaderCursorFree(pSegcsr); - pTok->pSegcsr = 0; - return rc; -} - -/* -** This function counts the total number of docids in the doclist stored -** in buffer aList[], size nList bytes. -** -** If the isPoslist argument is true, then it is assumed that the doclist -** contains a position-list following each docid. Otherwise, it is assumed -** that the doclist is simply a list of docids stored as delta encoded -** varints. -*/ -static int fts3DoclistCountDocids(char *aList, int nList){ - int nDoc = 0; /* Return value */ - if( aList ){ - char *aEnd = &aList[nList]; /* Pointer to one byte after EOF */ - char *p = aList; /* Cursor */ - while( p<aEnd ){ - nDoc++; - while( (*p++)&0x80 ); /* Skip docid varint */ - fts3PoslistCopy(0, &p); /* Skip over position list */ - } - } - - return nDoc; -} - -/* -** Advance the cursor to the next row in the %_content table that -** matches the search criteria. For a MATCH search, this will be -** the next row that matches. For a full-table scan, this will be -** simply the next row in the %_content table. For a docid lookup, -** this routine simply sets the EOF flag. -** -** Return SQLITE_OK if nothing goes wrong. SQLITE_OK is returned -** even if we reach end-of-file. The fts3EofMethod() will be called -** subsequently to determine whether or not an EOF was hit. -*/ -static int fts3NextMethod(sqlite3_vtab_cursor *pCursor){ - int rc; - Fts3Cursor *pCsr = (Fts3Cursor *)pCursor; - if( pCsr->eSearch==FTS3_DOCID_SEARCH || pCsr->eSearch==FTS3_FULLSCAN_SEARCH ){ - if( SQLITE_ROW!=sqlite3_step(pCsr->pStmt) ){ - pCsr->isEof = 1; - rc = sqlite3_reset(pCsr->pStmt); - }else{ - pCsr->iPrevId = sqlite3_column_int64(pCsr->pStmt, 0); - rc = SQLITE_OK; - } - }else{ - rc = fts3EvalNext((Fts3Cursor *)pCursor); - } - assert( ((Fts3Table *)pCsr->base.pVtab)->pSegments==0 ); - return rc; -} - -/* -** This is the xFilter interface for the virtual table. See -** the virtual table xFilter method documentation for additional -** information. -** -** If idxNum==FTS3_FULLSCAN_SEARCH then do a full table scan against -** the %_content table. -** -** If idxNum==FTS3_DOCID_SEARCH then do a docid lookup for a single entry -** in the %_content table. -** -** If idxNum>=FTS3_FULLTEXT_SEARCH then use the full text index. The -** column on the left-hand side of the MATCH operator is column -** number idxNum-FTS3_FULLTEXT_SEARCH, 0 indexed. argv[0] is the right-hand -** side of the MATCH operator. -*/ -static int fts3FilterMethod( - sqlite3_vtab_cursor *pCursor, /* The cursor used for this query */ - int idxNum, /* Strategy index */ - const char *idxStr, /* Unused */ - int nVal, /* Number of elements in apVal */ - sqlite3_value **apVal /* Arguments for the indexing scheme */ -){ - int rc; - char *zSql; /* SQL statement used to access %_content */ - Fts3Table *p = (Fts3Table *)pCursor->pVtab; - Fts3Cursor *pCsr = (Fts3Cursor *)pCursor; - - UNUSED_PARAMETER(idxStr); - UNUSED_PARAMETER(nVal); - - assert( idxNum>=0 && idxNum<=(FTS3_FULLTEXT_SEARCH+p->nColumn) ); - assert( nVal==0 || nVal==1 || nVal==2 ); - assert( (nVal==0)==(idxNum==FTS3_FULLSCAN_SEARCH) ); - assert( p->pSegments==0 ); - - /* In case the cursor has been used before, clear it now. */ - sqlite3_finalize(pCsr->pStmt); - sqlite3_free(pCsr->aDoclist); - sqlite3Fts3ExprFree(pCsr->pExpr); - memset(&pCursor[1], 0, sizeof(Fts3Cursor)-sizeof(sqlite3_vtab_cursor)); - - if( idxStr ){ - pCsr->bDesc = (idxStr[0]=='D'); - }else{ - pCsr->bDesc = p->bDescIdx; - } - pCsr->eSearch = (i16)idxNum; - - if( idxNum!=FTS3_DOCID_SEARCH && idxNum!=FTS3_FULLSCAN_SEARCH ){ - int iCol = idxNum-FTS3_FULLTEXT_SEARCH; - const char *zQuery = (const char *)sqlite3_value_text(apVal[0]); - - if( zQuery==0 && sqlite3_value_type(apVal[0])!=SQLITE_NULL ){ - return SQLITE_NOMEM; - } - - pCsr->iLangid = 0; - if( nVal==2 ) pCsr->iLangid = sqlite3_value_int(apVal[1]); - - rc = sqlite3Fts3ExprParse(p->pTokenizer, pCsr->iLangid, - p->azColumn, p->bFts4, p->nColumn, iCol, zQuery, -1, &pCsr->pExpr - ); - if( rc!=SQLITE_OK ){ - if( rc==SQLITE_ERROR ){ - static const char *zErr = "malformed MATCH expression: [%s]"; - p->base.zErrMsg = sqlite3_mprintf(zErr, zQuery); - } - return rc; - } - - rc = sqlite3Fts3ReadLock(p); - if( rc!=SQLITE_OK ) return rc; - - rc = fts3EvalStart(pCsr); - - sqlite3Fts3SegmentsClose(p); - if( rc!=SQLITE_OK ) return rc; - pCsr->pNextId = pCsr->aDoclist; - pCsr->iPrevId = 0; - } - - /* Compile a SELECT statement for this cursor. For a full-table-scan, the - ** statement loops through all rows of the %_content table. For a - ** full-text query or docid lookup, the statement retrieves a single - ** row by docid. - */ - if( idxNum==FTS3_FULLSCAN_SEARCH ){ - zSql = sqlite3_mprintf( - "SELECT %s ORDER BY rowid %s", - p->zReadExprlist, (pCsr->bDesc ? "DESC" : "ASC") - ); - if( zSql ){ - rc = sqlite3_prepare_v2(p->db, zSql, -1, &pCsr->pStmt, 0); - sqlite3_free(zSql); - }else{ - rc = SQLITE_NOMEM; - } - }else if( idxNum==FTS3_DOCID_SEARCH ){ - rc = fts3CursorSeekStmt(pCsr, &pCsr->pStmt); - if( rc==SQLITE_OK ){ - rc = sqlite3_bind_value(pCsr->pStmt, 1, apVal[0]); - } - } - if( rc!=SQLITE_OK ) return rc; - - return fts3NextMethod(pCursor); -} - -/* -** This is the xEof method of the virtual table. SQLite calls this -** routine to find out if it has reached the end of a result set. -*/ -static int fts3EofMethod(sqlite3_vtab_cursor *pCursor){ - return ((Fts3Cursor *)pCursor)->isEof; -} - -/* -** This is the xRowid method. The SQLite core calls this routine to -** retrieve the rowid for the current row of the result set. fts3 -** exposes %_content.docid as the rowid for the virtual table. The -** rowid should be written to *pRowid. -*/ -static int fts3RowidMethod(sqlite3_vtab_cursor *pCursor, sqlite_int64 *pRowid){ - Fts3Cursor *pCsr = (Fts3Cursor *) pCursor; - *pRowid = pCsr->iPrevId; - return SQLITE_OK; -} - -/* -** This is the xColumn method, called by SQLite to request a value from -** the row that the supplied cursor currently points to. -** -** If: -** -** (iCol < p->nColumn) -> The value of the iCol'th user column. -** (iCol == p->nColumn) -> Magic column with the same name as the table. -** (iCol == p->nColumn+1) -> Docid column -** (iCol == p->nColumn+2) -> Langid column -*/ -static int fts3ColumnMethod( - sqlite3_vtab_cursor *pCursor, /* Cursor to retrieve value from */ - sqlite3_context *pCtx, /* Context for sqlite3_result_xxx() calls */ - int iCol /* Index of column to read value from */ -){ - int rc = SQLITE_OK; /* Return Code */ - Fts3Cursor *pCsr = (Fts3Cursor *) pCursor; - Fts3Table *p = (Fts3Table *)pCursor->pVtab; - - /* The column value supplied by SQLite must be in range. */ - assert( iCol>=0 && iCol<=p->nColumn+2 ); - - if( iCol==p->nColumn+1 ){ - /* This call is a request for the "docid" column. Since "docid" is an - ** alias for "rowid", use the xRowid() method to obtain the value. - */ - sqlite3_result_int64(pCtx, pCsr->iPrevId); - }else if( iCol==p->nColumn ){ - /* The extra column whose name is the same as the table. - ** Return a blob which is a pointer to the cursor. */ - sqlite3_result_blob(pCtx, &pCsr, sizeof(pCsr), SQLITE_TRANSIENT); - }else if( iCol==p->nColumn+2 && pCsr->pExpr ){ - sqlite3_result_int64(pCtx, pCsr->iLangid); - }else{ - /* The requested column is either a user column (one that contains - ** indexed data), or the language-id column. */ - rc = fts3CursorSeek(0, pCsr); - - if( rc==SQLITE_OK ){ - if( iCol==p->nColumn+2 ){ - int iLangid = 0; - if( p->zLanguageid ){ - iLangid = sqlite3_column_int(pCsr->pStmt, p->nColumn+1); - } - sqlite3_result_int(pCtx, iLangid); - }else if( sqlite3_data_count(pCsr->pStmt)>(iCol+1) ){ - sqlite3_result_value(pCtx, sqlite3_column_value(pCsr->pStmt, iCol+1)); - } - } - } - - assert( ((Fts3Table *)pCsr->base.pVtab)->pSegments==0 ); - return rc; -} - -/* -** This function is the implementation of the xUpdate callback used by -** FTS3 virtual tables. It is invoked by SQLite each time a row is to be -** inserted, updated or deleted. -*/ -static int fts3UpdateMethod( - sqlite3_vtab *pVtab, /* Virtual table handle */ - int nArg, /* Size of argument array */ - sqlite3_value **apVal, /* Array of arguments */ - sqlite_int64 *pRowid /* OUT: The affected (or effected) rowid */ -){ - return sqlite3Fts3UpdateMethod(pVtab, nArg, apVal, pRowid); -} - -/* -** Implementation of xSync() method. Flush the contents of the pending-terms -** hash-table to the database. -*/ -static int fts3SyncMethod(sqlite3_vtab *pVtab){ - - /* Following an incremental-merge operation, assuming that the input - ** segments are not completely consumed (the usual case), they are updated - ** in place to remove the entries that have already been merged. This - ** involves updating the leaf block that contains the smallest unmerged - ** entry and each block (if any) between the leaf and the root node. So - ** if the height of the input segment b-trees is N, and input segments - ** are merged eight at a time, updating the input segments at the end - ** of an incremental-merge requires writing (8*(1+N)) blocks. N is usually - ** small - often between 0 and 2. So the overhead of the incremental - ** merge is somewhere between 8 and 24 blocks. To avoid this overhead - ** dwarfing the actual productive work accomplished, the incremental merge - ** is only attempted if it will write at least 64 leaf blocks. Hence - ** nMinMerge. - ** - ** Of course, updating the input segments also involves deleting a bunch - ** of blocks from the segments table. But this is not considered overhead - ** as it would also be required by a crisis-merge that used the same input - ** segments. - */ - const u32 nMinMerge = 64; /* Minimum amount of incr-merge work to do */ - - Fts3Table *p = (Fts3Table*)pVtab; - int rc = sqlite3Fts3PendingTermsFlush(p); - - if( rc==SQLITE_OK && p->bAutoincrmerge==1 && p->nLeafAdd>(nMinMerge/16) ){ - int mxLevel = 0; /* Maximum relative level value in db */ - int A; /* Incr-merge parameter A */ - - rc = sqlite3Fts3MaxLevel(p, &mxLevel); - assert( rc==SQLITE_OK || mxLevel==0 ); - A = p->nLeafAdd * mxLevel; - A += (A/2); - if( A>(int)nMinMerge ) rc = sqlite3Fts3Incrmerge(p, A, 8); - } - sqlite3Fts3SegmentsClose(p); - return rc; -} - -/* -** Implementation of xBegin() method. This is a no-op. -*/ -static int fts3BeginMethod(sqlite3_vtab *pVtab){ - Fts3Table *p = (Fts3Table*)pVtab; - UNUSED_PARAMETER(pVtab); - assert( p->pSegments==0 ); - assert( p->nPendingData==0 ); - assert( p->inTransaction!=1 ); - TESTONLY( p->inTransaction = 1 ); - TESTONLY( p->mxSavepoint = -1; ); - p->nLeafAdd = 0; - return SQLITE_OK; -} - -/* -** Implementation of xCommit() method. This is a no-op. The contents of -** the pending-terms hash-table have already been flushed into the database -** by fts3SyncMethod(). -*/ -static int fts3CommitMethod(sqlite3_vtab *pVtab){ - TESTONLY( Fts3Table *p = (Fts3Table*)pVtab ); - UNUSED_PARAMETER(pVtab); - assert( p->nPendingData==0 ); - assert( p->inTransaction!=0 ); - assert( p->pSegments==0 ); - TESTONLY( p->inTransaction = 0 ); - TESTONLY( p->mxSavepoint = -1; ); - return SQLITE_OK; -} - -/* -** Implementation of xRollback(). Discard the contents of the pending-terms -** hash-table. Any changes made to the database are reverted by SQLite. -*/ -static int fts3RollbackMethod(sqlite3_vtab *pVtab){ - Fts3Table *p = (Fts3Table*)pVtab; - sqlite3Fts3PendingTermsClear(p); - assert( p->inTransaction!=0 ); - TESTONLY( p->inTransaction = 0 ); - TESTONLY( p->mxSavepoint = -1; ); - return SQLITE_OK; -} - -/* -** When called, *ppPoslist must point to the byte immediately following the -** end of a position-list. i.e. ( (*ppPoslist)[-1]==POS_END ). This function -** moves *ppPoslist so that it instead points to the first byte of the -** same position list. -*/ -static void fts3ReversePoslist(char *pStart, char **ppPoslist){ - char *p = &(*ppPoslist)[-2]; - char c = 0; - - while( p>pStart && (c=*p--)==0 ); - while( p>pStart && (*p & 0x80) | c ){ - c = *p--; - } - if( p>pStart ){ p = &p[2]; } - while( *p++&0x80 ); - *ppPoslist = p; -} - -/* -** Helper function used by the implementation of the overloaded snippet(), -** offsets() and optimize() SQL functions. -** -** If the value passed as the third argument is a blob of size -** sizeof(Fts3Cursor*), then the blob contents are copied to the -** output variable *ppCsr and SQLITE_OK is returned. Otherwise, an error -** message is written to context pContext and SQLITE_ERROR returned. The -** string passed via zFunc is used as part of the error message. -*/ -static int fts3FunctionArg( - sqlite3_context *pContext, /* SQL function call context */ - const char *zFunc, /* Function name */ - sqlite3_value *pVal, /* argv[0] passed to function */ - Fts3Cursor **ppCsr /* OUT: Store cursor handle here */ -){ - Fts3Cursor *pRet; - if( sqlite3_value_type(pVal)!=SQLITE_BLOB - || sqlite3_value_bytes(pVal)!=sizeof(Fts3Cursor *) - ){ - char *zErr = sqlite3_mprintf("illegal first argument to %s", zFunc); - sqlite3_result_error(pContext, zErr, -1); - sqlite3_free(zErr); - return SQLITE_ERROR; - } - memcpy(&pRet, sqlite3_value_blob(pVal), sizeof(Fts3Cursor *)); - *ppCsr = pRet; - return SQLITE_OK; -} - -/* -** Implementation of the snippet() function for FTS3 -*/ -static void fts3SnippetFunc( - sqlite3_context *pContext, /* SQLite function call context */ - int nVal, /* Size of apVal[] array */ - sqlite3_value **apVal /* Array of arguments */ -){ - Fts3Cursor *pCsr; /* Cursor handle passed through apVal[0] */ - const char *zStart = "<b>"; - const char *zEnd = "</b>"; - const char *zEllipsis = "<b>...</b>"; - int iCol = -1; - int nToken = 15; /* Default number of tokens in snippet */ - - /* There must be at least one argument passed to this function (otherwise - ** the non-overloaded version would have been called instead of this one). - */ - assert( nVal>=1 ); - - if( nVal>6 ){ - sqlite3_result_error(pContext, - "wrong number of arguments to function snippet()", -1); - return; - } - if( fts3FunctionArg(pContext, "snippet", apVal[0], &pCsr) ) return; - - switch( nVal ){ - case 6: nToken = sqlite3_value_int(apVal[5]); - case 5: iCol = sqlite3_value_int(apVal[4]); - case 4: zEllipsis = (const char*)sqlite3_value_text(apVal[3]); - case 3: zEnd = (const char*)sqlite3_value_text(apVal[2]); - case 2: zStart = (const char*)sqlite3_value_text(apVal[1]); - } - if( !zEllipsis || !zEnd || !zStart ){ - sqlite3_result_error_nomem(pContext); - }else if( SQLITE_OK==fts3CursorSeek(pContext, pCsr) ){ - sqlite3Fts3Snippet(pContext, pCsr, zStart, zEnd, zEllipsis, iCol, nToken); - } -} - -/* -** Implementation of the offsets() function for FTS3 -*/ -static void fts3OffsetsFunc( - sqlite3_context *pContext, /* SQLite function call context */ - int nVal, /* Size of argument array */ - sqlite3_value **apVal /* Array of arguments */ -){ - Fts3Cursor *pCsr; /* Cursor handle passed through apVal[0] */ - - UNUSED_PARAMETER(nVal); - - assert( nVal==1 ); - if( fts3FunctionArg(pContext, "offsets", apVal[0], &pCsr) ) return; - assert( pCsr ); - if( SQLITE_OK==fts3CursorSeek(pContext, pCsr) ){ - sqlite3Fts3Offsets(pContext, pCsr); - } -} - -/* -** Implementation of the special optimize() function for FTS3. This -** function merges all segments in the database to a single segment. -** Example usage is: -** -** SELECT optimize(t) FROM t LIMIT 1; -** -** where 't' is the name of an FTS3 table. -*/ -static void fts3OptimizeFunc( - sqlite3_context *pContext, /* SQLite function call context */ - int nVal, /* Size of argument array */ - sqlite3_value **apVal /* Array of arguments */ -){ - int rc; /* Return code */ - Fts3Table *p; /* Virtual table handle */ - Fts3Cursor *pCursor; /* Cursor handle passed through apVal[0] */ - - UNUSED_PARAMETER(nVal); - - assert( nVal==1 ); - if( fts3FunctionArg(pContext, "optimize", apVal[0], &pCursor) ) return; - p = (Fts3Table *)pCursor->base.pVtab; - assert( p ); - - rc = sqlite3Fts3Optimize(p); - - switch( rc ){ - case SQLITE_OK: - sqlite3_result_text(pContext, "Index optimized", -1, SQLITE_STATIC); - break; - case SQLITE_DONE: - sqlite3_result_text(pContext, "Index already optimal", -1, SQLITE_STATIC); - break; - default: - sqlite3_result_error_code(pContext, rc); - break; - } -} - -/* -** Implementation of the matchinfo() function for FTS3 -*/ -static void fts3MatchinfoFunc( - sqlite3_context *pContext, /* SQLite function call context */ - int nVal, /* Size of argument array */ - sqlite3_value **apVal /* Array of arguments */ -){ - Fts3Cursor *pCsr; /* Cursor handle passed through apVal[0] */ - assert( nVal==1 || nVal==2 ); - if( SQLITE_OK==fts3FunctionArg(pContext, "matchinfo", apVal[0], &pCsr) ){ - const char *zArg = 0; - if( nVal>1 ){ - zArg = (const char *)sqlite3_value_text(apVal[1]); - } - sqlite3Fts3Matchinfo(pContext, pCsr, zArg); - } -} - -/* -** This routine implements the xFindFunction method for the FTS3 -** virtual table. -*/ -static int fts3FindFunctionMethod( - sqlite3_vtab *pVtab, /* Virtual table handle */ - int nArg, /* Number of SQL function arguments */ - const char *zName, /* Name of SQL function */ - void (**pxFunc)(sqlite3_context*,int,sqlite3_value**), /* OUT: Result */ - void **ppArg /* Unused */ -){ - struct Overloaded { - const char *zName; - void (*xFunc)(sqlite3_context*,int,sqlite3_value**); - } aOverload[] = { - { "snippet", fts3SnippetFunc }, - { "offsets", fts3OffsetsFunc }, - { "optimize", fts3OptimizeFunc }, - { "matchinfo", fts3MatchinfoFunc }, - }; - int i; /* Iterator variable */ - - UNUSED_PARAMETER(pVtab); - UNUSED_PARAMETER(nArg); - UNUSED_PARAMETER(ppArg); - - for(i=0; i<SizeofArray(aOverload); i++){ - if( strcmp(zName, aOverload[i].zName)==0 ){ - *pxFunc = aOverload[i].xFunc; - return 1; - } - } - - /* No function of the specified name was found. Return 0. */ - return 0; -} - -/* -** Implementation of FTS3 xRename method. Rename an fts3 table. -*/ -static int fts3RenameMethod( - sqlite3_vtab *pVtab, /* Virtual table handle */ - const char *zName /* New name of table */ -){ - Fts3Table *p = (Fts3Table *)pVtab; - sqlite3 *db = p->db; /* Database connection */ - int rc; /* Return Code */ - - /* As it happens, the pending terms table is always empty here. This is - ** because an "ALTER TABLE RENAME TABLE" statement inside a transaction - ** always opens a savepoint transaction. And the xSavepoint() method - ** flushes the pending terms table. But leave the (no-op) call to - ** PendingTermsFlush() in in case that changes. - */ - assert( p->nPendingData==0 ); - rc = sqlite3Fts3PendingTermsFlush(p); - - if( p->zContentTbl==0 ){ - fts3DbExec(&rc, db, - "ALTER TABLE %Q.'%q_content' RENAME TO '%q_content';", - p->zDb, p->zName, zName - ); - } - - if( p->bHasDocsize ){ - fts3DbExec(&rc, db, - "ALTER TABLE %Q.'%q_docsize' RENAME TO '%q_docsize';", - p->zDb, p->zName, zName - ); - } - if( p->bHasStat ){ - fts3DbExec(&rc, db, - "ALTER TABLE %Q.'%q_stat' RENAME TO '%q_stat';", - p->zDb, p->zName, zName - ); - } - fts3DbExec(&rc, db, - "ALTER TABLE %Q.'%q_segments' RENAME TO '%q_segments';", - p->zDb, p->zName, zName - ); - fts3DbExec(&rc, db, - "ALTER TABLE %Q.'%q_segdir' RENAME TO '%q_segdir';", - p->zDb, p->zName, zName - ); - return rc; -} - -/* -** The xSavepoint() method. -** -** Flush the contents of the pending-terms table to disk. -*/ -static int fts3SavepointMethod(sqlite3_vtab *pVtab, int iSavepoint){ - int rc = SQLITE_OK; - UNUSED_PARAMETER(iSavepoint); - assert( ((Fts3Table *)pVtab)->inTransaction ); - assert( ((Fts3Table *)pVtab)->mxSavepoint < iSavepoint ); - TESTONLY( ((Fts3Table *)pVtab)->mxSavepoint = iSavepoint ); - if( ((Fts3Table *)pVtab)->bIgnoreSavepoint==0 ){ - rc = fts3SyncMethod(pVtab); - } - return rc; -} - -/* -** The xRelease() method. -** -** This is a no-op. -*/ -static int fts3ReleaseMethod(sqlite3_vtab *pVtab, int iSavepoint){ - TESTONLY( Fts3Table *p = (Fts3Table*)pVtab ); - UNUSED_PARAMETER(iSavepoint); - UNUSED_PARAMETER(pVtab); - assert( p->inTransaction ); - assert( p->mxSavepoint >= iSavepoint ); - TESTONLY( p->mxSavepoint = iSavepoint-1 ); - return SQLITE_OK; -} - -/* -** The xRollbackTo() method. -** -** Discard the contents of the pending terms table. -*/ -static int fts3RollbackToMethod(sqlite3_vtab *pVtab, int iSavepoint){ - Fts3Table *p = (Fts3Table*)pVtab; - UNUSED_PARAMETER(iSavepoint); - assert( p->inTransaction ); - assert( p->mxSavepoint >= iSavepoint ); - TESTONLY( p->mxSavepoint = iSavepoint ); - sqlite3Fts3PendingTermsClear(p); - return SQLITE_OK; -} - -static const sqlite3_module fts3Module = { - /* iVersion */ 2, - /* xCreate */ fts3CreateMethod, - /* xConnect */ fts3ConnectMethod, - /* xBestIndex */ fts3BestIndexMethod, - /* xDisconnect */ fts3DisconnectMethod, - /* xDestroy */ fts3DestroyMethod, - /* xOpen */ fts3OpenMethod, - /* xClose */ fts3CloseMethod, - /* xFilter */ fts3FilterMethod, - /* xNext */ fts3NextMethod, - /* xEof */ fts3EofMethod, - /* xColumn */ fts3ColumnMethod, - /* xRowid */ fts3RowidMethod, - /* xUpdate */ fts3UpdateMethod, - /* xBegin */ fts3BeginMethod, - /* xSync */ fts3SyncMethod, - /* xCommit */ fts3CommitMethod, - /* xRollback */ fts3RollbackMethod, - /* xFindFunction */ fts3FindFunctionMethod, - /* xRename */ fts3RenameMethod, - /* xSavepoint */ fts3SavepointMethod, - /* xRelease */ fts3ReleaseMethod, - /* xRollbackTo */ fts3RollbackToMethod, -}; - -/* -** This function is registered as the module destructor (called when an -** FTS3 enabled database connection is closed). It frees the memory -** allocated for the tokenizer hash table. -*/ -static void hashDestroy(void *p){ - Fts3Hash *pHash = (Fts3Hash *)p; - sqlite3Fts3HashClear(pHash); - sqlite3_free(pHash); -} - -/* -** The fts3 built-in tokenizers - "simple", "porter" and "icu"- are -** implemented in files fts3_tokenizer1.c, fts3_porter.c and fts3_icu.c -** respectively. The following three forward declarations are for functions -** declared in these files used to retrieve the respective implementations. -** -** Calling sqlite3Fts3SimpleTokenizerModule() sets the value pointed -** to by the argument to point to the "simple" tokenizer implementation. -** And so on. -*/ -void sqlite3Fts3SimpleTokenizerModule(sqlite3_tokenizer_module const**ppModule); -void sqlite3Fts3PorterTokenizerModule(sqlite3_tokenizer_module const**ppModule); -#ifdef SQLITE_ENABLE_FTS4_UNICODE61 -void sqlite3Fts3UnicodeTokenizer(sqlite3_tokenizer_module const**ppModule); -#endif -#ifdef SQLITE_ENABLE_ICU -void sqlite3Fts3IcuTokenizerModule(sqlite3_tokenizer_module const**ppModule); -#endif - -/* -** Initialise the fts3 extension. If this extension is built as part -** of the sqlite library, then this function is called directly by -** SQLite. If fts3 is built as a dynamically loadable extension, this -** function is called by the sqlite3_extension_init() entry point. -*/ -int sqlite3Fts3Init(sqlite3 *db){ - int rc = SQLITE_OK; - Fts3Hash *pHash = 0; - const sqlite3_tokenizer_module *pSimple = 0; - const sqlite3_tokenizer_module *pPorter = 0; -#ifdef SQLITE_ENABLE_FTS4_UNICODE61 - const sqlite3_tokenizer_module *pUnicode = 0; -#endif - -#ifdef SQLITE_ENABLE_ICU - const sqlite3_tokenizer_module *pIcu = 0; - sqlite3Fts3IcuTokenizerModule(&pIcu); -#endif - -#ifdef SQLITE_ENABLE_FTS4_UNICODE61 - sqlite3Fts3UnicodeTokenizer(&pUnicode); -#endif - -#ifdef SQLITE_TEST - rc = sqlite3Fts3InitTerm(db); - if( rc!=SQLITE_OK ) return rc; -#endif - - rc = sqlite3Fts3InitAux(db); - if( rc!=SQLITE_OK ) return rc; - - sqlite3Fts3SimpleTokenizerModule(&pSimple); - sqlite3Fts3PorterTokenizerModule(&pPorter); - - /* Allocate and initialise the hash-table used to store tokenizers. */ - pHash = sqlite3_malloc(sizeof(Fts3Hash)); - if( !pHash ){ - rc = SQLITE_NOMEM; - }else{ - sqlite3Fts3HashInit(pHash, FTS3_HASH_STRING, 1); - } - - /* Load the built-in tokenizers into the hash table */ - if( rc==SQLITE_OK ){ - if( sqlite3Fts3HashInsert(pHash, "simple", 7, (void *)pSimple) - || sqlite3Fts3HashInsert(pHash, "porter", 7, (void *)pPorter) - -#ifdef SQLITE_ENABLE_FTS4_UNICODE61 - || sqlite3Fts3HashInsert(pHash, "unicode61", 10, (void *)pUnicode) -#endif -#ifdef SQLITE_ENABLE_ICU - || (pIcu && sqlite3Fts3HashInsert(pHash, "icu", 4, (void *)pIcu)) -#endif - ){ - rc = SQLITE_NOMEM; - } - } - -#ifdef SQLITE_TEST - if( rc==SQLITE_OK ){ - rc = sqlite3Fts3ExprInitTestInterface(db); - } -#endif - - /* Create the virtual table wrapper around the hash-table and overload - ** the two scalar functions. If this is successful, register the - ** module with sqlite. - */ - if( SQLITE_OK==rc - && SQLITE_OK==(rc = sqlite3Fts3InitHashTable(db, pHash, "fts3_tokenizer")) - && SQLITE_OK==(rc = sqlite3_overload_function(db, "snippet", -1)) - && SQLITE_OK==(rc = sqlite3_overload_function(db, "offsets", 1)) - && SQLITE_OK==(rc = sqlite3_overload_function(db, "matchinfo", 1)) - && SQLITE_OK==(rc = sqlite3_overload_function(db, "matchinfo", 2)) - && SQLITE_OK==(rc = sqlite3_overload_function(db, "optimize", 1)) - ){ - rc = sqlite3_create_module_v2( - db, "fts3", &fts3Module, (void *)pHash, hashDestroy - ); - if( rc==SQLITE_OK ){ - rc = sqlite3_create_module_v2( - db, "fts4", &fts3Module, (void *)pHash, 0 - ); - } - return rc; - } - - /* An error has occurred. Delete the hash table and return the error code. */ - assert( rc!=SQLITE_OK ); - if( pHash ){ - sqlite3Fts3HashClear(pHash); - sqlite3_free(pHash); - } - return rc; -} - -/* -** Allocate an Fts3MultiSegReader for each token in the expression headed -** by pExpr. -** -** An Fts3SegReader object is a cursor that can seek or scan a range of -** entries within a single segment b-tree. An Fts3MultiSegReader uses multiple -** Fts3SegReader objects internally to provide an interface to seek or scan -** within the union of all segments of a b-tree. Hence the name. -** -** If the allocated Fts3MultiSegReader just seeks to a single entry in a -** segment b-tree (if the term is not a prefix or it is a prefix for which -** there exists prefix b-tree of the right length) then it may be traversed -** and merged incrementally. Otherwise, it has to be merged into an in-memory -** doclist and then traversed. -*/ -static void fts3EvalAllocateReaders( - Fts3Cursor *pCsr, /* FTS cursor handle */ - Fts3Expr *pExpr, /* Allocate readers for this expression */ - int *pnToken, /* OUT: Total number of tokens in phrase. */ - int *pnOr, /* OUT: Total number of OR nodes in expr. */ - int *pRc /* IN/OUT: Error code */ -){ - if( pExpr && SQLITE_OK==*pRc ){ - if( pExpr->eType==FTSQUERY_PHRASE ){ - int i; - int nToken = pExpr->pPhrase->nToken; - *pnToken += nToken; - for(i=0; i<nToken; i++){ - Fts3PhraseToken *pToken = &pExpr->pPhrase->aToken[i]; - int rc = fts3TermSegReaderCursor(pCsr, - pToken->z, pToken->n, pToken->isPrefix, &pToken->pSegcsr - ); - if( rc!=SQLITE_OK ){ - *pRc = rc; - return; - } - } - assert( pExpr->pPhrase->iDoclistToken==0 ); - pExpr->pPhrase->iDoclistToken = -1; - }else{ - *pnOr += (pExpr->eType==FTSQUERY_OR); - fts3EvalAllocateReaders(pCsr, pExpr->pLeft, pnToken, pnOr, pRc); - fts3EvalAllocateReaders(pCsr, pExpr->pRight, pnToken, pnOr, pRc); - } - } -} - -/* -** Arguments pList/nList contain the doclist for token iToken of phrase p. -** It is merged into the main doclist stored in p->doclist.aAll/nAll. -** -** This function assumes that pList points to a buffer allocated using -** sqlite3_malloc(). This function takes responsibility for eventually -** freeing the buffer. -*/ -static void fts3EvalPhraseMergeToken( - Fts3Table *pTab, /* FTS Table pointer */ - Fts3Phrase *p, /* Phrase to merge pList/nList into */ - int iToken, /* Token pList/nList corresponds to */ - char *pList, /* Pointer to doclist */ - int nList /* Number of bytes in pList */ -){ - assert( iToken!=p->iDoclistToken ); - - if( pList==0 ){ - sqlite3_free(p->doclist.aAll); - p->doclist.aAll = 0; - p->doclist.nAll = 0; - } - - else if( p->iDoclistToken<0 ){ - p->doclist.aAll = pList; - p->doclist.nAll = nList; - } - - else if( p->doclist.aAll==0 ){ - sqlite3_free(pList); - } - - else { - char *pLeft; - char *pRight; - int nLeft; - int nRight; - int nDiff; - - if( p->iDoclistToken<iToken ){ - pLeft = p->doclist.aAll; - nLeft = p->doclist.nAll; - pRight = pList; - nRight = nList; - nDiff = iToken - p->iDoclistToken; - }else{ - pRight = p->doclist.aAll; - nRight = p->doclist.nAll; - pLeft = pList; - nLeft = nList; - nDiff = p->iDoclistToken - iToken; - } - - fts3DoclistPhraseMerge(pTab->bDescIdx, nDiff, pLeft, nLeft, pRight,&nRight); - sqlite3_free(pLeft); - p->doclist.aAll = pRight; - p->doclist.nAll = nRight; - } - - if( iToken>p->iDoclistToken ) p->iDoclistToken = iToken; -} - -/* -** Load the doclist for phrase p into p->doclist.aAll/nAll. The loaded doclist -** does not take deferred tokens into account. -** -** SQLITE_OK is returned if no error occurs, otherwise an SQLite error code. -*/ -static int fts3EvalPhraseLoad( - Fts3Cursor *pCsr, /* FTS Cursor handle */ - Fts3Phrase *p /* Phrase object */ -){ - Fts3Table *pTab = (Fts3Table *)pCsr->base.pVtab; - int iToken; - int rc = SQLITE_OK; - - for(iToken=0; rc==SQLITE_OK && iToken<p->nToken; iToken++){ - Fts3PhraseToken *pToken = &p->aToken[iToken]; - assert( pToken->pDeferred==0 || pToken->pSegcsr==0 ); - - if( pToken->pSegcsr ){ - int nThis = 0; - char *pThis = 0; - rc = fts3TermSelect(pTab, pToken, p->iColumn, &nThis, &pThis); - if( rc==SQLITE_OK ){ - fts3EvalPhraseMergeToken(pTab, p, iToken, pThis, nThis); - } - } - assert( pToken->pSegcsr==0 ); - } - - return rc; -} - -/* -** This function is called on each phrase after the position lists for -** any deferred tokens have been loaded into memory. It updates the phrases -** current position list to include only those positions that are really -** instances of the phrase (after considering deferred tokens). If this -** means that the phrase does not appear in the current row, doclist.pList -** and doclist.nList are both zeroed. -** -** SQLITE_OK is returned if no error occurs, otherwise an SQLite error code. -*/ -static int fts3EvalDeferredPhrase(Fts3Cursor *pCsr, Fts3Phrase *pPhrase){ - int iToken; /* Used to iterate through phrase tokens */ - char *aPoslist = 0; /* Position list for deferred tokens */ - int nPoslist = 0; /* Number of bytes in aPoslist */ - int iPrev = -1; /* Token number of previous deferred token */ - - assert( pPhrase->doclist.bFreeList==0 ); - - for(iToken=0; iToken<pPhrase->nToken; iToken++){ - Fts3PhraseToken *pToken = &pPhrase->aToken[iToken]; - Fts3DeferredToken *pDeferred = pToken->pDeferred; - - if( pDeferred ){ - char *pList; - int nList; - int rc = sqlite3Fts3DeferredTokenList(pDeferred, &pList, &nList); - if( rc!=SQLITE_OK ) return rc; - - if( pList==0 ){ - sqlite3_free(aPoslist); - pPhrase->doclist.pList = 0; - pPhrase->doclist.nList = 0; - return SQLITE_OK; - - }else if( aPoslist==0 ){ - aPoslist = pList; - nPoslist = nList; - - }else{ - char *aOut = pList; - char *p1 = aPoslist; - char *p2 = aOut; - - assert( iPrev>=0 ); - fts3PoslistPhraseMerge(&aOut, iToken-iPrev, 0, 1, &p1, &p2); - sqlite3_free(aPoslist); - aPoslist = pList; - nPoslist = (int)(aOut - aPoslist); - if( nPoslist==0 ){ - sqlite3_free(aPoslist); - pPhrase->doclist.pList = 0; - pPhrase->doclist.nList = 0; - return SQLITE_OK; - } - } - iPrev = iToken; - } - } - - if( iPrev>=0 ){ - int nMaxUndeferred = pPhrase->iDoclistToken; - if( nMaxUndeferred<0 ){ - pPhrase->doclist.pList = aPoslist; - pPhrase->doclist.nList = nPoslist; - pPhrase->doclist.iDocid = pCsr->iPrevId; - pPhrase->doclist.bFreeList = 1; - }else{ - int nDistance; - char *p1; - char *p2; - char *aOut; - - if( nMaxUndeferred>iPrev ){ - p1 = aPoslist; - p2 = pPhrase->doclist.pList; - nDistance = nMaxUndeferred - iPrev; - }else{ - p1 = pPhrase->doclist.pList; - p2 = aPoslist; - nDistance = iPrev - nMaxUndeferred; - } - - aOut = (char *)sqlite3_malloc(nPoslist+8); - if( !aOut ){ - sqlite3_free(aPoslist); - return SQLITE_NOMEM; - } - - pPhrase->doclist.pList = aOut; - if( fts3PoslistPhraseMerge(&aOut, nDistance, 0, 1, &p1, &p2) ){ - pPhrase->doclist.bFreeList = 1; - pPhrase->doclist.nList = (int)(aOut - pPhrase->doclist.pList); - }else{ - sqlite3_free(aOut); - pPhrase->doclist.pList = 0; - pPhrase->doclist.nList = 0; - } - sqlite3_free(aPoslist); - } - } - - return SQLITE_OK; -} - -/* -** This function is called for each Fts3Phrase in a full-text query -** expression to initialize the mechanism for returning rows. Once this -** function has been called successfully on an Fts3Phrase, it may be -** used with fts3EvalPhraseNext() to iterate through the matching docids. -** -** If parameter bOptOk is true, then the phrase may (or may not) use the -** incremental loading strategy. Otherwise, the entire doclist is loaded into -** memory within this call. -** -** SQLITE_OK is returned if no error occurs, otherwise an SQLite error code. -*/ -static int fts3EvalPhraseStart(Fts3Cursor *pCsr, int bOptOk, Fts3Phrase *p){ - int rc; /* Error code */ - Fts3PhraseToken *pFirst = &p->aToken[0]; - Fts3Table *pTab = (Fts3Table *)pCsr->base.pVtab; - - if( pCsr->bDesc==pTab->bDescIdx - && bOptOk==1 - && p->nToken==1 - && pFirst->pSegcsr - && pFirst->pSegcsr->bLookup - && pFirst->bFirst==0 - ){ - /* Use the incremental approach. */ - int iCol = (p->iColumn >= pTab->nColumn ? -1 : p->iColumn); - rc = sqlite3Fts3MsrIncrStart( - pTab, pFirst->pSegcsr, iCol, pFirst->z, pFirst->n); - p->bIncr = 1; - - }else{ - /* Load the full doclist for the phrase into memory. */ - rc = fts3EvalPhraseLoad(pCsr, p); - p->bIncr = 0; - } - - assert( rc!=SQLITE_OK || p->nToken<1 || p->aToken[0].pSegcsr==0 || p->bIncr ); - return rc; -} - -/* -** This function is used to iterate backwards (from the end to start) -** through doclists. It is used by this module to iterate through phrase -** doclists in reverse and by the fts3_write.c module to iterate through -** pending-terms lists when writing to databases with "order=desc". -** -** The doclist may be sorted in ascending (parameter bDescIdx==0) or -** descending (parameter bDescIdx==1) order of docid. Regardless, this -** function iterates from the end of the doclist to the beginning. -*/ -void sqlite3Fts3DoclistPrev( - int bDescIdx, /* True if the doclist is desc */ - char *aDoclist, /* Pointer to entire doclist */ - int nDoclist, /* Length of aDoclist in bytes */ - char **ppIter, /* IN/OUT: Iterator pointer */ - sqlite3_int64 *piDocid, /* IN/OUT: Docid pointer */ - int *pnList, /* OUT: List length pointer */ - u8 *pbEof /* OUT: End-of-file flag */ -){ - char *p = *ppIter; - - assert( nDoclist>0 ); - assert( *pbEof==0 ); - assert( p || *piDocid==0 ); - assert( !p || (p>aDoclist && p<&aDoclist[nDoclist]) ); - - if( p==0 ){ - sqlite3_int64 iDocid = 0; - char *pNext = 0; - char *pDocid = aDoclist; - char *pEnd = &aDoclist[nDoclist]; - int iMul = 1; - - while( pDocid<pEnd ){ - sqlite3_int64 iDelta; - pDocid += sqlite3Fts3GetVarint(pDocid, &iDelta); - iDocid += (iMul * iDelta); - pNext = pDocid; - fts3PoslistCopy(0, &pDocid); - while( pDocid<pEnd && *pDocid==0 ) pDocid++; - iMul = (bDescIdx ? -1 : 1); - } - - *pnList = (int)(pEnd - pNext); - *ppIter = pNext; - *piDocid = iDocid; - }else{ - int iMul = (bDescIdx ? -1 : 1); - sqlite3_int64 iDelta; - fts3GetReverseVarint(&p, aDoclist, &iDelta); - *piDocid -= (iMul * iDelta); - - if( p==aDoclist ){ - *pbEof = 1; - }else{ - char *pSave = p; - fts3ReversePoslist(aDoclist, &p); - *pnList = (int)(pSave - p); - } - *ppIter = p; - } -} - -/* -** Iterate forwards through a doclist. -*/ -void sqlite3Fts3DoclistNext( - int bDescIdx, /* True if the doclist is desc */ - char *aDoclist, /* Pointer to entire doclist */ - int nDoclist, /* Length of aDoclist in bytes */ - char **ppIter, /* IN/OUT: Iterator pointer */ - sqlite3_int64 *piDocid, /* IN/OUT: Docid pointer */ - u8 *pbEof /* OUT: End-of-file flag */ -){ - char *p = *ppIter; - - assert( nDoclist>0 ); - assert( *pbEof==0 ); - assert( p || *piDocid==0 ); - assert( !p || (p>=aDoclist && p<=&aDoclist[nDoclist]) ); - - if( p==0 ){ - p = aDoclist; - p += sqlite3Fts3GetVarint(p, piDocid); - }else{ - fts3PoslistCopy(0, &p); - if( p>=&aDoclist[nDoclist] ){ - *pbEof = 1; - }else{ - sqlite3_int64 iVar; - p += sqlite3Fts3GetVarint(p, &iVar); - *piDocid += ((bDescIdx ? -1 : 1) * iVar); - } - } - - *ppIter = p; -} - -/* -** Attempt to move the phrase iterator to point to the next matching docid. -** If an error occurs, return an SQLite error code. Otherwise, return -** SQLITE_OK. -** -** If there is no "next" entry and no error occurs, then *pbEof is set to -** 1 before returning. Otherwise, if no error occurs and the iterator is -** successfully advanced, *pbEof is set to 0. -*/ -static int fts3EvalPhraseNext( - Fts3Cursor *pCsr, /* FTS Cursor handle */ - Fts3Phrase *p, /* Phrase object to advance to next docid */ - u8 *pbEof /* OUT: Set to 1 if EOF */ -){ - int rc = SQLITE_OK; - Fts3Doclist *pDL = &p->doclist; - Fts3Table *pTab = (Fts3Table *)pCsr->base.pVtab; - - if( p->bIncr ){ - assert( p->nToken==1 ); - assert( pDL->pNextDocid==0 ); - rc = sqlite3Fts3MsrIncrNext(pTab, p->aToken[0].pSegcsr, - &pDL->iDocid, &pDL->pList, &pDL->nList - ); - if( rc==SQLITE_OK && !pDL->pList ){ - *pbEof = 1; - } - }else if( pCsr->bDesc!=pTab->bDescIdx && pDL->nAll ){ - sqlite3Fts3DoclistPrev(pTab->bDescIdx, pDL->aAll, pDL->nAll, - &pDL->pNextDocid, &pDL->iDocid, &pDL->nList, pbEof - ); - pDL->pList = pDL->pNextDocid; - }else{ - char *pIter; /* Used to iterate through aAll */ - char *pEnd = &pDL->aAll[pDL->nAll]; /* 1 byte past end of aAll */ - if( pDL->pNextDocid ){ - pIter = pDL->pNextDocid; - }else{ - pIter = pDL->aAll; - } - - if( pIter>=pEnd ){ - /* We have already reached the end of this doclist. EOF. */ - *pbEof = 1; - }else{ - sqlite3_int64 iDelta; - pIter += sqlite3Fts3GetVarint(pIter, &iDelta); - if( pTab->bDescIdx==0 || pDL->pNextDocid==0 ){ - pDL->iDocid += iDelta; - }else{ - pDL->iDocid -= iDelta; - } - pDL->pList = pIter; - fts3PoslistCopy(0, &pIter); - pDL->nList = (int)(pIter - pDL->pList); - - /* pIter now points just past the 0x00 that terminates the position- - ** list for document pDL->iDocid. However, if this position-list was - ** edited in place by fts3EvalNearTrim(), then pIter may not actually - ** point to the start of the next docid value. The following line deals - ** with this case by advancing pIter past the zero-padding added by - ** fts3EvalNearTrim(). */ - while( pIter<pEnd && *pIter==0 ) pIter++; - - pDL->pNextDocid = pIter; - assert( pIter>=&pDL->aAll[pDL->nAll] || *pIter ); - *pbEof = 0; - } - } - - return rc; -} - -/* -** -** If *pRc is not SQLITE_OK when this function is called, it is a no-op. -** Otherwise, fts3EvalPhraseStart() is called on all phrases within the -** expression. Also the Fts3Expr.bDeferred variable is set to true for any -** expressions for which all descendent tokens are deferred. -** -** If parameter bOptOk is zero, then it is guaranteed that the -** Fts3Phrase.doclist.aAll/nAll variables contain the entire doclist for -** each phrase in the expression (subject to deferred token processing). -** Or, if bOptOk is non-zero, then one or more tokens within the expression -** may be loaded incrementally, meaning doclist.aAll/nAll is not available. -** -** If an error occurs within this function, *pRc is set to an SQLite error -** code before returning. -*/ -static void fts3EvalStartReaders( - Fts3Cursor *pCsr, /* FTS Cursor handle */ - Fts3Expr *pExpr, /* Expression to initialize phrases in */ - int bOptOk, /* True to enable incremental loading */ - int *pRc /* IN/OUT: Error code */ -){ - if( pExpr && SQLITE_OK==*pRc ){ - if( pExpr->eType==FTSQUERY_PHRASE ){ - int i; - int nToken = pExpr->pPhrase->nToken; - for(i=0; i<nToken; i++){ - if( pExpr->pPhrase->aToken[i].pDeferred==0 ) break; - } - pExpr->bDeferred = (i==nToken); - *pRc = fts3EvalPhraseStart(pCsr, bOptOk, pExpr->pPhrase); - }else{ - fts3EvalStartReaders(pCsr, pExpr->pLeft, bOptOk, pRc); - fts3EvalStartReaders(pCsr, pExpr->pRight, bOptOk, pRc); - pExpr->bDeferred = (pExpr->pLeft->bDeferred && pExpr->pRight->bDeferred); - } - } -} - -/* -** An array of the following structures is assembled as part of the process -** of selecting tokens to defer before the query starts executing (as part -** of the xFilter() method). There is one element in the array for each -** token in the FTS expression. -** -** Tokens are divided into AND/NEAR clusters. All tokens in a cluster belong -** to phrases that are connected only by AND and NEAR operators (not OR or -** NOT). When determining tokens to defer, each AND/NEAR cluster is considered -** separately. The root of a tokens AND/NEAR cluster is stored in -** Fts3TokenAndCost.pRoot. -*/ -typedef struct Fts3TokenAndCost Fts3TokenAndCost; -struct Fts3TokenAndCost { - Fts3Phrase *pPhrase; /* The phrase the token belongs to */ - int iToken; /* Position of token in phrase */ - Fts3PhraseToken *pToken; /* The token itself */ - Fts3Expr *pRoot; /* Root of NEAR/AND cluster */ - int nOvfl; /* Number of overflow pages to load doclist */ - int iCol; /* The column the token must match */ -}; - -/* -** This function is used to populate an allocated Fts3TokenAndCost array. -** -** If *pRc is not SQLITE_OK when this function is called, it is a no-op. -** Otherwise, if an error occurs during execution, *pRc is set to an -** SQLite error code. -*/ -static void fts3EvalTokenCosts( - Fts3Cursor *pCsr, /* FTS Cursor handle */ - Fts3Expr *pRoot, /* Root of current AND/NEAR cluster */ - Fts3Expr *pExpr, /* Expression to consider */ - Fts3TokenAndCost **ppTC, /* Write new entries to *(*ppTC)++ */ - Fts3Expr ***ppOr, /* Write new OR root to *(*ppOr)++ */ - int *pRc /* IN/OUT: Error code */ -){ - if( *pRc==SQLITE_OK ){ - if( pExpr->eType==FTSQUERY_PHRASE ){ - Fts3Phrase *pPhrase = pExpr->pPhrase; - int i; - for(i=0; *pRc==SQLITE_OK && i<pPhrase->nToken; i++){ - Fts3TokenAndCost *pTC = (*ppTC)++; - pTC->pPhrase = pPhrase; - pTC->iToken = i; - pTC->pRoot = pRoot; - pTC->pToken = &pPhrase->aToken[i]; - pTC->iCol = pPhrase->iColumn; - *pRc = sqlite3Fts3MsrOvfl(pCsr, pTC->pToken->pSegcsr, &pTC->nOvfl); - } - }else if( pExpr->eType!=FTSQUERY_NOT ){ - assert( pExpr->eType==FTSQUERY_OR - || pExpr->eType==FTSQUERY_AND - || pExpr->eType==FTSQUERY_NEAR - ); - assert( pExpr->pLeft && pExpr->pRight ); - if( pExpr->eType==FTSQUERY_OR ){ - pRoot = pExpr->pLeft; - **ppOr = pRoot; - (*ppOr)++; - } - fts3EvalTokenCosts(pCsr, pRoot, pExpr->pLeft, ppTC, ppOr, pRc); - if( pExpr->eType==FTSQUERY_OR ){ - pRoot = pExpr->pRight; - **ppOr = pRoot; - (*ppOr)++; - } - fts3EvalTokenCosts(pCsr, pRoot, pExpr->pRight, ppTC, ppOr, pRc); - } - } -} - -/* -** Determine the average document (row) size in pages. If successful, -** write this value to *pnPage and return SQLITE_OK. Otherwise, return -** an SQLite error code. -** -** The average document size in pages is calculated by first calculating -** determining the average size in bytes, B. If B is less than the amount -** of data that will fit on a single leaf page of an intkey table in -** this database, then the average docsize is 1. Otherwise, it is 1 plus -** the number of overflow pages consumed by a record B bytes in size. -*/ -static int fts3EvalAverageDocsize(Fts3Cursor *pCsr, int *pnPage){ - if( pCsr->nRowAvg==0 ){ - /* The average document size, which is required to calculate the cost - ** of each doclist, has not yet been determined. Read the required - ** data from the %_stat table to calculate it. - ** - ** Entry 0 of the %_stat table is a blob containing (nCol+1) FTS3 - ** varints, where nCol is the number of columns in the FTS3 table. - ** The first varint is the number of documents currently stored in - ** the table. The following nCol varints contain the total amount of - ** data stored in all rows of each column of the table, from left - ** to right. - */ - int rc; - Fts3Table *p = (Fts3Table*)pCsr->base.pVtab; - sqlite3_stmt *pStmt; - sqlite3_int64 nDoc = 0; - sqlite3_int64 nByte = 0; - const char *pEnd; - const char *a; - - rc = sqlite3Fts3SelectDoctotal(p, &pStmt); - if( rc!=SQLITE_OK ) return rc; - a = sqlite3_column_blob(pStmt, 0); - assert( a ); - - pEnd = &a[sqlite3_column_bytes(pStmt, 0)]; - a += sqlite3Fts3GetVarint(a, &nDoc); - while( a<pEnd ){ - a += sqlite3Fts3GetVarint(a, &nByte); - } - if( nDoc==0 || nByte==0 ){ - sqlite3_reset(pStmt); - return FTS_CORRUPT_VTAB; - } - - pCsr->nDoc = nDoc; - pCsr->nRowAvg = (int)(((nByte / nDoc) + p->nPgsz) / p->nPgsz); - assert( pCsr->nRowAvg>0 ); - rc = sqlite3_reset(pStmt); - if( rc!=SQLITE_OK ) return rc; - } - - *pnPage = pCsr->nRowAvg; - return SQLITE_OK; -} - -/* -** This function is called to select the tokens (if any) that will be -** deferred. The array aTC[] has already been populated when this is -** called. -** -** This function is called once for each AND/NEAR cluster in the -** expression. Each invocation determines which tokens to defer within -** the cluster with root node pRoot. See comments above the definition -** of struct Fts3TokenAndCost for more details. -** -** If no error occurs, SQLITE_OK is returned and sqlite3Fts3DeferToken() -** called on each token to defer. Otherwise, an SQLite error code is -** returned. -*/ -static int fts3EvalSelectDeferred( - Fts3Cursor *pCsr, /* FTS Cursor handle */ - Fts3Expr *pRoot, /* Consider tokens with this root node */ - Fts3TokenAndCost *aTC, /* Array of expression tokens and costs */ - int nTC /* Number of entries in aTC[] */ -){ - Fts3Table *pTab = (Fts3Table *)pCsr->base.pVtab; - int nDocSize = 0; /* Number of pages per doc loaded */ - int rc = SQLITE_OK; /* Return code */ - int ii; /* Iterator variable for various purposes */ - int nOvfl = 0; /* Total overflow pages used by doclists */ - int nToken = 0; /* Total number of tokens in cluster */ - - int nMinEst = 0; /* The minimum count for any phrase so far. */ - int nLoad4 = 1; /* (Phrases that will be loaded)^4. */ - - /* Tokens are never deferred for FTS tables created using the content=xxx - ** option. The reason being that it is not guaranteed that the content - ** table actually contains the same data as the index. To prevent this from - ** causing any problems, the deferred token optimization is completely - ** disabled for content=xxx tables. */ - if( pTab->zContentTbl ){ - return SQLITE_OK; - } - - /* Count the tokens in this AND/NEAR cluster. If none of the doclists - ** associated with the tokens spill onto overflow pages, or if there is - ** only 1 token, exit early. No tokens to defer in this case. */ - for(ii=0; ii<nTC; ii++){ - if( aTC[ii].pRoot==pRoot ){ - nOvfl += aTC[ii].nOvfl; - nToken++; - } - } - if( nOvfl==0 || nToken<2 ) return SQLITE_OK; - - /* Obtain the average docsize (in pages). */ - rc = fts3EvalAverageDocsize(pCsr, &nDocSize); - assert( rc!=SQLITE_OK || nDocSize>0 ); - - - /* Iterate through all tokens in this AND/NEAR cluster, in ascending order - ** of the number of overflow pages that will be loaded by the pager layer - ** to retrieve the entire doclist for the token from the full-text index. - ** Load the doclists for tokens that are either: - ** - ** a. The cheapest token in the entire query (i.e. the one visited by the - ** first iteration of this loop), or - ** - ** b. Part of a multi-token phrase. - ** - ** After each token doclist is loaded, merge it with the others from the - ** same phrase and count the number of documents that the merged doclist - ** contains. Set variable "nMinEst" to the smallest number of documents in - ** any phrase doclist for which 1 or more token doclists have been loaded. - ** Let nOther be the number of other phrases for which it is certain that - ** one or more tokens will not be deferred. - ** - ** Then, for each token, defer it if loading the doclist would result in - ** loading N or more overflow pages into memory, where N is computed as: - ** - ** (nMinEst + 4^nOther - 1) / (4^nOther) - */ - for(ii=0; ii<nToken && rc==SQLITE_OK; ii++){ - int iTC; /* Used to iterate through aTC[] array. */ - Fts3TokenAndCost *pTC = 0; /* Set to cheapest remaining token. */ - - /* Set pTC to point to the cheapest remaining token. */ - for(iTC=0; iTC<nTC; iTC++){ - if( aTC[iTC].pToken && aTC[iTC].pRoot==pRoot - && (!pTC || aTC[iTC].nOvfl<pTC->nOvfl) - ){ - pTC = &aTC[iTC]; - } - } - assert( pTC ); - - if( ii && pTC->nOvfl>=((nMinEst+(nLoad4/4)-1)/(nLoad4/4))*nDocSize ){ - /* The number of overflow pages to load for this (and therefore all - ** subsequent) tokens is greater than the estimated number of pages - ** that will be loaded if all subsequent tokens are deferred. - */ - Fts3PhraseToken *pToken = pTC->pToken; - rc = sqlite3Fts3DeferToken(pCsr, pToken, pTC->iCol); - fts3SegReaderCursorFree(pToken->pSegcsr); - pToken->pSegcsr = 0; - }else{ - /* Set nLoad4 to the value of (4^nOther) for the next iteration of the - ** for-loop. Except, limit the value to 2^24 to prevent it from - ** overflowing the 32-bit integer it is stored in. */ - if( ii<12 ) nLoad4 = nLoad4*4; - - if( ii==0 || pTC->pPhrase->nToken>1 ){ - /* Either this is the cheapest token in the entire query, or it is - ** part of a multi-token phrase. Either way, the entire doclist will - ** (eventually) be loaded into memory. It may as well be now. */ - Fts3PhraseToken *pToken = pTC->pToken; - int nList = 0; - char *pList = 0; - rc = fts3TermSelect(pTab, pToken, pTC->iCol, &nList, &pList); - assert( rc==SQLITE_OK || pList==0 ); - if( rc==SQLITE_OK ){ - int nCount; - fts3EvalPhraseMergeToken(pTab, pTC->pPhrase, pTC->iToken,pList,nList); - nCount = fts3DoclistCountDocids( - pTC->pPhrase->doclist.aAll, pTC->pPhrase->doclist.nAll - ); - if( ii==0 || nCount<nMinEst ) nMinEst = nCount; - } - } - } - pTC->pToken = 0; - } - - return rc; -} - -/* -** This function is called from within the xFilter method. It initializes -** the full-text query currently stored in pCsr->pExpr. To iterate through -** the results of a query, the caller does: -** -** fts3EvalStart(pCsr); -** while( 1 ){ -** fts3EvalNext(pCsr); -** if( pCsr->bEof ) break; -** ... return row pCsr->iPrevId to the caller ... -** } -*/ -static int fts3EvalStart(Fts3Cursor *pCsr){ - Fts3Table *pTab = (Fts3Table *)pCsr->base.pVtab; - int rc = SQLITE_OK; - int nToken = 0; - int nOr = 0; - - /* Allocate a MultiSegReader for each token in the expression. */ - fts3EvalAllocateReaders(pCsr, pCsr->pExpr, &nToken, &nOr, &rc); - - /* Determine which, if any, tokens in the expression should be deferred. */ -#ifndef SQLITE_DISABLE_FTS4_DEFERRED - if( rc==SQLITE_OK && nToken>1 && pTab->bFts4 ){ - Fts3TokenAndCost *aTC; - Fts3Expr **apOr; - aTC = (Fts3TokenAndCost *)sqlite3_malloc( - sizeof(Fts3TokenAndCost) * nToken - + sizeof(Fts3Expr *) * nOr * 2 - ); - apOr = (Fts3Expr **)&aTC[nToken]; - - if( !aTC ){ - rc = SQLITE_NOMEM; - }else{ - int ii; - Fts3TokenAndCost *pTC = aTC; - Fts3Expr **ppOr = apOr; - - fts3EvalTokenCosts(pCsr, 0, pCsr->pExpr, &pTC, &ppOr, &rc); - nToken = (int)(pTC-aTC); - nOr = (int)(ppOr-apOr); - - if( rc==SQLITE_OK ){ - rc = fts3EvalSelectDeferred(pCsr, 0, aTC, nToken); - for(ii=0; rc==SQLITE_OK && ii<nOr; ii++){ - rc = fts3EvalSelectDeferred(pCsr, apOr[ii], aTC, nToken); - } - } - - sqlite3_free(aTC); - } - } -#endif - - fts3EvalStartReaders(pCsr, pCsr->pExpr, 1, &rc); - return rc; -} - -/* -** Invalidate the current position list for phrase pPhrase. -*/ -static void fts3EvalInvalidatePoslist(Fts3Phrase *pPhrase){ - if( pPhrase->doclist.bFreeList ){ - sqlite3_free(pPhrase->doclist.pList); - } - pPhrase->doclist.pList = 0; - pPhrase->doclist.nList = 0; - pPhrase->doclist.bFreeList = 0; -} - -/* -** This function is called to edit the position list associated with -** the phrase object passed as the fifth argument according to a NEAR -** condition. For example: -** -** abc NEAR/5 "def ghi" -** -** Parameter nNear is passed the NEAR distance of the expression (5 in -** the example above). When this function is called, *paPoslist points to -** the position list, and *pnToken is the number of phrase tokens in, the -** phrase on the other side of the NEAR operator to pPhrase. For example, -** if pPhrase refers to the "def ghi" phrase, then *paPoslist points to -** the position list associated with phrase "abc". -** -** All positions in the pPhrase position list that are not sufficiently -** close to a position in the *paPoslist position list are removed. If this -** leaves 0 positions, zero is returned. Otherwise, non-zero. -** -** Before returning, *paPoslist is set to point to the position lsit -** associated with pPhrase. And *pnToken is set to the number of tokens in -** pPhrase. -*/ -static int fts3EvalNearTrim( - int nNear, /* NEAR distance. As in "NEAR/nNear". */ - char *aTmp, /* Temporary space to use */ - char **paPoslist, /* IN/OUT: Position list */ - int *pnToken, /* IN/OUT: Tokens in phrase of *paPoslist */ - Fts3Phrase *pPhrase /* The phrase object to trim the doclist of */ -){ - int nParam1 = nNear + pPhrase->nToken; - int nParam2 = nNear + *pnToken; - int nNew; - char *p2; - char *pOut; - int res; - - assert( pPhrase->doclist.pList ); - - p2 = pOut = pPhrase->doclist.pList; - res = fts3PoslistNearMerge( - &pOut, aTmp, nParam1, nParam2, paPoslist, &p2 - ); - if( res ){ - nNew = (int)(pOut - pPhrase->doclist.pList) - 1; - assert( pPhrase->doclist.pList[nNew]=='\0' ); - assert( nNew<=pPhrase->doclist.nList && nNew>0 ); - memset(&pPhrase->doclist.pList[nNew], 0, pPhrase->doclist.nList - nNew); - pPhrase->doclist.nList = nNew; - *paPoslist = pPhrase->doclist.pList; - *pnToken = pPhrase->nToken; - } - - return res; -} - -/* -** This function is a no-op if *pRc is other than SQLITE_OK when it is called. -** Otherwise, it advances the expression passed as the second argument to -** point to the next matching row in the database. Expressions iterate through -** matching rows in docid order. Ascending order if Fts3Cursor.bDesc is zero, -** or descending if it is non-zero. -** -** If an error occurs, *pRc is set to an SQLite error code. Otherwise, if -** successful, the following variables in pExpr are set: -** -** Fts3Expr.bEof (non-zero if EOF - there is no next row) -** Fts3Expr.iDocid (valid if bEof==0. The docid of the next row) -** -** If the expression is of type FTSQUERY_PHRASE, and the expression is not -** at EOF, then the following variables are populated with the position list -** for the phrase for the visited row: -** -** FTs3Expr.pPhrase->doclist.nList (length of pList in bytes) -** FTs3Expr.pPhrase->doclist.pList (pointer to position list) -** -** It says above that this function advances the expression to the next -** matching row. This is usually true, but there are the following exceptions: -** -** 1. Deferred tokens are not taken into account. If a phrase consists -** entirely of deferred tokens, it is assumed to match every row in -** the db. In this case the position-list is not populated at all. -** -** Or, if a phrase contains one or more deferred tokens and one or -** more non-deferred tokens, then the expression is advanced to the -** next possible match, considering only non-deferred tokens. In other -** words, if the phrase is "A B C", and "B" is deferred, the expression -** is advanced to the next row that contains an instance of "A * C", -** where "*" may match any single token. The position list in this case -** is populated as for "A * C" before returning. -** -** 2. NEAR is treated as AND. If the expression is "x NEAR y", it is -** advanced to point to the next row that matches "x AND y". -** -** See fts3EvalTestDeferredAndNear() for details on testing if a row is -** really a match, taking into account deferred tokens and NEAR operators. -*/ -static void fts3EvalNextRow( - Fts3Cursor *pCsr, /* FTS Cursor handle */ - Fts3Expr *pExpr, /* Expr. to advance to next matching row */ - int *pRc /* IN/OUT: Error code */ -){ - if( *pRc==SQLITE_OK ){ - int bDescDoclist = pCsr->bDesc; /* Used by DOCID_CMP() macro */ - assert( pExpr->bEof==0 ); - pExpr->bStart = 1; - - switch( pExpr->eType ){ - case FTSQUERY_NEAR: - case FTSQUERY_AND: { - Fts3Expr *pLeft = pExpr->pLeft; - Fts3Expr *pRight = pExpr->pRight; - assert( !pLeft->bDeferred || !pRight->bDeferred ); - - if( pLeft->bDeferred ){ - /* LHS is entirely deferred. So we assume it matches every row. - ** Advance the RHS iterator to find the next row visited. */ - fts3EvalNextRow(pCsr, pRight, pRc); - pExpr->iDocid = pRight->iDocid; - pExpr->bEof = pRight->bEof; - }else if( pRight->bDeferred ){ - /* RHS is entirely deferred. So we assume it matches every row. - ** Advance the LHS iterator to find the next row visited. */ - fts3EvalNextRow(pCsr, pLeft, pRc); - pExpr->iDocid = pLeft->iDocid; - pExpr->bEof = pLeft->bEof; - }else{ - /* Neither the RHS or LHS are deferred. */ - fts3EvalNextRow(pCsr, pLeft, pRc); - fts3EvalNextRow(pCsr, pRight, pRc); - while( !pLeft->bEof && !pRight->bEof && *pRc==SQLITE_OK ){ - sqlite3_int64 iDiff = DOCID_CMP(pLeft->iDocid, pRight->iDocid); - if( iDiff==0 ) break; - if( iDiff<0 ){ - fts3EvalNextRow(pCsr, pLeft, pRc); - }else{ - fts3EvalNextRow(pCsr, pRight, pRc); - } - } - pExpr->iDocid = pLeft->iDocid; - pExpr->bEof = (pLeft->bEof || pRight->bEof); - } - break; - } - - case FTSQUERY_OR: { - Fts3Expr *pLeft = pExpr->pLeft; - Fts3Expr *pRight = pExpr->pRight; - sqlite3_int64 iCmp = DOCID_CMP(pLeft->iDocid, pRight->iDocid); - - assert( pLeft->bStart || pLeft->iDocid==pRight->iDocid ); - assert( pRight->bStart || pLeft->iDocid==pRight->iDocid ); - - if( pRight->bEof || (pLeft->bEof==0 && iCmp<0) ){ - fts3EvalNextRow(pCsr, pLeft, pRc); - }else if( pLeft->bEof || (pRight->bEof==0 && iCmp>0) ){ - fts3EvalNextRow(pCsr, pRight, pRc); - }else{ - fts3EvalNextRow(pCsr, pLeft, pRc); - fts3EvalNextRow(pCsr, pRight, pRc); - } - - pExpr->bEof = (pLeft->bEof && pRight->bEof); - iCmp = DOCID_CMP(pLeft->iDocid, pRight->iDocid); - if( pRight->bEof || (pLeft->bEof==0 && iCmp<0) ){ - pExpr->iDocid = pLeft->iDocid; - }else{ - pExpr->iDocid = pRight->iDocid; - } - - break; - } - - case FTSQUERY_NOT: { - Fts3Expr *pLeft = pExpr->pLeft; - Fts3Expr *pRight = pExpr->pRight; - - if( pRight->bStart==0 ){ - fts3EvalNextRow(pCsr, pRight, pRc); - assert( *pRc!=SQLITE_OK || pRight->bStart ); - } - - fts3EvalNextRow(pCsr, pLeft, pRc); - if( pLeft->bEof==0 ){ - while( !*pRc - && !pRight->bEof - && DOCID_CMP(pLeft->iDocid, pRight->iDocid)>0 - ){ - fts3EvalNextRow(pCsr, pRight, pRc); - } - } - pExpr->iDocid = pLeft->iDocid; - pExpr->bEof = pLeft->bEof; - break; - } - - default: { - Fts3Phrase *pPhrase = pExpr->pPhrase; - fts3EvalInvalidatePoslist(pPhrase); - *pRc = fts3EvalPhraseNext(pCsr, pPhrase, &pExpr->bEof); - pExpr->iDocid = pPhrase->doclist.iDocid; - break; - } - } - } -} - -/* -** If *pRc is not SQLITE_OK, or if pExpr is not the root node of a NEAR -** cluster, then this function returns 1 immediately. -** -** Otherwise, it checks if the current row really does match the NEAR -** expression, using the data currently stored in the position lists -** (Fts3Expr->pPhrase.doclist.pList/nList) for each phrase in the expression. -** -** If the current row is a match, the position list associated with each -** phrase in the NEAR expression is edited in place to contain only those -** phrase instances sufficiently close to their peers to satisfy all NEAR -** constraints. In this case it returns 1. If the NEAR expression does not -** match the current row, 0 is returned. The position lists may or may not -** be edited if 0 is returned. -*/ -static int fts3EvalNearTest(Fts3Expr *pExpr, int *pRc){ - int res = 1; - - /* The following block runs if pExpr is the root of a NEAR query. - ** For example, the query: - ** - ** "w" NEAR "x" NEAR "y" NEAR "z" - ** - ** which is represented in tree form as: - ** - ** | - ** +--NEAR--+ <-- root of NEAR query - ** | | - ** +--NEAR--+ "z" - ** | | - ** +--NEAR--+ "y" - ** | | - ** "w" "x" - ** - ** The right-hand child of a NEAR node is always a phrase. The - ** left-hand child may be either a phrase or a NEAR node. There are - ** no exceptions to this - it's the way the parser in fts3_expr.c works. - */ - if( *pRc==SQLITE_OK - && pExpr->eType==FTSQUERY_NEAR - && pExpr->bEof==0 - && (pExpr->pParent==0 || pExpr->pParent->eType!=FTSQUERY_NEAR) - ){ - Fts3Expr *p; - int nTmp = 0; /* Bytes of temp space */ - char *aTmp; /* Temp space for PoslistNearMerge() */ - - /* Allocate temporary working space. */ - for(p=pExpr; p->pLeft; p=p->pLeft){ - nTmp += p->pRight->pPhrase->doclist.nList; - } - nTmp += p->pPhrase->doclist.nList; - if( nTmp==0 ){ - res = 0; - }else{ - aTmp = sqlite3_malloc(nTmp*2); - if( !aTmp ){ - *pRc = SQLITE_NOMEM; - res = 0; - }else{ - char *aPoslist = p->pPhrase->doclist.pList; - int nToken = p->pPhrase->nToken; - - for(p=p->pParent;res && p && p->eType==FTSQUERY_NEAR; p=p->pParent){ - Fts3Phrase *pPhrase = p->pRight->pPhrase; - int nNear = p->nNear; - res = fts3EvalNearTrim(nNear, aTmp, &aPoslist, &nToken, pPhrase); - } - - aPoslist = pExpr->pRight->pPhrase->doclist.pList; - nToken = pExpr->pRight->pPhrase->nToken; - for(p=pExpr->pLeft; p && res; p=p->pLeft){ - int nNear; - Fts3Phrase *pPhrase; - assert( p->pParent && p->pParent->pLeft==p ); - nNear = p->pParent->nNear; - pPhrase = ( - p->eType==FTSQUERY_NEAR ? p->pRight->pPhrase : p->pPhrase - ); - res = fts3EvalNearTrim(nNear, aTmp, &aPoslist, &nToken, pPhrase); - } - } - - sqlite3_free(aTmp); - } - } - - return res; -} - -/* -** This function is a helper function for fts3EvalTestDeferredAndNear(). -** Assuming no error occurs or has occurred, It returns non-zero if the -** expression passed as the second argument matches the row that pCsr -** currently points to, or zero if it does not. -** -** If *pRc is not SQLITE_OK when this function is called, it is a no-op. -** If an error occurs during execution of this function, *pRc is set to -** the appropriate SQLite error code. In this case the returned value is -** undefined. -*/ -static int fts3EvalTestExpr( - Fts3Cursor *pCsr, /* FTS cursor handle */ - Fts3Expr *pExpr, /* Expr to test. May or may not be root. */ - int *pRc /* IN/OUT: Error code */ -){ - int bHit = 1; /* Return value */ - if( *pRc==SQLITE_OK ){ - switch( pExpr->eType ){ - case FTSQUERY_NEAR: - case FTSQUERY_AND: - bHit = ( - fts3EvalTestExpr(pCsr, pExpr->pLeft, pRc) - && fts3EvalTestExpr(pCsr, pExpr->pRight, pRc) - && fts3EvalNearTest(pExpr, pRc) - ); - - /* If the NEAR expression does not match any rows, zero the doclist for - ** all phrases involved in the NEAR. This is because the snippet(), - ** offsets() and matchinfo() functions are not supposed to recognize - ** any instances of phrases that are part of unmatched NEAR queries. - ** For example if this expression: - ** - ** ... MATCH 'a OR (b NEAR c)' - ** - ** is matched against a row containing: - ** - ** 'a b d e' - ** - ** then any snippet() should ony highlight the "a" term, not the "b" - ** (as "b" is part of a non-matching NEAR clause). - */ - if( bHit==0 - && pExpr->eType==FTSQUERY_NEAR - && (pExpr->pParent==0 || pExpr->pParent->eType!=FTSQUERY_NEAR) - ){ - Fts3Expr *p; - for(p=pExpr; p->pPhrase==0; p=p->pLeft){ - if( p->pRight->iDocid==pCsr->iPrevId ){ - fts3EvalInvalidatePoslist(p->pRight->pPhrase); - } - } - if( p->iDocid==pCsr->iPrevId ){ - fts3EvalInvalidatePoslist(p->pPhrase); - } - } - - break; - - case FTSQUERY_OR: { - int bHit1 = fts3EvalTestExpr(pCsr, pExpr->pLeft, pRc); - int bHit2 = fts3EvalTestExpr(pCsr, pExpr->pRight, pRc); - bHit = bHit1 || bHit2; - break; - } - - case FTSQUERY_NOT: - bHit = ( - fts3EvalTestExpr(pCsr, pExpr->pLeft, pRc) - && !fts3EvalTestExpr(pCsr, pExpr->pRight, pRc) - ); - break; - - default: { -#ifndef SQLITE_DISABLE_FTS4_DEFERRED - if( pCsr->pDeferred - && (pExpr->iDocid==pCsr->iPrevId || pExpr->bDeferred) - ){ - Fts3Phrase *pPhrase = pExpr->pPhrase; - assert( pExpr->bDeferred || pPhrase->doclist.bFreeList==0 ); - if( pExpr->bDeferred ){ - fts3EvalInvalidatePoslist(pPhrase); - } - *pRc = fts3EvalDeferredPhrase(pCsr, pPhrase); - bHit = (pPhrase->doclist.pList!=0); - pExpr->iDocid = pCsr->iPrevId; - }else -#endif - { - bHit = (pExpr->bEof==0 && pExpr->iDocid==pCsr->iPrevId); - } - break; - } - } - } - return bHit; -} - -/* -** This function is called as the second part of each xNext operation when -** iterating through the results of a full-text query. At this point the -** cursor points to a row that matches the query expression, with the -** following caveats: -** -** * Up until this point, "NEAR" operators in the expression have been -** treated as "AND". -** -** * Deferred tokens have not yet been considered. -** -** If *pRc is not SQLITE_OK when this function is called, it immediately -** returns 0. Otherwise, it tests whether or not after considering NEAR -** operators and deferred tokens the current row is still a match for the -** expression. It returns 1 if both of the following are true: -** -** 1. *pRc is SQLITE_OK when this function returns, and -** -** 2. After scanning the current FTS table row for the deferred tokens, -** it is determined that the row does *not* match the query. -** -** Or, if no error occurs and it seems the current row does match the FTS -** query, return 0. -*/ -static int fts3EvalTestDeferredAndNear(Fts3Cursor *pCsr, int *pRc){ - int rc = *pRc; - int bMiss = 0; - if( rc==SQLITE_OK ){ - - /* If there are one or more deferred tokens, load the current row into - ** memory and scan it to determine the position list for each deferred - ** token. Then, see if this row is really a match, considering deferred - ** tokens and NEAR operators (neither of which were taken into account - ** earlier, by fts3EvalNextRow()). - */ - if( pCsr->pDeferred ){ - rc = fts3CursorSeek(0, pCsr); - if( rc==SQLITE_OK ){ - rc = sqlite3Fts3CacheDeferredDoclists(pCsr); - } - } - bMiss = (0==fts3EvalTestExpr(pCsr, pCsr->pExpr, &rc)); - - /* Free the position-lists accumulated for each deferred token above. */ - sqlite3Fts3FreeDeferredDoclists(pCsr); - *pRc = rc; - } - return (rc==SQLITE_OK && bMiss); -} - -/* -** Advance to the next document that matches the FTS expression in -** Fts3Cursor.pExpr. -*/ -static int fts3EvalNext(Fts3Cursor *pCsr){ - int rc = SQLITE_OK; /* Return Code */ - Fts3Expr *pExpr = pCsr->pExpr; - assert( pCsr->isEof==0 ); - if( pExpr==0 ){ - pCsr->isEof = 1; - }else{ - do { - if( pCsr->isRequireSeek==0 ){ - sqlite3_reset(pCsr->pStmt); - } - assert( sqlite3_data_count(pCsr->pStmt)==0 ); - fts3EvalNextRow(pCsr, pExpr, &rc); - pCsr->isEof = pExpr->bEof; - pCsr->isRequireSeek = 1; - pCsr->isMatchinfoNeeded = 1; - pCsr->iPrevId = pExpr->iDocid; - }while( pCsr->isEof==0 && fts3EvalTestDeferredAndNear(pCsr, &rc) ); - } - return rc; -} - -/* -** Restart interation for expression pExpr so that the next call to -** fts3EvalNext() visits the first row. Do not allow incremental -** loading or merging of phrase doclists for this iteration. -** -** If *pRc is other than SQLITE_OK when this function is called, it is -** a no-op. If an error occurs within this function, *pRc is set to an -** SQLite error code before returning. -*/ -static void fts3EvalRestart( - Fts3Cursor *pCsr, - Fts3Expr *pExpr, - int *pRc -){ - if( pExpr && *pRc==SQLITE_OK ){ - Fts3Phrase *pPhrase = pExpr->pPhrase; - - if( pPhrase ){ - fts3EvalInvalidatePoslist(pPhrase); - if( pPhrase->bIncr ){ - assert( pPhrase->nToken==1 ); - assert( pPhrase->aToken[0].pSegcsr ); - sqlite3Fts3MsrIncrRestart(pPhrase->aToken[0].pSegcsr); - *pRc = fts3EvalPhraseStart(pCsr, 0, pPhrase); - } - - pPhrase->doclist.pNextDocid = 0; - pPhrase->doclist.iDocid = 0; - } - - pExpr->iDocid = 0; - pExpr->bEof = 0; - pExpr->bStart = 0; - - fts3EvalRestart(pCsr, pExpr->pLeft, pRc); - fts3EvalRestart(pCsr, pExpr->pRight, pRc); - } -} - -/* -** After allocating the Fts3Expr.aMI[] array for each phrase in the -** expression rooted at pExpr, the cursor iterates through all rows matched -** by pExpr, calling this function for each row. This function increments -** the values in Fts3Expr.aMI[] according to the position-list currently -** found in Fts3Expr.pPhrase->doclist.pList for each of the phrase -** expression nodes. -*/ -static void fts3EvalUpdateCounts(Fts3Expr *pExpr){ - if( pExpr ){ - Fts3Phrase *pPhrase = pExpr->pPhrase; - if( pPhrase && pPhrase->doclist.pList ){ - int iCol = 0; - char *p = pPhrase->doclist.pList; - - assert( *p ); - while( 1 ){ - u8 c = 0; - int iCnt = 0; - while( 0xFE & (*p | c) ){ - if( (c&0x80)==0 ) iCnt++; - c = *p++ & 0x80; - } - - /* aMI[iCol*3 + 1] = Number of occurrences - ** aMI[iCol*3 + 2] = Number of rows containing at least one instance - */ - pExpr->aMI[iCol*3 + 1] += iCnt; - pExpr->aMI[iCol*3 + 2] += (iCnt>0); - if( *p==0x00 ) break; - p++; - p += sqlite3Fts3GetVarint32(p, &iCol); - } - } - - fts3EvalUpdateCounts(pExpr->pLeft); - fts3EvalUpdateCounts(pExpr->pRight); - } -} - -/* -** Expression pExpr must be of type FTSQUERY_PHRASE. -** -** If it is not already allocated and populated, this function allocates and -** populates the Fts3Expr.aMI[] array for expression pExpr. If pExpr is part -** of a NEAR expression, then it also allocates and populates the same array -** for all other phrases that are part of the NEAR expression. -** -** SQLITE_OK is returned if the aMI[] array is successfully allocated and -** populated. Otherwise, if an error occurs, an SQLite error code is returned. -*/ -static int fts3EvalGatherStats( - Fts3Cursor *pCsr, /* Cursor object */ - Fts3Expr *pExpr /* FTSQUERY_PHRASE expression */ -){ - int rc = SQLITE_OK; /* Return code */ - - assert( pExpr->eType==FTSQUERY_PHRASE ); - if( pExpr->aMI==0 ){ - Fts3Table *pTab = (Fts3Table *)pCsr->base.pVtab; - Fts3Expr *pRoot; /* Root of NEAR expression */ - Fts3Expr *p; /* Iterator used for several purposes */ - - sqlite3_int64 iPrevId = pCsr->iPrevId; - sqlite3_int64 iDocid; - u8 bEof; - - /* Find the root of the NEAR expression */ - pRoot = pExpr; - while( pRoot->pParent && pRoot->pParent->eType==FTSQUERY_NEAR ){ - pRoot = pRoot->pParent; - } - iDocid = pRoot->iDocid; - bEof = pRoot->bEof; - assert( pRoot->bStart ); - - /* Allocate space for the aMSI[] array of each FTSQUERY_PHRASE node */ - for(p=pRoot; p; p=p->pLeft){ - Fts3Expr *pE = (p->eType==FTSQUERY_PHRASE?p:p->pRight); - assert( pE->aMI==0 ); - pE->aMI = (u32 *)sqlite3_malloc(pTab->nColumn * 3 * sizeof(u32)); - if( !pE->aMI ) return SQLITE_NOMEM; - memset(pE->aMI, 0, pTab->nColumn * 3 * sizeof(u32)); - } - - fts3EvalRestart(pCsr, pRoot, &rc); - - while( pCsr->isEof==0 && rc==SQLITE_OK ){ - - do { - /* Ensure the %_content statement is reset. */ - if( pCsr->isRequireSeek==0 ) sqlite3_reset(pCsr->pStmt); - assert( sqlite3_data_count(pCsr->pStmt)==0 ); - - /* Advance to the next document */ - fts3EvalNextRow(pCsr, pRoot, &rc); - pCsr->isEof = pRoot->bEof; - pCsr->isRequireSeek = 1; - pCsr->isMatchinfoNeeded = 1; - pCsr->iPrevId = pRoot->iDocid; - }while( pCsr->isEof==0 - && pRoot->eType==FTSQUERY_NEAR - && fts3EvalTestDeferredAndNear(pCsr, &rc) - ); - - if( rc==SQLITE_OK && pCsr->isEof==0 ){ - fts3EvalUpdateCounts(pRoot); - } - } - - pCsr->isEof = 0; - pCsr->iPrevId = iPrevId; - - if( bEof ){ - pRoot->bEof = bEof; - }else{ - /* Caution: pRoot may iterate through docids in ascending or descending - ** order. For this reason, even though it seems more defensive, the - ** do loop can not be written: - ** - ** do {...} while( pRoot->iDocid<iDocid && rc==SQLITE_OK ); - */ - fts3EvalRestart(pCsr, pRoot, &rc); - do { - fts3EvalNextRow(pCsr, pRoot, &rc); - assert( pRoot->bEof==0 ); - }while( pRoot->iDocid!=iDocid && rc==SQLITE_OK ); - fts3EvalTestDeferredAndNear(pCsr, &rc); - } - } - return rc; -} - -/* -** This function is used by the matchinfo() module to query a phrase -** expression node for the following information: -** -** 1. The total number of occurrences of the phrase in each column of -** the FTS table (considering all rows), and -** -** 2. For each column, the number of rows in the table for which the -** column contains at least one instance of the phrase. -** -** If no error occurs, SQLITE_OK is returned and the values for each column -** written into the array aiOut as follows: -** -** aiOut[iCol*3 + 1] = Number of occurrences -** aiOut[iCol*3 + 2] = Number of rows containing at least one instance -** -** Caveats: -** -** * If a phrase consists entirely of deferred tokens, then all output -** values are set to the number of documents in the table. In other -** words we assume that very common tokens occur exactly once in each -** column of each row of the table. -** -** * If a phrase contains some deferred tokens (and some non-deferred -** tokens), count the potential occurrence identified by considering -** the non-deferred tokens instead of actual phrase occurrences. -** -** * If the phrase is part of a NEAR expression, then only phrase instances -** that meet the NEAR constraint are included in the counts. -*/ -int sqlite3Fts3EvalPhraseStats( - Fts3Cursor *pCsr, /* FTS cursor handle */ - Fts3Expr *pExpr, /* Phrase expression */ - u32 *aiOut /* Array to write results into (see above) */ -){ - Fts3Table *pTab = (Fts3Table *)pCsr->base.pVtab; - int rc = SQLITE_OK; - int iCol; - - if( pExpr->bDeferred && pExpr->pParent->eType!=FTSQUERY_NEAR ){ - assert( pCsr->nDoc>0 ); - for(iCol=0; iCol<pTab->nColumn; iCol++){ - aiOut[iCol*3 + 1] = (u32)pCsr->nDoc; - aiOut[iCol*3 + 2] = (u32)pCsr->nDoc; - } - }else{ - rc = fts3EvalGatherStats(pCsr, pExpr); - if( rc==SQLITE_OK ){ - assert( pExpr->aMI ); - for(iCol=0; iCol<pTab->nColumn; iCol++){ - aiOut[iCol*3 + 1] = pExpr->aMI[iCol*3 + 1]; - aiOut[iCol*3 + 2] = pExpr->aMI[iCol*3 + 2]; - } - } - } - - return rc; -} - -/* -** The expression pExpr passed as the second argument to this function -** must be of type FTSQUERY_PHRASE. -** -** The returned value is either NULL or a pointer to a buffer containing -** a position-list indicating the occurrences of the phrase in column iCol -** of the current row. -** -** More specifically, the returned buffer contains 1 varint for each -** occurence of the phrase in the column, stored using the normal (delta+2) -** compression and is terminated by either an 0x01 or 0x00 byte. For example, -** if the requested column contains "a b X c d X X" and the position-list -** for 'X' is requested, the buffer returned may contain: -** -** 0x04 0x05 0x03 0x01 or 0x04 0x05 0x03 0x00 -** -** This function works regardless of whether or not the phrase is deferred, -** incremental, or neither. -*/ -int sqlite3Fts3EvalPhrasePoslist( - Fts3Cursor *pCsr, /* FTS3 cursor object */ - Fts3Expr *pExpr, /* Phrase to return doclist for */ - int iCol, /* Column to return position list for */ - char **ppOut /* OUT: Pointer to position list */ -){ - Fts3Phrase *pPhrase = pExpr->pPhrase; - Fts3Table *pTab = (Fts3Table *)pCsr->base.pVtab; - char *pIter; - int iThis; - sqlite3_int64 iDocid; - - /* If this phrase is applies specifically to some column other than - ** column iCol, return a NULL pointer. */ - *ppOut = 0; - assert( iCol>=0 && iCol<pTab->nColumn ); - if( (pPhrase->iColumn<pTab->nColumn && pPhrase->iColumn!=iCol) ){ - return SQLITE_OK; - } - - iDocid = pExpr->iDocid; - pIter = pPhrase->doclist.pList; - if( iDocid!=pCsr->iPrevId || pExpr->bEof ){ - int bDescDoclist = pTab->bDescIdx; /* For DOCID_CMP macro */ - int bOr = 0; - u8 bEof = 0; - Fts3Expr *p; - - /* Check if this phrase descends from an OR expression node. If not, - ** return NULL. Otherwise, the entry that corresponds to docid - ** pCsr->iPrevId may lie earlier in the doclist buffer. */ - for(p=pExpr->pParent; p; p=p->pParent){ - if( p->eType==FTSQUERY_OR ) bOr = 1; - } - if( bOr==0 ) return SQLITE_OK; - - /* This is the descendent of an OR node. In this case we cannot use - ** an incremental phrase. Load the entire doclist for the phrase - ** into memory in this case. */ - if( pPhrase->bIncr ){ - int rc = SQLITE_OK; - int bEofSave = pExpr->bEof; - fts3EvalRestart(pCsr, pExpr, &rc); - while( rc==SQLITE_OK && !pExpr->bEof ){ - fts3EvalNextRow(pCsr, pExpr, &rc); - if( bEofSave==0 && pExpr->iDocid==iDocid ) break; - } - pIter = pPhrase->doclist.pList; - assert( rc!=SQLITE_OK || pPhrase->bIncr==0 ); - if( rc!=SQLITE_OK ) return rc; - } - - if( pExpr->bEof ){ - pIter = 0; - iDocid = 0; - } - bEof = (pPhrase->doclist.nAll==0); - assert( bDescDoclist==0 || bDescDoclist==1 ); - assert( pCsr->bDesc==0 || pCsr->bDesc==1 ); - - if( pCsr->bDesc==bDescDoclist ){ - int dummy; - while( (pIter==0 || DOCID_CMP(iDocid, pCsr->iPrevId)>0 ) && bEof==0 ){ - sqlite3Fts3DoclistPrev( - bDescDoclist, pPhrase->doclist.aAll, pPhrase->doclist.nAll, - &pIter, &iDocid, &dummy, &bEof - ); - } - }else{ - while( (pIter==0 || DOCID_CMP(iDocid, pCsr->iPrevId)<0 ) && bEof==0 ){ - sqlite3Fts3DoclistNext( - bDescDoclist, pPhrase->doclist.aAll, pPhrase->doclist.nAll, - &pIter, &iDocid, &bEof - ); - } - } - - if( bEof || iDocid!=pCsr->iPrevId ) pIter = 0; - } - if( pIter==0 ) return SQLITE_OK; - - if( *pIter==0x01 ){ - pIter++; - pIter += sqlite3Fts3GetVarint32(pIter, &iThis); - }else{ - iThis = 0; - } - while( iThis<iCol ){ - fts3ColumnlistCopy(0, &pIter); - if( *pIter==0x00 ) return 0; - pIter++; - pIter += sqlite3Fts3GetVarint32(pIter, &iThis); - } - - *ppOut = ((iCol==iThis)?pIter:0); - return SQLITE_OK; -} - -/* -** Free all components of the Fts3Phrase structure that were allocated by -** the eval module. Specifically, this means to free: -** -** * the contents of pPhrase->doclist, and -** * any Fts3MultiSegReader objects held by phrase tokens. -*/ -void sqlite3Fts3EvalPhraseCleanup(Fts3Phrase *pPhrase){ - if( pPhrase ){ - int i; - sqlite3_free(pPhrase->doclist.aAll); - fts3EvalInvalidatePoslist(pPhrase); - memset(&pPhrase->doclist, 0, sizeof(Fts3Doclist)); - for(i=0; i<pPhrase->nToken; i++){ - fts3SegReaderCursorFree(pPhrase->aToken[i].pSegcsr); - pPhrase->aToken[i].pSegcsr = 0; - } - } -} - - -/* -** Return SQLITE_CORRUPT_VTAB. -*/ -#ifdef SQLITE_DEBUG -int sqlite3Fts3Corrupt(){ - return SQLITE_CORRUPT_VTAB; -} -#endif - -#if !SQLITE_CORE -/* -** Initialize API pointer table, if required. -*/ -int fts4_extension_init( - sqlite3 *db, - char **pzErrMsg, - void *pApi -){ - SQLITE_EXTENSION_INIT2(pApi) - return sqlite3Fts3Init(db); -} -#endif - -#endif diff --git a/src/libtracker-fts/fts3.h b/src/libtracker-fts/fts3.h deleted file mode 100644 index b0346826d..000000000 --- a/src/libtracker-fts/fts3.h +++ /dev/null @@ -1,32 +0,0 @@ -/* -** 2006 Oct 10 -** -** The author disclaims copyright to this source code. In place of -** a legal notice, here is a blessing: -** -** May you do good and not evil. -** May you find forgiveness for yourself and forgive others. -** May you share freely, never taking more than you give. -** -****************************************************************************** -** -** This header file is used by programs that want to link against the -** FTS3 library. All it does is declare the sqlite3Fts3Init() interface. -*/ -#include "sqlite3.h" - -#ifdef __cplusplus -extern "C" { -#endif /* __cplusplus */ - -int sqlite3Fts3Init(sqlite3 *db); - -int fts4_extension_init( - sqlite3 *db, - char **pzErrMsg, - void *pApi -); - -#ifdef __cplusplus -} /* extern "C" */ -#endif /* __cplusplus */ diff --git a/src/libtracker-fts/fts3Int.h b/src/libtracker-fts/fts3Int.h deleted file mode 100644 index 77ca4704e..000000000 --- a/src/libtracker-fts/fts3Int.h +++ /dev/null @@ -1,560 +0,0 @@ -/* -** 2009 Nov 12 -** -** The author disclaims copyright to this source code. In place of -** a legal notice, here is a blessing: -** -** May you do good and not evil. -** May you find forgiveness for yourself and forgive others. -** May you share freely, never taking more than you give. -** -****************************************************************************** -** -*/ -#ifndef _FTSINT_H -#define _FTSINT_H - -#if !defined(NDEBUG) && !defined(SQLITE_DEBUG) -# define NDEBUG 1 -#endif - -/* -** FTS4 is really an extension for FTS3. It is enabled using the -** SQLITE_ENABLE_FTS3 macro. But to avoid confusion we also all -** the SQLITE_ENABLE_FTS4 macro to serve as an alisse for SQLITE_ENABLE_FTS3. -*/ -#if defined(SQLITE_ENABLE_FTS4) && !defined(SQLITE_ENABLE_FTS3) -# define SQLITE_ENABLE_FTS3 -#endif - -#if !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3) - -/* If not building as part of the core, include sqlite3ext.h. */ -#ifndef SQLITE_CORE -# include "sqlite3ext.h" -extern const sqlite3_api_routines *sqlite3_api; -#endif - -#include "sqlite3.h" -#include "fts3_tokenizer.h" -#include "fts3_hash.h" - -/* -** This constant controls how often segments are merged. Once there are -** FTS3_MERGE_COUNT segments of level N, they are merged into a single -** segment of level N+1. -*/ -#define FTS3_MERGE_COUNT 16 - -/* -** This is the maximum amount of data (in bytes) to store in the -** Fts3Table.pendingTerms hash table. Normally, the hash table is -** populated as documents are inserted/updated/deleted in a transaction -** and used to create a new segment when the transaction is committed. -** However if this limit is reached midway through a transaction, a new -** segment is created and the hash table cleared immediately. -*/ -#define FTS3_MAX_PENDING_DATA (1*1024*1024) - -/* -** Macro to return the number of elements in an array. SQLite has a -** similar macro called ArraySize(). Use a different name to avoid -** a collision when building an amalgamation with built-in FTS3. -*/ -#define SizeofArray(X) ((int)(sizeof(X)/sizeof(X[0]))) - - -#ifndef MIN -# define MIN(x,y) ((x)<(y)?(x):(y)) -#endif -#ifndef MAX -# define MAX(x,y) ((x)>(y)?(x):(y)) -#endif - -/* -** Maximum length of a varint encoded integer. The varint format is different -** from that used by SQLite, so the maximum length is 10, not 9. -*/ -#define FTS3_VARINT_MAX 10 - -/* -** FTS4 virtual tables may maintain multiple indexes - one index of all terms -** in the document set and zero or more prefix indexes. All indexes are stored -** as one or more b+-trees in the %_segments and %_segdir tables. -** -** It is possible to determine which index a b+-tree belongs to based on the -** value stored in the "%_segdir.level" column. Given this value L, the index -** that the b+-tree belongs to is (L<<10). In other words, all b+-trees with -** level values between 0 and 1023 (inclusive) belong to index 0, all levels -** between 1024 and 2047 to index 1, and so on. -** -** It is considered impossible for an index to use more than 1024 levels. In -** theory though this may happen, but only after at least -** (FTS3_MERGE_COUNT^1024) separate flushes of the pending-terms tables. -*/ -#define FTS3_SEGDIR_MAXLEVEL 1024 -#define FTS3_SEGDIR_MAXLEVEL_STR "1024" - -/* -** The testcase() macro is only used by the amalgamation. If undefined, -** make it a no-op. -*/ -#ifndef testcase -# define testcase(X) -#endif - -/* -** Terminator values for position-lists and column-lists. -*/ -#define POS_COLUMN (1) /* Column-list terminator */ -#define POS_END (0) /* Position-list terminator */ - -/* -** This section provides definitions to allow the -** FTS3 extension to be compiled outside of the -** amalgamation. -*/ -#ifndef SQLITE_AMALGAMATION -/* -** Macros indicating that conditional expressions are always true or -** false. -*/ -#ifdef SQLITE_COVERAGE_TEST -# define ALWAYS(x) (1) -# define NEVER(X) (0) -#else -# define ALWAYS(x) (x) -# define NEVER(x) (x) -#endif - -/* -** Internal types used by SQLite. -*/ -typedef unsigned char u8; /* 1-byte (or larger) unsigned integer */ -typedef short int i16; /* 2-byte (or larger) signed integer */ -typedef unsigned int u32; /* 4-byte unsigned integer */ -typedef sqlite3_uint64 u64; /* 8-byte unsigned integer */ -typedef sqlite3_int64 i64; /* 8-byte signed integer */ - -/* -** Macro used to suppress compiler warnings for unused parameters. -*/ -#define UNUSED_PARAMETER(x) (void)(x) - -/* -** Activate assert() only if SQLITE_TEST is enabled. -*/ -#if !defined(NDEBUG) && !defined(SQLITE_DEBUG) -# define NDEBUG 1 -#endif - -/* -** The TESTONLY macro is used to enclose variable declarations or -** other bits of code that are needed to support the arguments -** within testcase() and assert() macros. -*/ -#if defined(SQLITE_DEBUG) || defined(SQLITE_COVERAGE_TEST) -# define TESTONLY(X) X -#else -# define TESTONLY(X) -#endif - -#endif /* SQLITE_AMALGAMATION */ - -#ifdef SQLITE_DEBUG -int sqlite3Fts3Corrupt(void); -# define FTS_CORRUPT_VTAB sqlite3Fts3Corrupt() -#else -# define FTS_CORRUPT_VTAB SQLITE_CORRUPT_VTAB -#endif - -typedef struct Fts3Table Fts3Table; -typedef struct Fts3Cursor Fts3Cursor; -typedef struct Fts3Expr Fts3Expr; -typedef struct Fts3Phrase Fts3Phrase; -typedef struct Fts3PhraseToken Fts3PhraseToken; - -typedef struct Fts3Doclist Fts3Doclist; -typedef struct Fts3SegFilter Fts3SegFilter; -typedef struct Fts3DeferredToken Fts3DeferredToken; -typedef struct Fts3SegReader Fts3SegReader; -typedef struct Fts3MultiSegReader Fts3MultiSegReader; - -/* -** A connection to a fulltext index is an instance of the following -** structure. The xCreate and xConnect methods create an instance -** of this structure and xDestroy and xDisconnect free that instance. -** All other methods receive a pointer to the structure as one of their -** arguments. -*/ -struct Fts3Table { - sqlite3_vtab base; /* Base class used by SQLite core */ - sqlite3 *db; /* The database connection */ - const char *zDb; /* logical database name */ - const char *zName; /* virtual table name */ - int nColumn; /* number of named columns in virtual table */ - char **azColumn; /* column names. malloced */ - sqlite3_tokenizer *pTokenizer; /* tokenizer for inserts and queries */ - char *zContentTbl; /* content=xxx option, or NULL */ - char *zLanguageid; /* languageid=xxx option, or NULL */ - u8 bAutoincrmerge; /* True if automerge=1 */ - u32 nLeafAdd; /* Number of leaf blocks added this trans */ - - /* Precompiled statements used by the implementation. Each of these - ** statements is run and reset within a single virtual table API call. - */ - sqlite3_stmt *aStmt[37]; - - char *zReadExprlist; - char *zWriteExprlist; - - int nNodeSize; /* Soft limit for node size */ - u8 bFts4; /* True for FTS4, false for FTS3 */ - u8 bHasStat; /* True if %_stat table exists */ - u8 bHasDocsize; /* True if %_docsize table exists */ - u8 bDescIdx; /* True if doclists are in reverse order */ - u8 bIgnoreSavepoint; /* True to ignore xSavepoint invocations */ - int nPgsz; /* Page size for host database */ - char *zSegmentsTbl; /* Name of %_segments table */ - sqlite3_blob *pSegments; /* Blob handle open on %_segments table */ - - /* - ** The following array of hash tables is used to buffer pending index - ** updates during transactions. All pending updates buffered at any one - ** time must share a common language-id (see the FTS4 langid= feature). - ** The current language id is stored in variable iPrevLangid. - ** - ** A single FTS4 table may have multiple full-text indexes. For each index - ** there is an entry in the aIndex[] array. Index 0 is an index of all the - ** terms that appear in the document set. Each subsequent index in aIndex[] - ** is an index of prefixes of a specific length. - ** - ** Variable nPendingData contains an estimate the memory consumed by the - ** pending data structures, including hash table overhead, but not including - ** malloc overhead. When nPendingData exceeds nMaxPendingData, all hash - ** tables are flushed to disk. Variable iPrevDocid is the docid of the most - ** recently inserted record. - */ - int nIndex; /* Size of aIndex[] */ - struct Fts3Index { - int nPrefix; /* Prefix length (0 for main terms index) */ - Fts3Hash hPending; /* Pending terms table for this index */ - } *aIndex; - int nMaxPendingData; /* Max pending data before flush to disk */ - int nPendingData; /* Current bytes of pending data */ - sqlite_int64 iPrevDocid; /* Docid of most recently inserted document */ - int iPrevLangid; /* Langid of recently inserted document */ - -#if defined(SQLITE_DEBUG) || defined(SQLITE_COVERAGE_TEST) - /* State variables used for validating that the transaction control - ** methods of the virtual table are called at appropriate times. These - ** values do not contribute to FTS functionality; they are used for - ** verifying the operation of the SQLite core. - */ - int inTransaction; /* True after xBegin but before xCommit/xRollback */ - int mxSavepoint; /* Largest valid xSavepoint integer */ -#endif -}; - -/* -** When the core wants to read from the virtual table, it creates a -** virtual table cursor (an instance of the following structure) using -** the xOpen method. Cursors are destroyed using the xClose method. -*/ -struct Fts3Cursor { - sqlite3_vtab_cursor base; /* Base class used by SQLite core */ - i16 eSearch; /* Search strategy (see below) */ - u8 isEof; /* True if at End Of Results */ - u8 isRequireSeek; /* True if must seek pStmt to %_content row */ - sqlite3_stmt *pStmt; /* Prepared statement in use by the cursor */ - Fts3Expr *pExpr; /* Parsed MATCH query string */ - int iLangid; /* Language being queried for */ - int nPhrase; /* Number of matchable phrases in query */ - Fts3DeferredToken *pDeferred; /* Deferred search tokens, if any */ - sqlite3_int64 iPrevId; /* Previous id read from aDoclist */ - char *pNextId; /* Pointer into the body of aDoclist */ - char *aDoclist; /* List of docids for full-text queries */ - int nDoclist; /* Size of buffer at aDoclist */ - u8 bDesc; /* True to sort in descending order */ - int eEvalmode; /* An FTS3_EVAL_XX constant */ - int nRowAvg; /* Average size of database rows, in pages */ - sqlite3_int64 nDoc; /* Documents in table */ - - int isMatchinfoNeeded; /* True when aMatchinfo[] needs filling in */ - u32 *aMatchinfo; /* Information about most recent match */ - int nMatchinfo; /* Number of elements in aMatchinfo[] */ - char *zMatchinfo; /* Matchinfo specification */ -}; - -#define FTS3_EVAL_FILTER 0 -#define FTS3_EVAL_NEXT 1 -#define FTS3_EVAL_MATCHINFO 2 - -/* -** The Fts3Cursor.eSearch member is always set to one of the following. -** Actualy, Fts3Cursor.eSearch can be greater than or equal to -** FTS3_FULLTEXT_SEARCH. If so, then Fts3Cursor.eSearch - 2 is the index -** of the column to be searched. For example, in -** -** CREATE VIRTUAL TABLE ex1 USING fts3(a,b,c,d); -** SELECT docid FROM ex1 WHERE b MATCH 'one two three'; -** -** Because the LHS of the MATCH operator is 2nd column "b", -** Fts3Cursor.eSearch will be set to FTS3_FULLTEXT_SEARCH+1. (+0 for a, -** +1 for b, +2 for c, +3 for d.) If the LHS of MATCH were "ex1" -** indicating that all columns should be searched, -** then eSearch would be set to FTS3_FULLTEXT_SEARCH+4. -*/ -#define FTS3_FULLSCAN_SEARCH 0 /* Linear scan of %_content table */ -#define FTS3_DOCID_SEARCH 1 /* Lookup by rowid on %_content table */ -#define FTS3_FULLTEXT_SEARCH 2 /* Full-text index search */ - - -struct Fts3Doclist { - char *aAll; /* Array containing doclist (or NULL) */ - int nAll; /* Size of a[] in bytes */ - char *pNextDocid; /* Pointer to next docid */ - - sqlite3_int64 iDocid; /* Current docid (if pList!=0) */ - int bFreeList; /* True if pList should be sqlite3_free()d */ - char *pList; /* Pointer to position list following iDocid */ - int nList; /* Length of position list */ -}; - -/* -** A "phrase" is a sequence of one or more tokens that must match in -** sequence. A single token is the base case and the most common case. -** For a sequence of tokens contained in double-quotes (i.e. "one two three") -** nToken will be the number of tokens in the string. -*/ -struct Fts3PhraseToken { - char *z; /* Text of the token */ - int n; /* Number of bytes in buffer z */ - int isPrefix; /* True if token ends with a "*" character */ - int bFirst; /* True if token must appear at position 0 */ - - /* Variables above this point are populated when the expression is - ** parsed (by code in fts3_expr.c). Below this point the variables are - ** used when evaluating the expression. */ - Fts3DeferredToken *pDeferred; /* Deferred token object for this token */ - Fts3MultiSegReader *pSegcsr; /* Segment-reader for this token */ -}; - -struct Fts3Phrase { - /* Cache of doclist for this phrase. */ - Fts3Doclist doclist; - int bIncr; /* True if doclist is loaded incrementally */ - int iDoclistToken; - - /* Variables below this point are populated by fts3_expr.c when parsing - ** a MATCH expression. Everything above is part of the evaluation phase. - */ - int nToken; /* Number of tokens in the phrase */ - int iColumn; /* Index of column this phrase must match */ - Fts3PhraseToken aToken[1]; /* One entry for each token in the phrase */ -}; - -/* -** A tree of these objects forms the RHS of a MATCH operator. -** -** If Fts3Expr.eType is FTSQUERY_PHRASE and isLoaded is true, then aDoclist -** points to a malloced buffer, size nDoclist bytes, containing the results -** of this phrase query in FTS3 doclist format. As usual, the initial -** "Length" field found in doclists stored on disk is omitted from this -** buffer. -** -** Variable aMI is used only for FTSQUERY_NEAR nodes to store the global -** matchinfo data. If it is not NULL, it points to an array of size nCol*3, -** where nCol is the number of columns in the queried FTS table. The array -** is populated as follows: -** -** aMI[iCol*3 + 0] = Undefined -** aMI[iCol*3 + 1] = Number of occurrences -** aMI[iCol*3 + 2] = Number of rows containing at least one instance -** -** The aMI array is allocated using sqlite3_malloc(). It should be freed -** when the expression node is. -*/ -struct Fts3Expr { - int eType; /* One of the FTSQUERY_XXX values defined below */ - int nNear; /* Valid if eType==FTSQUERY_NEAR */ - Fts3Expr *pParent; /* pParent->pLeft==this or pParent->pRight==this */ - Fts3Expr *pLeft; /* Left operand */ - Fts3Expr *pRight; /* Right operand */ - Fts3Phrase *pPhrase; /* Valid if eType==FTSQUERY_PHRASE */ - - /* The following are used by the fts3_eval.c module. */ - sqlite3_int64 iDocid; /* Current docid */ - u8 bEof; /* True this expression is at EOF already */ - u8 bStart; /* True if iDocid is valid */ - u8 bDeferred; /* True if this expression is entirely deferred */ - - u32 *aMI; -}; - -/* -** Candidate values for Fts3Query.eType. Note that the order of the first -** four values is in order of precedence when parsing expressions. For -** example, the following: -** -** "a OR b AND c NOT d NEAR e" -** -** is equivalent to: -** -** "a OR (b AND (c NOT (d NEAR e)))" -*/ -#define FTSQUERY_NEAR 1 -#define FTSQUERY_NOT 2 -#define FTSQUERY_AND 3 -#define FTSQUERY_OR 4 -#define FTSQUERY_PHRASE 5 - - -/* fts3_write.c */ -int sqlite3Fts3UpdateMethod(sqlite3_vtab*,int,sqlite3_value**,sqlite3_int64*); -int sqlite3Fts3PendingTermsFlush(Fts3Table *); -void sqlite3Fts3PendingTermsClear(Fts3Table *); -int sqlite3Fts3Optimize(Fts3Table *); -int sqlite3Fts3SegReaderNew(int, int, sqlite3_int64, - sqlite3_int64, sqlite3_int64, const char *, int, Fts3SegReader**); -int sqlite3Fts3SegReaderPending( - Fts3Table*,int,const char*,int,int,Fts3SegReader**); -void sqlite3Fts3SegReaderFree(Fts3SegReader *); -int sqlite3Fts3AllSegdirs(Fts3Table*, int, int, int, sqlite3_stmt **); -int sqlite3Fts3ReadLock(Fts3Table *); -int sqlite3Fts3ReadBlock(Fts3Table*, sqlite3_int64, char **, int*, int*); - -int sqlite3Fts3SelectDoctotal(Fts3Table *, sqlite3_stmt **); -int sqlite3Fts3SelectDocsize(Fts3Table *, sqlite3_int64, sqlite3_stmt **); - -#ifndef SQLITE_DISABLE_FTS4_DEFERRED -void sqlite3Fts3FreeDeferredTokens(Fts3Cursor *); -int sqlite3Fts3DeferToken(Fts3Cursor *, Fts3PhraseToken *, int); -int sqlite3Fts3CacheDeferredDoclists(Fts3Cursor *); -void sqlite3Fts3FreeDeferredDoclists(Fts3Cursor *); -int sqlite3Fts3DeferredTokenList(Fts3DeferredToken *, char **, int *); -#else -# define sqlite3Fts3FreeDeferredTokens(x) -# define sqlite3Fts3DeferToken(x,y,z) SQLITE_OK -# define sqlite3Fts3CacheDeferredDoclists(x) SQLITE_OK -# define sqlite3Fts3FreeDeferredDoclists(x) -# define sqlite3Fts3DeferredTokenList(x,y,z) SQLITE_OK -#endif - -void sqlite3Fts3SegmentsClose(Fts3Table *); -int sqlite3Fts3MaxLevel(Fts3Table *, int *); - -/* Special values interpreted by sqlite3SegReaderCursor() */ -#define FTS3_SEGCURSOR_PENDING -1 -#define FTS3_SEGCURSOR_ALL -2 - -int sqlite3Fts3SegReaderStart(Fts3Table*, Fts3MultiSegReader*, Fts3SegFilter*); -int sqlite3Fts3SegReaderStep(Fts3Table *, Fts3MultiSegReader *); -void sqlite3Fts3SegReaderFinish(Fts3MultiSegReader *); - -int sqlite3Fts3SegReaderCursor(Fts3Table *, - int, int, int, const char *, int, int, int, Fts3MultiSegReader *); - -/* Flags allowed as part of the 4th argument to SegmentReaderIterate() */ -#define FTS3_SEGMENT_REQUIRE_POS 0x00000001 -#define FTS3_SEGMENT_IGNORE_EMPTY 0x00000002 -#define FTS3_SEGMENT_COLUMN_FILTER 0x00000004 -#define FTS3_SEGMENT_PREFIX 0x00000008 -#define FTS3_SEGMENT_SCAN 0x00000010 -#define FTS3_SEGMENT_FIRST 0x00000020 - -/* Type passed as 4th argument to SegmentReaderIterate() */ -struct Fts3SegFilter { - const char *zTerm; - int nTerm; - int iCol; - int flags; -}; - -struct Fts3MultiSegReader { - /* Used internally by sqlite3Fts3SegReaderXXX() calls */ - Fts3SegReader **apSegment; /* Array of Fts3SegReader objects */ - int nSegment; /* Size of apSegment array */ - int nAdvance; /* How many seg-readers to advance */ - Fts3SegFilter *pFilter; /* Pointer to filter object */ - char *aBuffer; /* Buffer to merge doclists in */ - int nBuffer; /* Allocated size of aBuffer[] in bytes */ - - int iColFilter; /* If >=0, filter for this column */ - int bRestart; - - /* Used by fts3.c only. */ - int nCost; /* Cost of running iterator */ - int bLookup; /* True if a lookup of a single entry. */ - - /* Output values. Valid only after Fts3SegReaderStep() returns SQLITE_ROW. */ - char *zTerm; /* Pointer to term buffer */ - int nTerm; /* Size of zTerm in bytes */ - char *aDoclist; /* Pointer to doclist buffer */ - int nDoclist; /* Size of aDoclist[] in bytes */ -}; - -int sqlite3Fts3Incrmerge(Fts3Table*,int,int); - -/* fts3.c */ -int sqlite3Fts3PutVarint(char *, sqlite3_int64); -int sqlite3Fts3GetVarint(const char *, sqlite_int64 *); -int sqlite3Fts3GetVarint32(const char *, int *); -int sqlite3Fts3VarintLen(sqlite3_uint64); -void sqlite3Fts3Dequote(char *); -void sqlite3Fts3DoclistPrev(int,char*,int,char**,sqlite3_int64*,int*,u8*); -int sqlite3Fts3EvalPhraseStats(Fts3Cursor *, Fts3Expr *, u32 *); -int sqlite3Fts3FirstFilter(sqlite3_int64, char *, int, char *); -void sqlite3Fts3CreateStatTable(int*, Fts3Table*); - -/* fts3_tokenizer.c */ -const char *sqlite3Fts3NextToken(const char *, int *); -int sqlite3Fts3InitHashTable(sqlite3 *, Fts3Hash *, const char *); -int sqlite3Fts3InitTokenizer(Fts3Hash *pHash, const char *, - sqlite3_tokenizer **, char ** -); -int sqlite3Fts3IsIdChar(char); - -/* fts3_snippet.c */ -void sqlite3Fts3Offsets(sqlite3_context*, Fts3Cursor*); -void sqlite3Fts3Snippet(sqlite3_context *, Fts3Cursor *, const char *, - const char *, const char *, int, int -); -void sqlite3Fts3Matchinfo(sqlite3_context *, Fts3Cursor *, const char *); - -/* fts3_expr.c */ -int sqlite3Fts3ExprParse(sqlite3_tokenizer *, int, - char **, int, int, int, const char *, int, Fts3Expr ** -); -void sqlite3Fts3ExprFree(Fts3Expr *); -#ifdef SQLITE_TEST -int sqlite3Fts3ExprInitTestInterface(sqlite3 *db); -int sqlite3Fts3InitTerm(sqlite3 *db); -#endif - -int sqlite3Fts3OpenTokenizer(sqlite3_tokenizer *, int, const char *, int, - sqlite3_tokenizer_cursor ** -); - -/* fts3_aux.c */ -int sqlite3Fts3InitAux(sqlite3 *db); - -void sqlite3Fts3EvalPhraseCleanup(Fts3Phrase *); - -int sqlite3Fts3MsrIncrStart( - Fts3Table*, Fts3MultiSegReader*, int, const char*, int); -int sqlite3Fts3MsrIncrNext( - Fts3Table *, Fts3MultiSegReader *, sqlite3_int64 *, char **, int *); -int sqlite3Fts3EvalPhrasePoslist(Fts3Cursor *, Fts3Expr *, int iCol, char **); -int sqlite3Fts3MsrOvfl(Fts3Cursor *, Fts3MultiSegReader *, int *); -int sqlite3Fts3MsrIncrRestart(Fts3MultiSegReader *pCsr); - -/* fts3_unicode2.c (functions generated by parsing unicode text files) */ -#ifdef SQLITE_ENABLE_FTS4_UNICODE61 -int sqlite3FtsUnicodeFold(int, int); -int sqlite3FtsUnicodeIsalnum(int); -int sqlite3FtsUnicodeIsdiacritic(int); -#endif - -#endif /* !SQLITE_CORE || SQLITE_ENABLE_FTS3 */ -#endif /* _FTSINT_H */ diff --git a/src/libtracker-fts/fts3_aux.c b/src/libtracker-fts/fts3_aux.c deleted file mode 100644 index a2bff2e1d..000000000 --- a/src/libtracker-fts/fts3_aux.c +++ /dev/null @@ -1,474 +0,0 @@ -/* -** 2011 Jan 27 -** -** The author disclaims copyright to this source code. In place of -** a legal notice, here is a blessing: -** -** May you do good and not evil. -** May you find forgiveness for yourself and forgive others. -** May you share freely, never taking more than you give. -** -****************************************************************************** -** -*/ -#include "fts3Int.h" -#if !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3) - -#include <string.h> -#include <assert.h> - -typedef struct Fts3auxTable Fts3auxTable; -typedef struct Fts3auxCursor Fts3auxCursor; - -struct Fts3auxTable { - sqlite3_vtab base; /* Base class used by SQLite core */ - Fts3Table *pFts3Tab; -}; - -struct Fts3auxCursor { - sqlite3_vtab_cursor base; /* Base class used by SQLite core */ - Fts3MultiSegReader csr; /* Must be right after "base" */ - Fts3SegFilter filter; - char *zStop; - int nStop; /* Byte-length of string zStop */ - int isEof; /* True if cursor is at EOF */ - sqlite3_int64 iRowid; /* Current rowid */ - - int iCol; /* Current value of 'col' column */ - int nStat; /* Size of aStat[] array */ - struct Fts3auxColstats { - sqlite3_int64 nDoc; /* 'documents' values for current csr row */ - sqlite3_int64 nOcc; /* 'occurrences' values for current csr row */ - } *aStat; -}; - -/* -** Schema of the terms table. -*/ -#define FTS3_TERMS_SCHEMA "CREATE TABLE x(term, col, documents, occurrences)" - -/* -** This function does all the work for both the xConnect and xCreate methods. -** These tables have no persistent representation of their own, so xConnect -** and xCreate are identical operations. -*/ -static int fts3auxConnectMethod( - sqlite3 *db, /* Database connection */ - void *pUnused, /* Unused */ - int argc, /* Number of elements in argv array */ - const char * const *argv, /* xCreate/xConnect argument array */ - sqlite3_vtab **ppVtab, /* OUT: New sqlite3_vtab object */ - char **pzErr /* OUT: sqlite3_malloc'd error message */ -){ - char const *zDb; /* Name of database (e.g. "main") */ - char const *zFts3; /* Name of fts3 table */ - int nDb; /* Result of strlen(zDb) */ - int nFts3; /* Result of strlen(zFts3) */ - int nByte; /* Bytes of space to allocate here */ - int rc; /* value returned by declare_vtab() */ - Fts3auxTable *p; /* Virtual table object to return */ - - UNUSED_PARAMETER(pUnused); - - /* The user should specify a single argument - the name of an fts3 table. */ - if( argc!=4 ){ - *pzErr = sqlite3_mprintf( - "wrong number of arguments to fts4aux constructor" - ); - return SQLITE_ERROR; - } - - zDb = argv[1]; - nDb = (int)strlen(zDb); - zFts3 = argv[3]; - nFts3 = (int)strlen(zFts3); - - rc = sqlite3_declare_vtab(db, FTS3_TERMS_SCHEMA); - if( rc!=SQLITE_OK ) return rc; - - nByte = sizeof(Fts3auxTable) + sizeof(Fts3Table) + nDb + nFts3 + 2; - p = (Fts3auxTable *)sqlite3_malloc(nByte); - if( !p ) return SQLITE_NOMEM; - memset(p, 0, nByte); - - p->pFts3Tab = (Fts3Table *)&p[1]; - p->pFts3Tab->zDb = (char *)&p->pFts3Tab[1]; - p->pFts3Tab->zName = &p->pFts3Tab->zDb[nDb+1]; - p->pFts3Tab->db = db; - p->pFts3Tab->nIndex = 1; - - memcpy((char *)p->pFts3Tab->zDb, zDb, nDb); - memcpy((char *)p->pFts3Tab->zName, zFts3, nFts3); - sqlite3Fts3Dequote((char *)p->pFts3Tab->zName); - - *ppVtab = (sqlite3_vtab *)p; - return SQLITE_OK; -} - -/* -** This function does the work for both the xDisconnect and xDestroy methods. -** These tables have no persistent representation of their own, so xDisconnect -** and xDestroy are identical operations. -*/ -static int fts3auxDisconnectMethod(sqlite3_vtab *pVtab){ - Fts3auxTable *p = (Fts3auxTable *)pVtab; - Fts3Table *pFts3 = p->pFts3Tab; - int i; - - /* Free any prepared statements held */ - for(i=0; i<SizeofArray(pFts3->aStmt); i++){ - sqlite3_finalize(pFts3->aStmt[i]); - } - sqlite3_free(pFts3->zSegmentsTbl); - sqlite3_free(p); - return SQLITE_OK; -} - -#define FTS4AUX_EQ_CONSTRAINT 1 -#define FTS4AUX_GE_CONSTRAINT 2 -#define FTS4AUX_LE_CONSTRAINT 4 - -/* -** xBestIndex - Analyze a WHERE and ORDER BY clause. -*/ -static int fts3auxBestIndexMethod( - sqlite3_vtab *pVTab, - sqlite3_index_info *pInfo -){ - int i; - int iEq = -1; - int iGe = -1; - int iLe = -1; - - UNUSED_PARAMETER(pVTab); - - /* This vtab delivers always results in "ORDER BY term ASC" order. */ - if( pInfo->nOrderBy==1 - && pInfo->aOrderBy[0].iColumn==0 - && pInfo->aOrderBy[0].desc==0 - ){ - pInfo->orderByConsumed = 1; - } - - /* Search for equality and range constraints on the "term" column. */ - for(i=0; i<pInfo->nConstraint; i++){ - if( pInfo->aConstraint[i].usable && pInfo->aConstraint[i].iColumn==0 ){ - int op = pInfo->aConstraint[i].op; - if( op==SQLITE_INDEX_CONSTRAINT_EQ ) iEq = i; - if( op==SQLITE_INDEX_CONSTRAINT_LT ) iLe = i; - if( op==SQLITE_INDEX_CONSTRAINT_LE ) iLe = i; - if( op==SQLITE_INDEX_CONSTRAINT_GT ) iGe = i; - if( op==SQLITE_INDEX_CONSTRAINT_GE ) iGe = i; - } - } - - if( iEq>=0 ){ - pInfo->idxNum = FTS4AUX_EQ_CONSTRAINT; - pInfo->aConstraintUsage[iEq].argvIndex = 1; - pInfo->estimatedCost = 5; - }else{ - pInfo->idxNum = 0; - pInfo->estimatedCost = 20000; - if( iGe>=0 ){ - pInfo->idxNum += FTS4AUX_GE_CONSTRAINT; - pInfo->aConstraintUsage[iGe].argvIndex = 1; - pInfo->estimatedCost /= 2; - } - if( iLe>=0 ){ - pInfo->idxNum += FTS4AUX_LE_CONSTRAINT; - pInfo->aConstraintUsage[iLe].argvIndex = 1 + (iGe>=0); - pInfo->estimatedCost /= 2; - } - } - - return SQLITE_OK; -} - -/* -** xOpen - Open a cursor. -*/ -static int fts3auxOpenMethod(sqlite3_vtab *pVTab, sqlite3_vtab_cursor **ppCsr){ - Fts3auxCursor *pCsr; /* Pointer to cursor object to return */ - - UNUSED_PARAMETER(pVTab); - - pCsr = (Fts3auxCursor *)sqlite3_malloc(sizeof(Fts3auxCursor)); - if( !pCsr ) return SQLITE_NOMEM; - memset(pCsr, 0, sizeof(Fts3auxCursor)); - - *ppCsr = (sqlite3_vtab_cursor *)pCsr; - return SQLITE_OK; -} - -/* -** xClose - Close a cursor. -*/ -static int fts3auxCloseMethod(sqlite3_vtab_cursor *pCursor){ - Fts3Table *pFts3 = ((Fts3auxTable *)pCursor->pVtab)->pFts3Tab; - Fts3auxCursor *pCsr = (Fts3auxCursor *)pCursor; - - sqlite3Fts3SegmentsClose(pFts3); - sqlite3Fts3SegReaderFinish(&pCsr->csr); - sqlite3_free((void *)pCsr->filter.zTerm); - sqlite3_free(pCsr->zStop); - sqlite3_free(pCsr->aStat); - sqlite3_free(pCsr); - return SQLITE_OK; -} - -static int fts3auxGrowStatArray(Fts3auxCursor *pCsr, int nSize){ - if( nSize>pCsr->nStat ){ - struct Fts3auxColstats *aNew; - aNew = (struct Fts3auxColstats *)sqlite3_realloc(pCsr->aStat, - sizeof(struct Fts3auxColstats) * nSize - ); - if( aNew==0 ) return SQLITE_NOMEM; - memset(&aNew[pCsr->nStat], 0, - sizeof(struct Fts3auxColstats) * (nSize - pCsr->nStat) - ); - pCsr->aStat = aNew; - pCsr->nStat = nSize; - } - return SQLITE_OK; -} - -/* -** xNext - Advance the cursor to the next row, if any. -*/ -static int fts3auxNextMethod(sqlite3_vtab_cursor *pCursor){ - Fts3auxCursor *pCsr = (Fts3auxCursor *)pCursor; - Fts3Table *pFts3 = ((Fts3auxTable *)pCursor->pVtab)->pFts3Tab; - int rc; - - /* Increment our pretend rowid value. */ - pCsr->iRowid++; - - for(pCsr->iCol++; pCsr->iCol<pCsr->nStat; pCsr->iCol++){ - if( pCsr->aStat[pCsr->iCol].nDoc>0 ) return SQLITE_OK; - } - - rc = sqlite3Fts3SegReaderStep(pFts3, &pCsr->csr); - if( rc==SQLITE_ROW ){ - int i = 0; - int nDoclist = pCsr->csr.nDoclist; - char *aDoclist = pCsr->csr.aDoclist; - int iCol; - - int eState = 0; - - if( pCsr->zStop ){ - int n = (pCsr->nStop<pCsr->csr.nTerm) ? pCsr->nStop : pCsr->csr.nTerm; - int mc = memcmp(pCsr->zStop, pCsr->csr.zTerm, n); - if( mc<0 || (mc==0 && pCsr->csr.nTerm>pCsr->nStop) ){ - pCsr->isEof = 1; - return SQLITE_OK; - } - } - - if( fts3auxGrowStatArray(pCsr, 2) ) return SQLITE_NOMEM; - memset(pCsr->aStat, 0, sizeof(struct Fts3auxColstats) * pCsr->nStat); - iCol = 0; - - while( i<nDoclist ){ - sqlite3_int64 v = 0; - - i += sqlite3Fts3GetVarint(&aDoclist[i], &v); - switch( eState ){ - /* State 0. In this state the integer just read was a docid. */ - case 0: - pCsr->aStat[0].nDoc++; - eState = 1; - iCol = 0; - break; - - /* State 1. In this state we are expecting either a 1, indicating - ** that the following integer will be a column number, or the - ** start of a position list for column 0. - ** - ** The only difference between state 1 and state 2 is that if the - ** integer encountered in state 1 is not 0 or 1, then we need to - ** increment the column 0 "nDoc" count for this term. - */ - case 1: - assert( iCol==0 ); - if( v>1 ){ - pCsr->aStat[1].nDoc++; - } - eState = 2; - /* fall through */ - - case 2: - if( v==0 ){ /* 0x00. Next integer will be a docid. */ - eState = 0; - }else if( v==1 ){ /* 0x01. Next integer will be a column number. */ - eState = 3; - }else{ /* 2 or greater. A position. */ - pCsr->aStat[iCol+1].nOcc++; - pCsr->aStat[0].nOcc++; - } - break; - - /* State 3. The integer just read is a column number. */ - default: assert( eState==3 ); - iCol = (int)v; - if( fts3auxGrowStatArray(pCsr, iCol+2) ) return SQLITE_NOMEM; - pCsr->aStat[iCol+1].nDoc++; - eState = 2; - break; - } - } - - pCsr->iCol = 0; - rc = SQLITE_OK; - }else{ - pCsr->isEof = 1; - } - return rc; -} - -/* -** xFilter - Initialize a cursor to point at the start of its data. -*/ -static int fts3auxFilterMethod( - sqlite3_vtab_cursor *pCursor, /* The cursor used for this query */ - int idxNum, /* Strategy index */ - const char *idxStr, /* Unused */ - int nVal, /* Number of elements in apVal */ - sqlite3_value **apVal /* Arguments for the indexing scheme */ -){ - Fts3auxCursor *pCsr = (Fts3auxCursor *)pCursor; - Fts3Table *pFts3 = ((Fts3auxTable *)pCursor->pVtab)->pFts3Tab; - int rc; - int isScan; - - UNUSED_PARAMETER(nVal); - UNUSED_PARAMETER(idxStr); - - assert( idxStr==0 ); - assert( idxNum==FTS4AUX_EQ_CONSTRAINT || idxNum==0 - || idxNum==FTS4AUX_LE_CONSTRAINT || idxNum==FTS4AUX_GE_CONSTRAINT - || idxNum==(FTS4AUX_LE_CONSTRAINT|FTS4AUX_GE_CONSTRAINT) - ); - isScan = (idxNum!=FTS4AUX_EQ_CONSTRAINT); - - /* In case this cursor is being reused, close and zero it. */ - testcase(pCsr->filter.zTerm); - sqlite3Fts3SegReaderFinish(&pCsr->csr); - sqlite3_free((void *)pCsr->filter.zTerm); - sqlite3_free(pCsr->aStat); - memset(&pCsr->csr, 0, ((u8*)&pCsr[1]) - (u8*)&pCsr->csr); - - pCsr->filter.flags = FTS3_SEGMENT_REQUIRE_POS|FTS3_SEGMENT_IGNORE_EMPTY; - if( isScan ) pCsr->filter.flags |= FTS3_SEGMENT_SCAN; - - if( idxNum&(FTS4AUX_EQ_CONSTRAINT|FTS4AUX_GE_CONSTRAINT) ){ - const unsigned char *zStr = sqlite3_value_text(apVal[0]); - if( zStr ){ - pCsr->filter.zTerm = sqlite3_mprintf("%s", zStr); - pCsr->filter.nTerm = sqlite3_value_bytes(apVal[0]); - if( pCsr->filter.zTerm==0 ) return SQLITE_NOMEM; - } - } - if( idxNum&FTS4AUX_LE_CONSTRAINT ){ - int iIdx = (idxNum&FTS4AUX_GE_CONSTRAINT) ? 1 : 0; - pCsr->zStop = sqlite3_mprintf("%s", sqlite3_value_text(apVal[iIdx])); - pCsr->nStop = sqlite3_value_bytes(apVal[iIdx]); - if( pCsr->zStop==0 ) return SQLITE_NOMEM; - } - - rc = sqlite3Fts3SegReaderCursor(pFts3, 0, 0, FTS3_SEGCURSOR_ALL, - pCsr->filter.zTerm, pCsr->filter.nTerm, 0, isScan, &pCsr->csr - ); - if( rc==SQLITE_OK ){ - rc = sqlite3Fts3SegReaderStart(pFts3, &pCsr->csr, &pCsr->filter); - } - - if( rc==SQLITE_OK ) rc = fts3auxNextMethod(pCursor); - return rc; -} - -/* -** xEof - Return true if the cursor is at EOF, or false otherwise. -*/ -static int fts3auxEofMethod(sqlite3_vtab_cursor *pCursor){ - Fts3auxCursor *pCsr = (Fts3auxCursor *)pCursor; - return pCsr->isEof; -} - -/* -** xColumn - Return a column value. -*/ -static int fts3auxColumnMethod( - sqlite3_vtab_cursor *pCursor, /* Cursor to retrieve value from */ - sqlite3_context *pContext, /* Context for sqlite3_result_xxx() calls */ - int iCol /* Index of column to read value from */ -){ - Fts3auxCursor *p = (Fts3auxCursor *)pCursor; - - assert( p->isEof==0 ); - if( iCol==0 ){ /* Column "term" */ - sqlite3_result_text(pContext, p->csr.zTerm, p->csr.nTerm, SQLITE_TRANSIENT); - }else if( iCol==1 ){ /* Column "col" */ - if( p->iCol ){ - sqlite3_result_int(pContext, p->iCol-1); - }else{ - sqlite3_result_text(pContext, "*", -1, SQLITE_STATIC); - } - }else if( iCol==2 ){ /* Column "documents" */ - sqlite3_result_int64(pContext, p->aStat[p->iCol].nDoc); - }else{ /* Column "occurrences" */ - sqlite3_result_int64(pContext, p->aStat[p->iCol].nOcc); - } - - return SQLITE_OK; -} - -/* -** xRowid - Return the current rowid for the cursor. -*/ -static int fts3auxRowidMethod( - sqlite3_vtab_cursor *pCursor, /* Cursor to retrieve value from */ - sqlite_int64 *pRowid /* OUT: Rowid value */ -){ - Fts3auxCursor *pCsr = (Fts3auxCursor *)pCursor; - *pRowid = pCsr->iRowid; - return SQLITE_OK; -} - -/* -** Register the fts3aux module with database connection db. Return SQLITE_OK -** if successful or an error code if sqlite3_create_module() fails. -*/ -int sqlite3Fts3InitAux(sqlite3 *db){ - static const sqlite3_module fts3aux_module = { - 0, /* iVersion */ - fts3auxConnectMethod, /* xCreate */ - fts3auxConnectMethod, /* xConnect */ - fts3auxBestIndexMethod, /* xBestIndex */ - fts3auxDisconnectMethod, /* xDisconnect */ - fts3auxDisconnectMethod, /* xDestroy */ - fts3auxOpenMethod, /* xOpen */ - fts3auxCloseMethod, /* xClose */ - fts3auxFilterMethod, /* xFilter */ - fts3auxNextMethod, /* xNext */ - fts3auxEofMethod, /* xEof */ - fts3auxColumnMethod, /* xColumn */ - fts3auxRowidMethod, /* xRowid */ - 0, /* xUpdate */ - 0, /* xBegin */ - 0, /* xSync */ - 0, /* xCommit */ - 0, /* xRollback */ - 0, /* xFindFunction */ - 0, /* xRename */ - 0, /* xSavepoint */ - 0, /* xRelease */ - 0 /* xRollbackTo */ - }; - int rc; /* Return code */ - - rc = sqlite3_create_module(db, "fts4aux", &fts3aux_module, 0); - return rc; -} - -#endif /* !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3) */ diff --git a/src/libtracker-fts/fts3_expr.c b/src/libtracker-fts/fts3_expr.c deleted file mode 100644 index 7612789de..000000000 --- a/src/libtracker-fts/fts3_expr.c +++ /dev/null @@ -1,1013 +0,0 @@ -/* -** 2008 Nov 28 -** -** The author disclaims copyright to this source code. In place of -** a legal notice, here is a blessing: -** -** May you do good and not evil. -** May you find forgiveness for yourself and forgive others. -** May you share freely, never taking more than you give. -** -****************************************************************************** -** -** This module contains code that implements a parser for fts3 query strings -** (the right-hand argument to the MATCH operator). Because the supported -** syntax is relatively simple, the whole tokenizer/parser system is -** hand-coded. -*/ -#include "fts3Int.h" -#if !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3) - -/* -** By default, this module parses the legacy syntax that has been -** traditionally used by fts3. Or, if SQLITE_ENABLE_FTS3_PARENTHESIS -** is defined, then it uses the new syntax. The differences between -** the new and the old syntaxes are: -** -** a) The new syntax supports parenthesis. The old does not. -** -** b) The new syntax supports the AND and NOT operators. The old does not. -** -** c) The old syntax supports the "-" token qualifier. This is not -** supported by the new syntax (it is replaced by the NOT operator). -** -** d) When using the old syntax, the OR operator has a greater precedence -** than an implicit AND. When using the new, both implicity and explicit -** AND operators have a higher precedence than OR. -** -** If compiled with SQLITE_TEST defined, then this module exports the -** symbol "int sqlite3_fts3_enable_parentheses". Setting this variable -** to zero causes the module to use the old syntax. If it is set to -** non-zero the new syntax is activated. This is so both syntaxes can -** be tested using a single build of testfixture. -** -** The following describes the syntax supported by the fts3 MATCH -** operator in a similar format to that used by the lemon parser -** generator. This module does not use actually lemon, it uses a -** custom parser. -** -** query ::= andexpr (OR andexpr)*. -** -** andexpr ::= notexpr (AND? notexpr)*. -** -** notexpr ::= nearexpr (NOT nearexpr|-TOKEN)*. -** notexpr ::= LP query RP. -** -** nearexpr ::= phrase (NEAR distance_opt nearexpr)*. -** -** distance_opt ::= . -** distance_opt ::= / INTEGER. -** -** phrase ::= TOKEN. -** phrase ::= COLUMN:TOKEN. -** phrase ::= "TOKEN TOKEN TOKEN...". -*/ - -#ifdef SQLITE_TEST -int sqlite3_fts3_enable_parentheses = 0; -#else -# ifdef SQLITE_ENABLE_FTS3_PARENTHESIS -# define sqlite3_fts3_enable_parentheses 1 -# else -# define sqlite3_fts3_enable_parentheses 0 -# endif -#endif - -/* -** Default span for NEAR operators. -*/ -#define SQLITE_FTS3_DEFAULT_NEAR_PARAM 10 - -#include <string.h> -#include <assert.h> - -/* -** isNot: -** This variable is used by function getNextNode(). When getNextNode() is -** called, it sets ParseContext.isNot to true if the 'next node' is a -** FTSQUERY_PHRASE with a unary "-" attached to it. i.e. "mysql" in the -** FTS3 query "sqlite -mysql". Otherwise, ParseContext.isNot is set to -** zero. -*/ -typedef struct ParseContext ParseContext; -struct ParseContext { - sqlite3_tokenizer *pTokenizer; /* Tokenizer module */ - int iLangid; /* Language id used with tokenizer */ - const char **azCol; /* Array of column names for fts3 table */ - int bFts4; /* True to allow FTS4-only syntax */ - int nCol; /* Number of entries in azCol[] */ - int iDefaultCol; /* Default column to query */ - int isNot; /* True if getNextNode() sees a unary - */ - sqlite3_context *pCtx; /* Write error message here */ - int nNest; /* Number of nested brackets */ -}; - -/* -** This function is equivalent to the standard isspace() function. -** -** The standard isspace() can be awkward to use safely, because although it -** is defined to accept an argument of type int, its behaviour when passed -** an integer that falls outside of the range of the unsigned char type -** is undefined (and sometimes, "undefined" means segfault). This wrapper -** is defined to accept an argument of type char, and always returns 0 for -** any values that fall outside of the range of the unsigned char type (i.e. -** negative values). -*/ -static int fts3isspace(char c){ - return c==' ' || c=='\t' || c=='\n' || c=='\r' || c=='\v' || c=='\f'; -} - -/* -** Allocate nByte bytes of memory using sqlite3_malloc(). If successful, -** zero the memory before returning a pointer to it. If unsuccessful, -** return NULL. -*/ -static void *fts3MallocZero(int nByte){ - void *pRet = sqlite3_malloc(nByte); - if( pRet ) memset(pRet, 0, nByte); - return pRet; -} - -int sqlite3Fts3OpenTokenizer( - sqlite3_tokenizer *pTokenizer, - int iLangid, - const char *z, - int n, - sqlite3_tokenizer_cursor **ppCsr -){ - sqlite3_tokenizer_module const *pModule = pTokenizer->pModule; - sqlite3_tokenizer_cursor *pCsr = 0; - int rc; - - rc = pModule->xOpen(pTokenizer, z, n, &pCsr); - assert( rc==SQLITE_OK || pCsr==0 ); - if( rc==SQLITE_OK ){ - pCsr->pTokenizer = pTokenizer; - if( pModule->iVersion>=1 ){ - rc = pModule->xLanguageid(pCsr, iLangid); - if( rc!=SQLITE_OK ){ - pModule->xClose(pCsr); - pCsr = 0; - } - } - } - *ppCsr = pCsr; - return rc; -} - - -/* -** Extract the next token from buffer z (length n) using the tokenizer -** and other information (column names etc.) in pParse. Create an Fts3Expr -** structure of type FTSQUERY_PHRASE containing a phrase consisting of this -** single token and set *ppExpr to point to it. If the end of the buffer is -** reached before a token is found, set *ppExpr to zero. It is the -** responsibility of the caller to eventually deallocate the allocated -** Fts3Expr structure (if any) by passing it to sqlite3_free(). -** -** Return SQLITE_OK if successful, or SQLITE_NOMEM if a memory allocation -** fails. -*/ -static int getNextToken( - ParseContext *pParse, /* fts3 query parse context */ - int iCol, /* Value for Fts3Phrase.iColumn */ - const char *z, int n, /* Input string */ - Fts3Expr **ppExpr, /* OUT: expression */ - int *pnConsumed /* OUT: Number of bytes consumed */ -){ - sqlite3_tokenizer *pTokenizer = pParse->pTokenizer; - sqlite3_tokenizer_module const *pModule = pTokenizer->pModule; - int rc; - sqlite3_tokenizer_cursor *pCursor; - Fts3Expr *pRet = 0; - int nConsumed = 0; - - rc = sqlite3Fts3OpenTokenizer(pTokenizer, pParse->iLangid, z, n, &pCursor); - if( rc==SQLITE_OK ){ - const char *zToken; - int nToken = 0, iStart = 0, iEnd = 0, iPosition = 0; - int nByte; /* total space to allocate */ - - rc = pModule->xNext(pCursor, &zToken, &nToken, &iStart, &iEnd, &iPosition); - if( rc==SQLITE_OK ){ - nByte = sizeof(Fts3Expr) + sizeof(Fts3Phrase) + nToken; - pRet = (Fts3Expr *)fts3MallocZero(nByte); - if( !pRet ){ - rc = SQLITE_NOMEM; - }else{ - pRet->eType = FTSQUERY_PHRASE; - pRet->pPhrase = (Fts3Phrase *)&pRet[1]; - pRet->pPhrase->nToken = 1; - pRet->pPhrase->iColumn = iCol; - pRet->pPhrase->aToken[0].n = nToken; - pRet->pPhrase->aToken[0].z = (char *)&pRet->pPhrase[1]; - memcpy(pRet->pPhrase->aToken[0].z, zToken, nToken); - - if( iEnd<n && z[iEnd]=='*' ){ - pRet->pPhrase->aToken[0].isPrefix = 1; - iEnd++; - } - - while( 1 ){ - if( !sqlite3_fts3_enable_parentheses - && iStart>0 && z[iStart-1]=='-' - ){ - pParse->isNot = 1; - iStart--; - }else if( pParse->bFts4 && iStart>0 && z[iStart-1]=='^' ){ - pRet->pPhrase->aToken[0].bFirst = 1; - iStart--; - }else{ - break; - } - } - - } - nConsumed = iEnd; - } - - pModule->xClose(pCursor); - } - - *pnConsumed = nConsumed; - *ppExpr = pRet; - return rc; -} - - -/* -** Enlarge a memory allocation. If an out-of-memory allocation occurs, -** then free the old allocation. -*/ -static void *fts3ReallocOrFree(void *pOrig, int nNew){ - void *pRet = sqlite3_realloc(pOrig, nNew); - if( !pRet ){ - sqlite3_free(pOrig); - } - return pRet; -} - -/* -** Buffer zInput, length nInput, contains the contents of a quoted string -** that appeared as part of an fts3 query expression. Neither quote character -** is included in the buffer. This function attempts to tokenize the entire -** input buffer and create an Fts3Expr structure of type FTSQUERY_PHRASE -** containing the results. -** -** If successful, SQLITE_OK is returned and *ppExpr set to point at the -** allocated Fts3Expr structure. Otherwise, either SQLITE_NOMEM (out of memory -** error) or SQLITE_ERROR (tokenization error) is returned and *ppExpr set -** to 0. -*/ -static int getNextString( - ParseContext *pParse, /* fts3 query parse context */ - const char *zInput, int nInput, /* Input string */ - Fts3Expr **ppExpr /* OUT: expression */ -){ - sqlite3_tokenizer *pTokenizer = pParse->pTokenizer; - sqlite3_tokenizer_module const *pModule = pTokenizer->pModule; - int rc; - Fts3Expr *p = 0; - sqlite3_tokenizer_cursor *pCursor = 0; - char *zTemp = 0; - int nTemp = 0; - - const int nSpace = sizeof(Fts3Expr) + sizeof(Fts3Phrase); - int nToken = 0; - - /* The final Fts3Expr data structure, including the Fts3Phrase, - ** Fts3PhraseToken structures token buffers are all stored as a single - ** allocation so that the expression can be freed with a single call to - ** sqlite3_free(). Setting this up requires a two pass approach. - ** - ** The first pass, in the block below, uses a tokenizer cursor to iterate - ** through the tokens in the expression. This pass uses fts3ReallocOrFree() - ** to assemble data in two dynamic buffers: - ** - ** Buffer p: Points to the Fts3Expr structure, followed by the Fts3Phrase - ** structure, followed by the array of Fts3PhraseToken - ** structures. This pass only populates the Fts3PhraseToken array. - ** - ** Buffer zTemp: Contains copies of all tokens. - ** - ** The second pass, in the block that begins "if( rc==SQLITE_DONE )" below, - ** appends buffer zTemp to buffer p, and fills in the Fts3Expr and Fts3Phrase - ** structures. - */ - rc = sqlite3Fts3OpenTokenizer( - pTokenizer, pParse->iLangid, zInput, nInput, &pCursor); - if( rc==SQLITE_OK ){ - int ii; - for(ii=0; rc==SQLITE_OK; ii++){ - const char *zByte; - int nByte = 0, iBegin = 0, iEnd = 0, iPos = 0; - rc = pModule->xNext(pCursor, &zByte, &nByte, &iBegin, &iEnd, &iPos); - if( rc==SQLITE_OK ){ - Fts3PhraseToken *pToken; - - p = fts3ReallocOrFree(p, nSpace + ii*sizeof(Fts3PhraseToken)); - if( !p ) goto no_mem; - - zTemp = fts3ReallocOrFree(zTemp, nTemp + nByte); - if( !zTemp ) goto no_mem; - - assert( nToken==ii ); - pToken = &((Fts3Phrase *)(&p[1]))->aToken[ii]; - memset(pToken, 0, sizeof(Fts3PhraseToken)); - - memcpy(&zTemp[nTemp], zByte, nByte); - nTemp += nByte; - - pToken->n = nByte; - pToken->isPrefix = (iEnd<nInput && zInput[iEnd]=='*'); - pToken->bFirst = (iBegin>0 && zInput[iBegin-1]=='^'); - nToken = ii+1; - } - } - - pModule->xClose(pCursor); - pCursor = 0; - } - - if( rc==SQLITE_DONE ){ - int jj; - char *zBuf = 0; - - p = fts3ReallocOrFree(p, nSpace + nToken*sizeof(Fts3PhraseToken) + nTemp); - if( !p ) goto no_mem; - memset(p, 0, (char *)&(((Fts3Phrase *)&p[1])->aToken[0])-(char *)p); - p->eType = FTSQUERY_PHRASE; - p->pPhrase = (Fts3Phrase *)&p[1]; - p->pPhrase->iColumn = pParse->iDefaultCol; - p->pPhrase->nToken = nToken; - - zBuf = (char *)&p->pPhrase->aToken[nToken]; - if( zTemp ){ - memcpy(zBuf, zTemp, nTemp); - sqlite3_free(zTemp); - }else{ - assert( nTemp==0 ); - } - - for(jj=0; jj<p->pPhrase->nToken; jj++){ - p->pPhrase->aToken[jj].z = zBuf; - zBuf += p->pPhrase->aToken[jj].n; - } - rc = SQLITE_OK; - } - - *ppExpr = p; - return rc; -no_mem: - - if( pCursor ){ - pModule->xClose(pCursor); - } - sqlite3_free(zTemp); - sqlite3_free(p); - *ppExpr = 0; - return SQLITE_NOMEM; -} - -/* -** Function getNextNode(), which is called by fts3ExprParse(), may itself -** call fts3ExprParse(). So this forward declaration is required. -*/ -static int fts3ExprParse(ParseContext *, const char *, int, Fts3Expr **, int *); - -/* -** The output variable *ppExpr is populated with an allocated Fts3Expr -** structure, or set to 0 if the end of the input buffer is reached. -** -** Returns an SQLite error code. SQLITE_OK if everything works, SQLITE_NOMEM -** if a malloc failure occurs, or SQLITE_ERROR if a parse error is encountered. -** If SQLITE_ERROR is returned, pContext is populated with an error message. -*/ -static int getNextNode( - ParseContext *pParse, /* fts3 query parse context */ - const char *z, int n, /* Input string */ - Fts3Expr **ppExpr, /* OUT: expression */ - int *pnConsumed /* OUT: Number of bytes consumed */ -){ - static const struct Fts3Keyword { - char *z; /* Keyword text */ - unsigned char n; /* Length of the keyword */ - unsigned char parenOnly; /* Only valid in paren mode */ - unsigned char eType; /* Keyword code */ - } aKeyword[] = { - { "OR" , 2, 0, FTSQUERY_OR }, - { "AND", 3, 1, FTSQUERY_AND }, - { "NOT", 3, 1, FTSQUERY_NOT }, - { "NEAR", 4, 0, FTSQUERY_NEAR } - }; - int ii; - int iCol; - int iColLen; - int rc; - Fts3Expr *pRet = 0; - - const char *zInput = z; - int nInput = n; - - pParse->isNot = 0; - - /* Skip over any whitespace before checking for a keyword, an open or - ** close bracket, or a quoted string. - */ - while( nInput>0 && fts3isspace(*zInput) ){ - nInput--; - zInput++; - } - if( nInput==0 ){ - return SQLITE_DONE; - } - - /* See if we are dealing with a keyword. */ - for(ii=0; ii<(int)(sizeof(aKeyword)/sizeof(struct Fts3Keyword)); ii++){ - const struct Fts3Keyword *pKey = &aKeyword[ii]; - - if( (pKey->parenOnly & ~sqlite3_fts3_enable_parentheses)!=0 ){ - continue; - } - - if( nInput>=pKey->n && 0==memcmp(zInput, pKey->z, pKey->n) ){ - int nNear = SQLITE_FTS3_DEFAULT_NEAR_PARAM; - int nKey = pKey->n; - char cNext; - - /* If this is a "NEAR" keyword, check for an explicit nearness. */ - if( pKey->eType==FTSQUERY_NEAR ){ - assert( nKey==4 ); - if( zInput[4]=='/' && zInput[5]>='0' && zInput[5]<='9' ){ - nNear = 0; - for(nKey=5; zInput[nKey]>='0' && zInput[nKey]<='9'; nKey++){ - nNear = nNear * 10 + (zInput[nKey] - '0'); - } - } - } - - /* At this point this is probably a keyword. But for that to be true, - ** the next byte must contain either whitespace, an open or close - ** parenthesis, a quote character, or EOF. - */ - cNext = zInput[nKey]; - if( fts3isspace(cNext) - || cNext=='"' || cNext=='(' || cNext==')' || cNext==0 - ){ - pRet = (Fts3Expr *)fts3MallocZero(sizeof(Fts3Expr)); - if( !pRet ){ - return SQLITE_NOMEM; - } - pRet->eType = pKey->eType; - pRet->nNear = nNear; - *ppExpr = pRet; - *pnConsumed = (int)((zInput - z) + nKey); - return SQLITE_OK; - } - - /* Turns out that wasn't a keyword after all. This happens if the - ** user has supplied a token such as "ORacle". Continue. - */ - } - } - - /* Check for an open bracket. */ - if( sqlite3_fts3_enable_parentheses ){ - if( *zInput=='(' ){ - int nConsumed; - pParse->nNest++; - rc = fts3ExprParse(pParse, &zInput[1], nInput-1, ppExpr, &nConsumed); - if( rc==SQLITE_OK && !*ppExpr ){ - rc = SQLITE_DONE; - } - *pnConsumed = (int)((zInput - z) + 1 + nConsumed); - return rc; - } - - /* Check for a close bracket. */ - if( *zInput==')' ){ - pParse->nNest--; - *pnConsumed = (int)((zInput - z) + 1); - return SQLITE_DONE; - } - } - - /* See if we are dealing with a quoted phrase. If this is the case, then - ** search for the closing quote and pass the whole string to getNextString() - ** for processing. This is easy to do, as fts3 has no syntax for escaping - ** a quote character embedded in a string. - */ - if( *zInput=='"' ){ - for(ii=1; ii<nInput && zInput[ii]!='"'; ii++); - *pnConsumed = (int)((zInput - z) + ii + 1); - if( ii==nInput ){ - return SQLITE_ERROR; - } - return getNextString(pParse, &zInput[1], ii-1, ppExpr); - } - - - /* If control flows to this point, this must be a regular token, or - ** the end of the input. Read a regular token using the sqlite3_tokenizer - ** interface. Before doing so, figure out if there is an explicit - ** column specifier for the token. - ** - ** TODO: Strangely, it is not possible to associate a column specifier - ** with a quoted phrase, only with a single token. Not sure if this was - ** an implementation artifact or an intentional decision when fts3 was - ** first implemented. Whichever it was, this module duplicates the - ** limitation. - */ - iCol = pParse->iDefaultCol; - iColLen = 0; - for(ii=0; ii<pParse->nCol; ii++){ - const char *zStr = pParse->azCol[ii]; - int nStr = (int)strlen(zStr); - if( nInput>nStr && zInput[nStr]==':' - && sqlite3_strnicmp(zStr, zInput, nStr)==0 - ){ - iCol = ii; - iColLen = (int)((zInput - z) + nStr + 1); - break; - } - } - rc = getNextToken(pParse, iCol, &z[iColLen], n-iColLen, ppExpr, pnConsumed); - *pnConsumed += iColLen; - return rc; -} - -/* -** The argument is an Fts3Expr structure for a binary operator (any type -** except an FTSQUERY_PHRASE). Return an integer value representing the -** precedence of the operator. Lower values have a higher precedence (i.e. -** group more tightly). For example, in the C language, the == operator -** groups more tightly than ||, and would therefore have a higher precedence. -** -** When using the new fts3 query syntax (when SQLITE_ENABLE_FTS3_PARENTHESIS -** is defined), the order of the operators in precedence from highest to -** lowest is: -** -** NEAR -** NOT -** AND (including implicit ANDs) -** OR -** -** Note that when using the old query syntax, the OR operator has a higher -** precedence than the AND operator. -*/ -static int opPrecedence(Fts3Expr *p){ - assert( p->eType!=FTSQUERY_PHRASE ); - if( sqlite3_fts3_enable_parentheses ){ - return p->eType; - }else if( p->eType==FTSQUERY_NEAR ){ - return 1; - }else if( p->eType==FTSQUERY_OR ){ - return 2; - } - assert( p->eType==FTSQUERY_AND ); - return 3; -} - -/* -** Argument ppHead contains a pointer to the current head of a query -** expression tree being parsed. pPrev is the expression node most recently -** inserted into the tree. This function adds pNew, which is always a binary -** operator node, into the expression tree based on the relative precedence -** of pNew and the existing nodes of the tree. This may result in the head -** of the tree changing, in which case *ppHead is set to the new root node. -*/ -static void insertBinaryOperator( - Fts3Expr **ppHead, /* Pointer to the root node of a tree */ - Fts3Expr *pPrev, /* Node most recently inserted into the tree */ - Fts3Expr *pNew /* New binary node to insert into expression tree */ -){ - Fts3Expr *pSplit = pPrev; - while( pSplit->pParent && opPrecedence(pSplit->pParent)<=opPrecedence(pNew) ){ - pSplit = pSplit->pParent; - } - - if( pSplit->pParent ){ - assert( pSplit->pParent->pRight==pSplit ); - pSplit->pParent->pRight = pNew; - pNew->pParent = pSplit->pParent; - }else{ - *ppHead = pNew; - } - pNew->pLeft = pSplit; - pSplit->pParent = pNew; -} - -/* -** Parse the fts3 query expression found in buffer z, length n. This function -** returns either when the end of the buffer is reached or an unmatched -** closing bracket - ')' - is encountered. -** -** If successful, SQLITE_OK is returned, *ppExpr is set to point to the -** parsed form of the expression and *pnConsumed is set to the number of -** bytes read from buffer z. Otherwise, *ppExpr is set to 0 and SQLITE_NOMEM -** (out of memory error) or SQLITE_ERROR (parse error) is returned. -*/ -static int fts3ExprParse( - ParseContext *pParse, /* fts3 query parse context */ - const char *z, int n, /* Text of MATCH query */ - Fts3Expr **ppExpr, /* OUT: Parsed query structure */ - int *pnConsumed /* OUT: Number of bytes consumed */ -){ - Fts3Expr *pRet = 0; - Fts3Expr *pPrev = 0; - Fts3Expr *pNotBranch = 0; /* Only used in legacy parse mode */ - int nIn = n; - const char *zIn = z; - int rc = SQLITE_OK; - int isRequirePhrase = 1; - - while( rc==SQLITE_OK ){ - Fts3Expr *p = 0; - int nByte = 0; - rc = getNextNode(pParse, zIn, nIn, &p, &nByte); - if( rc==SQLITE_OK ){ - int isPhrase; - - if( !sqlite3_fts3_enable_parentheses - && p->eType==FTSQUERY_PHRASE && pParse->isNot - ){ - /* Create an implicit NOT operator. */ - Fts3Expr *pNot = fts3MallocZero(sizeof(Fts3Expr)); - if( !pNot ){ - sqlite3Fts3ExprFree(p); - rc = SQLITE_NOMEM; - goto exprparse_out; - } - pNot->eType = FTSQUERY_NOT; - pNot->pRight = p; - if( pNotBranch ){ - pNot->pLeft = pNotBranch; - } - pNotBranch = pNot; - p = pPrev; - }else{ - int eType = p->eType; - isPhrase = (eType==FTSQUERY_PHRASE || p->pLeft); - - /* The isRequirePhrase variable is set to true if a phrase or - ** an expression contained in parenthesis is required. If a - ** binary operator (AND, OR, NOT or NEAR) is encounted when - ** isRequirePhrase is set, this is a syntax error. - */ - if( !isPhrase && isRequirePhrase ){ - sqlite3Fts3ExprFree(p); - rc = SQLITE_ERROR; - goto exprparse_out; - } - - if( isPhrase && !isRequirePhrase ){ - /* Insert an implicit AND operator. */ - Fts3Expr *pAnd; - assert( pRet && pPrev ); - pAnd = fts3MallocZero(sizeof(Fts3Expr)); - if( !pAnd ){ - sqlite3Fts3ExprFree(p); - rc = SQLITE_NOMEM; - goto exprparse_out; - } - pAnd->eType = FTSQUERY_AND; - insertBinaryOperator(&pRet, pPrev, pAnd); - pPrev = pAnd; - } - - /* This test catches attempts to make either operand of a NEAR - ** operator something other than a phrase. For example, either of - ** the following: - ** - ** (bracketed expression) NEAR phrase - ** phrase NEAR (bracketed expression) - ** - ** Return an error in either case. - */ - if( pPrev && ( - (eType==FTSQUERY_NEAR && !isPhrase && pPrev->eType!=FTSQUERY_PHRASE) - || (eType!=FTSQUERY_PHRASE && isPhrase && pPrev->eType==FTSQUERY_NEAR) - )){ - sqlite3Fts3ExprFree(p); - rc = SQLITE_ERROR; - goto exprparse_out; - } - - if( isPhrase ){ - if( pRet ){ - assert( pPrev && pPrev->pLeft && pPrev->pRight==0 ); - pPrev->pRight = p; - p->pParent = pPrev; - }else{ - pRet = p; - } - }else{ - insertBinaryOperator(&pRet, pPrev, p); - } - isRequirePhrase = !isPhrase; - } - assert( nByte>0 ); - } - assert( rc!=SQLITE_OK || (nByte>0 && nByte<=nIn) ); - nIn -= nByte; - zIn += nByte; - pPrev = p; - } - - if( rc==SQLITE_DONE && pRet && isRequirePhrase ){ - rc = SQLITE_ERROR; - } - - if( rc==SQLITE_DONE ){ - rc = SQLITE_OK; - if( !sqlite3_fts3_enable_parentheses && pNotBranch ){ - if( !pRet ){ - rc = SQLITE_ERROR; - }else{ - Fts3Expr *pIter = pNotBranch; - while( pIter->pLeft ){ - pIter = pIter->pLeft; - } - pIter->pLeft = pRet; - pRet = pNotBranch; - } - } - } - *pnConsumed = n - nIn; - -exprparse_out: - if( rc!=SQLITE_OK ){ - sqlite3Fts3ExprFree(pRet); - sqlite3Fts3ExprFree(pNotBranch); - pRet = 0; - } - *ppExpr = pRet; - return rc; -} - -/* -** Parameters z and n contain a pointer to and length of a buffer containing -** an fts3 query expression, respectively. This function attempts to parse the -** query expression and create a tree of Fts3Expr structures representing the -** parsed expression. If successful, *ppExpr is set to point to the head -** of the parsed expression tree and SQLITE_OK is returned. If an error -** occurs, either SQLITE_NOMEM (out-of-memory error) or SQLITE_ERROR (parse -** error) is returned and *ppExpr is set to 0. -** -** If parameter n is a negative number, then z is assumed to point to a -** nul-terminated string and the length is determined using strlen(). -** -** The first parameter, pTokenizer, is passed the fts3 tokenizer module to -** use to normalize query tokens while parsing the expression. The azCol[] -** array, which is assumed to contain nCol entries, should contain the names -** of each column in the target fts3 table, in order from left to right. -** Column names must be nul-terminated strings. -** -** The iDefaultCol parameter should be passed the index of the table column -** that appears on the left-hand-side of the MATCH operator (the default -** column to match against for tokens for which a column name is not explicitly -** specified as part of the query string), or -1 if tokens may by default -** match any table column. -*/ -int sqlite3Fts3ExprParse( - sqlite3_tokenizer *pTokenizer, /* Tokenizer module */ - int iLangid, /* Language id for tokenizer */ - char **azCol, /* Array of column names for fts3 table */ - int bFts4, /* True to allow FTS4-only syntax */ - int nCol, /* Number of entries in azCol[] */ - int iDefaultCol, /* Default column to query */ - const char *z, int n, /* Text of MATCH query */ - Fts3Expr **ppExpr /* OUT: Parsed query structure */ -){ - int nParsed; - int rc; - ParseContext sParse; - - memset(&sParse, 0, sizeof(ParseContext)); - sParse.pTokenizer = pTokenizer; - sParse.iLangid = iLangid; - sParse.azCol = (const char **)azCol; - sParse.nCol = nCol; - sParse.iDefaultCol = iDefaultCol; - sParse.bFts4 = bFts4; - if( z==0 ){ - *ppExpr = 0; - return SQLITE_OK; - } - if( n<0 ){ - n = (int)strlen(z); - } - rc = fts3ExprParse(&sParse, z, n, ppExpr, &nParsed); - - /* Check for mismatched parenthesis */ - if( rc==SQLITE_OK && sParse.nNest ){ - rc = SQLITE_ERROR; - sqlite3Fts3ExprFree(*ppExpr); - *ppExpr = 0; - } - - return rc; -} - -/* -** Free a parsed fts3 query expression allocated by sqlite3Fts3ExprParse(). -*/ -void sqlite3Fts3ExprFree(Fts3Expr *p){ - if( p ){ - assert( p->eType==FTSQUERY_PHRASE || p->pPhrase==0 ); - sqlite3Fts3ExprFree(p->pLeft); - sqlite3Fts3ExprFree(p->pRight); - sqlite3Fts3EvalPhraseCleanup(p->pPhrase); - sqlite3_free(p->aMI); - sqlite3_free(p); - } -} - -/**************************************************************************** -***************************************************************************** -** Everything after this point is just test code. -*/ - -#ifdef SQLITE_TEST - -#include <stdio.h> - -/* -** Function to query the hash-table of tokenizers (see README.tokenizers). -*/ -static int queryTestTokenizer( - sqlite3 *db, - const char *zName, - const sqlite3_tokenizer_module **pp -){ - int rc; - sqlite3_stmt *pStmt; - const char zSql[] = "SELECT fts3_tokenizer(?)"; - - *pp = 0; - rc = sqlite3_prepare_v2(db, zSql, -1, &pStmt, 0); - if( rc!=SQLITE_OK ){ - return rc; - } - - sqlite3_bind_text(pStmt, 1, zName, -1, SQLITE_STATIC); - if( SQLITE_ROW==sqlite3_step(pStmt) ){ - if( sqlite3_column_type(pStmt, 0)==SQLITE_BLOB ){ - memcpy((void *)pp, sqlite3_column_blob(pStmt, 0), sizeof(*pp)); - } - } - - return sqlite3_finalize(pStmt); -} - -/* -** Return a pointer to a buffer containing a text representation of the -** expression passed as the first argument. The buffer is obtained from -** sqlite3_malloc(). It is the responsibility of the caller to use -** sqlite3_free() to release the memory. If an OOM condition is encountered, -** NULL is returned. -** -** If the second argument is not NULL, then its contents are prepended to -** the returned expression text and then freed using sqlite3_free(). -*/ -static char *exprToString(Fts3Expr *pExpr, char *zBuf){ - switch( pExpr->eType ){ - case FTSQUERY_PHRASE: { - Fts3Phrase *pPhrase = pExpr->pPhrase; - int i; - zBuf = sqlite3_mprintf( - "%zPHRASE %d 0", zBuf, pPhrase->iColumn); - for(i=0; zBuf && i<pPhrase->nToken; i++){ - zBuf = sqlite3_mprintf("%z %.*s%s", zBuf, - pPhrase->aToken[i].n, pPhrase->aToken[i].z, - (pPhrase->aToken[i].isPrefix?"+":"") - ); - } - return zBuf; - } - - case FTSQUERY_NEAR: - zBuf = sqlite3_mprintf("%zNEAR/%d ", zBuf, pExpr->nNear); - break; - case FTSQUERY_NOT: - zBuf = sqlite3_mprintf("%zNOT ", zBuf); - break; - case FTSQUERY_AND: - zBuf = sqlite3_mprintf("%zAND ", zBuf); - break; - case FTSQUERY_OR: - zBuf = sqlite3_mprintf("%zOR ", zBuf); - break; - } - - if( zBuf ) zBuf = sqlite3_mprintf("%z{", zBuf); - if( zBuf ) zBuf = exprToString(pExpr->pLeft, zBuf); - if( zBuf ) zBuf = sqlite3_mprintf("%z} {", zBuf); - - if( zBuf ) zBuf = exprToString(pExpr->pRight, zBuf); - if( zBuf ) zBuf = sqlite3_mprintf("%z}", zBuf); - - return zBuf; -} - -/* -** This is the implementation of a scalar SQL function used to test the -** expression parser. It should be called as follows: -** -** fts3_exprtest(<tokenizer>, <expr>, <column 1>, ...); -** -** The first argument, <tokenizer>, is the name of the fts3 tokenizer used -** to parse the query expression (see README.tokenizers). The second argument -** is the query expression to parse. Each subsequent argument is the name -** of a column of the fts3 table that the query expression may refer to. -** For example: -** -** SELECT fts3_exprtest('simple', 'Bill col2:Bloggs', 'col1', 'col2'); -*/ -static void fts3ExprTest( - sqlite3_context *context, - int argc, - sqlite3_value **argv -){ - sqlite3_tokenizer_module const *pModule = 0; - sqlite3_tokenizer *pTokenizer = 0; - int rc; - char **azCol = 0; - const char *zExpr; - int nExpr; - int nCol; - int ii; - Fts3Expr *pExpr; - char *zBuf = 0; - sqlite3 *db = sqlite3_context_db_handle(context); - - if( argc<3 ){ - sqlite3_result_error(context, - "Usage: fts3_exprtest(tokenizer, expr, col1, ...", -1 - ); - return; - } - - rc = queryTestTokenizer(db, - (const char *)sqlite3_value_text(argv[0]), &pModule); - if( rc==SQLITE_NOMEM ){ - sqlite3_result_error_nomem(context); - goto exprtest_out; - }else if( !pModule ){ - sqlite3_result_error(context, "No such tokenizer module", -1); - goto exprtest_out; - } - - rc = pModule->xCreate(0, 0, &pTokenizer); - assert( rc==SQLITE_NOMEM || rc==SQLITE_OK ); - if( rc==SQLITE_NOMEM ){ - sqlite3_result_error_nomem(context); - goto exprtest_out; - } - pTokenizer->pModule = pModule; - - zExpr = (const char *)sqlite3_value_text(argv[1]); - nExpr = sqlite3_value_bytes(argv[1]); - nCol = argc-2; - azCol = (char **)sqlite3_malloc(nCol*sizeof(char *)); - if( !azCol ){ - sqlite3_result_error_nomem(context); - goto exprtest_out; - } - for(ii=0; ii<nCol; ii++){ - azCol[ii] = (char *)sqlite3_value_text(argv[ii+2]); - } - - rc = sqlite3Fts3ExprParse( - pTokenizer, 0, azCol, 0, nCol, nCol, zExpr, nExpr, &pExpr - ); - if( rc!=SQLITE_OK && rc!=SQLITE_NOMEM ){ - sqlite3_result_error(context, "Error parsing expression", -1); - }else if( rc==SQLITE_NOMEM || !(zBuf = exprToString(pExpr, 0)) ){ - sqlite3_result_error_nomem(context); - }else{ - sqlite3_result_text(context, zBuf, -1, SQLITE_TRANSIENT); - sqlite3_free(zBuf); - } - - sqlite3Fts3ExprFree(pExpr); - -exprtest_out: - if( pModule && pTokenizer ){ - rc = pModule->xDestroy(pTokenizer); - } - sqlite3_free(azCol); -} - -/* -** Register the query expression parser test function fts3_exprtest() -** with database connection db. -*/ -int sqlite3Fts3ExprInitTestInterface(sqlite3* db){ - return sqlite3_create_function( - db, "fts3_exprtest", -1, SQLITE_UTF8, 0, fts3ExprTest, 0, 0 - ); -} - -#endif -#endif /* !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3) */ diff --git a/src/libtracker-fts/fts3_hash.c b/src/libtracker-fts/fts3_hash.c deleted file mode 100644 index 57c59b587..000000000 --- a/src/libtracker-fts/fts3_hash.c +++ /dev/null @@ -1,383 +0,0 @@ -/* -** 2001 September 22 -** -** The author disclaims copyright to this source code. In place of -** a legal notice, here is a blessing: -** -** May you do good and not evil. -** May you find forgiveness for yourself and forgive others. -** May you share freely, never taking more than you give. -** -************************************************************************* -** This is the implementation of generic hash-tables used in SQLite. -** We've modified it slightly to serve as a standalone hash table -** implementation for the full-text indexing module. -*/ - -/* -** The code in this file is only compiled if: -** -** * The FTS3 module is being built as an extension -** (in which case SQLITE_CORE is not defined), or -** -** * The FTS3 module is being built into the core of -** SQLite (in which case SQLITE_ENABLE_FTS3 is defined). -*/ -#include "fts3Int.h" -#if !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3) - -#include <assert.h> -#include <stdlib.h> -#include <string.h> - -#include "fts3_hash.h" - -/* -** Malloc and Free functions -*/ -static void *fts3HashMalloc(int n){ - void *p = sqlite3_malloc(n); - if( p ){ - memset(p, 0, n); - } - return p; -} -static void fts3HashFree(void *p){ - sqlite3_free(p); -} - -/* Turn bulk memory into a hash table object by initializing the -** fields of the Hash structure. -** -** "pNew" is a pointer to the hash table that is to be initialized. -** keyClass is one of the constants -** FTS3_HASH_BINARY or FTS3_HASH_STRING. The value of keyClass -** determines what kind of key the hash table will use. "copyKey" is -** true if the hash table should make its own private copy of keys and -** false if it should just use the supplied pointer. -*/ -void sqlite3Fts3HashInit(Fts3Hash *pNew, char keyClass, char copyKey){ - assert( pNew!=0 ); - assert( keyClass>=FTS3_HASH_STRING && keyClass<=FTS3_HASH_BINARY ); - pNew->keyClass = keyClass; - pNew->copyKey = copyKey; - pNew->first = 0; - pNew->count = 0; - pNew->htsize = 0; - pNew->ht = 0; -} - -/* Remove all entries from a hash table. Reclaim all memory. -** Call this routine to delete a hash table or to reset a hash table -** to the empty state. -*/ -void sqlite3Fts3HashClear(Fts3Hash *pH){ - Fts3HashElem *elem; /* For looping over all elements of the table */ - - assert( pH!=0 ); - elem = pH->first; - pH->first = 0; - fts3HashFree(pH->ht); - pH->ht = 0; - pH->htsize = 0; - while( elem ){ - Fts3HashElem *next_elem = elem->next; - if( pH->copyKey && elem->pKey ){ - fts3HashFree(elem->pKey); - } - fts3HashFree(elem); - elem = next_elem; - } - pH->count = 0; -} - -/* -** Hash and comparison functions when the mode is FTS3_HASH_STRING -*/ -static int fts3StrHash(const void *pKey, int nKey){ - const char *z = (const char *)pKey; - int h = 0; - if( nKey<=0 ) nKey = (int) strlen(z); - while( nKey > 0 ){ - h = (h<<3) ^ h ^ *z++; - nKey--; - } - return h & 0x7fffffff; -} -static int fts3StrCompare(const void *pKey1, int n1, const void *pKey2, int n2){ - if( n1!=n2 ) return 1; - return strncmp((const char*)pKey1,(const char*)pKey2,n1); -} - -/* -** Hash and comparison functions when the mode is FTS3_HASH_BINARY -*/ -static int fts3BinHash(const void *pKey, int nKey){ - int h = 0; - const char *z = (const char *)pKey; - while( nKey-- > 0 ){ - h = (h<<3) ^ h ^ *(z++); - } - return h & 0x7fffffff; -} -static int fts3BinCompare(const void *pKey1, int n1, const void *pKey2, int n2){ - if( n1!=n2 ) return 1; - return memcmp(pKey1,pKey2,n1); -} - -/* -** Return a pointer to the appropriate hash function given the key class. -** -** The C syntax in this function definition may be unfamilar to some -** programmers, so we provide the following additional explanation: -** -** The name of the function is "ftsHashFunction". The function takes a -** single parameter "keyClass". The return value of ftsHashFunction() -** is a pointer to another function. Specifically, the return value -** of ftsHashFunction() is a pointer to a function that takes two parameters -** with types "const void*" and "int" and returns an "int". -*/ -static int (*ftsHashFunction(int keyClass))(const void*,int){ - if( keyClass==FTS3_HASH_STRING ){ - return &fts3StrHash; - }else{ - assert( keyClass==FTS3_HASH_BINARY ); - return &fts3BinHash; - } -} - -/* -** Return a pointer to the appropriate hash function given the key class. -** -** For help in interpreted the obscure C code in the function definition, -** see the header comment on the previous function. -*/ -static int (*ftsCompareFunction(int keyClass))(const void*,int,const void*,int){ - if( keyClass==FTS3_HASH_STRING ){ - return &fts3StrCompare; - }else{ - assert( keyClass==FTS3_HASH_BINARY ); - return &fts3BinCompare; - } -} - -/* Link an element into the hash table -*/ -static void fts3HashInsertElement( - Fts3Hash *pH, /* The complete hash table */ - struct _fts3ht *pEntry, /* The entry into which pNew is inserted */ - Fts3HashElem *pNew /* The element to be inserted */ -){ - Fts3HashElem *pHead; /* First element already in pEntry */ - pHead = pEntry->chain; - if( pHead ){ - pNew->next = pHead; - pNew->prev = pHead->prev; - if( pHead->prev ){ pHead->prev->next = pNew; } - else { pH->first = pNew; } - pHead->prev = pNew; - }else{ - pNew->next = pH->first; - if( pH->first ){ pH->first->prev = pNew; } - pNew->prev = 0; - pH->first = pNew; - } - pEntry->count++; - pEntry->chain = pNew; -} - - -/* Resize the hash table so that it cantains "new_size" buckets. -** "new_size" must be a power of 2. The hash table might fail -** to resize if sqliteMalloc() fails. -** -** Return non-zero if a memory allocation error occurs. -*/ -static int fts3Rehash(Fts3Hash *pH, int new_size){ - struct _fts3ht *new_ht; /* The new hash table */ - Fts3HashElem *elem, *next_elem; /* For looping over existing elements */ - int (*xHash)(const void*,int); /* The hash function */ - - assert( (new_size & (new_size-1))==0 ); - new_ht = (struct _fts3ht *)fts3HashMalloc( new_size*sizeof(struct _fts3ht) ); - if( new_ht==0 ) return 1; - fts3HashFree(pH->ht); - pH->ht = new_ht; - pH->htsize = new_size; - xHash = ftsHashFunction(pH->keyClass); - for(elem=pH->first, pH->first=0; elem; elem = next_elem){ - int h = (*xHash)(elem->pKey, elem->nKey) & (new_size-1); - next_elem = elem->next; - fts3HashInsertElement(pH, &new_ht[h], elem); - } - return 0; -} - -/* This function (for internal use only) locates an element in an -** hash table that matches the given key. The hash for this key has -** already been computed and is passed as the 4th parameter. -*/ -static Fts3HashElem *fts3FindElementByHash( - const Fts3Hash *pH, /* The pH to be searched */ - const void *pKey, /* The key we are searching for */ - int nKey, - int h /* The hash for this key. */ -){ - Fts3HashElem *elem; /* Used to loop thru the element list */ - int count; /* Number of elements left to test */ - int (*xCompare)(const void*,int,const void*,int); /* comparison function */ - - if( pH->ht ){ - struct _fts3ht *pEntry = &pH->ht[h]; - elem = pEntry->chain; - count = pEntry->count; - xCompare = ftsCompareFunction(pH->keyClass); - while( count-- && elem ){ - if( (*xCompare)(elem->pKey,elem->nKey,pKey,nKey)==0 ){ - return elem; - } - elem = elem->next; - } - } - return 0; -} - -/* Remove a single entry from the hash table given a pointer to that -** element and a hash on the element's key. -*/ -static void fts3RemoveElementByHash( - Fts3Hash *pH, /* The pH containing "elem" */ - Fts3HashElem* elem, /* The element to be removed from the pH */ - int h /* Hash value for the element */ -){ - struct _fts3ht *pEntry; - if( elem->prev ){ - elem->prev->next = elem->next; - }else{ - pH->first = elem->next; - } - if( elem->next ){ - elem->next->prev = elem->prev; - } - pEntry = &pH->ht[h]; - if( pEntry->chain==elem ){ - pEntry->chain = elem->next; - } - pEntry->count--; - if( pEntry->count<=0 ){ - pEntry->chain = 0; - } - if( pH->copyKey && elem->pKey ){ - fts3HashFree(elem->pKey); - } - fts3HashFree( elem ); - pH->count--; - if( pH->count<=0 ){ - assert( pH->first==0 ); - assert( pH->count==0 ); - fts3HashClear(pH); - } -} - -Fts3HashElem *sqlite3Fts3HashFindElem( - const Fts3Hash *pH, - const void *pKey, - int nKey -){ - int h; /* A hash on key */ - int (*xHash)(const void*,int); /* The hash function */ - - if( pH==0 || pH->ht==0 ) return 0; - xHash = ftsHashFunction(pH->keyClass); - assert( xHash!=0 ); - h = (*xHash)(pKey,nKey); - assert( (pH->htsize & (pH->htsize-1))==0 ); - return fts3FindElementByHash(pH,pKey,nKey, h & (pH->htsize-1)); -} - -/* -** Attempt to locate an element of the hash table pH with a key -** that matches pKey,nKey. Return the data for this element if it is -** found, or NULL if there is no match. -*/ -void *sqlite3Fts3HashFind(const Fts3Hash *pH, const void *pKey, int nKey){ - Fts3HashElem *pElem; /* The element that matches key (if any) */ - - pElem = sqlite3Fts3HashFindElem(pH, pKey, nKey); - return pElem ? pElem->data : 0; -} - -/* Insert an element into the hash table pH. The key is pKey,nKey -** and the data is "data". -** -** If no element exists with a matching key, then a new -** element is created. A copy of the key is made if the copyKey -** flag is set. NULL is returned. -** -** If another element already exists with the same key, then the -** new data replaces the old data and the old data is returned. -** The key is not copied in this instance. If a malloc fails, then -** the new data is returned and the hash table is unchanged. -** -** If the "data" parameter to this function is NULL, then the -** element corresponding to "key" is removed from the hash table. -*/ -void *sqlite3Fts3HashInsert( - Fts3Hash *pH, /* The hash table to insert into */ - const void *pKey, /* The key */ - int nKey, /* Number of bytes in the key */ - void *data /* The data */ -){ - int hraw; /* Raw hash value of the key */ - int h; /* the hash of the key modulo hash table size */ - Fts3HashElem *elem; /* Used to loop thru the element list */ - Fts3HashElem *new_elem; /* New element added to the pH */ - int (*xHash)(const void*,int); /* The hash function */ - - assert( pH!=0 ); - xHash = ftsHashFunction(pH->keyClass); - assert( xHash!=0 ); - hraw = (*xHash)(pKey, nKey); - assert( (pH->htsize & (pH->htsize-1))==0 ); - h = hraw & (pH->htsize-1); - elem = fts3FindElementByHash(pH,pKey,nKey,h); - if( elem ){ - void *old_data = elem->data; - if( data==0 ){ - fts3RemoveElementByHash(pH,elem,h); - }else{ - elem->data = data; - } - return old_data; - } - if( data==0 ) return 0; - if( (pH->htsize==0 && fts3Rehash(pH,8)) - || (pH->count>=pH->htsize && fts3Rehash(pH, pH->htsize*2)) - ){ - pH->count = 0; - return data; - } - assert( pH->htsize>0 ); - new_elem = (Fts3HashElem*)fts3HashMalloc( sizeof(Fts3HashElem) ); - if( new_elem==0 ) return data; - if( pH->copyKey && pKey!=0 ){ - new_elem->pKey = fts3HashMalloc( nKey ); - if( new_elem->pKey==0 ){ - fts3HashFree(new_elem); - return data; - } - memcpy((void*)new_elem->pKey, pKey, nKey); - }else{ - new_elem->pKey = (void*)pKey; - } - new_elem->nKey = nKey; - pH->count++; - assert( pH->htsize>0 ); - assert( (pH->htsize & (pH->htsize-1))==0 ); - h = hraw & (pH->htsize-1); - fts3HashInsertElement(pH, &pH->ht[h], new_elem); - new_elem->data = data; - return 0; -} - -#endif /* !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3) */ diff --git a/src/libtracker-fts/fts3_hash.h b/src/libtracker-fts/fts3_hash.h deleted file mode 100644 index 399f51544..000000000 --- a/src/libtracker-fts/fts3_hash.h +++ /dev/null @@ -1,112 +0,0 @@ -/* -** 2001 September 22 -** -** The author disclaims copyright to this source code. In place of -** a legal notice, here is a blessing: -** -** May you do good and not evil. -** May you find forgiveness for yourself and forgive others. -** May you share freely, never taking more than you give. -** -************************************************************************* -** This is the header file for the generic hash-table implemenation -** used in SQLite. We've modified it slightly to serve as a standalone -** hash table implementation for the full-text indexing module. -** -*/ -#ifndef _FTS3_HASH_H_ -#define _FTS3_HASH_H_ - -/* Forward declarations of structures. */ -typedef struct Fts3Hash Fts3Hash; -typedef struct Fts3HashElem Fts3HashElem; - -/* A complete hash table is an instance of the following structure. -** The internals of this structure are intended to be opaque -- client -** code should not attempt to access or modify the fields of this structure -** directly. Change this structure only by using the routines below. -** However, many of the "procedures" and "functions" for modifying and -** accessing this structure are really macros, so we can't really make -** this structure opaque. -*/ -struct Fts3Hash { - char keyClass; /* HASH_INT, _POINTER, _STRING, _BINARY */ - char copyKey; /* True if copy of key made on insert */ - int count; /* Number of entries in this table */ - Fts3HashElem *first; /* The first element of the array */ - int htsize; /* Number of buckets in the hash table */ - struct _fts3ht { /* the hash table */ - int count; /* Number of entries with this hash */ - Fts3HashElem *chain; /* Pointer to first entry with this hash */ - } *ht; -}; - -/* Each element in the hash table is an instance of the following -** structure. All elements are stored on a single doubly-linked list. -** -** Again, this structure is intended to be opaque, but it can't really -** be opaque because it is used by macros. -*/ -struct Fts3HashElem { - Fts3HashElem *next, *prev; /* Next and previous elements in the table */ - void *data; /* Data associated with this element */ - void *pKey; int nKey; /* Key associated with this element */ -}; - -/* -** There are 2 different modes of operation for a hash table: -** -** FTS3_HASH_STRING pKey points to a string that is nKey bytes long -** (including the null-terminator, if any). Case -** is respected in comparisons. -** -** FTS3_HASH_BINARY pKey points to binary data nKey bytes long. -** memcmp() is used to compare keys. -** -** A copy of the key is made if the copyKey parameter to fts3HashInit is 1. -*/ -#define FTS3_HASH_STRING 1 -#define FTS3_HASH_BINARY 2 - -/* -** Access routines. To delete, insert a NULL pointer. -*/ -void sqlite3Fts3HashInit(Fts3Hash *pNew, char keyClass, char copyKey); -void *sqlite3Fts3HashInsert(Fts3Hash*, const void *pKey, int nKey, void *pData); -void *sqlite3Fts3HashFind(const Fts3Hash*, const void *pKey, int nKey); -void sqlite3Fts3HashClear(Fts3Hash*); -Fts3HashElem *sqlite3Fts3HashFindElem(const Fts3Hash *, const void *, int); - -/* -** Shorthand for the functions above -*/ -#define fts3HashInit sqlite3Fts3HashInit -#define fts3HashInsert sqlite3Fts3HashInsert -#define fts3HashFind sqlite3Fts3HashFind -#define fts3HashClear sqlite3Fts3HashClear -#define fts3HashFindElem sqlite3Fts3HashFindElem - -/* -** Macros for looping over all elements of a hash table. The idiom is -** like this: -** -** Fts3Hash h; -** Fts3HashElem *p; -** ... -** for(p=fts3HashFirst(&h); p; p=fts3HashNext(p)){ -** SomeStructure *pData = fts3HashData(p); -** // do something with pData -** } -*/ -#define fts3HashFirst(H) ((H)->first) -#define fts3HashNext(E) ((E)->next) -#define fts3HashData(E) ((E)->data) -#define fts3HashKey(E) ((E)->pKey) -#define fts3HashKeysize(E) ((E)->nKey) - -/* -** Number of entries in a hash table -*/ -#define fts3HashCount(H) ((H)->count) - -#endif /* _FTS3_HASH_H_ */ diff --git a/src/libtracker-fts/fts3_icu.c b/src/libtracker-fts/fts3_icu.c deleted file mode 100644 index 52df8c7d8..000000000 --- a/src/libtracker-fts/fts3_icu.c +++ /dev/null @@ -1,261 +0,0 @@ -/* -** 2007 June 22 -** -** The author disclaims copyright to this source code. In place of -** a legal notice, here is a blessing: -** -** May you do good and not evil. -** May you find forgiveness for yourself and forgive others. -** May you share freely, never taking more than you give. -** -************************************************************************* -** This file implements a tokenizer for fts3 based on the ICU library. -*/ -#include "fts3Int.h" -#if !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3) -#ifdef SQLITE_ENABLE_ICU - -#include <assert.h> -#include <string.h> -#include "fts3_tokenizer.h" - -#include <unicode/ubrk.h> -#include <unicode/ucol.h> -#include <unicode/ustring.h> -#include <unicode/utf16.h> - -typedef struct IcuTokenizer IcuTokenizer; -typedef struct IcuCursor IcuCursor; - -struct IcuTokenizer { - sqlite3_tokenizer base; - char *zLocale; -}; - -struct IcuCursor { - sqlite3_tokenizer_cursor base; - - UBreakIterator *pIter; /* ICU break-iterator object */ - int nChar; /* Number of UChar elements in pInput */ - UChar *aChar; /* Copy of input using utf-16 encoding */ - int *aOffset; /* Offsets of each character in utf-8 input */ - - int nBuffer; - char *zBuffer; - - int iToken; -}; - -/* -** Create a new tokenizer instance. -*/ -static int icuCreate( - int argc, /* Number of entries in argv[] */ - const char * const *argv, /* Tokenizer creation arguments */ - sqlite3_tokenizer **ppTokenizer /* OUT: Created tokenizer */ -){ - IcuTokenizer *p; - int n = 0; - - if( argc>0 ){ - n = strlen(argv[0])+1; - } - p = (IcuTokenizer *)sqlite3_malloc(sizeof(IcuTokenizer)+n); - if( !p ){ - return SQLITE_NOMEM; - } - memset(p, 0, sizeof(IcuTokenizer)); - - if( n ){ - p->zLocale = (char *)&p[1]; - memcpy(p->zLocale, argv[0], n); - } - - *ppTokenizer = (sqlite3_tokenizer *)p; - - return SQLITE_OK; -} - -/* -** Destroy a tokenizer -*/ -static int icuDestroy(sqlite3_tokenizer *pTokenizer){ - IcuTokenizer *p = (IcuTokenizer *)pTokenizer; - sqlite3_free(p); - return SQLITE_OK; -} - -/* -** Prepare to begin tokenizing a particular string. The input -** string to be tokenized is pInput[0..nBytes-1]. A cursor -** used to incrementally tokenize this string is returned in -** *ppCursor. -*/ -static int icuOpen( - sqlite3_tokenizer *pTokenizer, /* The tokenizer */ - const char *zInput, /* Input string */ - int nInput, /* Length of zInput in bytes */ - sqlite3_tokenizer_cursor **ppCursor /* OUT: Tokenization cursor */ -){ - IcuTokenizer *p = (IcuTokenizer *)pTokenizer; - IcuCursor *pCsr; - - const int32_t opt = U_FOLD_CASE_DEFAULT; - UErrorCode status = U_ZERO_ERROR; - int nChar; - - UChar32 c; - int iInput = 0; - int iOut = 0; - - *ppCursor = 0; - - if( zInput==0 ){ - nInput = 0; - zInput = ""; - }else if( nInput<0 ){ - nInput = strlen(zInput); - } - nChar = nInput+1; - pCsr = (IcuCursor *)sqlite3_malloc( - sizeof(IcuCursor) + /* IcuCursor */ - ((nChar+3)&~3) * sizeof(UChar) + /* IcuCursor.aChar[] */ - (nChar+1) * sizeof(int) /* IcuCursor.aOffset[] */ - ); - if( !pCsr ){ - return SQLITE_NOMEM; - } - memset(pCsr, 0, sizeof(IcuCursor)); - pCsr->aChar = (UChar *)&pCsr[1]; - pCsr->aOffset = (int *)&pCsr->aChar[(nChar+3)&~3]; - - pCsr->aOffset[iOut] = iInput; - U8_NEXT(zInput, iInput, nInput, c); - while( c>0 ){ - int isError = 0; - c = u_foldCase(c, opt); - U16_APPEND(pCsr->aChar, iOut, nChar, c, isError); - if( isError ){ - sqlite3_free(pCsr); - return SQLITE_ERROR; - } - pCsr->aOffset[iOut] = iInput; - - if( iInput<nInput ){ - U8_NEXT(zInput, iInput, nInput, c); - }else{ - c = 0; - } - } - - pCsr->pIter = ubrk_open(UBRK_WORD, p->zLocale, pCsr->aChar, iOut, &status); - if( !U_SUCCESS(status) ){ - sqlite3_free(pCsr); - return SQLITE_ERROR; - } - pCsr->nChar = iOut; - - ubrk_first(pCsr->pIter); - *ppCursor = (sqlite3_tokenizer_cursor *)pCsr; - return SQLITE_OK; -} - -/* -** Close a tokenization cursor previously opened by a call to icuOpen(). -*/ -static int icuClose(sqlite3_tokenizer_cursor *pCursor){ - IcuCursor *pCsr = (IcuCursor *)pCursor; - ubrk_close(pCsr->pIter); - sqlite3_free(pCsr->zBuffer); - sqlite3_free(pCsr); - return SQLITE_OK; -} - -/* -** Extract the next token from a tokenization cursor. -*/ -static int icuNext( - sqlite3_tokenizer_cursor *pCursor, /* Cursor returned by simpleOpen */ - const char **ppToken, /* OUT: *ppToken is the token text */ - int *pnBytes, /* OUT: Number of bytes in token */ - int *piStartOffset, /* OUT: Starting offset of token */ - int *piEndOffset, /* OUT: Ending offset of token */ - int *piPosition /* OUT: Position integer of token */ -){ - IcuCursor *pCsr = (IcuCursor *)pCursor; - - int iStart = 0; - int iEnd = 0; - int nByte = 0; - - while( iStart==iEnd ){ - UChar32 c; - - iStart = ubrk_current(pCsr->pIter); - iEnd = ubrk_next(pCsr->pIter); - if( iEnd==UBRK_DONE ){ - return SQLITE_DONE; - } - - while( iStart<iEnd ){ - int iWhite = iStart; - U16_NEXT(pCsr->aChar, iWhite, pCsr->nChar, c); - if( u_isspace(c) ){ - iStart = iWhite; - }else{ - break; - } - } - assert(iStart<=iEnd); - } - - do { - UErrorCode status = U_ZERO_ERROR; - if( nByte ){ - char *zNew = sqlite3_realloc(pCsr->zBuffer, nByte); - if( !zNew ){ - return SQLITE_NOMEM; - } - pCsr->zBuffer = zNew; - pCsr->nBuffer = nByte; - } - - u_strToUTF8( - pCsr->zBuffer, pCsr->nBuffer, &nByte, /* Output vars */ - &pCsr->aChar[iStart], iEnd-iStart, /* Input vars */ - &status /* Output success/failure */ - ); - } while( nByte>pCsr->nBuffer ); - - *ppToken = pCsr->zBuffer; - *pnBytes = nByte; - *piStartOffset = pCsr->aOffset[iStart]; - *piEndOffset = pCsr->aOffset[iEnd]; - *piPosition = pCsr->iToken++; - - return SQLITE_OK; -} - -/* -** The set of routines that implement the simple tokenizer -*/ -static const sqlite3_tokenizer_module icuTokenizerModule = { - 0, /* iVersion */ - icuCreate, /* xCreate */ - icuDestroy, /* xCreate */ - icuOpen, /* xOpen */ - icuClose, /* xClose */ - icuNext, /* xNext */ -}; - -/* -** Set *ppModule to point at the implementation of the ICU tokenizer. -*/ -void sqlite3Fts3IcuTokenizerModule( - sqlite3_tokenizer_module const**ppModule -){ - *ppModule = &icuTokenizerModule; -} - -#endif /* defined(SQLITE_ENABLE_ICU) */ -#endif /* !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3) */ diff --git a/src/libtracker-fts/fts3_porter.c b/src/libtracker-fts/fts3_porter.c deleted file mode 100644 index 579745b85..000000000 --- a/src/libtracker-fts/fts3_porter.c +++ /dev/null @@ -1,646 +0,0 @@ -/* -** 2006 September 30 -** -** The author disclaims copyright to this source code. In place of -** a legal notice, here is a blessing: -** -** May you do good and not evil. -** May you find forgiveness for yourself and forgive others. -** May you share freely, never taking more than you give. -** -************************************************************************* -** Implementation of the full-text-search tokenizer that implements -** a Porter stemmer. -*/ - -/* -** The code in this file is only compiled if: -** -** * The FTS3 module is being built as an extension -** (in which case SQLITE_CORE is not defined), or -** -** * The FTS3 module is being built into the core of -** SQLite (in which case SQLITE_ENABLE_FTS3 is defined). -*/ -#include "fts3Int.h" -#if !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3) - -#include <assert.h> -#include <stdlib.h> -#include <stdio.h> -#include <string.h> - -#include "fts3_tokenizer.h" - -/* -** Class derived from sqlite3_tokenizer -*/ -typedef struct porter_tokenizer { - sqlite3_tokenizer base; /* Base class */ -} porter_tokenizer; - -/* -** Class derived from sqlite3_tokenizer_cursor -*/ -typedef struct porter_tokenizer_cursor { - sqlite3_tokenizer_cursor base; - const char *zInput; /* input we are tokenizing */ - int nInput; /* size of the input */ - int iOffset; /* current position in zInput */ - int iToken; /* index of next token to be returned */ - char *zToken; /* storage for current token */ - int nAllocated; /* space allocated to zToken buffer */ -} porter_tokenizer_cursor; - - -/* -** Create a new tokenizer instance. -*/ -static int porterCreate( - int argc, const char * const *argv, - sqlite3_tokenizer **ppTokenizer -){ - porter_tokenizer *t; - - UNUSED_PARAMETER(argc); - UNUSED_PARAMETER(argv); - - t = (porter_tokenizer *) sqlite3_malloc(sizeof(*t)); - if( t==NULL ) return SQLITE_NOMEM; - memset(t, 0, sizeof(*t)); - *ppTokenizer = &t->base; - return SQLITE_OK; -} - -/* -** Destroy a tokenizer -*/ -static int porterDestroy(sqlite3_tokenizer *pTokenizer){ - sqlite3_free(pTokenizer); - return SQLITE_OK; -} - -/* -** Prepare to begin tokenizing a particular string. The input -** string to be tokenized is zInput[0..nInput-1]. A cursor -** used to incrementally tokenize this string is returned in -** *ppCursor. -*/ -static int porterOpen( - sqlite3_tokenizer *pTokenizer, /* The tokenizer */ - const char *zInput, int nInput, /* String to be tokenized */ - sqlite3_tokenizer_cursor **ppCursor /* OUT: Tokenization cursor */ -){ - porter_tokenizer_cursor *c; - - UNUSED_PARAMETER(pTokenizer); - - c = (porter_tokenizer_cursor *) sqlite3_malloc(sizeof(*c)); - if( c==NULL ) return SQLITE_NOMEM; - - c->zInput = zInput; - if( zInput==0 ){ - c->nInput = 0; - }else if( nInput<0 ){ - c->nInput = (int)strlen(zInput); - }else{ - c->nInput = nInput; - } - c->iOffset = 0; /* start tokenizing at the beginning */ - c->iToken = 0; - c->zToken = NULL; /* no space allocated, yet. */ - c->nAllocated = 0; - - *ppCursor = &c->base; - return SQLITE_OK; -} - -/* -** Close a tokenization cursor previously opened by a call to -** porterOpen() above. -*/ -static int porterClose(sqlite3_tokenizer_cursor *pCursor){ - porter_tokenizer_cursor *c = (porter_tokenizer_cursor *) pCursor; - sqlite3_free(c->zToken); - sqlite3_free(c); - return SQLITE_OK; -} -/* -** Vowel or consonant -*/ -static const char cType[] = { - 0, 1, 1, 1, 0, 1, 1, 1, 0, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 0, - 1, 1, 1, 2, 1 -}; - -/* -** isConsonant() and isVowel() determine if their first character in -** the string they point to is a consonant or a vowel, according -** to Porter ruls. -** -** A consonate is any letter other than 'a', 'e', 'i', 'o', or 'u'. -** 'Y' is a consonant unless it follows another consonant, -** in which case it is a vowel. -** -** In these routine, the letters are in reverse order. So the 'y' rule -** is that 'y' is a consonant unless it is followed by another -** consonent. -*/ -static int isVowel(const char*); -static int isConsonant(const char *z){ - int j; - char x = *z; - if( x==0 ) return 0; - assert( x>='a' && x<='z' ); - j = cType[x-'a']; - if( j<2 ) return j; - return z[1]==0 || isVowel(z + 1); -} -static int isVowel(const char *z){ - int j; - char x = *z; - if( x==0 ) return 0; - assert( x>='a' && x<='z' ); - j = cType[x-'a']; - if( j<2 ) return 1-j; - return isConsonant(z + 1); -} - -/* -** Let any sequence of one or more vowels be represented by V and let -** C be sequence of one or more consonants. Then every word can be -** represented as: -** -** [C] (VC){m} [V] -** -** In prose: A word is an optional consonant followed by zero or -** vowel-consonant pairs followed by an optional vowel. "m" is the -** number of vowel consonant pairs. This routine computes the value -** of m for the first i bytes of a word. -** -** Return true if the m-value for z is 1 or more. In other words, -** return true if z contains at least one vowel that is followed -** by a consonant. -** -** In this routine z[] is in reverse order. So we are really looking -** for an instance of of a consonant followed by a vowel. -*/ -static int m_gt_0(const char *z){ - while( isVowel(z) ){ z++; } - if( *z==0 ) return 0; - while( isConsonant(z) ){ z++; } - return *z!=0; -} - -/* Like mgt0 above except we are looking for a value of m which is -** exactly 1 -*/ -static int m_eq_1(const char *z){ - while( isVowel(z) ){ z++; } - if( *z==0 ) return 0; - while( isConsonant(z) ){ z++; } - if( *z==0 ) return 0; - while( isVowel(z) ){ z++; } - if( *z==0 ) return 1; - while( isConsonant(z) ){ z++; } - return *z==0; -} - -/* Like mgt0 above except we are looking for a value of m>1 instead -** or m>0 -*/ -static int m_gt_1(const char *z){ - while( isVowel(z) ){ z++; } - if( *z==0 ) return 0; - while( isConsonant(z) ){ z++; } - if( *z==0 ) return 0; - while( isVowel(z) ){ z++; } - if( *z==0 ) return 0; - while( isConsonant(z) ){ z++; } - return *z!=0; -} - -/* -** Return TRUE if there is a vowel anywhere within z[0..n-1] -*/ -static int hasVowel(const char *z){ - while( isConsonant(z) ){ z++; } - return *z!=0; -} - -/* -** Return TRUE if the word ends in a double consonant. -** -** The text is reversed here. So we are really looking at -** the first two characters of z[]. -*/ -static int doubleConsonant(const char *z){ - return isConsonant(z) && z[0]==z[1]; -} - -/* -** Return TRUE if the word ends with three letters which -** are consonant-vowel-consonent and where the final consonant -** is not 'w', 'x', or 'y'. -** -** The word is reversed here. So we are really checking the -** first three letters and the first one cannot be in [wxy]. -*/ -static int star_oh(const char *z){ - return - isConsonant(z) && - z[0]!='w' && z[0]!='x' && z[0]!='y' && - isVowel(z+1) && - isConsonant(z+2); -} - -/* -** If the word ends with zFrom and xCond() is true for the stem -** of the word that preceeds the zFrom ending, then change the -** ending to zTo. -** -** The input word *pz and zFrom are both in reverse order. zTo -** is in normal order. -** -** Return TRUE if zFrom matches. Return FALSE if zFrom does not -** match. Not that TRUE is returned even if xCond() fails and -** no substitution occurs. -*/ -static int stem( - char **pz, /* The word being stemmed (Reversed) */ - const char *zFrom, /* If the ending matches this... (Reversed) */ - const char *zTo, /* ... change the ending to this (not reversed) */ - int (*xCond)(const char*) /* Condition that must be true */ -){ - char *z = *pz; - while( *zFrom && *zFrom==*z ){ z++; zFrom++; } - if( *zFrom!=0 ) return 0; - if( xCond && !xCond(z) ) return 1; - while( *zTo ){ - *(--z) = *(zTo++); - } - *pz = z; - return 1; -} - -/* -** This is the fallback stemmer used when the porter stemmer is -** inappropriate. The input word is copied into the output with -** US-ASCII case folding. If the input word is too long (more -** than 20 bytes if it contains no digits or more than 6 bytes if -** it contains digits) then word is truncated to 20 or 6 bytes -** by taking 10 or 3 bytes from the beginning and end. -*/ -static void copy_stemmer(const char *zIn, int nIn, char *zOut, int *pnOut){ - int i, mx, j; - int hasDigit = 0; - for(i=0; i<nIn; i++){ - char c = zIn[i]; - if( c>='A' && c<='Z' ){ - zOut[i] = c - 'A' + 'a'; - }else{ - if( c>='0' && c<='9' ) hasDigit = 1; - zOut[i] = c; - } - } - mx = hasDigit ? 3 : 10; - if( nIn>mx*2 ){ - for(j=mx, i=nIn-mx; i<nIn; i++, j++){ - zOut[j] = zOut[i]; - } - i = j; - } - zOut[i] = 0; - *pnOut = i; -} - - -/* -** Stem the input word zIn[0..nIn-1]. Store the output in zOut. -** zOut is at least big enough to hold nIn bytes. Write the actual -** size of the output word (exclusive of the '\0' terminator) into *pnOut. -** -** Any upper-case characters in the US-ASCII character set ([A-Z]) -** are converted to lower case. Upper-case UTF characters are -** unchanged. -** -** Words that are longer than about 20 bytes are stemmed by retaining -** a few bytes from the beginning and the end of the word. If the -** word contains digits, 3 bytes are taken from the beginning and -** 3 bytes from the end. For long words without digits, 10 bytes -** are taken from each end. US-ASCII case folding still applies. -** -** If the input word contains not digits but does characters not -** in [a-zA-Z] then no stemming is attempted and this routine just -** copies the input into the input into the output with US-ASCII -** case folding. -** -** Stemming never increases the length of the word. So there is -** no chance of overflowing the zOut buffer. -*/ -static void porter_stemmer(const char *zIn, int nIn, char *zOut, int *pnOut){ - int i, j; - char zReverse[28]; - char *z, *z2; - if( nIn<3 || nIn>=(int)sizeof(zReverse)-7 ){ - /* The word is too big or too small for the porter stemmer. - ** Fallback to the copy stemmer */ - copy_stemmer(zIn, nIn, zOut, pnOut); - return; - } - for(i=0, j=sizeof(zReverse)-6; i<nIn; i++, j--){ - char c = zIn[i]; - if( c>='A' && c<='Z' ){ - zReverse[j] = c + 'a' - 'A'; - }else if( c>='a' && c<='z' ){ - zReverse[j] = c; - }else{ - /* The use of a character not in [a-zA-Z] means that we fallback - ** to the copy stemmer */ - copy_stemmer(zIn, nIn, zOut, pnOut); - return; - } - } - memset(&zReverse[sizeof(zReverse)-5], 0, 5); - z = &zReverse[j+1]; - - - /* Step 1a */ - if( z[0]=='s' ){ - if( - !stem(&z, "sess", "ss", 0) && - !stem(&z, "sei", "i", 0) && - !stem(&z, "ss", "ss", 0) - ){ - z++; - } - } - - /* Step 1b */ - z2 = z; - if( stem(&z, "dee", "ee", m_gt_0) ){ - /* Do nothing. The work was all in the test */ - }else if( - (stem(&z, "gni", "", hasVowel) || stem(&z, "de", "", hasVowel)) - && z!=z2 - ){ - if( stem(&z, "ta", "ate", 0) || - stem(&z, "lb", "ble", 0) || - stem(&z, "zi", "ize", 0) ){ - /* Do nothing. The work was all in the test */ - }else if( doubleConsonant(z) && (*z!='l' && *z!='s' && *z!='z') ){ - z++; - }else if( m_eq_1(z) && star_oh(z) ){ - *(--z) = 'e'; - } - } - - /* Step 1c */ - if( z[0]=='y' && hasVowel(z+1) ){ - z[0] = 'i'; - } - - /* Step 2 */ - switch( z[1] ){ - case 'a': - stem(&z, "lanoita", "ate", m_gt_0) || - stem(&z, "lanoit", "tion", m_gt_0); - break; - case 'c': - stem(&z, "icne", "ence", m_gt_0) || - stem(&z, "icna", "ance", m_gt_0); - break; - case 'e': - stem(&z, "rezi", "ize", m_gt_0); - break; - case 'g': - stem(&z, "igol", "log", m_gt_0); - break; - case 'l': - stem(&z, "ilb", "ble", m_gt_0) || - stem(&z, "illa", "al", m_gt_0) || - stem(&z, "iltne", "ent", m_gt_0) || - stem(&z, "ile", "e", m_gt_0) || - stem(&z, "ilsuo", "ous", m_gt_0); - break; - case 'o': - stem(&z, "noitazi", "ize", m_gt_0) || - stem(&z, "noita", "ate", m_gt_0) || - stem(&z, "rota", "ate", m_gt_0); - break; - case 's': - stem(&z, "msila", "al", m_gt_0) || - stem(&z, "ssenevi", "ive", m_gt_0) || - stem(&z, "ssenluf", "ful", m_gt_0) || - stem(&z, "ssensuo", "ous", m_gt_0); - break; - case 't': - stem(&z, "itila", "al", m_gt_0) || - stem(&z, "itivi", "ive", m_gt_0) || - stem(&z, "itilib", "ble", m_gt_0); - break; - } - - /* Step 3 */ - switch( z[0] ){ - case 'e': - stem(&z, "etaci", "ic", m_gt_0) || - stem(&z, "evita", "", m_gt_0) || - stem(&z, "ezila", "al", m_gt_0); - break; - case 'i': - stem(&z, "itici", "ic", m_gt_0); - break; - case 'l': - stem(&z, "laci", "ic", m_gt_0) || - stem(&z, "luf", "", m_gt_0); - break; - case 's': - stem(&z, "ssen", "", m_gt_0); - break; - } - - /* Step 4 */ - switch( z[1] ){ - case 'a': - if( z[0]=='l' && m_gt_1(z+2) ){ - z += 2; - } - break; - case 'c': - if( z[0]=='e' && z[2]=='n' && (z[3]=='a' || z[3]=='e') && m_gt_1(z+4) ){ - z += 4; - } - break; - case 'e': - if( z[0]=='r' && m_gt_1(z+2) ){ - z += 2; - } - break; - case 'i': - if( z[0]=='c' && m_gt_1(z+2) ){ - z += 2; - } - break; - case 'l': - if( z[0]=='e' && z[2]=='b' && (z[3]=='a' || z[3]=='i') && m_gt_1(z+4) ){ - z += 4; - } - break; - case 'n': - if( z[0]=='t' ){ - if( z[2]=='a' ){ - if( m_gt_1(z+3) ){ - z += 3; - } - }else if( z[2]=='e' ){ - stem(&z, "tneme", "", m_gt_1) || - stem(&z, "tnem", "", m_gt_1) || - stem(&z, "tne", "", m_gt_1); - } - } - break; - case 'o': - if( z[0]=='u' ){ - if( m_gt_1(z+2) ){ - z += 2; - } - }else if( z[3]=='s' || z[3]=='t' ){ - stem(&z, "noi", "", m_gt_1); - } - break; - case 's': - if( z[0]=='m' && z[2]=='i' && m_gt_1(z+3) ){ - z += 3; - } - break; - case 't': - stem(&z, "eta", "", m_gt_1) || - stem(&z, "iti", "", m_gt_1); - break; - case 'u': - if( z[0]=='s' && z[2]=='o' && m_gt_1(z+3) ){ - z += 3; - } - break; - case 'v': - case 'z': - if( z[0]=='e' && z[2]=='i' && m_gt_1(z+3) ){ - z += 3; - } - break; - } - - /* Step 5a */ - if( z[0]=='e' ){ - if( m_gt_1(z+1) ){ - z++; - }else if( m_eq_1(z+1) && !star_oh(z+1) ){ - z++; - } - } - - /* Step 5b */ - if( m_gt_1(z) && z[0]=='l' && z[1]=='l' ){ - z++; - } - - /* z[] is now the stemmed word in reverse order. Flip it back - ** around into forward order and return. - */ - *pnOut = i = (int)strlen(z); - zOut[i] = 0; - while( *z ){ - zOut[--i] = *(z++); - } -} - -/* -** Characters that can be part of a token. We assume any character -** whose value is greater than 0x80 (any UTF character) can be -** part of a token. In other words, delimiters all must have -** values of 0x7f or lower. -*/ -static const char porterIdChar[] = { -/* x0 x1 x2 x3 x4 x5 x6 x7 x8 x9 xA xB xC xD xE xF */ - 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, /* 3x */ - 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, /* 4x */ - 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 1, /* 5x */ - 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, /* 6x */ - 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, /* 7x */ -}; -#define isDelim(C) (((ch=C)&0x80)==0 && (ch<0x30 || !porterIdChar[ch-0x30])) - -/* -** Extract the next token from a tokenization cursor. The cursor must -** have been opened by a prior call to porterOpen(). -*/ -static int porterNext( - sqlite3_tokenizer_cursor *pCursor, /* Cursor returned by porterOpen */ - const char **pzToken, /* OUT: *pzToken is the token text */ - int *pnBytes, /* OUT: Number of bytes in token */ - int *piStartOffset, /* OUT: Starting offset of token */ - int *piEndOffset, /* OUT: Ending offset of token */ - int *piPosition /* OUT: Position integer of token */ -){ - porter_tokenizer_cursor *c = (porter_tokenizer_cursor *) pCursor; - const char *z = c->zInput; - - while( c->iOffset<c->nInput ){ - int iStartOffset, ch; - - /* Scan past delimiter characters */ - while( c->iOffset<c->nInput && isDelim(z[c->iOffset]) ){ - c->iOffset++; - } - - /* Count non-delimiter characters. */ - iStartOffset = c->iOffset; - while( c->iOffset<c->nInput && !isDelim(z[c->iOffset]) ){ - c->iOffset++; - } - - if( c->iOffset>iStartOffset ){ - int n = c->iOffset-iStartOffset; - if( n>c->nAllocated ){ - char *pNew; - c->nAllocated = n+20; - pNew = sqlite3_realloc(c->zToken, c->nAllocated); - if( !pNew ) return SQLITE_NOMEM; - c->zToken = pNew; - } - porter_stemmer(&z[iStartOffset], n, c->zToken, pnBytes); - *pzToken = c->zToken; - *piStartOffset = iStartOffset; - *piEndOffset = c->iOffset; - *piPosition = c->iToken++; - return SQLITE_OK; - } - } - return SQLITE_DONE; -} - -/* -** The set of routines that implement the porter-stemmer tokenizer -*/ -static const sqlite3_tokenizer_module porterTokenizerModule = { - 0, - porterCreate, - porterDestroy, - porterOpen, - porterClose, - porterNext, - 0 -}; - -/* -** Allocate a new porter tokenizer. Return a pointer to the new -** tokenizer in *ppModule -*/ -void sqlite3Fts3PorterTokenizerModule( - sqlite3_tokenizer_module const**ppModule -){ - *ppModule = &porterTokenizerModule; -} - -#endif /* !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3) */ diff --git a/src/libtracker-fts/fts3_snippet.c b/src/libtracker-fts/fts3_snippet.c deleted file mode 100644 index 4bee014dc..000000000 --- a/src/libtracker-fts/fts3_snippet.c +++ /dev/null @@ -1,1520 +0,0 @@ -/* -** 2009 Oct 23 -** -** The author disclaims copyright to this source code. In place of -** a legal notice, here is a blessing: -** -** May you do good and not evil. -** May you find forgiveness for yourself and forgive others. -** May you share freely, never taking more than you give. -** -****************************************************************************** -*/ - -#include "fts3Int.h" -#if !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3) - -#include <string.h> -#include <assert.h> - -/* -** Characters that may appear in the second argument to matchinfo(). -*/ -#define FTS3_MATCHINFO_NPHRASE 'p' /* 1 value */ -#define FTS3_MATCHINFO_NCOL 'c' /* 1 value */ -#define FTS3_MATCHINFO_NDOC 'n' /* 1 value */ -#define FTS3_MATCHINFO_AVGLENGTH 'a' /* nCol values */ -#define FTS3_MATCHINFO_LENGTH 'l' /* nCol values */ -#define FTS3_MATCHINFO_LCS 's' /* nCol values */ -#define FTS3_MATCHINFO_HITS 'x' /* 3*nCol*nPhrase values */ - -/* -** The default value for the second argument to matchinfo(). -*/ -#define FTS3_MATCHINFO_DEFAULT "pcx" - - -/* -** Used as an fts3ExprIterate() context when loading phrase doclists to -** Fts3Expr.aDoclist[]/nDoclist. -*/ -typedef struct LoadDoclistCtx LoadDoclistCtx; -struct LoadDoclistCtx { - Fts3Cursor *pCsr; /* FTS3 Cursor */ - int nPhrase; /* Number of phrases seen so far */ - int nToken; /* Number of tokens seen so far */ -}; - -/* -** The following types are used as part of the implementation of the -** fts3BestSnippet() routine. -*/ -typedef struct SnippetIter SnippetIter; -typedef struct SnippetPhrase SnippetPhrase; -typedef struct SnippetFragment SnippetFragment; - -struct SnippetIter { - Fts3Cursor *pCsr; /* Cursor snippet is being generated from */ - int iCol; /* Extract snippet from this column */ - int nSnippet; /* Requested snippet length (in tokens) */ - int nPhrase; /* Number of phrases in query */ - SnippetPhrase *aPhrase; /* Array of size nPhrase */ - int iCurrent; /* First token of current snippet */ -}; - -struct SnippetPhrase { - int nToken; /* Number of tokens in phrase */ - char *pList; /* Pointer to start of phrase position list */ - int iHead; /* Next value in position list */ - char *pHead; /* Position list data following iHead */ - int iTail; /* Next value in trailing position list */ - char *pTail; /* Position list data following iTail */ -}; - -struct SnippetFragment { - int iCol; /* Column snippet is extracted from */ - int iPos; /* Index of first token in snippet */ - u64 covered; /* Mask of query phrases covered */ - u64 hlmask; /* Mask of snippet terms to highlight */ -}; - -/* -** This type is used as an fts3ExprIterate() context object while -** accumulating the data returned by the matchinfo() function. -*/ -typedef struct MatchInfo MatchInfo; -struct MatchInfo { - Fts3Cursor *pCursor; /* FTS3 Cursor */ - int nCol; /* Number of columns in table */ - int nPhrase; /* Number of matchable phrases in query */ - sqlite3_int64 nDoc; /* Number of docs in database */ - u32 *aMatchinfo; /* Pre-allocated buffer */ -}; - - - -/* -** The snippet() and offsets() functions both return text values. An instance -** of the following structure is used to accumulate those values while the -** functions are running. See fts3StringAppend() for details. -*/ -typedef struct StrBuffer StrBuffer; -struct StrBuffer { - char *z; /* Pointer to buffer containing string */ - int n; /* Length of z in bytes (excl. nul-term) */ - int nAlloc; /* Allocated size of buffer z in bytes */ -}; - - -/* -** This function is used to help iterate through a position-list. A position -** list is a list of unique integers, sorted from smallest to largest. Each -** element of the list is represented by an FTS3 varint that takes the value -** of the difference between the current element and the previous one plus -** two. For example, to store the position-list: -** -** 4 9 113 -** -** the three varints: -** -** 6 7 106 -** -** are encoded. -** -** When this function is called, *pp points to the start of an element of -** the list. *piPos contains the value of the previous entry in the list. -** After it returns, *piPos contains the value of the next element of the -** list and *pp is advanced to the following varint. -*/ -static void fts3GetDeltaPosition(char **pp, int *piPos){ - int iVal; - *pp += sqlite3Fts3GetVarint32(*pp, &iVal); - *piPos += (iVal-2); -} - -/* -** Helper function for fts3ExprIterate() (see below). -*/ -static int fts3ExprIterate2( - Fts3Expr *pExpr, /* Expression to iterate phrases of */ - int *piPhrase, /* Pointer to phrase counter */ - int (*x)(Fts3Expr*,int,void*), /* Callback function to invoke for phrases */ - void *pCtx /* Second argument to pass to callback */ -){ - int rc; /* Return code */ - int eType = pExpr->eType; /* Type of expression node pExpr */ - - if( eType!=FTSQUERY_PHRASE ){ - assert( pExpr->pLeft && pExpr->pRight ); - rc = fts3ExprIterate2(pExpr->pLeft, piPhrase, x, pCtx); - if( rc==SQLITE_OK && eType!=FTSQUERY_NOT ){ - rc = fts3ExprIterate2(pExpr->pRight, piPhrase, x, pCtx); - } - }else{ - rc = x(pExpr, *piPhrase, pCtx); - (*piPhrase)++; - } - return rc; -} - -/* -** Iterate through all phrase nodes in an FTS3 query, except those that -** are part of a sub-tree that is the right-hand-side of a NOT operator. -** For each phrase node found, the supplied callback function is invoked. -** -** If the callback function returns anything other than SQLITE_OK, -** the iteration is abandoned and the error code returned immediately. -** Otherwise, SQLITE_OK is returned after a callback has been made for -** all eligible phrase nodes. -*/ -static int fts3ExprIterate( - Fts3Expr *pExpr, /* Expression to iterate phrases of */ - int (*x)(Fts3Expr*,int,void*), /* Callback function to invoke for phrases */ - void *pCtx /* Second argument to pass to callback */ -){ - int iPhrase = 0; /* Variable used as the phrase counter */ - return fts3ExprIterate2(pExpr, &iPhrase, x, pCtx); -} - -/* -** This is an fts3ExprIterate() callback used while loading the doclists -** for each phrase into Fts3Expr.aDoclist[]/nDoclist. See also -** fts3ExprLoadDoclists(). -*/ -static int fts3ExprLoadDoclistsCb(Fts3Expr *pExpr, int iPhrase, void *ctx){ - int rc = SQLITE_OK; - Fts3Phrase *pPhrase = pExpr->pPhrase; - LoadDoclistCtx *p = (LoadDoclistCtx *)ctx; - - UNUSED_PARAMETER(iPhrase); - - p->nPhrase++; - p->nToken += pPhrase->nToken; - - return rc; -} - -/* -** Load the doclists for each phrase in the query associated with FTS3 cursor -** pCsr. -** -** If pnPhrase is not NULL, then *pnPhrase is set to the number of matchable -** phrases in the expression (all phrases except those directly or -** indirectly descended from the right-hand-side of a NOT operator). If -** pnToken is not NULL, then it is set to the number of tokens in all -** matchable phrases of the expression. -*/ -static int fts3ExprLoadDoclists( - Fts3Cursor *pCsr, /* Fts3 cursor for current query */ - int *pnPhrase, /* OUT: Number of phrases in query */ - int *pnToken /* OUT: Number of tokens in query */ -){ - int rc; /* Return Code */ - LoadDoclistCtx sCtx = {0,0,0}; /* Context for fts3ExprIterate() */ - sCtx.pCsr = pCsr; - rc = fts3ExprIterate(pCsr->pExpr, fts3ExprLoadDoclistsCb, (void *)&sCtx); - if( pnPhrase ) *pnPhrase = sCtx.nPhrase; - if( pnToken ) *pnToken = sCtx.nToken; - return rc; -} - -static int fts3ExprPhraseCountCb(Fts3Expr *pExpr, int iPhrase, void *ctx){ - (*(int *)ctx)++; - UNUSED_PARAMETER(pExpr); - UNUSED_PARAMETER(iPhrase); - return SQLITE_OK; -} -static int fts3ExprPhraseCount(Fts3Expr *pExpr){ - int nPhrase = 0; - (void)fts3ExprIterate(pExpr, fts3ExprPhraseCountCb, (void *)&nPhrase); - return nPhrase; -} - -/* -** Advance the position list iterator specified by the first two -** arguments so that it points to the first element with a value greater -** than or equal to parameter iNext. -*/ -static void fts3SnippetAdvance(char **ppIter, int *piIter, int iNext){ - char *pIter = *ppIter; - if( pIter ){ - int iIter = *piIter; - - while( iIter<iNext ){ - if( 0==(*pIter & 0xFE) ){ - iIter = -1; - pIter = 0; - break; - } - fts3GetDeltaPosition(&pIter, &iIter); - } - - *piIter = iIter; - *ppIter = pIter; - } -} - -/* -** Advance the snippet iterator to the next candidate snippet. -*/ -static int fts3SnippetNextCandidate(SnippetIter *pIter){ - int i; /* Loop counter */ - - if( pIter->iCurrent<0 ){ - /* The SnippetIter object has just been initialized. The first snippet - ** candidate always starts at offset 0 (even if this candidate has a - ** score of 0.0). - */ - pIter->iCurrent = 0; - - /* Advance the 'head' iterator of each phrase to the first offset that - ** is greater than or equal to (iNext+nSnippet). - */ - for(i=0; i<pIter->nPhrase; i++){ - SnippetPhrase *pPhrase = &pIter->aPhrase[i]; - fts3SnippetAdvance(&pPhrase->pHead, &pPhrase->iHead, pIter->nSnippet); - } - }else{ - int iStart; - int iEnd = 0x7FFFFFFF; - - for(i=0; i<pIter->nPhrase; i++){ - SnippetPhrase *pPhrase = &pIter->aPhrase[i]; - if( pPhrase->pHead && pPhrase->iHead<iEnd ){ - iEnd = pPhrase->iHead; - } - } - if( iEnd==0x7FFFFFFF ){ - return 1; - } - - pIter->iCurrent = iStart = iEnd - pIter->nSnippet + 1; - for(i=0; i<pIter->nPhrase; i++){ - SnippetPhrase *pPhrase = &pIter->aPhrase[i]; - fts3SnippetAdvance(&pPhrase->pHead, &pPhrase->iHead, iEnd+1); - fts3SnippetAdvance(&pPhrase->pTail, &pPhrase->iTail, iStart); - } - } - - return 0; -} - -/* -** Retrieve information about the current candidate snippet of snippet -** iterator pIter. -*/ -static void fts3SnippetDetails( - SnippetIter *pIter, /* Snippet iterator */ - u64 mCovered, /* Bitmask of phrases already covered */ - int *piToken, /* OUT: First token of proposed snippet */ - int *piScore, /* OUT: "Score" for this snippet */ - u64 *pmCover, /* OUT: Bitmask of phrases covered */ - u64 *pmHighlight /* OUT: Bitmask of terms to highlight */ -){ - int iStart = pIter->iCurrent; /* First token of snippet */ - int iScore = 0; /* Score of this snippet */ - int i; /* Loop counter */ - u64 mCover = 0; /* Mask of phrases covered by this snippet */ - u64 mHighlight = 0; /* Mask of tokens to highlight in snippet */ - - for(i=0; i<pIter->nPhrase; i++){ - SnippetPhrase *pPhrase = &pIter->aPhrase[i]; - if( pPhrase->pTail ){ - char *pCsr = pPhrase->pTail; - int iCsr = pPhrase->iTail; - - while( iCsr<(iStart+pIter->nSnippet) ){ - int j; - u64 mPhrase = (u64)1 << i; - u64 mPos = (u64)1 << (iCsr - iStart); - assert( iCsr>=iStart ); - if( (mCover|mCovered)&mPhrase ){ - iScore++; - }else{ - iScore += 1000; - } - mCover |= mPhrase; - - for(j=0; j<pPhrase->nToken; j++){ - mHighlight |= (mPos>>j); - } - - if( 0==(*pCsr & 0x0FE) ) break; - fts3GetDeltaPosition(&pCsr, &iCsr); - } - } - } - - /* Set the output variables before returning. */ - *piToken = iStart; - *piScore = iScore; - *pmCover = mCover; - *pmHighlight = mHighlight; -} - -/* -** This function is an fts3ExprIterate() callback used by fts3BestSnippet(). -** Each invocation populates an element of the SnippetIter.aPhrase[] array. -*/ -static int fts3SnippetFindPositions(Fts3Expr *pExpr, int iPhrase, void *ctx){ - SnippetIter *p = (SnippetIter *)ctx; - SnippetPhrase *pPhrase = &p->aPhrase[iPhrase]; - char *pCsr; - int rc; - - pPhrase->nToken = pExpr->pPhrase->nToken; - rc = sqlite3Fts3EvalPhrasePoslist(p->pCsr, pExpr, p->iCol, &pCsr); - assert( rc==SQLITE_OK || pCsr==0 ); - if( pCsr ){ - int iFirst = 0; - pPhrase->pList = pCsr; - fts3GetDeltaPosition(&pCsr, &iFirst); - assert( iFirst>=0 ); - pPhrase->pHead = pCsr; - pPhrase->pTail = pCsr; - pPhrase->iHead = iFirst; - pPhrase->iTail = iFirst; - }else{ - assert( rc!=SQLITE_OK || ( - pPhrase->pList==0 && pPhrase->pHead==0 && pPhrase->pTail==0 - )); - } - - return rc; -} - -/* -** Select the fragment of text consisting of nFragment contiguous tokens -** from column iCol that represent the "best" snippet. The best snippet -** is the snippet with the highest score, where scores are calculated -** by adding: -** -** (a) +1 point for each occurence of a matchable phrase in the snippet. -** -** (b) +1000 points for the first occurence of each matchable phrase in -** the snippet for which the corresponding mCovered bit is not set. -** -** The selected snippet parameters are stored in structure *pFragment before -** returning. The score of the selected snippet is stored in *piScore -** before returning. -*/ -static int fts3BestSnippet( - int nSnippet, /* Desired snippet length */ - Fts3Cursor *pCsr, /* Cursor to create snippet for */ - int iCol, /* Index of column to create snippet from */ - u64 mCovered, /* Mask of phrases already covered */ - u64 *pmSeen, /* IN/OUT: Mask of phrases seen */ - SnippetFragment *pFragment, /* OUT: Best snippet found */ - int *piScore /* OUT: Score of snippet pFragment */ -){ - int rc; /* Return Code */ - int nList; /* Number of phrases in expression */ - SnippetIter sIter; /* Iterates through snippet candidates */ - int nByte; /* Number of bytes of space to allocate */ - int iBestScore = -1; /* Best snippet score found so far */ - int i; /* Loop counter */ - - memset(&sIter, 0, sizeof(sIter)); - - /* Iterate through the phrases in the expression to count them. The same - ** callback makes sure the doclists are loaded for each phrase. - */ - rc = fts3ExprLoadDoclists(pCsr, &nList, 0); - if( rc!=SQLITE_OK ){ - return rc; - } - - /* Now that it is known how many phrases there are, allocate and zero - ** the required space using malloc(). - */ - nByte = sizeof(SnippetPhrase) * nList; - sIter.aPhrase = (SnippetPhrase *)sqlite3_malloc(nByte); - if( !sIter.aPhrase ){ - return SQLITE_NOMEM; - } - memset(sIter.aPhrase, 0, nByte); - - /* Initialize the contents of the SnippetIter object. Then iterate through - ** the set of phrases in the expression to populate the aPhrase[] array. - */ - sIter.pCsr = pCsr; - sIter.iCol = iCol; - sIter.nSnippet = nSnippet; - sIter.nPhrase = nList; - sIter.iCurrent = -1; - (void)fts3ExprIterate(pCsr->pExpr, fts3SnippetFindPositions, (void *)&sIter); - - /* Set the *pmSeen output variable. */ - for(i=0; i<nList; i++){ - if( sIter.aPhrase[i].pHead ){ - *pmSeen |= (u64)1 << i; - } - } - - /* Loop through all candidate snippets. Store the best snippet in - ** *pFragment. Store its associated 'score' in iBestScore. - */ - pFragment->iCol = iCol; - while( !fts3SnippetNextCandidate(&sIter) ){ - int iPos; - int iScore; - u64 mCover; - u64 mHighlight; - fts3SnippetDetails(&sIter, mCovered, &iPos, &iScore, &mCover, &mHighlight); - assert( iScore>=0 ); - if( iScore>iBestScore ){ - pFragment->iPos = iPos; - pFragment->hlmask = mHighlight; - pFragment->covered = mCover; - iBestScore = iScore; - } - } - - sqlite3_free(sIter.aPhrase); - *piScore = iBestScore; - return SQLITE_OK; -} - - -/* -** Append a string to the string-buffer passed as the first argument. -** -** If nAppend is negative, then the length of the string zAppend is -** determined using strlen(). -*/ -static int fts3StringAppend( - StrBuffer *pStr, /* Buffer to append to */ - const char *zAppend, /* Pointer to data to append to buffer */ - int nAppend /* Size of zAppend in bytes (or -1) */ -){ - if( nAppend<0 ){ - nAppend = (int)strlen(zAppend); - } - - /* If there is insufficient space allocated at StrBuffer.z, use realloc() - ** to grow the buffer until so that it is big enough to accomadate the - ** appended data. - */ - if( pStr->n+nAppend+1>=pStr->nAlloc ){ - int nAlloc = pStr->nAlloc+nAppend+100; - char *zNew = sqlite3_realloc(pStr->z, nAlloc); - if( !zNew ){ - return SQLITE_NOMEM; - } - pStr->z = zNew; - pStr->nAlloc = nAlloc; - } - - /* Append the data to the string buffer. */ - memcpy(&pStr->z[pStr->n], zAppend, nAppend); - pStr->n += nAppend; - pStr->z[pStr->n] = '\0'; - - return SQLITE_OK; -} - -/* -** The fts3BestSnippet() function often selects snippets that end with a -** query term. That is, the final term of the snippet is always a term -** that requires highlighting. For example, if 'X' is a highlighted term -** and '.' is a non-highlighted term, BestSnippet() may select: -** -** ........X.....X -** -** This function "shifts" the beginning of the snippet forward in the -** document so that there are approximately the same number of -** non-highlighted terms to the right of the final highlighted term as there -** are to the left of the first highlighted term. For example, to this: -** -** ....X.....X.... -** -** This is done as part of extracting the snippet text, not when selecting -** the snippet. Snippet selection is done based on doclists only, so there -** is no way for fts3BestSnippet() to know whether or not the document -** actually contains terms that follow the final highlighted term. -*/ -static int fts3SnippetShift( - Fts3Table *pTab, /* FTS3 table snippet comes from */ - int iLangid, /* Language id to use in tokenizing */ - int nSnippet, /* Number of tokens desired for snippet */ - const char *zDoc, /* Document text to extract snippet from */ - int nDoc, /* Size of buffer zDoc in bytes */ - int *piPos, /* IN/OUT: First token of snippet */ - u64 *pHlmask /* IN/OUT: Mask of tokens to highlight */ -){ - u64 hlmask = *pHlmask; /* Local copy of initial highlight-mask */ - - if( hlmask ){ - int nLeft; /* Tokens to the left of first highlight */ - int nRight; /* Tokens to the right of last highlight */ - int nDesired; /* Ideal number of tokens to shift forward */ - - for(nLeft=0; !(hlmask & ((u64)1 << nLeft)); nLeft++); - for(nRight=0; !(hlmask & ((u64)1 << (nSnippet-1-nRight))); nRight++); - nDesired = (nLeft-nRight)/2; - - /* Ideally, the start of the snippet should be pushed forward in the - ** document nDesired tokens. This block checks if there are actually - ** nDesired tokens to the right of the snippet. If so, *piPos and - ** *pHlMask are updated to shift the snippet nDesired tokens to the - ** right. Otherwise, the snippet is shifted by the number of tokens - ** available. - */ - if( nDesired>0 ){ - int nShift; /* Number of tokens to shift snippet by */ - int iCurrent = 0; /* Token counter */ - int rc; /* Return Code */ - sqlite3_tokenizer_module *pMod; - sqlite3_tokenizer_cursor *pC; - pMod = (sqlite3_tokenizer_module *)pTab->pTokenizer->pModule; - - /* Open a cursor on zDoc/nDoc. Check if there are (nSnippet+nDesired) - ** or more tokens in zDoc/nDoc. - */ - rc = sqlite3Fts3OpenTokenizer(pTab->pTokenizer, iLangid, zDoc, nDoc, &pC); - if( rc!=SQLITE_OK ){ - return rc; - } - while( rc==SQLITE_OK && iCurrent<(nSnippet+nDesired) ){ - const char *ZDUMMY; int DUMMY1 = 0, DUMMY2 = 0, DUMMY3 = 0; - rc = pMod->xNext(pC, &ZDUMMY, &DUMMY1, &DUMMY2, &DUMMY3, &iCurrent); - } - pMod->xClose(pC); - if( rc!=SQLITE_OK && rc!=SQLITE_DONE ){ return rc; } - - nShift = (rc==SQLITE_DONE)+iCurrent-nSnippet; - assert( nShift<=nDesired ); - if( nShift>0 ){ - *piPos += nShift; - *pHlmask = hlmask >> nShift; - } - } - } - return SQLITE_OK; -} - -/* -** Extract the snippet text for fragment pFragment from cursor pCsr and -** append it to string buffer pOut. -*/ -static int fts3SnippetText( - Fts3Cursor *pCsr, /* FTS3 Cursor */ - SnippetFragment *pFragment, /* Snippet to extract */ - int iFragment, /* Fragment number */ - int isLast, /* True for final fragment in snippet */ - int nSnippet, /* Number of tokens in extracted snippet */ - const char *zOpen, /* String inserted before highlighted term */ - const char *zClose, /* String inserted after highlighted term */ - const char *zEllipsis, /* String inserted between snippets */ - StrBuffer *pOut /* Write output here */ -){ - Fts3Table *pTab = (Fts3Table *)pCsr->base.pVtab; - int rc; /* Return code */ - const char *zDoc; /* Document text to extract snippet from */ - int nDoc; /* Size of zDoc in bytes */ - int iCurrent = 0; /* Current token number of document */ - int iEnd = 0; /* Byte offset of end of current token */ - int isShiftDone = 0; /* True after snippet is shifted */ - int iPos = pFragment->iPos; /* First token of snippet */ - u64 hlmask = pFragment->hlmask; /* Highlight-mask for snippet */ - int iCol = pFragment->iCol+1; /* Query column to extract text from */ - sqlite3_tokenizer_module *pMod; /* Tokenizer module methods object */ - sqlite3_tokenizer_cursor *pC; /* Tokenizer cursor open on zDoc/nDoc */ - - zDoc = (const char *)sqlite3_column_text(pCsr->pStmt, iCol); - if( zDoc==0 ){ - if( sqlite3_column_type(pCsr->pStmt, iCol)!=SQLITE_NULL ){ - return SQLITE_NOMEM; - } - return SQLITE_OK; - } - nDoc = sqlite3_column_bytes(pCsr->pStmt, iCol); - - /* Open a token cursor on the document. */ - pMod = (sqlite3_tokenizer_module *)pTab->pTokenizer->pModule; - rc = sqlite3Fts3OpenTokenizer(pTab->pTokenizer, pCsr->iLangid, zDoc,nDoc,&pC); - if( rc!=SQLITE_OK ){ - return rc; - } - - while( rc==SQLITE_OK ){ - const char *ZDUMMY; /* Dummy argument used with tokenizer */ - int DUMMY1 = -1; /* Dummy argument used with tokenizer */ - int iBegin = 0; /* Offset in zDoc of start of token */ - int iFin = 0; /* Offset in zDoc of end of token */ - int isHighlight = 0; /* True for highlighted terms */ - - /* Variable DUMMY1 is initialized to a negative value above. Elsewhere - ** in the FTS code the variable that the third argument to xNext points to - ** is initialized to zero before the first (*but not necessarily - ** subsequent*) call to xNext(). This is done for a particular application - ** that needs to know whether or not the tokenizer is being used for - ** snippet generation or for some other purpose. - ** - ** Extreme care is required when writing code to depend on this - ** initialization. It is not a documented part of the tokenizer interface. - ** If a tokenizer is used directly by any code outside of FTS, this - ** convention might not be respected. */ - rc = pMod->xNext(pC, &ZDUMMY, &DUMMY1, &iBegin, &iFin, &iCurrent); - if( rc!=SQLITE_OK ){ - if( rc==SQLITE_DONE ){ - /* Special case - the last token of the snippet is also the last token - ** of the column. Append any punctuation that occurred between the end - ** of the previous token and the end of the document to the output. - ** Then break out of the loop. */ - rc = fts3StringAppend(pOut, &zDoc[iEnd], -1); - } - break; - } - if( iCurrent<iPos ){ continue; } - - if( !isShiftDone ){ - int n = nDoc - iBegin; - rc = fts3SnippetShift( - pTab, pCsr->iLangid, nSnippet, &zDoc[iBegin], n, &iPos, &hlmask - ); - isShiftDone = 1; - - /* Now that the shift has been done, check if the initial "..." are - ** required. They are required if (a) this is not the first fragment, - ** or (b) this fragment does not begin at position 0 of its column. - */ - if( rc==SQLITE_OK && (iPos>0 || iFragment>0) ){ - rc = fts3StringAppend(pOut, zEllipsis, -1); - } - if( rc!=SQLITE_OK || iCurrent<iPos ) continue; - } - - if( iCurrent>=(iPos+nSnippet) ){ - if( isLast ){ - rc = fts3StringAppend(pOut, zEllipsis, -1); - } - break; - } - - /* Set isHighlight to true if this term should be highlighted. */ - isHighlight = (hlmask & ((u64)1 << (iCurrent-iPos)))!=0; - - if( iCurrent>iPos ) rc = fts3StringAppend(pOut, &zDoc[iEnd], iBegin-iEnd); - if( rc==SQLITE_OK && isHighlight ) rc = fts3StringAppend(pOut, zOpen, -1); - if( rc==SQLITE_OK ) rc = fts3StringAppend(pOut, &zDoc[iBegin], iFin-iBegin); - if( rc==SQLITE_OK && isHighlight ) rc = fts3StringAppend(pOut, zClose, -1); - - iEnd = iFin; - } - - pMod->xClose(pC); - return rc; -} - - -/* -** This function is used to count the entries in a column-list (a -** delta-encoded list of term offsets within a single column of a single -** row). When this function is called, *ppCollist should point to the -** beginning of the first varint in the column-list (the varint that -** contains the position of the first matching term in the column data). -** Before returning, *ppCollist is set to point to the first byte after -** the last varint in the column-list (either the 0x00 signifying the end -** of the position-list, or the 0x01 that precedes the column number of -** the next column in the position-list). -** -** The number of elements in the column-list is returned. -*/ -static int fts3ColumnlistCount(char **ppCollist){ - char *pEnd = *ppCollist; - char c = 0; - int nEntry = 0; - - /* A column-list is terminated by either a 0x01 or 0x00. */ - while( 0xFE & (*pEnd | c) ){ - c = *pEnd++ & 0x80; - if( !c ) nEntry++; - } - - *ppCollist = pEnd; - return nEntry; -} - -/* -** fts3ExprIterate() callback used to collect the "global" matchinfo stats -** for a single query. -** -** fts3ExprIterate() callback to load the 'global' elements of a -** FTS3_MATCHINFO_HITS matchinfo array. The global stats are those elements -** of the matchinfo array that are constant for all rows returned by the -** current query. -** -** Argument pCtx is actually a pointer to a struct of type MatchInfo. This -** function populates Matchinfo.aMatchinfo[] as follows: -** -** for(iCol=0; iCol<nCol; iCol++){ -** aMatchinfo[3*iPhrase*nCol + 3*iCol + 1] = X; -** aMatchinfo[3*iPhrase*nCol + 3*iCol + 2] = Y; -** } -** -** where X is the number of matches for phrase iPhrase is column iCol of all -** rows of the table. Y is the number of rows for which column iCol contains -** at least one instance of phrase iPhrase. -** -** If the phrase pExpr consists entirely of deferred tokens, then all X and -** Y values are set to nDoc, where nDoc is the number of documents in the -** file system. This is done because the full-text index doclist is required -** to calculate these values properly, and the full-text index doclist is -** not available for deferred tokens. -*/ -static int fts3ExprGlobalHitsCb( - Fts3Expr *pExpr, /* Phrase expression node */ - int iPhrase, /* Phrase number (numbered from zero) */ - void *pCtx /* Pointer to MatchInfo structure */ -){ - MatchInfo *p = (MatchInfo *)pCtx; - return sqlite3Fts3EvalPhraseStats( - p->pCursor, pExpr, &p->aMatchinfo[3*iPhrase*p->nCol] - ); -} - -/* -** fts3ExprIterate() callback used to collect the "local" part of the -** FTS3_MATCHINFO_HITS array. The local stats are those elements of the -** array that are different for each row returned by the query. -*/ -static int fts3ExprLocalHitsCb( - Fts3Expr *pExpr, /* Phrase expression node */ - int iPhrase, /* Phrase number */ - void *pCtx /* Pointer to MatchInfo structure */ -){ - int rc = SQLITE_OK; - MatchInfo *p = (MatchInfo *)pCtx; - int iStart = iPhrase * p->nCol * 3; - int i; - - for(i=0; i<p->nCol && rc==SQLITE_OK; i++){ - char *pCsr; - rc = sqlite3Fts3EvalPhrasePoslist(p->pCursor, pExpr, i, &pCsr); - if( pCsr ){ - p->aMatchinfo[iStart+i*3] = fts3ColumnlistCount(&pCsr); - }else{ - p->aMatchinfo[iStart+i*3] = 0; - } - } - - return rc; -} - -static int fts3MatchinfoCheck( - Fts3Table *pTab, - char cArg, - char **pzErr -){ - if( (cArg==FTS3_MATCHINFO_NPHRASE) - || (cArg==FTS3_MATCHINFO_NCOL) - || (cArg==FTS3_MATCHINFO_NDOC && pTab->bFts4) - || (cArg==FTS3_MATCHINFO_AVGLENGTH && pTab->bFts4) - || (cArg==FTS3_MATCHINFO_LENGTH && pTab->bHasDocsize) - || (cArg==FTS3_MATCHINFO_LCS) - || (cArg==FTS3_MATCHINFO_HITS) - ){ - return SQLITE_OK; - } - *pzErr = sqlite3_mprintf("unrecognized matchinfo request: %c", cArg); - return SQLITE_ERROR; -} - -static int fts3MatchinfoSize(MatchInfo *pInfo, char cArg){ - int nVal; /* Number of integers output by cArg */ - - switch( cArg ){ - case FTS3_MATCHINFO_NDOC: - case FTS3_MATCHINFO_NPHRASE: - case FTS3_MATCHINFO_NCOL: - nVal = 1; - break; - - case FTS3_MATCHINFO_AVGLENGTH: - case FTS3_MATCHINFO_LENGTH: - case FTS3_MATCHINFO_LCS: - nVal = pInfo->nCol; - break; - - default: - assert( cArg==FTS3_MATCHINFO_HITS ); - nVal = pInfo->nCol * pInfo->nPhrase * 3; - break; - } - - return nVal; -} - -static int fts3MatchinfoSelectDoctotal( - Fts3Table *pTab, - sqlite3_stmt **ppStmt, - sqlite3_int64 *pnDoc, - const char **paLen -){ - sqlite3_stmt *pStmt; - const char *a; - sqlite3_int64 nDoc; - - if( !*ppStmt ){ - int rc = sqlite3Fts3SelectDoctotal(pTab, ppStmt); - if( rc!=SQLITE_OK ) return rc; - } - pStmt = *ppStmt; - assert( sqlite3_data_count(pStmt)==1 ); - - a = sqlite3_column_blob(pStmt, 0); - a += sqlite3Fts3GetVarint(a, &nDoc); - if( nDoc==0 ) return FTS_CORRUPT_VTAB; - *pnDoc = (u32)nDoc; - - if( paLen ) *paLen = a; - return SQLITE_OK; -} - -/* -** An instance of the following structure is used to store state while -** iterating through a multi-column position-list corresponding to the -** hits for a single phrase on a single row in order to calculate the -** values for a matchinfo() FTS3_MATCHINFO_LCS request. -*/ -typedef struct LcsIterator LcsIterator; -struct LcsIterator { - Fts3Expr *pExpr; /* Pointer to phrase expression */ - int iPosOffset; /* Tokens count up to end of this phrase */ - char *pRead; /* Cursor used to iterate through aDoclist */ - int iPos; /* Current position */ -}; - -/* -** If LcsIterator.iCol is set to the following value, the iterator has -** finished iterating through all offsets for all columns. -*/ -#define LCS_ITERATOR_FINISHED 0x7FFFFFFF; - -static int fts3MatchinfoLcsCb( - Fts3Expr *pExpr, /* Phrase expression node */ - int iPhrase, /* Phrase number (numbered from zero) */ - void *pCtx /* Pointer to MatchInfo structure */ -){ - LcsIterator *aIter = (LcsIterator *)pCtx; - aIter[iPhrase].pExpr = pExpr; - return SQLITE_OK; -} - -/* -** Advance the iterator passed as an argument to the next position. Return -** 1 if the iterator is at EOF or if it now points to the start of the -** position list for the next column. -*/ -static int fts3LcsIteratorAdvance(LcsIterator *pIter){ - char *pRead = pIter->pRead; - sqlite3_int64 iRead; - int rc = 0; - - pRead += sqlite3Fts3GetVarint(pRead, &iRead); - if( iRead==0 || iRead==1 ){ - pRead = 0; - rc = 1; - }else{ - pIter->iPos += (int)(iRead-2); - } - - pIter->pRead = pRead; - return rc; -} - -/* -** This function implements the FTS3_MATCHINFO_LCS matchinfo() flag. -** -** If the call is successful, the longest-common-substring lengths for each -** column are written into the first nCol elements of the pInfo->aMatchinfo[] -** array before returning. SQLITE_OK is returned in this case. -** -** Otherwise, if an error occurs, an SQLite error code is returned and the -** data written to the first nCol elements of pInfo->aMatchinfo[] is -** undefined. -*/ -static int fts3MatchinfoLcs(Fts3Cursor *pCsr, MatchInfo *pInfo){ - LcsIterator *aIter; - int i; - int iCol; - int nToken = 0; - - /* Allocate and populate the array of LcsIterator objects. The array - ** contains one element for each matchable phrase in the query. - **/ - aIter = sqlite3_malloc(sizeof(LcsIterator) * pCsr->nPhrase); - if( !aIter ) return SQLITE_NOMEM; - memset(aIter, 0, sizeof(LcsIterator) * pCsr->nPhrase); - (void)fts3ExprIterate(pCsr->pExpr, fts3MatchinfoLcsCb, (void*)aIter); - - for(i=0; i<pInfo->nPhrase; i++){ - LcsIterator *pIter = &aIter[i]; - nToken -= pIter->pExpr->pPhrase->nToken; - pIter->iPosOffset = nToken; - } - - for(iCol=0; iCol<pInfo->nCol; iCol++){ - int nLcs = 0; /* LCS value for this column */ - int nLive = 0; /* Number of iterators in aIter not at EOF */ - - for(i=0; i<pInfo->nPhrase; i++){ - int rc; - LcsIterator *pIt = &aIter[i]; - rc = sqlite3Fts3EvalPhrasePoslist(pCsr, pIt->pExpr, iCol, &pIt->pRead); - if( rc!=SQLITE_OK ) return rc; - if( pIt->pRead ){ - pIt->iPos = pIt->iPosOffset; - fts3LcsIteratorAdvance(&aIter[i]); - nLive++; - } - } - - while( nLive>0 ){ - LcsIterator *pAdv = 0; /* The iterator to advance by one position */ - int nThisLcs = 0; /* LCS for the current iterator positions */ - - for(i=0; i<pInfo->nPhrase; i++){ - LcsIterator *pIter = &aIter[i]; - if( pIter->pRead==0 ){ - /* This iterator is already at EOF for this column. */ - nThisLcs = 0; - }else{ - if( pAdv==0 || pIter->iPos<pAdv->iPos ){ - pAdv = pIter; - } - if( nThisLcs==0 || pIter->iPos==pIter[-1].iPos ){ - nThisLcs++; - }else{ - nThisLcs = 1; - } - if( nThisLcs>nLcs ) nLcs = nThisLcs; - } - } - if( fts3LcsIteratorAdvance(pAdv) ) nLive--; - } - - pInfo->aMatchinfo[iCol] = nLcs; - } - - sqlite3_free(aIter); - return SQLITE_OK; -} - -/* -** Populate the buffer pInfo->aMatchinfo[] with an array of integers to -** be returned by the matchinfo() function. Argument zArg contains the -** format string passed as the second argument to matchinfo (or the -** default value "pcx" if no second argument was specified). The format -** string has already been validated and the pInfo->aMatchinfo[] array -** is guaranteed to be large enough for the output. -** -** If bGlobal is true, then populate all fields of the matchinfo() output. -** If it is false, then assume that those fields that do not change between -** rows (i.e. FTS3_MATCHINFO_NPHRASE, NCOL, NDOC, AVGLENGTH and part of HITS) -** have already been populated. -** -** Return SQLITE_OK if successful, or an SQLite error code if an error -** occurs. If a value other than SQLITE_OK is returned, the state the -** pInfo->aMatchinfo[] buffer is left in is undefined. -*/ -static int fts3MatchinfoValues( - Fts3Cursor *pCsr, /* FTS3 cursor object */ - int bGlobal, /* True to grab the global stats */ - MatchInfo *pInfo, /* Matchinfo context object */ - const char *zArg /* Matchinfo format string */ -){ - int rc = SQLITE_OK; - int i; - Fts3Table *pTab = (Fts3Table *)pCsr->base.pVtab; - sqlite3_stmt *pSelect = 0; - - for(i=0; rc==SQLITE_OK && zArg[i]; i++){ - - switch( zArg[i] ){ - case FTS3_MATCHINFO_NPHRASE: - if( bGlobal ) pInfo->aMatchinfo[0] = pInfo->nPhrase; - break; - - case FTS3_MATCHINFO_NCOL: - if( bGlobal ) pInfo->aMatchinfo[0] = pInfo->nCol; - break; - - case FTS3_MATCHINFO_NDOC: - if( bGlobal ){ - sqlite3_int64 nDoc = 0; - rc = fts3MatchinfoSelectDoctotal(pTab, &pSelect, &nDoc, 0); - pInfo->aMatchinfo[0] = (u32)nDoc; - } - break; - - case FTS3_MATCHINFO_AVGLENGTH: - if( bGlobal ){ - sqlite3_int64 nDoc; /* Number of rows in table */ - const char *a; /* Aggregate column length array */ - - rc = fts3MatchinfoSelectDoctotal(pTab, &pSelect, &nDoc, &a); - if( rc==SQLITE_OK ){ - int iCol; - for(iCol=0; iCol<pInfo->nCol; iCol++){ - u32 iVal; - sqlite3_int64 nToken; - a += sqlite3Fts3GetVarint(a, &nToken); - iVal = (u32)(((u32)(nToken&0xffffffff)+nDoc/2)/nDoc); - pInfo->aMatchinfo[iCol] = iVal; - } - } - } - break; - - case FTS3_MATCHINFO_LENGTH: { - sqlite3_stmt *pSelectDocsize = 0; - rc = sqlite3Fts3SelectDocsize(pTab, pCsr->iPrevId, &pSelectDocsize); - if( rc==SQLITE_OK ){ - int iCol; - const char *a = sqlite3_column_blob(pSelectDocsize, 0); - for(iCol=0; iCol<pInfo->nCol; iCol++){ - sqlite3_int64 nToken; - a += sqlite3Fts3GetVarint(a, &nToken); - pInfo->aMatchinfo[iCol] = (u32)nToken; - } - } - sqlite3_reset(pSelectDocsize); - break; - } - - case FTS3_MATCHINFO_LCS: - rc = fts3ExprLoadDoclists(pCsr, 0, 0); - if( rc==SQLITE_OK ){ - rc = fts3MatchinfoLcs(pCsr, pInfo); - } - break; - - default: { - Fts3Expr *pExpr; - assert( zArg[i]==FTS3_MATCHINFO_HITS ); - pExpr = pCsr->pExpr; - rc = fts3ExprLoadDoclists(pCsr, 0, 0); - if( rc!=SQLITE_OK ) break; - if( bGlobal ){ - if( pCsr->pDeferred ){ - rc = fts3MatchinfoSelectDoctotal(pTab, &pSelect, &pInfo->nDoc, 0); - if( rc!=SQLITE_OK ) break; - } - rc = fts3ExprIterate(pExpr, fts3ExprGlobalHitsCb,(void*)pInfo); - if( rc!=SQLITE_OK ) break; - } - (void)fts3ExprIterate(pExpr, fts3ExprLocalHitsCb,(void*)pInfo); - break; - } - } - - pInfo->aMatchinfo += fts3MatchinfoSize(pInfo, zArg[i]); - } - - sqlite3_reset(pSelect); - return rc; -} - - -/* -** Populate pCsr->aMatchinfo[] with data for the current row. The -** 'matchinfo' data is an array of 32-bit unsigned integers (C type u32). -*/ -static int fts3GetMatchinfo( - Fts3Cursor *pCsr, /* FTS3 Cursor object */ - const char *zArg /* Second argument to matchinfo() function */ -){ - MatchInfo sInfo; - Fts3Table *pTab = (Fts3Table *)pCsr->base.pVtab; - int rc = SQLITE_OK; - int bGlobal = 0; /* Collect 'global' stats as well as local */ - - memset(&sInfo, 0, sizeof(MatchInfo)); - sInfo.pCursor = pCsr; - sInfo.nCol = pTab->nColumn; - - /* If there is cached matchinfo() data, but the format string for the - ** cache does not match the format string for this request, discard - ** the cached data. */ - if( pCsr->zMatchinfo && strcmp(pCsr->zMatchinfo, zArg) ){ - assert( pCsr->aMatchinfo ); - sqlite3_free(pCsr->aMatchinfo); - pCsr->zMatchinfo = 0; - pCsr->aMatchinfo = 0; - } - - /* If Fts3Cursor.aMatchinfo[] is NULL, then this is the first time the - ** matchinfo function has been called for this query. In this case - ** allocate the array used to accumulate the matchinfo data and - ** initialize those elements that are constant for every row. - */ - if( pCsr->aMatchinfo==0 ){ - int nMatchinfo = 0; /* Number of u32 elements in match-info */ - int nArg; /* Bytes in zArg */ - int i; /* Used to iterate through zArg */ - - /* Determine the number of phrases in the query */ - pCsr->nPhrase = fts3ExprPhraseCount(pCsr->pExpr); - sInfo.nPhrase = pCsr->nPhrase; - - /* Determine the number of integers in the buffer returned by this call. */ - for(i=0; zArg[i]; i++){ - nMatchinfo += fts3MatchinfoSize(&sInfo, zArg[i]); - } - - /* Allocate space for Fts3Cursor.aMatchinfo[] and Fts3Cursor.zMatchinfo. */ - nArg = (int)strlen(zArg); - pCsr->aMatchinfo = (u32 *)sqlite3_malloc(sizeof(u32)*nMatchinfo + nArg + 1); - if( !pCsr->aMatchinfo ) return SQLITE_NOMEM; - - pCsr->zMatchinfo = (char *)&pCsr->aMatchinfo[nMatchinfo]; - pCsr->nMatchinfo = nMatchinfo; - memcpy(pCsr->zMatchinfo, zArg, nArg+1); - memset(pCsr->aMatchinfo, 0, sizeof(u32)*nMatchinfo); - pCsr->isMatchinfoNeeded = 1; - bGlobal = 1; - } - - sInfo.aMatchinfo = pCsr->aMatchinfo; - sInfo.nPhrase = pCsr->nPhrase; - if( pCsr->isMatchinfoNeeded ){ - rc = fts3MatchinfoValues(pCsr, bGlobal, &sInfo, zArg); - pCsr->isMatchinfoNeeded = 0; - } - - return rc; -} - -/* -** Implementation of snippet() function. -*/ -void sqlite3Fts3Snippet( - sqlite3_context *pCtx, /* SQLite function call context */ - Fts3Cursor *pCsr, /* Cursor object */ - const char *zStart, /* Snippet start text - "<b>" */ - const char *zEnd, /* Snippet end text - "</b>" */ - const char *zEllipsis, /* Snippet ellipsis text - "<b>...</b>" */ - int iCol, /* Extract snippet from this column */ - int nToken /* Approximate number of tokens in snippet */ -){ - Fts3Table *pTab = (Fts3Table *)pCsr->base.pVtab; - int rc = SQLITE_OK; - int i; - StrBuffer res = {0, 0, 0}; - - /* The returned text includes up to four fragments of text extracted from - ** the data in the current row. The first iteration of the for(...) loop - ** below attempts to locate a single fragment of text nToken tokens in - ** size that contains at least one instance of all phrases in the query - ** expression that appear in the current row. If such a fragment of text - ** cannot be found, the second iteration of the loop attempts to locate - ** a pair of fragments, and so on. - */ - int nSnippet = 0; /* Number of fragments in this snippet */ - SnippetFragment aSnippet[4]; /* Maximum of 4 fragments per snippet */ - int nFToken = -1; /* Number of tokens in each fragment */ - - if( !pCsr->pExpr ){ - sqlite3_result_text(pCtx, "", 0, SQLITE_STATIC); - return; - } - - for(nSnippet=1; 1; nSnippet++){ - - int iSnip; /* Loop counter 0..nSnippet-1 */ - u64 mCovered = 0; /* Bitmask of phrases covered by snippet */ - u64 mSeen = 0; /* Bitmask of phrases seen by BestSnippet() */ - - if( nToken>=0 ){ - nFToken = (nToken+nSnippet-1) / nSnippet; - }else{ - nFToken = -1 * nToken; - } - - for(iSnip=0; iSnip<nSnippet; iSnip++){ - int iBestScore = -1; /* Best score of columns checked so far */ - int iRead; /* Used to iterate through columns */ - SnippetFragment *pFragment = &aSnippet[iSnip]; - - memset(pFragment, 0, sizeof(*pFragment)); - - /* Loop through all columns of the table being considered for snippets. - ** If the iCol argument to this function was negative, this means all - ** columns of the FTS3 table. Otherwise, only column iCol is considered. - */ - for(iRead=0; iRead<pTab->nColumn; iRead++){ - SnippetFragment sF = {0, 0, 0, 0}; - int iS; - if( iCol>=0 && iRead!=iCol ) continue; - - /* Find the best snippet of nFToken tokens in column iRead. */ - rc = fts3BestSnippet(nFToken, pCsr, iRead, mCovered, &mSeen, &sF, &iS); - if( rc!=SQLITE_OK ){ - goto snippet_out; - } - if( iS>iBestScore ){ - *pFragment = sF; - iBestScore = iS; - } - } - - mCovered |= pFragment->covered; - } - - /* If all query phrases seen by fts3BestSnippet() are present in at least - ** one of the nSnippet snippet fragments, break out of the loop. - */ - assert( (mCovered&mSeen)==mCovered ); - if( mSeen==mCovered || nSnippet==SizeofArray(aSnippet) ) break; - } - - assert( nFToken>0 ); - - for(i=0; i<nSnippet && rc==SQLITE_OK; i++){ - rc = fts3SnippetText(pCsr, &aSnippet[i], - i, (i==nSnippet-1), nFToken, zStart, zEnd, zEllipsis, &res - ); - } - - snippet_out: - sqlite3Fts3SegmentsClose(pTab); - if( rc!=SQLITE_OK ){ - sqlite3_result_error_code(pCtx, rc); - sqlite3_free(res.z); - }else{ - sqlite3_result_text(pCtx, res.z, -1, sqlite3_free); - } -} - - -typedef struct TermOffset TermOffset; -typedef struct TermOffsetCtx TermOffsetCtx; - -struct TermOffset { - char *pList; /* Position-list */ - int iPos; /* Position just read from pList */ - int iOff; /* Offset of this term from read positions */ -}; - -struct TermOffsetCtx { - Fts3Cursor *pCsr; - int iCol; /* Column of table to populate aTerm for */ - int iTerm; - sqlite3_int64 iDocid; - TermOffset *aTerm; -}; - -/* -** This function is an fts3ExprIterate() callback used by sqlite3Fts3Offsets(). -*/ -static int fts3ExprTermOffsetInit(Fts3Expr *pExpr, int iPhrase, void *ctx){ - TermOffsetCtx *p = (TermOffsetCtx *)ctx; - int nTerm; /* Number of tokens in phrase */ - int iTerm; /* For looping through nTerm phrase terms */ - char *pList; /* Pointer to position list for phrase */ - int iPos = 0; /* First position in position-list */ - int rc; - - UNUSED_PARAMETER(iPhrase); - rc = sqlite3Fts3EvalPhrasePoslist(p->pCsr, pExpr, p->iCol, &pList); - nTerm = pExpr->pPhrase->nToken; - if( pList ){ - fts3GetDeltaPosition(&pList, &iPos); - assert( iPos>=0 ); - } - - for(iTerm=0; iTerm<nTerm; iTerm++){ - TermOffset *pT = &p->aTerm[p->iTerm++]; - pT->iOff = nTerm-iTerm-1; - pT->pList = pList; - pT->iPos = iPos; - } - - return rc; -} - -/* -** Implementation of offsets() function. -*/ -void sqlite3Fts3Offsets( - sqlite3_context *pCtx, /* SQLite function call context */ - Fts3Cursor *pCsr /* Cursor object */ -){ - Fts3Table *pTab = (Fts3Table *)pCsr->base.pVtab; - sqlite3_tokenizer_module const *pMod = pTab->pTokenizer->pModule; - int rc; /* Return Code */ - int nToken; /* Number of tokens in query */ - int iCol; /* Column currently being processed */ - StrBuffer res = {0, 0, 0}; /* Result string */ - TermOffsetCtx sCtx; /* Context for fts3ExprTermOffsetInit() */ - - if( !pCsr->pExpr ){ - sqlite3_result_text(pCtx, "", 0, SQLITE_STATIC); - return; - } - - memset(&sCtx, 0, sizeof(sCtx)); - assert( pCsr->isRequireSeek==0 ); - - /* Count the number of terms in the query */ - rc = fts3ExprLoadDoclists(pCsr, 0, &nToken); - if( rc!=SQLITE_OK ) goto offsets_out; - - /* Allocate the array of TermOffset iterators. */ - sCtx.aTerm = (TermOffset *)sqlite3_malloc(sizeof(TermOffset)*nToken); - if( 0==sCtx.aTerm ){ - rc = SQLITE_NOMEM; - goto offsets_out; - } - sCtx.iDocid = pCsr->iPrevId; - sCtx.pCsr = pCsr; - - /* Loop through the table columns, appending offset information to - ** string-buffer res for each column. - */ - for(iCol=0; iCol<pTab->nColumn; iCol++){ - sqlite3_tokenizer_cursor *pC; /* Tokenizer cursor */ - const char *ZDUMMY; /* Dummy argument used with xNext() */ - int NDUMMY = 0; /* Dummy argument used with xNext() */ - int iStart = 0; - int iEnd = 0; - int iCurrent = 0; - const char *zDoc; - int nDoc; - - /* Initialize the contents of sCtx.aTerm[] for column iCol. There is - ** no way that this operation can fail, so the return code from - ** fts3ExprIterate() can be discarded. - */ - sCtx.iCol = iCol; - sCtx.iTerm = 0; - (void)fts3ExprIterate(pCsr->pExpr, fts3ExprTermOffsetInit, (void *)&sCtx); - - /* Retreive the text stored in column iCol. If an SQL NULL is stored - ** in column iCol, jump immediately to the next iteration of the loop. - ** If an OOM occurs while retrieving the data (this can happen if SQLite - ** needs to transform the data from utf-16 to utf-8), return SQLITE_NOMEM - ** to the caller. - */ - zDoc = (const char *)sqlite3_column_text(pCsr->pStmt, iCol+1); - nDoc = sqlite3_column_bytes(pCsr->pStmt, iCol+1); - if( zDoc==0 ){ - if( sqlite3_column_type(pCsr->pStmt, iCol+1)==SQLITE_NULL ){ - continue; - } - rc = SQLITE_NOMEM; - goto offsets_out; - } - - /* Initialize a tokenizer iterator to iterate through column iCol. */ - rc = sqlite3Fts3OpenTokenizer(pTab->pTokenizer, pCsr->iLangid, - zDoc, nDoc, &pC - ); - if( rc!=SQLITE_OK ) goto offsets_out; - - rc = pMod->xNext(pC, &ZDUMMY, &NDUMMY, &iStart, &iEnd, &iCurrent); - while( rc==SQLITE_OK ){ - int i; /* Used to loop through terms */ - int iMinPos = 0x7FFFFFFF; /* Position of next token */ - TermOffset *pTerm = 0; /* TermOffset associated with next token */ - - for(i=0; i<nToken; i++){ - TermOffset *pT = &sCtx.aTerm[i]; - if( pT->pList && (pT->iPos-pT->iOff)<iMinPos ){ - iMinPos = pT->iPos-pT->iOff; - pTerm = pT; - } - } - - if( !pTerm ){ - /* All offsets for this column have been gathered. */ - rc = SQLITE_DONE; - }else{ - assert( iCurrent<=iMinPos ); - if( 0==(0xFE&*pTerm->pList) ){ - pTerm->pList = 0; - }else{ - fts3GetDeltaPosition(&pTerm->pList, &pTerm->iPos); - } - while( rc==SQLITE_OK && iCurrent<iMinPos ){ - rc = pMod->xNext(pC, &ZDUMMY, &NDUMMY, &iStart, &iEnd, &iCurrent); - } - if( rc==SQLITE_OK ){ - char aBuffer[64]; - sqlite3_snprintf(sizeof(aBuffer), aBuffer, - "%d %d %d %d ", iCol, pTerm-sCtx.aTerm, iStart, iEnd-iStart - ); - rc = fts3StringAppend(&res, aBuffer, -1); - }else if( rc==SQLITE_DONE && pTab->zContentTbl==0 ){ - rc = FTS_CORRUPT_VTAB; - } - } - } - if( rc==SQLITE_DONE ){ - rc = SQLITE_OK; - } - - pMod->xClose(pC); - if( rc!=SQLITE_OK ) goto offsets_out; - } - - offsets_out: - sqlite3_free(sCtx.aTerm); - assert( rc!=SQLITE_DONE ); - sqlite3Fts3SegmentsClose(pTab); - if( rc!=SQLITE_OK ){ - sqlite3_result_error_code(pCtx, rc); - sqlite3_free(res.z); - }else{ - sqlite3_result_text(pCtx, res.z, res.n-1, sqlite3_free); - } - return; -} - -/* -** Implementation of matchinfo() function. -*/ -void sqlite3Fts3Matchinfo( - sqlite3_context *pContext, /* Function call context */ - Fts3Cursor *pCsr, /* FTS3 table cursor */ - const char *zArg /* Second arg to matchinfo() function */ -){ - Fts3Table *pTab = (Fts3Table *)pCsr->base.pVtab; - int rc; - int i; - const char *zFormat; - - if( zArg ){ - for(i=0; zArg[i]; i++){ - char *zErr = 0; - if( fts3MatchinfoCheck(pTab, zArg[i], &zErr) ){ - sqlite3_result_error(pContext, zErr, -1); - sqlite3_free(zErr); - return; - } - } - zFormat = zArg; - }else{ - zFormat = FTS3_MATCHINFO_DEFAULT; - } - - if( !pCsr->pExpr ){ - sqlite3_result_blob(pContext, "", 0, SQLITE_STATIC); - return; - } - - /* Retrieve matchinfo() data. */ - rc = fts3GetMatchinfo(pCsr, zFormat); - sqlite3Fts3SegmentsClose(pTab); - - if( rc!=SQLITE_OK ){ - sqlite3_result_error_code(pContext, rc); - }else{ - int n = pCsr->nMatchinfo * sizeof(u32); - sqlite3_result_blob(pContext, pCsr->aMatchinfo, n, SQLITE_TRANSIENT); - } -} - -#endif diff --git a/src/libtracker-fts/fts3_term.c b/src/libtracker-fts/fts3_term.c deleted file mode 100644 index c49d5cb65..000000000 --- a/src/libtracker-fts/fts3_term.c +++ /dev/null @@ -1,373 +0,0 @@ -/* -** 2011 Jan 27 -** -** The author disclaims copyright to this source code. In place of -** a legal notice, here is a blessing: -** -** May you do good and not evil. -** May you find forgiveness for yourself and forgive others. -** May you share freely, never taking more than you give. -** -****************************************************************************** -** -** This file is not part of the production FTS code. It is only used for -** testing. It contains a virtual table implementation that provides direct -** access to the full-text index of an FTS table. -*/ - -#include "fts3Int.h" -#if !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3) -#ifdef SQLITE_TEST - -#include <string.h> -#include <assert.h> -#include <stdlib.h> - -typedef struct Fts3termTable Fts3termTable; -typedef struct Fts3termCursor Fts3termCursor; - -struct Fts3termTable { - sqlite3_vtab base; /* Base class used by SQLite core */ - int iIndex; /* Index for Fts3Table.aIndex[] */ - Fts3Table *pFts3Tab; -}; - -struct Fts3termCursor { - sqlite3_vtab_cursor base; /* Base class used by SQLite core */ - Fts3MultiSegReader csr; /* Must be right after "base" */ - Fts3SegFilter filter; - - int isEof; /* True if cursor is at EOF */ - char *pNext; - - sqlite3_int64 iRowid; /* Current 'rowid' value */ - sqlite3_int64 iDocid; /* Current 'docid' value */ - int iCol; /* Current 'col' value */ - int iPos; /* Current 'pos' value */ -}; - -/* -** Schema of the terms table. -*/ -#define FTS3_TERMS_SCHEMA "CREATE TABLE x(term, docid, col, pos)" - -/* -** This function does all the work for both the xConnect and xCreate methods. -** These tables have no persistent representation of their own, so xConnect -** and xCreate are identical operations. -*/ -static int fts3termConnectMethod( - sqlite3 *db, /* Database connection */ - void *pCtx, /* Non-zero for an fts4prefix table */ - int argc, /* Number of elements in argv array */ - const char * const *argv, /* xCreate/xConnect argument array */ - sqlite3_vtab **ppVtab, /* OUT: New sqlite3_vtab object */ - char **pzErr /* OUT: sqlite3_malloc'd error message */ -){ - char const *zDb; /* Name of database (e.g. "main") */ - char const *zFts3; /* Name of fts3 table */ - int nDb; /* Result of strlen(zDb) */ - int nFts3; /* Result of strlen(zFts3) */ - int nByte; /* Bytes of space to allocate here */ - int rc; /* value returned by declare_vtab() */ - Fts3termTable *p; /* Virtual table object to return */ - int iIndex = 0; - - UNUSED_PARAMETER(pCtx); - if( argc==5 ){ - iIndex = atoi(argv[4]); - argc--; - } - - /* The user should specify a single argument - the name of an fts3 table. */ - if( argc!=4 ){ - *pzErr = sqlite3_mprintf( - "wrong number of arguments to fts4term constructor" - ); - return SQLITE_ERROR; - } - - zDb = argv[1]; - nDb = (int)strlen(zDb); - zFts3 = argv[3]; - nFts3 = (int)strlen(zFts3); - - rc = sqlite3_declare_vtab(db, FTS3_TERMS_SCHEMA); - if( rc!=SQLITE_OK ) return rc; - - nByte = sizeof(Fts3termTable) + sizeof(Fts3Table) + nDb + nFts3 + 2; - p = (Fts3termTable *)sqlite3_malloc(nByte); - if( !p ) return SQLITE_NOMEM; - memset(p, 0, nByte); - - p->pFts3Tab = (Fts3Table *)&p[1]; - p->pFts3Tab->zDb = (char *)&p->pFts3Tab[1]; - p->pFts3Tab->zName = &p->pFts3Tab->zDb[nDb+1]; - p->pFts3Tab->db = db; - p->pFts3Tab->nIndex = iIndex+1; - p->iIndex = iIndex; - - memcpy((char *)p->pFts3Tab->zDb, zDb, nDb); - memcpy((char *)p->pFts3Tab->zName, zFts3, nFts3); - sqlite3Fts3Dequote((char *)p->pFts3Tab->zName); - - *ppVtab = (sqlite3_vtab *)p; - return SQLITE_OK; -} - -/* -** This function does the work for both the xDisconnect and xDestroy methods. -** These tables have no persistent representation of their own, so xDisconnect -** and xDestroy are identical operations. -*/ -static int fts3termDisconnectMethod(sqlite3_vtab *pVtab){ - Fts3termTable *p = (Fts3termTable *)pVtab; - Fts3Table *pFts3 = p->pFts3Tab; - int i; - - /* Free any prepared statements held */ - for(i=0; i<SizeofArray(pFts3->aStmt); i++){ - sqlite3_finalize(pFts3->aStmt[i]); - } - sqlite3_free(pFts3->zSegmentsTbl); - sqlite3_free(p); - return SQLITE_OK; -} - -#define FTS4AUX_EQ_CONSTRAINT 1 -#define FTS4AUX_GE_CONSTRAINT 2 -#define FTS4AUX_LE_CONSTRAINT 4 - -/* -** xBestIndex - Analyze a WHERE and ORDER BY clause. -*/ -static int fts3termBestIndexMethod( - sqlite3_vtab *pVTab, - sqlite3_index_info *pInfo -){ - UNUSED_PARAMETER(pVTab); - - /* This vtab naturally does "ORDER BY term, docid, col, pos". */ - if( pInfo->nOrderBy ){ - int i; - for(i=0; i<pInfo->nOrderBy; i++){ - if( pInfo->aOrderBy[i].iColumn!=i || pInfo->aOrderBy[i].desc ) break; - } - if( i==pInfo->nOrderBy ){ - pInfo->orderByConsumed = 1; - } - } - - return SQLITE_OK; -} - -/* -** xOpen - Open a cursor. -*/ -static int fts3termOpenMethod(sqlite3_vtab *pVTab, sqlite3_vtab_cursor **ppCsr){ - Fts3termCursor *pCsr; /* Pointer to cursor object to return */ - - UNUSED_PARAMETER(pVTab); - - pCsr = (Fts3termCursor *)sqlite3_malloc(sizeof(Fts3termCursor)); - if( !pCsr ) return SQLITE_NOMEM; - memset(pCsr, 0, sizeof(Fts3termCursor)); - - *ppCsr = (sqlite3_vtab_cursor *)pCsr; - return SQLITE_OK; -} - -/* -** xClose - Close a cursor. -*/ -static int fts3termCloseMethod(sqlite3_vtab_cursor *pCursor){ - Fts3Table *pFts3 = ((Fts3termTable *)pCursor->pVtab)->pFts3Tab; - Fts3termCursor *pCsr = (Fts3termCursor *)pCursor; - - sqlite3Fts3SegmentsClose(pFts3); - sqlite3Fts3SegReaderFinish(&pCsr->csr); - sqlite3_free(pCsr); - return SQLITE_OK; -} - -/* -** xNext - Advance the cursor to the next row, if any. -*/ -static int fts3termNextMethod(sqlite3_vtab_cursor *pCursor){ - Fts3termCursor *pCsr = (Fts3termCursor *)pCursor; - Fts3Table *pFts3 = ((Fts3termTable *)pCursor->pVtab)->pFts3Tab; - int rc; - sqlite3_int64 v; - - /* Increment our pretend rowid value. */ - pCsr->iRowid++; - - /* Advance to the next term in the full-text index. */ - if( pCsr->csr.aDoclist==0 - || pCsr->pNext>=&pCsr->csr.aDoclist[pCsr->csr.nDoclist-1] - ){ - rc = sqlite3Fts3SegReaderStep(pFts3, &pCsr->csr); - if( rc!=SQLITE_ROW ){ - pCsr->isEof = 1; - return rc; - } - - pCsr->iCol = 0; - pCsr->iPos = 0; - pCsr->iDocid = 0; - pCsr->pNext = pCsr->csr.aDoclist; - - /* Read docid */ - pCsr->pNext += sqlite3Fts3GetVarint(pCsr->pNext, &pCsr->iDocid); - } - - pCsr->pNext += sqlite3Fts3GetVarint(pCsr->pNext, &v); - if( v==0 ){ - pCsr->pNext += sqlite3Fts3GetVarint(pCsr->pNext, &v); - pCsr->iDocid += v; - pCsr->pNext += sqlite3Fts3GetVarint(pCsr->pNext, &v); - pCsr->iCol = 0; - pCsr->iPos = 0; - } - - if( v==1 ){ - pCsr->pNext += sqlite3Fts3GetVarint(pCsr->pNext, &v); - pCsr->iCol += (int)v; - pCsr->iPos = 0; - pCsr->pNext += sqlite3Fts3GetVarint(pCsr->pNext, &v); - } - - pCsr->iPos += (int)(v - 2); - - return SQLITE_OK; -} - -/* -** xFilter - Initialize a cursor to point at the start of its data. -*/ -static int fts3termFilterMethod( - sqlite3_vtab_cursor *pCursor, /* The cursor used for this query */ - int idxNum, /* Strategy index */ - const char *idxStr, /* Unused */ - int nVal, /* Number of elements in apVal */ - sqlite3_value **apVal /* Arguments for the indexing scheme */ -){ - Fts3termCursor *pCsr = (Fts3termCursor *)pCursor; - Fts3termTable *p = (Fts3termTable *)pCursor->pVtab; - Fts3Table *pFts3 = p->pFts3Tab; - int rc; - - UNUSED_PARAMETER(nVal); - UNUSED_PARAMETER(idxNum); - UNUSED_PARAMETER(idxStr); - UNUSED_PARAMETER(apVal); - - assert( idxStr==0 && idxNum==0 ); - - /* In case this cursor is being reused, close and zero it. */ - testcase(pCsr->filter.zTerm); - sqlite3Fts3SegReaderFinish(&pCsr->csr); - memset(&pCsr->csr, 0, ((u8*)&pCsr[1]) - (u8*)&pCsr->csr); - - pCsr->filter.flags = FTS3_SEGMENT_REQUIRE_POS|FTS3_SEGMENT_IGNORE_EMPTY; - pCsr->filter.flags |= FTS3_SEGMENT_SCAN; - - rc = sqlite3Fts3SegReaderCursor(pFts3, 0, p->iIndex, FTS3_SEGCURSOR_ALL, - pCsr->filter.zTerm, pCsr->filter.nTerm, 0, 1, &pCsr->csr - ); - if( rc==SQLITE_OK ){ - rc = sqlite3Fts3SegReaderStart(pFts3, &pCsr->csr, &pCsr->filter); - } - if( rc==SQLITE_OK ){ - rc = fts3termNextMethod(pCursor); - } - return rc; -} - -/* -** xEof - Return true if the cursor is at EOF, or false otherwise. -*/ -static int fts3termEofMethod(sqlite3_vtab_cursor *pCursor){ - Fts3termCursor *pCsr = (Fts3termCursor *)pCursor; - return pCsr->isEof; -} - -/* -** xColumn - Return a column value. -*/ -static int fts3termColumnMethod( - sqlite3_vtab_cursor *pCursor, /* Cursor to retrieve value from */ - sqlite3_context *pCtx, /* Context for sqlite3_result_xxx() calls */ - int iCol /* Index of column to read value from */ -){ - Fts3termCursor *p = (Fts3termCursor *)pCursor; - - assert( iCol>=0 && iCol<=3 ); - switch( iCol ){ - case 0: - sqlite3_result_text(pCtx, p->csr.zTerm, p->csr.nTerm, SQLITE_TRANSIENT); - break; - case 1: - sqlite3_result_int64(pCtx, p->iDocid); - break; - case 2: - sqlite3_result_int64(pCtx, p->iCol); - break; - default: - sqlite3_result_int64(pCtx, p->iPos); - break; - } - - return SQLITE_OK; -} - -/* -** xRowid - Return the current rowid for the cursor. -*/ -static int fts3termRowidMethod( - sqlite3_vtab_cursor *pCursor, /* Cursor to retrieve value from */ - sqlite_int64 *pRowid /* OUT: Rowid value */ -){ - Fts3termCursor *pCsr = (Fts3termCursor *)pCursor; - *pRowid = pCsr->iRowid; - return SQLITE_OK; -} - -/* -** Register the fts3term module with database connection db. Return SQLITE_OK -** if successful or an error code if sqlite3_create_module() fails. -*/ -int sqlite3Fts3InitTerm(sqlite3 *db){ - static const sqlite3_module fts3term_module = { - 0, /* iVersion */ - fts3termConnectMethod, /* xCreate */ - fts3termConnectMethod, /* xConnect */ - fts3termBestIndexMethod, /* xBestIndex */ - fts3termDisconnectMethod, /* xDisconnect */ - fts3termDisconnectMethod, /* xDestroy */ - fts3termOpenMethod, /* xOpen */ - fts3termCloseMethod, /* xClose */ - fts3termFilterMethod, /* xFilter */ - fts3termNextMethod, /* xNext */ - fts3termEofMethod, /* xEof */ - fts3termColumnMethod, /* xColumn */ - fts3termRowidMethod, /* xRowid */ - 0, /* xUpdate */ - 0, /* xBegin */ - 0, /* xSync */ - 0, /* xCommit */ - 0, /* xRollback */ - 0, /* xFindFunction */ - 0, /* xRename */ - 0, /* xSavepoint */ - 0, /* xRelease */ - 0 /* xRollbackTo */ - }; - int rc; /* Return code */ - - rc = sqlite3_create_module(db, "fts4term", &fts3term_module, 0); - return rc; -} - -#endif -#endif /* !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3) */ diff --git a/src/libtracker-fts/fts3_test.c b/src/libtracker-fts/fts3_test.c deleted file mode 100644 index 4da0b8f13..000000000 --- a/src/libtracker-fts/fts3_test.c +++ /dev/null @@ -1,535 +0,0 @@ -/* -** 2011 Jun 13 -** -** The author disclaims copyright to this source code. In place of -** a legal notice, here is a blessing: -** -** May you do good and not evil. -** May you find forgiveness for yourself and forgive others. -** May you share freely, never taking more than you give. -** -****************************************************************************** -** -** This file is not part of the production FTS code. It is only used for -** testing. It contains a Tcl command that can be used to test if a document -** matches an FTS NEAR expression. -** -** As of March 2012, it also contains a version 1 tokenizer used for testing -** that the sqlite3_tokenizer_module.xLanguage() method is invoked correctly. -*/ - -#include <tcl.h> -#include <string.h> -#include <assert.h> - -#if defined(SQLITE_TEST) -#if defined(SQLITE_ENABLE_FTS3) || defined(SQLITE_ENABLE_FTS4) - -/* Required so that the "ifdef SQLITE_ENABLE_FTS3" below works */ -#include "fts3Int.h" - -#define NM_MAX_TOKEN 12 - -typedef struct NearPhrase NearPhrase; -typedef struct NearDocument NearDocument; -typedef struct NearToken NearToken; - -struct NearDocument { - int nToken; /* Length of token in bytes */ - NearToken *aToken; /* Token array */ -}; - -struct NearToken { - int n; /* Length of token in bytes */ - const char *z; /* Pointer to token string */ -}; - -struct NearPhrase { - int nNear; /* Preceding NEAR value */ - int nToken; /* Number of tokens in this phrase */ - NearToken aToken[NM_MAX_TOKEN]; /* Array of tokens in this phrase */ -}; - -static int nm_phrase_match( - NearPhrase *p, - NearToken *aToken -){ - int ii; - - for(ii=0; ii<p->nToken; ii++){ - NearToken *pToken = &p->aToken[ii]; - if( pToken->n>0 && pToken->z[pToken->n-1]=='*' ){ - if( aToken[ii].n<(pToken->n-1) ) return 0; - if( memcmp(aToken[ii].z, pToken->z, pToken->n-1) ) return 0; - }else{ - if( aToken[ii].n!=pToken->n ) return 0; - if( memcmp(aToken[ii].z, pToken->z, pToken->n) ) return 0; - } - } - - return 1; -} - -static int nm_near_chain( - int iDir, /* Direction to iterate through aPhrase[] */ - NearDocument *pDoc, /* Document to match against */ - int iPos, /* Position at which iPhrase was found */ - int nPhrase, /* Size of phrase array */ - NearPhrase *aPhrase, /* Phrase array */ - int iPhrase /* Index of phrase found */ -){ - int iStart; - int iStop; - int ii; - int nNear; - int iPhrase2; - NearPhrase *p; - NearPhrase *pPrev; - - assert( iDir==1 || iDir==-1 ); - - if( iDir==1 ){ - if( (iPhrase+1)==nPhrase ) return 1; - nNear = aPhrase[iPhrase+1].nNear; - }else{ - if( iPhrase==0 ) return 1; - nNear = aPhrase[iPhrase].nNear; - } - pPrev = &aPhrase[iPhrase]; - iPhrase2 = iPhrase+iDir; - p = &aPhrase[iPhrase2]; - - iStart = iPos - nNear - p->nToken; - iStop = iPos + nNear + pPrev->nToken; - - if( iStart<0 ) iStart = 0; - if( iStop > pDoc->nToken - p->nToken ) iStop = pDoc->nToken - p->nToken; - - for(ii=iStart; ii<=iStop; ii++){ - if( nm_phrase_match(p, &pDoc->aToken[ii]) ){ - if( nm_near_chain(iDir, pDoc, ii, nPhrase, aPhrase, iPhrase2) ) return 1; - } - } - - return 0; -} - -static int nm_match_count( - NearDocument *pDoc, /* Document to match against */ - int nPhrase, /* Size of phrase array */ - NearPhrase *aPhrase, /* Phrase array */ - int iPhrase /* Index of phrase to count matches for */ -){ - int nOcc = 0; - int ii; - NearPhrase *p = &aPhrase[iPhrase]; - - for(ii=0; ii<(pDoc->nToken + 1 - p->nToken); ii++){ - if( nm_phrase_match(p, &pDoc->aToken[ii]) ){ - /* Test forward NEAR chain (i>iPhrase) */ - if( 0==nm_near_chain(1, pDoc, ii, nPhrase, aPhrase, iPhrase) ) continue; - - /* Test reverse NEAR chain (i<iPhrase) */ - if( 0==nm_near_chain(-1, pDoc, ii, nPhrase, aPhrase, iPhrase) ) continue; - - /* This is a real match. Increment the counter. */ - nOcc++; - } - } - - return nOcc; -} - -/* -** Tclcmd: fts3_near_match DOCUMENT EXPR ?OPTIONS? -*/ -static int fts3_near_match_cmd( - ClientData clientData, - Tcl_Interp *interp, - int objc, - Tcl_Obj *CONST objv[] -){ - int nTotal = 0; - int rc; - int ii; - int nPhrase; - NearPhrase *aPhrase = 0; - NearDocument doc = {0, 0}; - Tcl_Obj **apDocToken; - Tcl_Obj *pRet; - Tcl_Obj *pPhrasecount = 0; - - Tcl_Obj **apExprToken; - int nExprToken; - - UNUSED_PARAMETER(clientData); - - /* Must have 3 or more arguments. */ - if( objc<3 || (objc%2)==0 ){ - Tcl_WrongNumArgs(interp, 1, objv, "DOCUMENT EXPR ?OPTION VALUE?..."); - rc = TCL_ERROR; - goto near_match_out; - } - - for(ii=3; ii<objc; ii+=2){ - enum NM_enum { NM_PHRASECOUNTS }; - struct TestnmSubcmd { - char *zName; - enum NM_enum eOpt; - } aOpt[] = { - { "-phrasecountvar", NM_PHRASECOUNTS }, - { 0, 0 } - }; - int iOpt; - if( Tcl_GetIndexFromObjStruct( - interp, objv[ii], aOpt, sizeof(aOpt[0]), "option", 0, &iOpt) - ){ - return TCL_ERROR; - } - - switch( aOpt[iOpt].eOpt ){ - case NM_PHRASECOUNTS: - pPhrasecount = objv[ii+1]; - break; - } - } - - rc = Tcl_ListObjGetElements(interp, objv[1], &doc.nToken, &apDocToken); - if( rc!=TCL_OK ) goto near_match_out; - doc.aToken = (NearToken *)ckalloc(doc.nToken*sizeof(NearToken)); - for(ii=0; ii<doc.nToken; ii++){ - doc.aToken[ii].z = Tcl_GetStringFromObj(apDocToken[ii], &doc.aToken[ii].n); - } - - rc = Tcl_ListObjGetElements(interp, objv[2], &nExprToken, &apExprToken); - if( rc!=TCL_OK ) goto near_match_out; - - nPhrase = (nExprToken + 1) / 2; - aPhrase = (NearPhrase *)ckalloc(nPhrase * sizeof(NearPhrase)); - memset(aPhrase, 0, nPhrase * sizeof(NearPhrase)); - for(ii=0; ii<nPhrase; ii++){ - Tcl_Obj *pPhrase = apExprToken[ii*2]; - Tcl_Obj **apToken; - int nToken; - int jj; - - rc = Tcl_ListObjGetElements(interp, pPhrase, &nToken, &apToken); - if( rc!=TCL_OK ) goto near_match_out; - if( nToken>NM_MAX_TOKEN ){ - Tcl_AppendResult(interp, "Too many tokens in phrase", 0); - rc = TCL_ERROR; - goto near_match_out; - } - for(jj=0; jj<nToken; jj++){ - NearToken *pT = &aPhrase[ii].aToken[jj]; - pT->z = Tcl_GetStringFromObj(apToken[jj], &pT->n); - } - aPhrase[ii].nToken = nToken; - } - for(ii=1; ii<nPhrase; ii++){ - Tcl_Obj *pNear = apExprToken[2*ii-1]; - int nNear; - rc = Tcl_GetIntFromObj(interp, pNear, &nNear); - if( rc!=TCL_OK ) goto near_match_out; - aPhrase[ii].nNear = nNear; - } - - pRet = Tcl_NewObj(); - Tcl_IncrRefCount(pRet); - for(ii=0; ii<nPhrase; ii++){ - int nOcc = nm_match_count(&doc, nPhrase, aPhrase, ii); - Tcl_ListObjAppendElement(interp, pRet, Tcl_NewIntObj(nOcc)); - nTotal += nOcc; - } - if( pPhrasecount ){ - Tcl_ObjSetVar2(interp, pPhrasecount, 0, pRet, 0); - } - Tcl_DecrRefCount(pRet); - Tcl_SetObjResult(interp, Tcl_NewBooleanObj(nTotal>0)); - - near_match_out: - ckfree((char *)aPhrase); - ckfree((char *)doc.aToken); - return rc; -} - -/* -** Tclcmd: fts3_configure_incr_load ?CHUNKSIZE THRESHOLD? -** -** Normally, FTS uses hard-coded values to determine the minimum doclist -** size eligible for incremental loading, and the size of the chunks loaded -** when a doclist is incrementally loaded. This command allows the built-in -** values to be overridden for testing purposes. -** -** If present, the first argument is the chunksize in bytes to load doclists -** in. The second argument is the minimum doclist size in bytes to use -** incremental loading with. -** -** Whether or not the arguments are present, this command returns a list of -** two integers - the initial chunksize and threshold when the command is -** invoked. This can be used to restore the default behaviour after running -** tests. For example: -** -** # Override incr-load settings for testing: -** set cfg [fts3_configure_incr_load $new_chunksize $new_threshold] -** -** .... run tests .... -** -** # Restore initial incr-load settings: -** eval fts3_configure_incr_load $cfg -*/ -static int fts3_configure_incr_load_cmd( - ClientData clientData, - Tcl_Interp *interp, - int objc, - Tcl_Obj *CONST objv[] -){ -#ifdef SQLITE_ENABLE_FTS3 - extern int test_fts3_node_chunksize; - extern int test_fts3_node_chunk_threshold; - Tcl_Obj *pRet; - - if( objc!=1 && objc!=3 ){ - Tcl_WrongNumArgs(interp, 1, objv, "?CHUNKSIZE THRESHOLD?"); - return TCL_ERROR; - } - - pRet = Tcl_NewObj(); - Tcl_IncrRefCount(pRet); - Tcl_ListObjAppendElement( - interp, pRet, Tcl_NewIntObj(test_fts3_node_chunksize)); - Tcl_ListObjAppendElement( - interp, pRet, Tcl_NewIntObj(test_fts3_node_chunk_threshold)); - - if( objc==3 ){ - int iArg1; - int iArg2; - if( Tcl_GetIntFromObj(interp, objv[1], &iArg1) - || Tcl_GetIntFromObj(interp, objv[2], &iArg2) - ){ - Tcl_DecrRefCount(pRet); - return TCL_ERROR; - } - test_fts3_node_chunksize = iArg1; - test_fts3_node_chunk_threshold = iArg2; - } - - Tcl_SetObjResult(interp, pRet); - Tcl_DecrRefCount(pRet); -#endif - UNUSED_PARAMETER(clientData); - return TCL_OK; -} - -#ifdef SQLITE_ENABLE_FTS3 -/************************************************************************** -** Beginning of test tokenizer code. -** -** For language 0, this tokenizer is similar to the default 'simple' -** tokenizer. For other languages L, the following: -** -** * Odd numbered languages are case-sensitive. Even numbered -** languages are not. -** -** * Language ids 100 or greater are considered an error. -** -** The implementation assumes that the input contains only ASCII characters -** (i.e. those that may be encoded in UTF-8 using a single byte). -*/ -typedef struct test_tokenizer { - sqlite3_tokenizer base; -} test_tokenizer; - -typedef struct test_tokenizer_cursor { - sqlite3_tokenizer_cursor base; - const char *aInput; /* Input being tokenized */ - int nInput; /* Size of the input in bytes */ - int iInput; /* Current offset in aInput */ - int iToken; /* Index of next token to be returned */ - char *aBuffer; /* Buffer containing current token */ - int nBuffer; /* Number of bytes allocated at pToken */ - int iLangid; /* Configured language id */ -} test_tokenizer_cursor; - -static int testTokenizerCreate( - int argc, const char * const *argv, - sqlite3_tokenizer **ppTokenizer -){ - test_tokenizer *pNew; - UNUSED_PARAMETER(argc); - UNUSED_PARAMETER(argv); - - pNew = sqlite3_malloc(sizeof(test_tokenizer)); - if( !pNew ) return SQLITE_NOMEM; - memset(pNew, 0, sizeof(test_tokenizer)); - - *ppTokenizer = (sqlite3_tokenizer *)pNew; - return SQLITE_OK; -} - -static int testTokenizerDestroy(sqlite3_tokenizer *pTokenizer){ - test_tokenizer *p = (test_tokenizer *)pTokenizer; - sqlite3_free(p); - return SQLITE_OK; -} - -static int testTokenizerOpen( - sqlite3_tokenizer *pTokenizer, /* The tokenizer */ - const char *pInput, int nBytes, /* String to be tokenized */ - sqlite3_tokenizer_cursor **ppCursor /* OUT: Tokenization cursor */ -){ - int rc = SQLITE_OK; /* Return code */ - test_tokenizer_cursor *pCsr; /* New cursor object */ - - UNUSED_PARAMETER(pTokenizer); - - pCsr = (test_tokenizer_cursor *)sqlite3_malloc(sizeof(test_tokenizer_cursor)); - if( pCsr==0 ){ - rc = SQLITE_NOMEM; - }else{ - memset(pCsr, 0, sizeof(test_tokenizer_cursor)); - pCsr->aInput = pInput; - if( nBytes<0 ){ - pCsr->nInput = (int)strlen(pInput); - }else{ - pCsr->nInput = nBytes; - } - } - - *ppCursor = (sqlite3_tokenizer_cursor *)pCsr; - return rc; -} - -static int testTokenizerClose(sqlite3_tokenizer_cursor *pCursor){ - test_tokenizer_cursor *pCsr = (test_tokenizer_cursor *)pCursor; - sqlite3_free(pCsr->aBuffer); - sqlite3_free(pCsr); - return SQLITE_OK; -} - -static int testIsTokenChar(char c){ - return (c>='a' && c<='z') || (c>='A' && c<='Z'); -} -static int testTolower(char c){ - char ret = c; - if( ret>='A' && ret<='Z') ret = ret - ('A'-'a'); - return ret; -} - -static int testTokenizerNext( - sqlite3_tokenizer_cursor *pCursor, /* Cursor returned by testTokenizerOpen */ - const char **ppToken, /* OUT: *ppToken is the token text */ - int *pnBytes, /* OUT: Number of bytes in token */ - int *piStartOffset, /* OUT: Starting offset of token */ - int *piEndOffset, /* OUT: Ending offset of token */ - int *piPosition /* OUT: Position integer of token */ -){ - test_tokenizer_cursor *pCsr = (test_tokenizer_cursor *)pCursor; - int rc = SQLITE_OK; - const char *p; - const char *pEnd; - - p = &pCsr->aInput[pCsr->iInput]; - pEnd = &pCsr->aInput[pCsr->nInput]; - - /* Skip past any white-space */ - assert( p<=pEnd ); - while( p<pEnd && testIsTokenChar(*p)==0 ) p++; - - if( p==pEnd ){ - rc = SQLITE_DONE; - }else{ - /* Advance to the end of the token */ - const char *pToken = p; - int nToken; - while( p<pEnd && testIsTokenChar(*p) ) p++; - nToken = (int)(p-pToken); - - /* Copy the token into the buffer */ - if( nToken>pCsr->nBuffer ){ - sqlite3_free(pCsr->aBuffer); - pCsr->aBuffer = sqlite3_malloc(nToken); - } - if( pCsr->aBuffer==0 ){ - rc = SQLITE_NOMEM; - }else{ - int i; - - if( pCsr->iLangid & 0x00000001 ){ - for(i=0; i<nToken; i++) pCsr->aBuffer[i] = pToken[i]; - }else{ - for(i=0; i<nToken; i++) pCsr->aBuffer[i] = testTolower(pToken[i]); - } - pCsr->iToken++; - pCsr->iInput = (int)(p - pCsr->aInput); - - *ppToken = pCsr->aBuffer; - *pnBytes = nToken; - *piStartOffset = (int)(pToken - pCsr->aInput); - *piEndOffset = (int)(p - pCsr->aInput); - *piPosition = pCsr->iToken; - } - } - - return rc; -} - -static int testTokenizerLanguage( - sqlite3_tokenizer_cursor *pCursor, - int iLangid -){ - int rc = SQLITE_OK; - test_tokenizer_cursor *pCsr = (test_tokenizer_cursor *)pCursor; - pCsr->iLangid = iLangid; - if( pCsr->iLangid>=100 ){ - rc = SQLITE_ERROR; - } - return rc; -} -#endif - -static int fts3_test_tokenizer_cmd( - ClientData clientData, - Tcl_Interp *interp, - int objc, - Tcl_Obj *CONST objv[] -){ -#ifdef SQLITE_ENABLE_FTS3 - static const sqlite3_tokenizer_module testTokenizerModule = { - 1, - testTokenizerCreate, - testTokenizerDestroy, - testTokenizerOpen, - testTokenizerClose, - testTokenizerNext, - testTokenizerLanguage - }; - const sqlite3_tokenizer_module *pPtr = &testTokenizerModule; - if( objc!=1 ){ - Tcl_WrongNumArgs(interp, 1, objv, ""); - return TCL_ERROR; - } - Tcl_SetObjResult(interp, Tcl_NewByteArrayObj( - (const unsigned char *)&pPtr, sizeof(sqlite3_tokenizer_module *) - )); -#endif - UNUSED_PARAMETER(clientData); - return TCL_OK; -} - -/* -** End of tokenizer code. -**************************************************************************/ - -int Sqlitetestfts3_Init(Tcl_Interp *interp){ - Tcl_CreateObjCommand(interp, "fts3_near_match", fts3_near_match_cmd, 0, 0); - Tcl_CreateObjCommand(interp, - "fts3_configure_incr_load", fts3_configure_incr_load_cmd, 0, 0 - ); - Tcl_CreateObjCommand( - interp, "fts3_test_tokenizer", fts3_test_tokenizer_cmd, 0, 0 - ); - return TCL_OK; -} -#endif /* SQLITE_ENABLE_FTS3 || SQLITE_ENABLE_FTS4 */ -#endif /* ifdef SQLITE_TEST */ diff --git a/src/libtracker-fts/fts3_tokenizer.c b/src/libtracker-fts/fts3_tokenizer.c deleted file mode 100644 index 8241be81f..000000000 --- a/src/libtracker-fts/fts3_tokenizer.c +++ /dev/null @@ -1,488 +0,0 @@ -/* -** 2007 June 22 -** -** The author disclaims copyright to this source code. In place of -** a legal notice, here is a blessing: -** -** May you do good and not evil. -** May you find forgiveness for yourself and forgive others. -** May you share freely, never taking more than you give. -** -****************************************************************************** -** -** This is part of an SQLite module implementing full-text search. -** This particular file implements the generic tokenizer interface. -*/ - -/* -** The code in this file is only compiled if: -** -** * The FTS3 module is being built as an extension -** (in which case SQLITE_CORE is not defined), or -** -** * The FTS3 module is being built into the core of -** SQLite (in which case SQLITE_ENABLE_FTS3 is defined). -*/ -#include "fts3Int.h" -#if !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3) - -#include <assert.h> -#include <string.h> - -/* -** Implementation of the SQL scalar function for accessing the underlying -** hash table. This function may be called as follows: -** -** SELECT <function-name>(<key-name>); -** SELECT <function-name>(<key-name>, <pointer>); -** -** where <function-name> is the name passed as the second argument -** to the sqlite3Fts3InitHashTable() function (e.g. 'fts3_tokenizer'). -** -** If the <pointer> argument is specified, it must be a blob value -** containing a pointer to be stored as the hash data corresponding -** to the string <key-name>. If <pointer> is not specified, then -** the string <key-name> must already exist in the has table. Otherwise, -** an error is returned. -** -** Whether or not the <pointer> argument is specified, the value returned -** is a blob containing the pointer stored as the hash data corresponding -** to string <key-name> (after the hash-table is updated, if applicable). -*/ -static void scalarFunc( - sqlite3_context *context, - int argc, - sqlite3_value **argv -){ - Fts3Hash *pHash; - void *pPtr = 0; - const unsigned char *zName; - int nName; - - assert( argc==1 || argc==2 ); - - pHash = (Fts3Hash *)sqlite3_user_data(context); - - zName = sqlite3_value_text(argv[0]); - nName = sqlite3_value_bytes(argv[0])+1; - - if( argc==2 ){ - void *pOld; - int n = sqlite3_value_bytes(argv[1]); - if( n!=sizeof(pPtr) ){ - sqlite3_result_error(context, "argument type mismatch", -1); - return; - } - pPtr = *(void **)sqlite3_value_blob(argv[1]); - pOld = sqlite3Fts3HashInsert(pHash, (void *)zName, nName, pPtr); - if( pOld==pPtr ){ - sqlite3_result_error(context, "out of memory", -1); - return; - } - }else{ - pPtr = sqlite3Fts3HashFind(pHash, zName, nName); - if( !pPtr ){ - char *zErr = sqlite3_mprintf("unknown tokenizer: %s", zName); - sqlite3_result_error(context, zErr, -1); - sqlite3_free(zErr); - return; - } - } - - sqlite3_result_blob(context, (void *)&pPtr, sizeof(pPtr), SQLITE_TRANSIENT); -} - -int sqlite3Fts3IsIdChar(char c){ - static const char isFtsIdChar[] = { - 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 0x */ - 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 1x */ - 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 2x */ - 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, /* 3x */ - 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, /* 4x */ - 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 1, /* 5x */ - 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, /* 6x */ - 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, /* 7x */ - }; - return (c&0x80 || isFtsIdChar[(int)(c)]); -} - -const char *sqlite3Fts3NextToken(const char *zStr, int *pn){ - const char *z1; - const char *z2 = 0; - - /* Find the start of the next token. */ - z1 = zStr; - while( z2==0 ){ - char c = *z1; - switch( c ){ - case '\0': return 0; /* No more tokens here */ - case '\'': - case '"': - case '`': { - z2 = z1; - while( *++z2 && (*z2!=c || *++z2==c) ); - break; - } - case '[': - z2 = &z1[1]; - while( *z2 && z2[0]!=']' ) z2++; - if( *z2 ) z2++; - break; - - default: - if( sqlite3Fts3IsIdChar(*z1) ){ - z2 = &z1[1]; - while( sqlite3Fts3IsIdChar(*z2) ) z2++; - }else{ - z1++; - } - } - } - - *pn = (int)(z2-z1); - return z1; -} - -int sqlite3Fts3InitTokenizer( - Fts3Hash *pHash, /* Tokenizer hash table */ - const char *zArg, /* Tokenizer name */ - sqlite3_tokenizer **ppTok, /* OUT: Tokenizer (if applicable) */ - char **pzErr /* OUT: Set to malloced error message */ -){ - int rc; - char *z = (char *)zArg; - int n = 0; - char *zCopy; - char *zEnd; /* Pointer to nul-term of zCopy */ - sqlite3_tokenizer_module *m; - - zCopy = sqlite3_mprintf("%s", zArg); - if( !zCopy ) return SQLITE_NOMEM; - zEnd = &zCopy[strlen(zCopy)]; - - z = (char *)sqlite3Fts3NextToken(zCopy, &n); - z[n] = '\0'; - sqlite3Fts3Dequote(z); - - m = (sqlite3_tokenizer_module *)sqlite3Fts3HashFind(pHash,z,(int)strlen(z)+1); - if( !m ){ - *pzErr = sqlite3_mprintf("unknown tokenizer: %s", z); - rc = SQLITE_ERROR; - }else{ - char const **aArg = 0; - int iArg = 0; - z = &z[n+1]; - while( z<zEnd && (NULL!=(z = (char *)sqlite3Fts3NextToken(z, &n))) ){ - int nNew = sizeof(char *)*(iArg+1); - char const **aNew = (const char **)sqlite3_realloc((void *)aArg, nNew); - if( !aNew ){ - sqlite3_free(zCopy); - sqlite3_free((void *)aArg); - return SQLITE_NOMEM; - } - aArg = aNew; - aArg[iArg++] = z; - z[n] = '\0'; - sqlite3Fts3Dequote(z); - z = &z[n+1]; - } - rc = m->xCreate(iArg, aArg, ppTok); - assert( rc!=SQLITE_OK || *ppTok ); - if( rc!=SQLITE_OK ){ - *pzErr = sqlite3_mprintf("unknown tokenizer"); - }else{ - (*ppTok)->pModule = m; - } - sqlite3_free((void *)aArg); - } - - sqlite3_free(zCopy); - return rc; -} - - -#ifdef SQLITE_TEST - -#include <tcl.h> -#include <string.h> - -/* -** Implementation of a special SQL scalar function for testing tokenizers -** designed to be used in concert with the Tcl testing framework. This -** function must be called with two or more arguments: -** -** SELECT <function-name>(<key-name>, ..., <input-string>); -** -** where <function-name> is the name passed as the second argument -** to the sqlite3Fts3InitHashTable() function (e.g. 'fts3_tokenizer') -** concatenated with the string '_test' (e.g. 'fts3_tokenizer_test'). -** -** The return value is a string that may be interpreted as a Tcl -** list. For each token in the <input-string>, three elements are -** added to the returned list. The first is the token position, the -** second is the token text (folded, stemmed, etc.) and the third is the -** substring of <input-string> associated with the token. For example, -** using the built-in "simple" tokenizer: -** -** SELECT fts_tokenizer_test('simple', 'I don't see how'); -** -** will return the string: -** -** "{0 i I 1 dont don't 2 see see 3 how how}" -** -*/ -static void testFunc( - sqlite3_context *context, - int argc, - sqlite3_value **argv -){ - Fts3Hash *pHash; - sqlite3_tokenizer_module *p; - sqlite3_tokenizer *pTokenizer = 0; - sqlite3_tokenizer_cursor *pCsr = 0; - - const char *zErr = 0; - - const char *zName; - int nName; - const char *zInput; - int nInput; - - const char *azArg[64]; - - const char *zToken; - int nToken = 0; - int iStart = 0; - int iEnd = 0; - int iPos = 0; - int i; - - Tcl_Obj *pRet; - - if( argc<2 ){ - sqlite3_result_error(context, "insufficient arguments", -1); - return; - } - - nName = sqlite3_value_bytes(argv[0]); - zName = (const char *)sqlite3_value_text(argv[0]); - nInput = sqlite3_value_bytes(argv[argc-1]); - zInput = (const char *)sqlite3_value_text(argv[argc-1]); - - pHash = (Fts3Hash *)sqlite3_user_data(context); - p = (sqlite3_tokenizer_module *)sqlite3Fts3HashFind(pHash, zName, nName+1); - - if( !p ){ - char *zErr = sqlite3_mprintf("unknown tokenizer: %s", zName); - sqlite3_result_error(context, zErr, -1); - sqlite3_free(zErr); - return; - } - - pRet = Tcl_NewObj(); - Tcl_IncrRefCount(pRet); - - for(i=1; i<argc-1; i++){ - azArg[i-1] = (const char *)sqlite3_value_text(argv[i]); - } - - if( SQLITE_OK!=p->xCreate(argc-2, azArg, &pTokenizer) ){ - zErr = "error in xCreate()"; - goto finish; - } - pTokenizer->pModule = p; - if( sqlite3Fts3OpenTokenizer(pTokenizer, 0, zInput, nInput, &pCsr) ){ - zErr = "error in xOpen()"; - goto finish; - } - - while( SQLITE_OK==p->xNext(pCsr, &zToken, &nToken, &iStart, &iEnd, &iPos) ){ - Tcl_ListObjAppendElement(0, pRet, Tcl_NewIntObj(iPos)); - Tcl_ListObjAppendElement(0, pRet, Tcl_NewStringObj(zToken, nToken)); - zToken = &zInput[iStart]; - nToken = iEnd-iStart; - Tcl_ListObjAppendElement(0, pRet, Tcl_NewStringObj(zToken, nToken)); - } - - if( SQLITE_OK!=p->xClose(pCsr) ){ - zErr = "error in xClose()"; - goto finish; - } - if( SQLITE_OK!=p->xDestroy(pTokenizer) ){ - zErr = "error in xDestroy()"; - goto finish; - } - -finish: - if( zErr ){ - sqlite3_result_error(context, zErr, -1); - }else{ - sqlite3_result_text(context, Tcl_GetString(pRet), -1, SQLITE_TRANSIENT); - } - Tcl_DecrRefCount(pRet); -} - -static -int registerTokenizer( - sqlite3 *db, - char *zName, - const sqlite3_tokenizer_module *p -){ - int rc; - sqlite3_stmt *pStmt; - const char zSql[] = "SELECT fts3_tokenizer(?, ?)"; - - rc = sqlite3_prepare_v2(db, zSql, -1, &pStmt, 0); - if( rc!=SQLITE_OK ){ - return rc; - } - - sqlite3_bind_text(pStmt, 1, zName, -1, SQLITE_STATIC); - sqlite3_bind_blob(pStmt, 2, &p, sizeof(p), SQLITE_STATIC); - sqlite3_step(pStmt); - - return sqlite3_finalize(pStmt); -} - -static -int queryTokenizer( - sqlite3 *db, - char *zName, - const sqlite3_tokenizer_module **pp -){ - int rc; - sqlite3_stmt *pStmt; - const char zSql[] = "SELECT fts3_tokenizer(?)"; - - *pp = 0; - rc = sqlite3_prepare_v2(db, zSql, -1, &pStmt, 0); - if( rc!=SQLITE_OK ){ - return rc; - } - - sqlite3_bind_text(pStmt, 1, zName, -1, SQLITE_STATIC); - if( SQLITE_ROW==sqlite3_step(pStmt) ){ - if( sqlite3_column_type(pStmt, 0)==SQLITE_BLOB ){ - memcpy((void *)pp, sqlite3_column_blob(pStmt, 0), sizeof(*pp)); - } - } - - return sqlite3_finalize(pStmt); -} - -void sqlite3Fts3SimpleTokenizerModule(sqlite3_tokenizer_module const**ppModule); - -/* -** Implementation of the scalar function fts3_tokenizer_internal_test(). -** This function is used for testing only, it is not included in the -** build unless SQLITE_TEST is defined. -** -** The purpose of this is to test that the fts3_tokenizer() function -** can be used as designed by the C-code in the queryTokenizer and -** registerTokenizer() functions above. These two functions are repeated -** in the README.tokenizer file as an example, so it is important to -** test them. -** -** To run the tests, evaluate the fts3_tokenizer_internal_test() scalar -** function with no arguments. An assert() will fail if a problem is -** detected. i.e.: -** -** SELECT fts3_tokenizer_internal_test(); -** -*/ -static void intTestFunc( - sqlite3_context *context, - int argc, - sqlite3_value **argv -){ - int rc; - const sqlite3_tokenizer_module *p1; - const sqlite3_tokenizer_module *p2; - sqlite3 *db = (sqlite3 *)sqlite3_user_data(context); - - UNUSED_PARAMETER(argc); - UNUSED_PARAMETER(argv); - - /* Test the query function */ - sqlite3Fts3SimpleTokenizerModule(&p1); - rc = queryTokenizer(db, "simple", &p2); - assert( rc==SQLITE_OK ); - assert( p1==p2 ); - rc = queryTokenizer(db, "nosuchtokenizer", &p2); - assert( rc==SQLITE_ERROR ); - assert( p2==0 ); - assert( 0==strcmp(sqlite3_errmsg(db), "unknown tokenizer: nosuchtokenizer") ); - - /* Test the storage function */ - rc = registerTokenizer(db, "nosuchtokenizer", p1); - assert( rc==SQLITE_OK ); - rc = queryTokenizer(db, "nosuchtokenizer", &p2); - assert( rc==SQLITE_OK ); - assert( p2==p1 ); - - sqlite3_result_text(context, "ok", -1, SQLITE_STATIC); -} - -#endif - -/* -** Set up SQL objects in database db used to access the contents of -** the hash table pointed to by argument pHash. The hash table must -** been initialised to use string keys, and to take a private copy -** of the key when a value is inserted. i.e. by a call similar to: -** -** sqlite3Fts3HashInit(pHash, FTS3_HASH_STRING, 1); -** -** This function adds a scalar function (see header comment above -** scalarFunc() in this file for details) and, if ENABLE_TABLE is -** defined at compilation time, a temporary virtual table (see header -** comment above struct HashTableVtab) to the database schema. Both -** provide read/write access to the contents of *pHash. -** -** The third argument to this function, zName, is used as the name -** of both the scalar and, if created, the virtual table. -*/ -int sqlite3Fts3InitHashTable( - sqlite3 *db, - Fts3Hash *pHash, - const char *zName -){ - int rc = SQLITE_OK; - void *p = (void *)pHash; - const int any = SQLITE_ANY; - -#ifdef SQLITE_TEST - char *zTest = 0; - char *zTest2 = 0; - void *pdb = (void *)db; - zTest = sqlite3_mprintf("%s_test", zName); - zTest2 = sqlite3_mprintf("%s_internal_test", zName); - if( !zTest || !zTest2 ){ - rc = SQLITE_NOMEM; - } -#endif - - if( SQLITE_OK==rc ){ - rc = sqlite3_create_function(db, zName, 1, any, p, scalarFunc, 0, 0); - } - if( SQLITE_OK==rc ){ - rc = sqlite3_create_function(db, zName, 2, any, p, scalarFunc, 0, 0); - } -#ifdef SQLITE_TEST - if( SQLITE_OK==rc ){ - rc = sqlite3_create_function(db, zTest, -1, any, p, testFunc, 0, 0); - } - if( SQLITE_OK==rc ){ - rc = sqlite3_create_function(db, zTest2, 0, any, pdb, intTestFunc, 0, 0); - } -#endif - -#ifdef SQLITE_TEST - sqlite3_free(zTest); - sqlite3_free(zTest2); -#endif - - return rc; -} - -#endif /* !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3) */ diff --git a/src/libtracker-fts/fts3_tokenizer.h b/src/libtracker-fts/fts3_tokenizer.h deleted file mode 100644 index c91c7ed79..000000000 --- a/src/libtracker-fts/fts3_tokenizer.h +++ /dev/null @@ -1,161 +0,0 @@ -/* -** 2006 July 10 -** -** The author disclaims copyright to this source code. -** -************************************************************************* -** Defines the interface to tokenizers used by fulltext-search. There -** are three basic components: -** -** sqlite3_tokenizer_module is a singleton defining the tokenizer -** interface functions. This is essentially the class structure for -** tokenizers. -** -** sqlite3_tokenizer is used to define a particular tokenizer, perhaps -** including customization information defined at creation time. -** -** sqlite3_tokenizer_cursor is generated by a tokenizer to generate -** tokens from a particular input. -*/ -#ifndef _FTS3_TOKENIZER_H_ -#define _FTS3_TOKENIZER_H_ - -/* TODO(shess) Only used for SQLITE_OK and SQLITE_DONE at this time. -** If tokenizers are to be allowed to call sqlite3_*() functions, then -** we will need a way to register the API consistently. -*/ -#include "sqlite3.h" - -/* -** Structures used by the tokenizer interface. When a new tokenizer -** implementation is registered, the caller provides a pointer to -** an sqlite3_tokenizer_module containing pointers to the callback -** functions that make up an implementation. -** -** When an fts3 table is created, it passes any arguments passed to -** the tokenizer clause of the CREATE VIRTUAL TABLE statement to the -** sqlite3_tokenizer_module.xCreate() function of the requested tokenizer -** implementation. The xCreate() function in turn returns an -** sqlite3_tokenizer structure representing the specific tokenizer to -** be used for the fts3 table (customized by the tokenizer clause arguments). -** -** To tokenize an input buffer, the sqlite3_tokenizer_module.xOpen() -** method is called. It returns an sqlite3_tokenizer_cursor object -** that may be used to tokenize a specific input buffer based on -** the tokenization rules supplied by a specific sqlite3_tokenizer -** object. -*/ -typedef struct sqlite3_tokenizer_module sqlite3_tokenizer_module; -typedef struct sqlite3_tokenizer sqlite3_tokenizer; -typedef struct sqlite3_tokenizer_cursor sqlite3_tokenizer_cursor; - -struct sqlite3_tokenizer_module { - - /* - ** Structure version. Should always be set to 0 or 1. - */ - int iVersion; - - /* - ** Create a new tokenizer. The values in the argv[] array are the - ** arguments passed to the "tokenizer" clause of the CREATE VIRTUAL - ** TABLE statement that created the fts3 table. For example, if - ** the following SQL is executed: - ** - ** CREATE .. USING fts3( ... , tokenizer <tokenizer-name> arg1 arg2) - ** - ** then argc is set to 2, and the argv[] array contains pointers - ** to the strings "arg1" and "arg2". - ** - ** This method should return either SQLITE_OK (0), or an SQLite error - ** code. If SQLITE_OK is returned, then *ppTokenizer should be set - ** to point at the newly created tokenizer structure. The generic - ** sqlite3_tokenizer.pModule variable should not be initialised by - ** this callback. The caller will do so. - */ - int (*xCreate)( - int argc, /* Size of argv array */ - const char *const*argv, /* Tokenizer argument strings */ - sqlite3_tokenizer **ppTokenizer /* OUT: Created tokenizer */ - ); - - /* - ** Destroy an existing tokenizer. The fts3 module calls this method - ** exactly once for each successful call to xCreate(). - */ - int (*xDestroy)(sqlite3_tokenizer *pTokenizer); - - /* - ** Create a tokenizer cursor to tokenize an input buffer. The caller - ** is responsible for ensuring that the input buffer remains valid - ** until the cursor is closed (using the xClose() method). - */ - int (*xOpen)( - sqlite3_tokenizer *pTokenizer, /* Tokenizer object */ - const char *pInput, int nBytes, /* Input buffer */ - sqlite3_tokenizer_cursor **ppCursor /* OUT: Created tokenizer cursor */ - ); - - /* - ** Destroy an existing tokenizer cursor. The fts3 module calls this - ** method exactly once for each successful call to xOpen(). - */ - int (*xClose)(sqlite3_tokenizer_cursor *pCursor); - - /* - ** Retrieve the next token from the tokenizer cursor pCursor. This - ** method should either return SQLITE_OK and set the values of the - ** "OUT" variables identified below, or SQLITE_DONE to indicate that - ** the end of the buffer has been reached, or an SQLite error code. - ** - ** *ppToken should be set to point at a buffer containing the - ** normalized version of the token (i.e. after any case-folding and/or - ** stemming has been performed). *pnBytes should be set to the length - ** of this buffer in bytes. The input text that generated the token is - ** identified by the byte offsets returned in *piStartOffset and - ** *piEndOffset. *piStartOffset should be set to the index of the first - ** byte of the token in the input buffer. *piEndOffset should be set - ** to the index of the first byte just past the end of the token in - ** the input buffer. - ** - ** The buffer *ppToken is set to point at is managed by the tokenizer - ** implementation. It is only required to be valid until the next call - ** to xNext() or xClose(). - */ - /* TODO(shess) current implementation requires pInput to be - ** nul-terminated. This should either be fixed, or pInput/nBytes - ** should be converted to zInput. - */ - int (*xNext)( - sqlite3_tokenizer_cursor *pCursor, /* Tokenizer cursor */ - const char **ppToken, int *pnBytes, /* OUT: Normalized text for token */ - int *piStartOffset, /* OUT: Byte offset of token in input buffer */ - int *piEndOffset, /* OUT: Byte offset of end of token in input buffer */ - int *piPosition /* OUT: Number of tokens returned before this one */ - ); - - /*********************************************************************** - ** Methods below this point are only available if iVersion>=1. - */ - - /* - ** Configure the language id of a tokenizer cursor. - */ - int (*xLanguageid)(sqlite3_tokenizer_cursor *pCsr, int iLangid); -}; - -struct sqlite3_tokenizer { - const sqlite3_tokenizer_module *pModule; /* The module for this tokenizer */ - /* Tokenizer implementations will typically add additional fields */ -}; - -struct sqlite3_tokenizer_cursor { - sqlite3_tokenizer *pTokenizer; /* Tokenizer for this cursor. */ - /* Tokenizer implementations will typically add additional fields */ -}; - -int fts3_global_term_cnt(int iTerm, int iCol); -int fts3_term_cnt(int iTerm, int iCol); - - -#endif /* _FTS3_TOKENIZER_H_ */ diff --git a/src/libtracker-fts/fts3_tokenizer1.c b/src/libtracker-fts/fts3_tokenizer1.c deleted file mode 100644 index deea06d92..000000000 --- a/src/libtracker-fts/fts3_tokenizer1.c +++ /dev/null @@ -1,234 +0,0 @@ -/* -** 2006 Oct 10 -** -** The author disclaims copyright to this source code. In place of -** a legal notice, here is a blessing: -** -** May you do good and not evil. -** May you find forgiveness for yourself and forgive others. -** May you share freely, never taking more than you give. -** -****************************************************************************** -** -** Implementation of the "simple" full-text-search tokenizer. -*/ - -/* -** The code in this file is only compiled if: -** -** * The FTS3 module is being built as an extension -** (in which case SQLITE_CORE is not defined), or -** -** * The FTS3 module is being built into the core of -** SQLite (in which case SQLITE_ENABLE_FTS3 is defined). -*/ -#include "fts3Int.h" -#if !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3) - -#include <assert.h> -#include <stdlib.h> -#include <stdio.h> -#include <string.h> - -#include "fts3_tokenizer.h" - -typedef struct simple_tokenizer { - sqlite3_tokenizer base; - char delim[128]; /* flag ASCII delimiters */ -} simple_tokenizer; - -typedef struct simple_tokenizer_cursor { - sqlite3_tokenizer_cursor base; - const char *pInput; /* input we are tokenizing */ - int nBytes; /* size of the input */ - int iOffset; /* current position in pInput */ - int iToken; /* index of next token to be returned */ - char *pToken; /* storage for current token */ - int nTokenAllocated; /* space allocated to zToken buffer */ -} simple_tokenizer_cursor; - - -static int simpleDelim(simple_tokenizer *t, unsigned char c){ - return c<0x80 && t->delim[c]; -} -static int fts3_isalnum(int x){ - return (x>='0' && x<='9') || (x>='A' && x<='Z') || (x>='a' && x<='z'); -} - -/* -** Create a new tokenizer instance. -*/ -static int simpleCreate( - int argc, const char * const *argv, - sqlite3_tokenizer **ppTokenizer -){ - simple_tokenizer *t; - - t = (simple_tokenizer *) sqlite3_malloc(sizeof(*t)); - if( t==NULL ) return SQLITE_NOMEM; - memset(t, 0, sizeof(*t)); - - /* TODO(shess) Delimiters need to remain the same from run to run, - ** else we need to reindex. One solution would be a meta-table to - ** track such information in the database, then we'd only want this - ** information on the initial create. - */ - if( argc>1 ){ - int i, n = (int)strlen(argv[1]); - for(i=0; i<n; i++){ - unsigned char ch = argv[1][i]; - /* We explicitly don't support UTF-8 delimiters for now. */ - if( ch>=0x80 ){ - sqlite3_free(t); - return SQLITE_ERROR; - } - t->delim[ch] = 1; - } - } else { - /* Mark non-alphanumeric ASCII characters as delimiters */ - int i; - for(i=1; i<0x80; i++){ - t->delim[i] = !fts3_isalnum(i) ? -1 : 0; - } - } - - *ppTokenizer = &t->base; - return SQLITE_OK; -} - -/* -** Destroy a tokenizer -*/ -static int simpleDestroy(sqlite3_tokenizer *pTokenizer){ - sqlite3_free(pTokenizer); - return SQLITE_OK; -} - -/* -** Prepare to begin tokenizing a particular string. The input -** string to be tokenized is pInput[0..nBytes-1]. A cursor -** used to incrementally tokenize this string is returned in -** *ppCursor. -*/ -static int simpleOpen( - sqlite3_tokenizer *pTokenizer, /* The tokenizer */ - const char *pInput, int nBytes, /* String to be tokenized */ - sqlite3_tokenizer_cursor **ppCursor /* OUT: Tokenization cursor */ -){ - simple_tokenizer_cursor *c; - - UNUSED_PARAMETER(pTokenizer); - - c = (simple_tokenizer_cursor *) sqlite3_malloc(sizeof(*c)); - if( c==NULL ) return SQLITE_NOMEM; - - c->pInput = pInput; - if( pInput==0 ){ - c->nBytes = 0; - }else if( nBytes<0 ){ - c->nBytes = (int)strlen(pInput); - }else{ - c->nBytes = nBytes; - } - c->iOffset = 0; /* start tokenizing at the beginning */ - c->iToken = 0; - c->pToken = NULL; /* no space allocated, yet. */ - c->nTokenAllocated = 0; - - *ppCursor = &c->base; - return SQLITE_OK; -} - -/* -** Close a tokenization cursor previously opened by a call to -** simpleOpen() above. -*/ -static int simpleClose(sqlite3_tokenizer_cursor *pCursor){ - simple_tokenizer_cursor *c = (simple_tokenizer_cursor *) pCursor; - sqlite3_free(c->pToken); - sqlite3_free(c); - return SQLITE_OK; -} - -/* -** Extract the next token from a tokenization cursor. The cursor must -** have been opened by a prior call to simpleOpen(). -*/ -static int simpleNext( - sqlite3_tokenizer_cursor *pCursor, /* Cursor returned by simpleOpen */ - const char **ppToken, /* OUT: *ppToken is the token text */ - int *pnBytes, /* OUT: Number of bytes in token */ - int *piStartOffset, /* OUT: Starting offset of token */ - int *piEndOffset, /* OUT: Ending offset of token */ - int *piPosition /* OUT: Position integer of token */ -){ - simple_tokenizer_cursor *c = (simple_tokenizer_cursor *) pCursor; - simple_tokenizer *t = (simple_tokenizer *) pCursor->pTokenizer; - unsigned char *p = (unsigned char *)c->pInput; - - while( c->iOffset<c->nBytes ){ - int iStartOffset; - - /* Scan past delimiter characters */ - while( c->iOffset<c->nBytes && simpleDelim(t, p[c->iOffset]) ){ - c->iOffset++; - } - - /* Count non-delimiter characters. */ - iStartOffset = c->iOffset; - while( c->iOffset<c->nBytes && !simpleDelim(t, p[c->iOffset]) ){ - c->iOffset++; - } - - if( c->iOffset>iStartOffset ){ - int i, n = c->iOffset-iStartOffset; - if( n>c->nTokenAllocated ){ - char *pNew; - c->nTokenAllocated = n+20; - pNew = sqlite3_realloc(c->pToken, c->nTokenAllocated); - if( !pNew ) return SQLITE_NOMEM; - c->pToken = pNew; - } - for(i=0; i<n; i++){ - /* TODO(shess) This needs expansion to handle UTF-8 - ** case-insensitivity. - */ - unsigned char ch = p[iStartOffset+i]; - c->pToken[i] = (char)((ch>='A' && ch<='Z') ? ch-'A'+'a' : ch); - } - *ppToken = c->pToken; - *pnBytes = n; - *piStartOffset = iStartOffset; - *piEndOffset = c->iOffset; - *piPosition = c->iToken++; - - return SQLITE_OK; - } - } - return SQLITE_DONE; -} - -/* -** The set of routines that implement the simple tokenizer -*/ -static const sqlite3_tokenizer_module simpleTokenizerModule = { - 0, - simpleCreate, - simpleDestroy, - simpleOpen, - simpleClose, - simpleNext, - 0, -}; - -/* -** Allocate a new simple tokenizer. Return a pointer to the new -** tokenizer in *ppModule -*/ -void sqlite3Fts3SimpleTokenizerModule( - sqlite3_tokenizer_module const**ppModule -){ - *ppModule = &simpleTokenizerModule; -} - -#endif /* !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3) */ diff --git a/src/libtracker-fts/fts3_unicode.c b/src/libtracker-fts/fts3_unicode.c deleted file mode 100644 index 79941edbb..000000000 --- a/src/libtracker-fts/fts3_unicode.c +++ /dev/null @@ -1,393 +0,0 @@ -/* -** 2012 May 24 -** -** The author disclaims copyright to this source code. In place of -** a legal notice, here is a blessing: -** -** May you do good and not evil. -** May you find forgiveness for yourself and forgive others. -** May you share freely, never taking more than you give. -** -****************************************************************************** -** -** Implementation of the "unicode" full-text-search tokenizer. -*/ - -#ifdef SQLITE_ENABLE_FTS4_UNICODE61 - -#include "fts3Int.h" -#if !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3) - -#include <assert.h> -#include <stdlib.h> -#include <stdio.h> -#include <string.h> - -#include "fts3_tokenizer.h" - -/* -** The following two macros - READ_UTF8 and WRITE_UTF8 - have been copied -** from the sqlite3 source file utf.c. If this file is compiled as part -** of the amalgamation, they are not required. -*/ -#ifndef SQLITE_AMALGAMATION - -static const unsigned char sqlite3Utf8Trans1[] = { - 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, - 0x08, 0x09, 0x0a, 0x0b, 0x0c, 0x0d, 0x0e, 0x0f, - 0x10, 0x11, 0x12, 0x13, 0x14, 0x15, 0x16, 0x17, - 0x18, 0x19, 0x1a, 0x1b, 0x1c, 0x1d, 0x1e, 0x1f, - 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, - 0x08, 0x09, 0x0a, 0x0b, 0x0c, 0x0d, 0x0e, 0x0f, - 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, - 0x00, 0x01, 0x02, 0x03, 0x00, 0x01, 0x00, 0x00, -}; - -#define READ_UTF8(zIn, zTerm, c) \ - c = *(zIn++); \ - if( c>=0xc0 ){ \ - c = sqlite3Utf8Trans1[c-0xc0]; \ - while( zIn!=zTerm && (*zIn & 0xc0)==0x80 ){ \ - c = (c<<6) + (0x3f & *(zIn++)); \ - } \ - if( c<0x80 \ - || (c&0xFFFFF800)==0xD800 \ - || (c&0xFFFFFFFE)==0xFFFE ){ c = 0xFFFD; } \ - } - -#define WRITE_UTF8(zOut, c) { \ - if( c<0x00080 ){ \ - *zOut++ = (u8)(c&0xFF); \ - } \ - else if( c<0x00800 ){ \ - *zOut++ = 0xC0 + (u8)((c>>6)&0x1F); \ - *zOut++ = 0x80 + (u8)(c & 0x3F); \ - } \ - else if( c<0x10000 ){ \ - *zOut++ = 0xE0 + (u8)((c>>12)&0x0F); \ - *zOut++ = 0x80 + (u8)((c>>6) & 0x3F); \ - *zOut++ = 0x80 + (u8)(c & 0x3F); \ - }else{ \ - *zOut++ = 0xF0 + (u8)((c>>18) & 0x07); \ - *zOut++ = 0x80 + (u8)((c>>12) & 0x3F); \ - *zOut++ = 0x80 + (u8)((c>>6) & 0x3F); \ - *zOut++ = 0x80 + (u8)(c & 0x3F); \ - } \ -} - -#endif /* ifndef SQLITE_AMALGAMATION */ - -typedef struct unicode_tokenizer unicode_tokenizer; -typedef struct unicode_cursor unicode_cursor; - -struct unicode_tokenizer { - sqlite3_tokenizer base; - int bRemoveDiacritic; - int nException; - int *aiException; -}; - -struct unicode_cursor { - sqlite3_tokenizer_cursor base; - const unsigned char *aInput; /* Input text being tokenized */ - int nInput; /* Size of aInput[] in bytes */ - int iOff; /* Current offset within aInput[] */ - int iToken; /* Index of next token to be returned */ - char *zToken; /* storage for current token */ - int nAlloc; /* space allocated at zToken */ -}; - - -/* -** Destroy a tokenizer allocated by unicodeCreate(). -*/ -static int unicodeDestroy(sqlite3_tokenizer *pTokenizer){ - if( pTokenizer ){ - unicode_tokenizer *p = (unicode_tokenizer *)pTokenizer; - sqlite3_free(p->aiException); - sqlite3_free(p); - } - return SQLITE_OK; -} - -/* -** As part of a tokenchars= or separators= option, the CREATE VIRTUAL TABLE -** statement has specified that the tokenizer for this table shall consider -** all characters in string zIn/nIn to be separators (if bAlnum==0) or -** token characters (if bAlnum==1). -** -** For each codepoint in the zIn/nIn string, this function checks if the -** sqlite3FtsUnicodeIsalnum() function already returns the desired result. -** If so, no action is taken. Otherwise, the codepoint is added to the -** unicode_tokenizer.aiException[] array. For the purposes of tokenization, -** the return value of sqlite3FtsUnicodeIsalnum() is inverted for all -** codepoints in the aiException[] array. -** -** If a standalone diacritic mark (one that sqlite3FtsUnicodeIsdiacritic() -** identifies as a diacritic) occurs in the zIn/nIn string it is ignored. -** It is not possible to change the behaviour of the tokenizer with respect -** to these codepoints. -*/ -static int unicodeAddExceptions( - unicode_tokenizer *p, /* Tokenizer to add exceptions to */ - int bAlnum, /* Replace Isalnum() return value with this */ - const char *zIn, /* Array of characters to make exceptions */ - int nIn /* Length of z in bytes */ -){ - const unsigned char *z = (const unsigned char *)zIn; - const unsigned char *zTerm = &z[nIn]; - int iCode; - int nEntry = 0; - - assert( bAlnum==0 || bAlnum==1 ); - - while( z<zTerm ){ - READ_UTF8(z, zTerm, iCode); - assert( (sqlite3FtsUnicodeIsalnum(iCode) & 0xFFFFFFFE)==0 ); - if( sqlite3FtsUnicodeIsalnum(iCode)!=bAlnum - && sqlite3FtsUnicodeIsdiacritic(iCode)==0 - ){ - nEntry++; - } - } - - if( nEntry ){ - int *aNew; /* New aiException[] array */ - int nNew; /* Number of valid entries in array aNew[] */ - - aNew = sqlite3_realloc(p->aiException, (p->nException+nEntry)*sizeof(int)); - if( aNew==0 ) return SQLITE_NOMEM; - nNew = p->nException; - - z = (const unsigned char *)zIn; - while( z<zTerm ){ - READ_UTF8(z, zTerm, iCode); - if( sqlite3FtsUnicodeIsalnum(iCode)!=bAlnum - && sqlite3FtsUnicodeIsdiacritic(iCode)==0 - ){ - int i, j; - for(i=0; i<nNew && aNew[i]<iCode; i++); - for(j=nNew; j>i; j--) aNew[j] = aNew[j-1]; - aNew[i] = iCode; - nNew++; - } - } - p->aiException = aNew; - p->nException = nNew; - } - - return SQLITE_OK; -} - -/* -** Return true if the p->aiException[] array contains the value iCode. -*/ -static int unicodeIsException(unicode_tokenizer *p, int iCode){ - if( p->nException>0 ){ - int *a = p->aiException; - int iLo = 0; - int iHi = p->nException-1; - - while( iHi>=iLo ){ - int iTest = (iHi + iLo) / 2; - if( iCode==a[iTest] ){ - return 1; - }else if( iCode>a[iTest] ){ - iLo = iTest+1; - }else{ - iHi = iTest-1; - } - } - } - - return 0; -} - -/* -** Return true if, for the purposes of tokenization, codepoint iCode is -** considered a token character (not a separator). -*/ -static int unicodeIsAlnum(unicode_tokenizer *p, int iCode){ - assert( (sqlite3FtsUnicodeIsalnum(iCode) & 0xFFFFFFFE)==0 ); - return sqlite3FtsUnicodeIsalnum(iCode) ^ unicodeIsException(p, iCode); -} - -/* -** Create a new tokenizer instance. -*/ -static int unicodeCreate( - int nArg, /* Size of array argv[] */ - const char * const *azArg, /* Tokenizer creation arguments */ - sqlite3_tokenizer **pp /* OUT: New tokenizer handle */ -){ - unicode_tokenizer *pNew; /* New tokenizer object */ - int i; - int rc = SQLITE_OK; - - pNew = (unicode_tokenizer *) sqlite3_malloc(sizeof(unicode_tokenizer)); - if( pNew==NULL ) return SQLITE_NOMEM; - memset(pNew, 0, sizeof(unicode_tokenizer)); - pNew->bRemoveDiacritic = 1; - - for(i=0; rc==SQLITE_OK && i<nArg; i++){ - const char *z = azArg[i]; - int n = strlen(z); - - if( n==19 && memcmp("remove_diacritics=1", z, 19)==0 ){ - pNew->bRemoveDiacritic = 1; - } - else if( n==19 && memcmp("remove_diacritics=0", z, 19)==0 ){ - pNew->bRemoveDiacritic = 0; - } - else if( n>=11 && memcmp("tokenchars=", z, 11)==0 ){ - rc = unicodeAddExceptions(pNew, 1, &z[11], n-11); - } - else if( n>=11 && memcmp("separators=", z, 11)==0 ){ - rc = unicodeAddExceptions(pNew, 0, &z[11], n-11); - } - else{ - /* Unrecognized argument */ - rc = SQLITE_ERROR; - } - } - - if( rc!=SQLITE_OK ){ - unicodeDestroy((sqlite3_tokenizer *)pNew); - pNew = 0; - } - *pp = (sqlite3_tokenizer *)pNew; - return rc; -} - -/* -** Prepare to begin tokenizing a particular string. The input -** string to be tokenized is pInput[0..nBytes-1]. A cursor -** used to incrementally tokenize this string is returned in -** *ppCursor. -*/ -static int unicodeOpen( - sqlite3_tokenizer *p, /* The tokenizer */ - const char *aInput, /* Input string */ - int nInput, /* Size of string aInput in bytes */ - sqlite3_tokenizer_cursor **pp /* OUT: New cursor object */ -){ - unicode_cursor *pCsr; - - pCsr = (unicode_cursor *)sqlite3_malloc(sizeof(unicode_cursor)); - if( pCsr==0 ){ - return SQLITE_NOMEM; - } - memset(pCsr, 0, sizeof(unicode_cursor)); - - pCsr->aInput = (const unsigned char *)aInput; - if( aInput==0 ){ - pCsr->nInput = 0; - }else if( nInput<0 ){ - pCsr->nInput = (int)strlen(aInput); - }else{ - pCsr->nInput = nInput; - } - - *pp = &pCsr->base; - UNUSED_PARAMETER(p); - return SQLITE_OK; -} - -/* -** Close a tokenization cursor previously opened by a call to -** simpleOpen() above. -*/ -static int unicodeClose(sqlite3_tokenizer_cursor *pCursor){ - unicode_cursor *pCsr = (unicode_cursor *) pCursor; - sqlite3_free(pCsr->zToken); - sqlite3_free(pCsr); - return SQLITE_OK; -} - -/* -** Extract the next token from a tokenization cursor. The cursor must -** have been opened by a prior call to simpleOpen(). -*/ -static int unicodeNext( - sqlite3_tokenizer_cursor *pC, /* Cursor returned by simpleOpen */ - const char **paToken, /* OUT: Token text */ - int *pnToken, /* OUT: Number of bytes at *paToken */ - int *piStart, /* OUT: Starting offset of token */ - int *piEnd, /* OUT: Ending offset of token */ - int *piPos /* OUT: Position integer of token */ -){ - unicode_cursor *pCsr = (unicode_cursor *)pC; - unicode_tokenizer *p = ((unicode_tokenizer *)pCsr->base.pTokenizer); - int iCode; - char *zOut; - const unsigned char *z = &pCsr->aInput[pCsr->iOff]; - const unsigned char *zStart = z; - const unsigned char *zEnd; - const unsigned char *zTerm = &pCsr->aInput[pCsr->nInput]; - - /* Scan past any delimiter characters before the start of the next token. - ** Return SQLITE_DONE early if this takes us all the way to the end of - ** the input. */ - while( z<zTerm ){ - READ_UTF8(z, zTerm, iCode); - if( unicodeIsAlnum(p, iCode) ) break; - zStart = z; - } - if( zStart>=zTerm ) return SQLITE_DONE; - - zOut = pCsr->zToken; - do { - int iOut; - - /* Grow the output buffer if required. */ - if( (zOut-pCsr->zToken)>=(pCsr->nAlloc-4) ){ - char *zNew = sqlite3_realloc(pCsr->zToken, pCsr->nAlloc+64); - if( !zNew ) return SQLITE_NOMEM; - zOut = &zNew[zOut - pCsr->zToken]; - pCsr->zToken = zNew; - pCsr->nAlloc += 64; - } - - /* Write the folded case of the last character read to the output */ - zEnd = z; - iOut = sqlite3FtsUnicodeFold(iCode, p->bRemoveDiacritic); - if( iOut ){ - WRITE_UTF8(zOut, iOut); - } - - /* If the cursor is not at EOF, read the next character */ - if( z>=zTerm ) break; - READ_UTF8(z, zTerm, iCode); - }while( unicodeIsAlnum(p, iCode) - || sqlite3FtsUnicodeIsdiacritic(iCode) - ); - - /* Set the output variables and return. */ - pCsr->iOff = (z - pCsr->aInput); - *paToken = pCsr->zToken; - *pnToken = zOut - pCsr->zToken; - *piStart = (zStart - pCsr->aInput); - *piEnd = (zEnd - pCsr->aInput); - *piPos = pCsr->iToken++; - return SQLITE_OK; -} - -/* -** Set *ppModule to a pointer to the sqlite3_tokenizer_module -** structure for the unicode tokenizer. -*/ -void sqlite3Fts3UnicodeTokenizer(sqlite3_tokenizer_module const **ppModule){ - static const sqlite3_tokenizer_module module = { - 0, - unicodeCreate, - unicodeDestroy, - unicodeOpen, - unicodeClose, - unicodeNext, - 0, - }; - *ppModule = &module; -} - -#endif /* !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3) */ -#endif /* ifndef SQLITE_ENABLE_FTS4_UNICODE61 */ diff --git a/src/libtracker-fts/fts3_unicode2.c b/src/libtracker-fts/fts3_unicode2.c deleted file mode 100644 index 3c2456902..000000000 --- a/src/libtracker-fts/fts3_unicode2.c +++ /dev/null @@ -1,366 +0,0 @@ -/* -** 2012 May 25 -** -** The author disclaims copyright to this source code. In place of -** a legal notice, here is a blessing: -** -** May you do good and not evil. -** May you find forgiveness for yourself and forgive others. -** May you share freely, never taking more than you give. -** -****************************************************************************** -*/ - -/* -** DO NOT EDIT THIS MACHINE GENERATED FILE. -*/ - -#if defined(SQLITE_ENABLE_FTS4_UNICODE61) -#if defined(SQLITE_ENABLE_FTS3) || defined(SQLITE_ENABLE_FTS4) - -#include <assert.h> - -/* -** Return true if the argument corresponds to a unicode codepoint -** classified as either a letter or a number. Otherwise false. -** -** The results are undefined if the value passed to this function -** is less than zero. -*/ -int sqlite3FtsUnicodeIsalnum(int c){ - /* Each unsigned integer in the following array corresponds to a contiguous - ** range of unicode codepoints that are not either letters or numbers (i.e. - ** codepoints for which this function should return 0). - ** - ** The most significant 22 bits in each 32-bit value contain the first - ** codepoint in the range. The least significant 10 bits are used to store - ** the size of the range (always at least 1). In other words, the value - ** ((C<<22) + N) represents a range of N codepoints starting with codepoint - ** C. It is not possible to represent a range larger than 1023 codepoints - ** using this format. - */ - const static unsigned int aEntry[] = { - 0x00000030, 0x0000E807, 0x00016C06, 0x0001EC2F, 0x0002AC07, - 0x0002D001, 0x0002D803, 0x0002EC01, 0x0002FC01, 0x00035C01, - 0x0003DC01, 0x000B0804, 0x000B480E, 0x000B9407, 0x000BB401, - 0x000BBC81, 0x000DD401, 0x000DF801, 0x000E1002, 0x000E1C01, - 0x000FD801, 0x00120808, 0x00156806, 0x00162402, 0x00163C01, - 0x00164437, 0x0017CC02, 0x00180005, 0x00181816, 0x00187802, - 0x00192C15, 0x0019A804, 0x0019C001, 0x001B5001, 0x001B580F, - 0x001B9C07, 0x001BF402, 0x001C000E, 0x001C3C01, 0x001C4401, - 0x001CC01B, 0x001E980B, 0x001FAC09, 0x001FD804, 0x00205804, - 0x00206C09, 0x00209403, 0x0020A405, 0x0020C00F, 0x00216403, - 0x00217801, 0x0023901B, 0x00240004, 0x0024E803, 0x0024F812, - 0x00254407, 0x00258804, 0x0025C001, 0x00260403, 0x0026F001, - 0x0026F807, 0x00271C02, 0x00272C03, 0x00275C01, 0x00278802, - 0x0027C802, 0x0027E802, 0x00280403, 0x0028F001, 0x0028F805, - 0x00291C02, 0x00292C03, 0x00294401, 0x0029C002, 0x0029D401, - 0x002A0403, 0x002AF001, 0x002AF808, 0x002B1C03, 0x002B2C03, - 0x002B8802, 0x002BC002, 0x002C0403, 0x002CF001, 0x002CF807, - 0x002D1C02, 0x002D2C03, 0x002D5802, 0x002D8802, 0x002DC001, - 0x002E0801, 0x002EF805, 0x002F1803, 0x002F2804, 0x002F5C01, - 0x002FCC08, 0x00300403, 0x0030F807, 0x00311803, 0x00312804, - 0x00315402, 0x00318802, 0x0031FC01, 0x00320802, 0x0032F001, - 0x0032F807, 0x00331803, 0x00332804, 0x00335402, 0x00338802, - 0x00340802, 0x0034F807, 0x00351803, 0x00352804, 0x00355C01, - 0x00358802, 0x0035E401, 0x00360802, 0x00372801, 0x00373C06, - 0x00375801, 0x00376008, 0x0037C803, 0x0038C401, 0x0038D007, - 0x0038FC01, 0x00391C09, 0x00396802, 0x003AC401, 0x003AD006, - 0x003AEC02, 0x003B2006, 0x003C041F, 0x003CD00C, 0x003DC417, - 0x003E340B, 0x003E6424, 0x003EF80F, 0x003F380D, 0x0040AC14, - 0x00412806, 0x00415804, 0x00417803, 0x00418803, 0x00419C07, - 0x0041C404, 0x0042080C, 0x00423C01, 0x00426806, 0x0043EC01, - 0x004D740C, 0x004E400A, 0x00500001, 0x0059B402, 0x005A0001, - 0x005A6C02, 0x005BAC03, 0x005C4803, 0x005CC805, 0x005D4802, - 0x005DC802, 0x005ED023, 0x005F6004, 0x005F7401, 0x0060000F, - 0x0062A401, 0x0064800C, 0x0064C00C, 0x00650001, 0x00651002, - 0x0066C011, 0x00672002, 0x00677822, 0x00685C05, 0x00687802, - 0x0069540A, 0x0069801D, 0x0069FC01, 0x006A8007, 0x006AA006, - 0x006C0005, 0x006CD011, 0x006D6823, 0x006E0003, 0x006E840D, - 0x006F980E, 0x006FF004, 0x00709014, 0x0070EC05, 0x0071F802, - 0x00730008, 0x00734019, 0x0073B401, 0x0073C803, 0x00770027, - 0x0077F004, 0x007EF401, 0x007EFC03, 0x007F3403, 0x007F7403, - 0x007FB403, 0x007FF402, 0x00800065, 0x0081A806, 0x0081E805, - 0x00822805, 0x0082801A, 0x00834021, 0x00840002, 0x00840C04, - 0x00842002, 0x00845001, 0x00845803, 0x00847806, 0x00849401, - 0x00849C01, 0x0084A401, 0x0084B801, 0x0084E802, 0x00850005, - 0x00852804, 0x00853C01, 0x00864264, 0x00900027, 0x0091000B, - 0x0092704E, 0x00940200, 0x009C0475, 0x009E53B9, 0x00AD400A, - 0x00B39406, 0x00B3BC03, 0x00B3E404, 0x00B3F802, 0x00B5C001, - 0x00B5FC01, 0x00B7804F, 0x00B8C00C, 0x00BA001A, 0x00BA6C59, - 0x00BC00D6, 0x00BFC00C, 0x00C00005, 0x00C02019, 0x00C0A807, - 0x00C0D802, 0x00C0F403, 0x00C26404, 0x00C28001, 0x00C3EC01, - 0x00C64002, 0x00C6580A, 0x00C70024, 0x00C8001F, 0x00C8A81E, - 0x00C94001, 0x00C98020, 0x00CA2827, 0x00CB003F, 0x00CC0100, - 0x01370040, 0x02924037, 0x0293F802, 0x02983403, 0x0299BC10, - 0x029A7C01, 0x029BC008, 0x029C0017, 0x029C8002, 0x029E2402, - 0x02A00801, 0x02A01801, 0x02A02C01, 0x02A08C09, 0x02A0D804, - 0x02A1D004, 0x02A20002, 0x02A2D011, 0x02A33802, 0x02A38012, - 0x02A3E003, 0x02A4980A, 0x02A51C0D, 0x02A57C01, 0x02A60004, - 0x02A6CC1B, 0x02A77802, 0x02A8A40E, 0x02A90C01, 0x02A93002, - 0x02A97004, 0x02A9DC03, 0x02A9EC01, 0x02AAC001, 0x02AAC803, - 0x02AADC02, 0x02AAF802, 0x02AB0401, 0x02AB7802, 0x02ABAC07, - 0x02ABD402, 0x02AF8C0B, 0x03600001, 0x036DFC02, 0x036FFC02, - 0x037FFC02, 0x03E3FC01, 0x03EC7801, 0x03ECA401, 0x03EEC810, - 0x03F4F802, 0x03F7F002, 0x03F8001A, 0x03F88007, 0x03F8C023, - 0x03F95013, 0x03F9A004, 0x03FBFC01, 0x03FC040F, 0x03FC6807, - 0x03FCEC06, 0x03FD6C0B, 0x03FF8007, 0x03FFA007, 0x03FFE405, - 0x04040003, 0x0404DC09, 0x0405E411, 0x0406400C, 0x0407402E, - 0x040E7C01, 0x040F4001, 0x04215C01, 0x04247C01, 0x0424FC01, - 0x04280403, 0x04281402, 0x04283004, 0x0428E003, 0x0428FC01, - 0x04294009, 0x0429FC01, 0x042CE407, 0x04400003, 0x0440E016, - 0x04420003, 0x0442C012, 0x04440003, 0x04449C0E, 0x04450004, - 0x04460003, 0x0446CC0E, 0x04471404, 0x045AAC0D, 0x0491C004, - 0x05BD442E, 0x05BE3C04, 0x074000F6, 0x07440027, 0x0744A4B5, - 0x07480046, 0x074C0057, 0x075B0401, 0x075B6C01, 0x075BEC01, - 0x075C5401, 0x075CD401, 0x075D3C01, 0x075DBC01, 0x075E2401, - 0x075EA401, 0x075F0C01, 0x07BBC002, 0x07C0002C, 0x07C0C064, - 0x07C2800F, 0x07C2C40E, 0x07C3040F, 0x07C3440F, 0x07C4401F, - 0x07C4C03C, 0x07C5C02B, 0x07C7981D, 0x07C8402B, 0x07C90009, - 0x07C94002, 0x07CC0021, 0x07CCC006, 0x07CCDC46, 0x07CE0014, - 0x07CE8025, 0x07CF1805, 0x07CF8011, 0x07D0003F, 0x07D10001, - 0x07D108B6, 0x07D3E404, 0x07D4003E, 0x07D50004, 0x07D54018, - 0x07D7EC46, 0x07D9140B, 0x07DA0046, 0x07DC0074, 0x38000401, - 0x38008060, 0x380400F0, 0x3C000001, 0x3FFFF401, 0x40000001, - 0x43FFF401, - }; - static const unsigned int aAscii[4] = { - 0xFFFFFFFF, 0xFC00FFFF, 0xF8000001, 0xF8000001, - }; - - if( c<128 ){ - return ( (aAscii[c >> 5] & (1 << (c & 0x001F)))==0 ); - }else if( c<(1<<22) ){ - unsigned int key = (((unsigned int)c)<<10) | 0x000003FF; - int iRes; - int iHi = sizeof(aEntry)/sizeof(aEntry[0]) - 1; - int iLo = 0; - while( iHi>=iLo ){ - int iTest = (iHi + iLo) / 2; - if( key >= aEntry[iTest] ){ - iRes = iTest; - iLo = iTest+1; - }else{ - iHi = iTest-1; - } - } - assert( aEntry[0]<key ); - assert( key>=aEntry[iRes] ); - return (((unsigned int)c) >= ((aEntry[iRes]>>10) + (aEntry[iRes]&0x3FF))); - } - return 1; -} - - -/* -** If the argument is a codepoint corresponding to a lowercase letter -** in the ASCII range with a diacritic added, return the codepoint -** of the ASCII letter only. For example, if passed 235 - "LATIN -** SMALL LETTER E WITH DIAERESIS" - return 65 ("LATIN SMALL LETTER -** E"). The resuls of passing a codepoint that corresponds to an -** uppercase letter are undefined. -*/ -static int remove_diacritic(int c){ - unsigned short aDia[] = { - 0, 1797, 1848, 1859, 1891, 1928, 1940, 1995, - 2024, 2040, 2060, 2110, 2168, 2206, 2264, 2286, - 2344, 2383, 2472, 2488, 2516, 2596, 2668, 2732, - 2782, 2842, 2894, 2954, 2984, 3000, 3028, 3336, - 3456, 3696, 3712, 3728, 3744, 3896, 3912, 3928, - 3968, 4008, 4040, 4106, 4138, 4170, 4202, 4234, - 4266, 4296, 4312, 4344, 4408, 4424, 4472, 4504, - 6148, 6198, 6264, 6280, 6360, 6429, 6505, 6529, - 61448, 61468, 61534, 61592, 61642, 61688, 61704, 61726, - 61784, 61800, 61836, 61880, 61914, 61948, 61998, 62122, - 62154, 62200, 62218, 62302, 62364, 62442, 62478, 62536, - 62554, 62584, 62604, 62640, 62648, 62656, 62664, 62730, - 62924, 63050, 63082, 63274, 63390, - }; - char aChar[] = { - '\0', 'a', 'c', 'e', 'i', 'n', 'o', 'u', 'y', 'y', 'a', 'c', - 'd', 'e', 'e', 'g', 'h', 'i', 'j', 'k', 'l', 'n', 'o', 'r', - 's', 't', 'u', 'u', 'w', 'y', 'z', 'o', 'u', 'a', 'i', 'o', - 'u', 'g', 'k', 'o', 'j', 'g', 'n', 'a', 'e', 'i', 'o', 'r', - 'u', 's', 't', 'h', 'a', 'e', 'o', 'y', '\0', '\0', '\0', '\0', - '\0', '\0', '\0', '\0', 'a', 'b', 'd', 'd', 'e', 'f', 'g', 'h', - 'h', 'i', 'k', 'l', 'l', 'm', 'n', 'p', 'r', 'r', 's', 't', - 'u', 'v', 'w', 'w', 'x', 'y', 'z', 'h', 't', 'w', 'y', 'a', - 'e', 'i', 'o', 'u', 'y', - }; - - unsigned int key = (((unsigned int)c)<<3) | 0x00000007; - int iRes = 0; - int iHi = sizeof(aDia)/sizeof(aDia[0]) - 1; - int iLo = 0; - while( iHi>=iLo ){ - int iTest = (iHi + iLo) / 2; - if( key >= aDia[iTest] ){ - iRes = iTest; - iLo = iTest+1; - }else{ - iHi = iTest-1; - } - } - assert( key>=aDia[iRes] ); - return ((c > (aDia[iRes]>>3) + (aDia[iRes]&0x07)) ? c : (int)aChar[iRes]); -}; - - -/* -** Return true if the argument interpreted as a unicode codepoint -** is a diacritical modifier character. -*/ -int sqlite3FtsUnicodeIsdiacritic(int c){ - unsigned int mask0 = 0x08029FDF; - unsigned int mask1 = 0x000361F8; - if( c<768 || c>817 ) return 0; - return (c < 768+32) ? - (mask0 & (1 << (c-768))) : - (mask1 & (1 << (c-768-32))); -} - - -/* -** Interpret the argument as a unicode codepoint. If the codepoint -** is an upper case character that has a lower case equivalent, -** return the codepoint corresponding to the lower case version. -** Otherwise, return a copy of the argument. -** -** The results are undefined if the value passed to this function -** is less than zero. -*/ -int sqlite3FtsUnicodeFold(int c, int bRemoveDiacritic){ - /* Each entry in the following array defines a rule for folding a range - ** of codepoints to lower case. The rule applies to a range of nRange - ** codepoints starting at codepoint iCode. - ** - ** If the least significant bit in flags is clear, then the rule applies - ** to all nRange codepoints (i.e. all nRange codepoints are upper case and - ** need to be folded). Or, if it is set, then the rule only applies to - ** every second codepoint in the range, starting with codepoint C. - ** - ** The 7 most significant bits in flags are an index into the aiOff[] - ** array. If a specific codepoint C does require folding, then its lower - ** case equivalent is ((C + aiOff[flags>>1]) & 0xFFFF). - ** - ** The contents of this array are generated by parsing the CaseFolding.txt - ** file distributed as part of the "Unicode Character Database". See - ** http://www.unicode.org for details. - */ - static const struct TableEntry { - unsigned short iCode; - unsigned char flags; - unsigned char nRange; - } aEntry[] = { - {65, 14, 26}, {181, 64, 1}, {192, 14, 23}, - {216, 14, 7}, {256, 1, 48}, {306, 1, 6}, - {313, 1, 16}, {330, 1, 46}, {376, 116, 1}, - {377, 1, 6}, {383, 104, 1}, {385, 50, 1}, - {386, 1, 4}, {390, 44, 1}, {391, 0, 1}, - {393, 42, 2}, {395, 0, 1}, {398, 32, 1}, - {399, 38, 1}, {400, 40, 1}, {401, 0, 1}, - {403, 42, 1}, {404, 46, 1}, {406, 52, 1}, - {407, 48, 1}, {408, 0, 1}, {412, 52, 1}, - {413, 54, 1}, {415, 56, 1}, {416, 1, 6}, - {422, 60, 1}, {423, 0, 1}, {425, 60, 1}, - {428, 0, 1}, {430, 60, 1}, {431, 0, 1}, - {433, 58, 2}, {435, 1, 4}, {439, 62, 1}, - {440, 0, 1}, {444, 0, 1}, {452, 2, 1}, - {453, 0, 1}, {455, 2, 1}, {456, 0, 1}, - {458, 2, 1}, {459, 1, 18}, {478, 1, 18}, - {497, 2, 1}, {498, 1, 4}, {502, 122, 1}, - {503, 134, 1}, {504, 1, 40}, {544, 110, 1}, - {546, 1, 18}, {570, 70, 1}, {571, 0, 1}, - {573, 108, 1}, {574, 68, 1}, {577, 0, 1}, - {579, 106, 1}, {580, 28, 1}, {581, 30, 1}, - {582, 1, 10}, {837, 36, 1}, {880, 1, 4}, - {886, 0, 1}, {902, 18, 1}, {904, 16, 3}, - {908, 26, 1}, {910, 24, 2}, {913, 14, 17}, - {931, 14, 9}, {962, 0, 1}, {975, 4, 1}, - {976, 140, 1}, {977, 142, 1}, {981, 146, 1}, - {982, 144, 1}, {984, 1, 24}, {1008, 136, 1}, - {1009, 138, 1}, {1012, 130, 1}, {1013, 128, 1}, - {1015, 0, 1}, {1017, 152, 1}, {1018, 0, 1}, - {1021, 110, 3}, {1024, 34, 16}, {1040, 14, 32}, - {1120, 1, 34}, {1162, 1, 54}, {1216, 6, 1}, - {1217, 1, 14}, {1232, 1, 88}, {1329, 22, 38}, - {4256, 66, 38}, {4295, 66, 1}, {4301, 66, 1}, - {7680, 1, 150}, {7835, 132, 1}, {7838, 96, 1}, - {7840, 1, 96}, {7944, 150, 8}, {7960, 150, 6}, - {7976, 150, 8}, {7992, 150, 8}, {8008, 150, 6}, - {8025, 151, 8}, {8040, 150, 8}, {8072, 150, 8}, - {8088, 150, 8}, {8104, 150, 8}, {8120, 150, 2}, - {8122, 126, 2}, {8124, 148, 1}, {8126, 100, 1}, - {8136, 124, 4}, {8140, 148, 1}, {8152, 150, 2}, - {8154, 120, 2}, {8168, 150, 2}, {8170, 118, 2}, - {8172, 152, 1}, {8184, 112, 2}, {8186, 114, 2}, - {8188, 148, 1}, {8486, 98, 1}, {8490, 92, 1}, - {8491, 94, 1}, {8498, 12, 1}, {8544, 8, 16}, - {8579, 0, 1}, {9398, 10, 26}, {11264, 22, 47}, - {11360, 0, 1}, {11362, 88, 1}, {11363, 102, 1}, - {11364, 90, 1}, {11367, 1, 6}, {11373, 84, 1}, - {11374, 86, 1}, {11375, 80, 1}, {11376, 82, 1}, - {11378, 0, 1}, {11381, 0, 1}, {11390, 78, 2}, - {11392, 1, 100}, {11499, 1, 4}, {11506, 0, 1}, - {42560, 1, 46}, {42624, 1, 24}, {42786, 1, 14}, - {42802, 1, 62}, {42873, 1, 4}, {42877, 76, 1}, - {42878, 1, 10}, {42891, 0, 1}, {42893, 74, 1}, - {42896, 1, 4}, {42912, 1, 10}, {42922, 72, 1}, - {65313, 14, 26}, - }; - static const unsigned short aiOff[] = { - 1, 2, 8, 15, 16, 26, 28, 32, - 37, 38, 40, 48, 63, 64, 69, 71, - 79, 80, 116, 202, 203, 205, 206, 207, - 209, 210, 211, 213, 214, 217, 218, 219, - 775, 7264, 10792, 10795, 23228, 23256, 30204, 54721, - 54753, 54754, 54756, 54787, 54793, 54809, 57153, 57274, - 57921, 58019, 58363, 61722, 65268, 65341, 65373, 65406, - 65408, 65410, 65415, 65424, 65436, 65439, 65450, 65462, - 65472, 65476, 65478, 65480, 65482, 65488, 65506, 65511, - 65514, 65521, 65527, 65528, 65529, - }; - - int ret = c; - - assert( c>=0 ); - assert( sizeof(unsigned short)==2 && sizeof(unsigned char)==1 ); - - if( c<128 ){ - if( c>='A' && c<='Z' ) ret = c + ('a' - 'A'); - }else if( c<65536 ){ - int iHi = sizeof(aEntry)/sizeof(aEntry[0]) - 1; - int iLo = 0; - int iRes = -1; - - while( iHi>=iLo ){ - int iTest = (iHi + iLo) / 2; - int cmp = (c - aEntry[iTest].iCode); - if( cmp>=0 ){ - iRes = iTest; - iLo = iTest+1; - }else{ - iHi = iTest-1; - } - } - assert( iRes<0 || c>=aEntry[iRes].iCode ); - - if( iRes>=0 ){ - const struct TableEntry *p = &aEntry[iRes]; - if( c<(p->iCode + p->nRange) && 0==(0x01 & p->flags & (p->iCode ^ c)) ){ - ret = (c + (aiOff[p->flags>>1])) & 0x0000FFFF; - assert( ret>0 ); - } - } - - if( bRemoveDiacritic ) ret = remove_diacritic(ret); - } - - else if( c>=66560 && c<66600 ){ - ret = c + 40; - } - - return ret; -} -#endif /* defined(SQLITE_ENABLE_FTS3) || defined(SQLITE_ENABLE_FTS4) */ -#endif /* !defined(SQLITE_ENABLE_FTS4_UNICODE61) */ diff --git a/src/libtracker-fts/fts3_write.c b/src/libtracker-fts/fts3_write.c deleted file mode 100644 index bda7fbbe1..000000000 --- a/src/libtracker-fts/fts3_write.c +++ /dev/null @@ -1,5402 +0,0 @@ -/* -** 2009 Oct 23 -** -** The author disclaims copyright to this source code. In place of -** a legal notice, here is a blessing: -** -** May you do good and not evil. -** May you find forgiveness for yourself and forgive others. -** May you share freely, never taking more than you give. -** -****************************************************************************** -** -** This file is part of the SQLite FTS3 extension module. Specifically, -** this file contains code to insert, update and delete rows from FTS3 -** tables. It also contains code to merge FTS3 b-tree segments. Some -** of the sub-routines used to merge segments are also used by the query -** code in fts3.c. -*/ - -#include "fts3Int.h" -#if !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3) - -#include <string.h> -#include <assert.h> -#include <stdlib.h> - - -#define FTS_MAX_APPENDABLE_HEIGHT 16 - -/* -** When full-text index nodes are loaded from disk, the buffer that they -** are loaded into has the following number of bytes of padding at the end -** of it. i.e. if a full-text index node is 900 bytes in size, then a buffer -** of 920 bytes is allocated for it. -** -** This means that if we have a pointer into a buffer containing node data, -** it is always safe to read up to two varints from it without risking an -** overread, even if the node data is corrupted. -*/ -#define FTS3_NODE_PADDING (FTS3_VARINT_MAX*2) - -/* -** Under certain circumstances, b-tree nodes (doclists) can be loaded into -** memory incrementally instead of all at once. This can be a big performance -** win (reduced IO and CPU) if SQLite stops calling the virtual table xNext() -** method before retrieving all query results (as may happen, for example, -** if a query has a LIMIT clause). -** -** Incremental loading is used for b-tree nodes FTS3_NODE_CHUNK_THRESHOLD -** bytes and larger. Nodes are loaded in chunks of FTS3_NODE_CHUNKSIZE bytes. -** The code is written so that the hard lower-limit for each of these values -** is 1. Clearly such small values would be inefficient, but can be useful -** for testing purposes. -** -** If this module is built with SQLITE_TEST defined, these constants may -** be overridden at runtime for testing purposes. File fts3_test.c contains -** a Tcl interface to read and write the values. -*/ -#ifdef SQLITE_TEST -int test_fts3_node_chunksize = (4*1024); -int test_fts3_node_chunk_threshold = (4*1024)*4; -# define FTS3_NODE_CHUNKSIZE test_fts3_node_chunksize -# define FTS3_NODE_CHUNK_THRESHOLD test_fts3_node_chunk_threshold -#else -# define FTS3_NODE_CHUNKSIZE (4*1024) -# define FTS3_NODE_CHUNK_THRESHOLD (FTS3_NODE_CHUNKSIZE*4) -#endif - -/* -** The two values that may be meaningfully bound to the :1 parameter in -** statements SQL_REPLACE_STAT and SQL_SELECT_STAT. -*/ -#define FTS_STAT_DOCTOTAL 0 -#define FTS_STAT_INCRMERGEHINT 1 -#define FTS_STAT_AUTOINCRMERGE 2 - -/* -** If FTS_LOG_MERGES is defined, call sqlite3_log() to report each automatic -** and incremental merge operation that takes place. This is used for -** debugging FTS only, it should not usually be turned on in production -** systems. -*/ -#ifdef FTS3_LOG_MERGES -static void fts3LogMerge(int nMerge, sqlite3_int64 iAbsLevel){ - sqlite3_log(SQLITE_OK, "%d-way merge from level %d", nMerge, (int)iAbsLevel); -} -#else -#define fts3LogMerge(x, y) -#endif - - -typedef struct PendingList PendingList; -typedef struct SegmentNode SegmentNode; -typedef struct SegmentWriter SegmentWriter; - -/* -** An instance of the following data structure is used to build doclists -** incrementally. See function fts3PendingListAppend() for details. -*/ -struct PendingList { - int nData; - char *aData; - int nSpace; - sqlite3_int64 iLastDocid; - sqlite3_int64 iLastCol; - sqlite3_int64 iLastPos; -}; - - -/* -** Each cursor has a (possibly empty) linked list of the following objects. -*/ -struct Fts3DeferredToken { - Fts3PhraseToken *pToken; /* Pointer to corresponding expr token */ - int iCol; /* Column token must occur in */ - Fts3DeferredToken *pNext; /* Next in list of deferred tokens */ - PendingList *pList; /* Doclist is assembled here */ -}; - -/* -** An instance of this structure is used to iterate through the terms on -** a contiguous set of segment b-tree leaf nodes. Although the details of -** this structure are only manipulated by code in this file, opaque handles -** of type Fts3SegReader* are also used by code in fts3.c to iterate through -** terms when querying the full-text index. See functions: -** -** sqlite3Fts3SegReaderNew() -** sqlite3Fts3SegReaderFree() -** sqlite3Fts3SegReaderIterate() -** -** Methods used to manipulate Fts3SegReader structures: -** -** fts3SegReaderNext() -** fts3SegReaderFirstDocid() -** fts3SegReaderNextDocid() -*/ -struct Fts3SegReader { - int iIdx; /* Index within level, or 0x7FFFFFFF for PT */ - u8 bLookup; /* True for a lookup only */ - u8 rootOnly; /* True for a root-only reader */ - - sqlite3_int64 iStartBlock; /* Rowid of first leaf block to traverse */ - sqlite3_int64 iLeafEndBlock; /* Rowid of final leaf block to traverse */ - sqlite3_int64 iEndBlock; /* Rowid of final block in segment (or 0) */ - sqlite3_int64 iCurrentBlock; /* Current leaf block (or 0) */ - - char *aNode; /* Pointer to node data (or NULL) */ - int nNode; /* Size of buffer at aNode (or 0) */ - int nPopulate; /* If >0, bytes of buffer aNode[] loaded */ - sqlite3_blob *pBlob; /* If not NULL, blob handle to read node */ - - Fts3HashElem **ppNextElem; - - /* Variables set by fts3SegReaderNext(). These may be read directly - ** by the caller. They are valid from the time SegmentReaderNew() returns - ** until SegmentReaderNext() returns something other than SQLITE_OK - ** (i.e. SQLITE_DONE). - */ - int nTerm; /* Number of bytes in current term */ - char *zTerm; /* Pointer to current term */ - int nTermAlloc; /* Allocated size of zTerm buffer */ - char *aDoclist; /* Pointer to doclist of current entry */ - int nDoclist; /* Size of doclist in current entry */ - - /* The following variables are used by fts3SegReaderNextDocid() to iterate - ** through the current doclist (aDoclist/nDoclist). - */ - char *pOffsetList; - int nOffsetList; /* For descending pending seg-readers only */ - sqlite3_int64 iDocid; -}; - -#define fts3SegReaderIsPending(p) ((p)->ppNextElem!=0) -#define fts3SegReaderIsRootOnly(p) ((p)->rootOnly!=0) - -/* -** An instance of this structure is used to create a segment b-tree in the -** database. The internal details of this type are only accessed by the -** following functions: -** -** fts3SegWriterAdd() -** fts3SegWriterFlush() -** fts3SegWriterFree() -*/ -struct SegmentWriter { - SegmentNode *pTree; /* Pointer to interior tree structure */ - sqlite3_int64 iFirst; /* First slot in %_segments written */ - sqlite3_int64 iFree; /* Next free slot in %_segments */ - char *zTerm; /* Pointer to previous term buffer */ - int nTerm; /* Number of bytes in zTerm */ - int nMalloc; /* Size of malloc'd buffer at zMalloc */ - char *zMalloc; /* Malloc'd space (possibly) used for zTerm */ - int nSize; /* Size of allocation at aData */ - int nData; /* Bytes of data in aData */ - char *aData; /* Pointer to block from malloc() */ -}; - -/* -** Type SegmentNode is used by the following three functions to create -** the interior part of the segment b+-tree structures (everything except -** the leaf nodes). These functions and type are only ever used by code -** within the fts3SegWriterXXX() family of functions described above. -** -** fts3NodeAddTerm() -** fts3NodeWrite() -** fts3NodeFree() -** -** When a b+tree is written to the database (either as a result of a merge -** or the pending-terms table being flushed), leaves are written into the -** database file as soon as they are completely populated. The interior of -** the tree is assembled in memory and written out only once all leaves have -** been populated and stored. This is Ok, as the b+-tree fanout is usually -** very large, meaning that the interior of the tree consumes relatively -** little memory. -*/ -struct SegmentNode { - SegmentNode *pParent; /* Parent node (or NULL for root node) */ - SegmentNode *pRight; /* Pointer to right-sibling */ - SegmentNode *pLeftmost; /* Pointer to left-most node of this depth */ - int nEntry; /* Number of terms written to node so far */ - char *zTerm; /* Pointer to previous term buffer */ - int nTerm; /* Number of bytes in zTerm */ - int nMalloc; /* Size of malloc'd buffer at zMalloc */ - char *zMalloc; /* Malloc'd space (possibly) used for zTerm */ - int nData; /* Bytes of valid data so far */ - char *aData; /* Node data */ -}; - -/* -** Valid values for the second argument to fts3SqlStmt(). -*/ -#define SQL_DELETE_CONTENT 0 -#define SQL_IS_EMPTY 1 -#define SQL_DELETE_ALL_CONTENT 2 -#define SQL_DELETE_ALL_SEGMENTS 3 -#define SQL_DELETE_ALL_SEGDIR 4 -#define SQL_DELETE_ALL_DOCSIZE 5 -#define SQL_DELETE_ALL_STAT 6 -#define SQL_SELECT_CONTENT_BY_ROWID 7 -#define SQL_NEXT_SEGMENT_INDEX 8 -#define SQL_INSERT_SEGMENTS 9 -#define SQL_NEXT_SEGMENTS_ID 10 -#define SQL_INSERT_SEGDIR 11 -#define SQL_SELECT_LEVEL 12 -#define SQL_SELECT_LEVEL_RANGE 13 -#define SQL_SELECT_LEVEL_COUNT 14 -#define SQL_SELECT_SEGDIR_MAX_LEVEL 15 -#define SQL_DELETE_SEGDIR_LEVEL 16 -#define SQL_DELETE_SEGMENTS_RANGE 17 -#define SQL_CONTENT_INSERT 18 -#define SQL_DELETE_DOCSIZE 19 -#define SQL_REPLACE_DOCSIZE 20 -#define SQL_SELECT_DOCSIZE 21 -#define SQL_SELECT_STAT 22 -#define SQL_REPLACE_STAT 23 - -#define SQL_SELECT_ALL_PREFIX_LEVEL 24 -#define SQL_DELETE_ALL_TERMS_SEGDIR 25 -#define SQL_DELETE_SEGDIR_RANGE 26 -#define SQL_SELECT_ALL_LANGID 27 -#define SQL_FIND_MERGE_LEVEL 28 -#define SQL_MAX_LEAF_NODE_ESTIMATE 29 -#define SQL_DELETE_SEGDIR_ENTRY 30 -#define SQL_SHIFT_SEGDIR_ENTRY 31 -#define SQL_SELECT_SEGDIR 32 -#define SQL_CHOMP_SEGDIR 33 -#define SQL_SEGMENT_IS_APPENDABLE 34 -#define SQL_SELECT_INDEXES 35 -#define SQL_SELECT_MXLEVEL 36 - -/* -** This function is used to obtain an SQLite prepared statement handle -** for the statement identified by the second argument. If successful, -** *pp is set to the requested statement handle and SQLITE_OK returned. -** Otherwise, an SQLite error code is returned and *pp is set to 0. -** -** If argument apVal is not NULL, then it must point to an array with -** at least as many entries as the requested statement has bound -** parameters. The values are bound to the statements parameters before -** returning. -*/ -static int fts3SqlStmt( - Fts3Table *p, /* Virtual table handle */ - int eStmt, /* One of the SQL_XXX constants above */ - sqlite3_stmt **pp, /* OUT: Statement handle */ - sqlite3_value **apVal /* Values to bind to statement */ -){ - const char *azSql[] = { -/* 0 */ "DELETE FROM %Q.'%q_content' WHERE rowid = ?", -/* 1 */ "SELECT NOT EXISTS(SELECT docid FROM %Q.'%q_content' WHERE rowid!=?)", -/* 2 */ "DELETE FROM %Q.'%q_content'", -/* 3 */ "DELETE FROM %Q.'%q_segments'", -/* 4 */ "DELETE FROM %Q.'%q_segdir'", -/* 5 */ "DELETE FROM %Q.'%q_docsize'", -/* 6 */ "DELETE FROM %Q.'%q_stat'", -/* 7 */ "SELECT %s WHERE rowid=?", -/* 8 */ "SELECT (SELECT max(idx) FROM %Q.'%q_segdir' WHERE level = ?) + 1", -/* 9 */ "REPLACE INTO %Q.'%q_segments'(blockid, block) VALUES(?, ?)", -/* 10 */ "SELECT coalesce((SELECT max(blockid) FROM %Q.'%q_segments') + 1, 1)", -/* 11 */ "REPLACE INTO %Q.'%q_segdir' VALUES(?,?,?,?,?,?)", - - /* Return segments in order from oldest to newest.*/ -/* 12 */ "SELECT idx, start_block, leaves_end_block, end_block, root " - "FROM %Q.'%q_segdir' WHERE level = ? ORDER BY idx ASC", -/* 13 */ "SELECT idx, start_block, leaves_end_block, end_block, root " - "FROM %Q.'%q_segdir' WHERE level BETWEEN ? AND ?" - "ORDER BY level DESC, idx ASC", - -/* 14 */ "SELECT count(*) FROM %Q.'%q_segdir' WHERE level = ?", -/* 15 */ "SELECT max(level) FROM %Q.'%q_segdir' WHERE level BETWEEN ? AND ?", - -/* 16 */ "DELETE FROM %Q.'%q_segdir' WHERE level = ?", -/* 17 */ "DELETE FROM %Q.'%q_segments' WHERE blockid BETWEEN ? AND ?", -/* 18 */ "INSERT INTO %Q.'%q_content' VALUES(%s)", -/* 19 */ "DELETE FROM %Q.'%q_docsize' WHERE docid = ?", -/* 20 */ "REPLACE INTO %Q.'%q_docsize' VALUES(?,?)", -/* 21 */ "SELECT size FROM %Q.'%q_docsize' WHERE docid=?", -/* 22 */ "SELECT value FROM %Q.'%q_stat' WHERE id=?", -/* 23 */ "REPLACE INTO %Q.'%q_stat' VALUES(?,?)", -/* 24 */ "", -/* 25 */ "", - -/* 26 */ "DELETE FROM %Q.'%q_segdir' WHERE level BETWEEN ? AND ?", -/* 27 */ "SELECT DISTINCT level / (1024 * ?) FROM %Q.'%q_segdir'", - -/* This statement is used to determine which level to read the input from -** when performing an incremental merge. It returns the absolute level number -** of the oldest level in the db that contains at least ? segments. Or, -** if no level in the FTS index contains more than ? segments, the statement -** returns zero rows. */ -/* 28 */ "SELECT level FROM %Q.'%q_segdir' GROUP BY level HAVING count(*)>=?" - " ORDER BY (level %% 1024) ASC LIMIT 1", - -/* Estimate the upper limit on the number of leaf nodes in a new segment -** created by merging the oldest :2 segments from absolute level :1. See -** function sqlite3Fts3Incrmerge() for details. */ -/* 29 */ "SELECT 2 * total(1 + leaves_end_block - start_block) " - " FROM %Q.'%q_segdir' WHERE level = ? AND idx < ?", - -/* SQL_DELETE_SEGDIR_ENTRY -** Delete the %_segdir entry on absolute level :1 with index :2. */ -/* 30 */ "DELETE FROM %Q.'%q_segdir' WHERE level = ? AND idx = ?", - -/* SQL_SHIFT_SEGDIR_ENTRY -** Modify the idx value for the segment with idx=:3 on absolute level :2 -** to :1. */ -/* 31 */ "UPDATE %Q.'%q_segdir' SET idx = ? WHERE level=? AND idx=?", - -/* SQL_SELECT_SEGDIR -** Read a single entry from the %_segdir table. The entry from absolute -** level :1 with index value :2. */ -/* 32 */ "SELECT idx, start_block, leaves_end_block, end_block, root " - "FROM %Q.'%q_segdir' WHERE level = ? AND idx = ?", - -/* SQL_CHOMP_SEGDIR -** Update the start_block (:1) and root (:2) fields of the %_segdir -** entry located on absolute level :3 with index :4. */ -/* 33 */ "UPDATE %Q.'%q_segdir' SET start_block = ?, root = ?" - "WHERE level = ? AND idx = ?", - -/* SQL_SEGMENT_IS_APPENDABLE -** Return a single row if the segment with end_block=? is appendable. Or -** no rows otherwise. */ -/* 34 */ "SELECT 1 FROM %Q.'%q_segments' WHERE blockid=? AND block IS NULL", - -/* SQL_SELECT_INDEXES -** Return the list of valid segment indexes for absolute level ? */ -/* 35 */ "SELECT idx FROM %Q.'%q_segdir' WHERE level=? ORDER BY 1 ASC", - -/* SQL_SELECT_MXLEVEL -** Return the largest relative level in the FTS index or indexes. */ -/* 36 */ "SELECT max( level %% 1024 ) FROM %Q.'%q_segdir'" - }; - int rc = SQLITE_OK; - sqlite3_stmt *pStmt; - - assert( SizeofArray(azSql)==SizeofArray(p->aStmt) ); - assert( eStmt<SizeofArray(azSql) && eStmt>=0 ); - - pStmt = p->aStmt[eStmt]; - if( !pStmt ){ - char *zSql; - if( eStmt==SQL_CONTENT_INSERT ){ - zSql = sqlite3_mprintf(azSql[eStmt], p->zDb, p->zName, p->zWriteExprlist); - }else if( eStmt==SQL_SELECT_CONTENT_BY_ROWID ){ - zSql = sqlite3_mprintf(azSql[eStmt], p->zReadExprlist); - }else{ - zSql = sqlite3_mprintf(azSql[eStmt], p->zDb, p->zName); - } - if( !zSql ){ - rc = SQLITE_NOMEM; - }else{ - rc = sqlite3_prepare_v2(p->db, zSql, -1, &pStmt, NULL); - sqlite3_free(zSql); - assert( rc==SQLITE_OK || pStmt==0 ); - p->aStmt[eStmt] = pStmt; - } - } - if( apVal ){ - int i; - int nParam = sqlite3_bind_parameter_count(pStmt); - for(i=0; rc==SQLITE_OK && i<nParam; i++){ - rc = sqlite3_bind_value(pStmt, i+1, apVal[i]); - } - } - *pp = pStmt; - return rc; -} - - -static int fts3SelectDocsize( - Fts3Table *pTab, /* FTS3 table handle */ - sqlite3_int64 iDocid, /* Docid to bind for SQL_SELECT_DOCSIZE */ - sqlite3_stmt **ppStmt /* OUT: Statement handle */ -){ - sqlite3_stmt *pStmt = 0; /* Statement requested from fts3SqlStmt() */ - int rc; /* Return code */ - - rc = fts3SqlStmt(pTab, SQL_SELECT_DOCSIZE, &pStmt, 0); - if( rc==SQLITE_OK ){ - sqlite3_bind_int64(pStmt, 1, iDocid); - rc = sqlite3_step(pStmt); - if( rc!=SQLITE_ROW || sqlite3_column_type(pStmt, 0)!=SQLITE_BLOB ){ - rc = sqlite3_reset(pStmt); - if( rc==SQLITE_OK ) rc = FTS_CORRUPT_VTAB; - pStmt = 0; - }else{ - rc = SQLITE_OK; - } - } - - *ppStmt = pStmt; - return rc; -} - -int sqlite3Fts3SelectDoctotal( - Fts3Table *pTab, /* Fts3 table handle */ - sqlite3_stmt **ppStmt /* OUT: Statement handle */ -){ - sqlite3_stmt *pStmt = 0; - int rc; - rc = fts3SqlStmt(pTab, SQL_SELECT_STAT, &pStmt, 0); - if( rc==SQLITE_OK ){ - sqlite3_bind_int(pStmt, 1, FTS_STAT_DOCTOTAL); - if( sqlite3_step(pStmt)!=SQLITE_ROW - || sqlite3_column_type(pStmt, 0)!=SQLITE_BLOB - ){ - rc = sqlite3_reset(pStmt); - if( rc==SQLITE_OK ) rc = FTS_CORRUPT_VTAB; - pStmt = 0; - } - } - *ppStmt = pStmt; - return rc; -} - -int sqlite3Fts3SelectDocsize( - Fts3Table *pTab, /* Fts3 table handle */ - sqlite3_int64 iDocid, /* Docid to read size data for */ - sqlite3_stmt **ppStmt /* OUT: Statement handle */ -){ - return fts3SelectDocsize(pTab, iDocid, ppStmt); -} - -/* -** Similar to fts3SqlStmt(). Except, after binding the parameters in -** array apVal[] to the SQL statement identified by eStmt, the statement -** is executed. -** -** Returns SQLITE_OK if the statement is successfully executed, or an -** SQLite error code otherwise. -*/ -static void fts3SqlExec( - int *pRC, /* Result code */ - Fts3Table *p, /* The FTS3 table */ - int eStmt, /* Index of statement to evaluate */ - sqlite3_value **apVal /* Parameters to bind */ -){ - sqlite3_stmt *pStmt; - int rc; - if( *pRC ) return; - rc = fts3SqlStmt(p, eStmt, &pStmt, apVal); - if( rc==SQLITE_OK ){ - sqlite3_step(pStmt); - rc = sqlite3_reset(pStmt); - } - *pRC = rc; -} - - -/* -** This function ensures that the caller has obtained a shared-cache -** table-lock on the %_content table. This is required before reading -** data from the fts3 table. If this lock is not acquired first, then -** the caller may end up holding read-locks on the %_segments and %_segdir -** tables, but no read-lock on the %_content table. If this happens -** a second connection will be able to write to the fts3 table, but -** attempting to commit those writes might return SQLITE_LOCKED or -** SQLITE_LOCKED_SHAREDCACHE (because the commit attempts to obtain -** write-locks on the %_segments and %_segdir ** tables). -** -** We try to avoid this because if FTS3 returns any error when committing -** a transaction, the whole transaction will be rolled back. And this is -** not what users expect when they get SQLITE_LOCKED_SHAREDCACHE. It can -** still happen if the user reads data directly from the %_segments or -** %_segdir tables instead of going through FTS3 though. -** -** This reasoning does not apply to a content=xxx table. -*/ -int sqlite3Fts3ReadLock(Fts3Table *p){ - int rc; /* Return code */ - sqlite3_stmt *pStmt; /* Statement used to obtain lock */ - - if( p->zContentTbl==0 ){ - rc = fts3SqlStmt(p, SQL_SELECT_CONTENT_BY_ROWID, &pStmt, 0); - if( rc==SQLITE_OK ){ - sqlite3_bind_null(pStmt, 1); - sqlite3_step(pStmt); - rc = sqlite3_reset(pStmt); - } - }else{ - rc = SQLITE_OK; - } - - return rc; -} - -/* -** FTS maintains a separate indexes for each language-id (a 32-bit integer). -** Within each language id, a separate index is maintained to store the -** document terms, and each configured prefix size (configured the FTS -** "prefix=" option). And each index consists of multiple levels ("relative -** levels"). -** -** All three of these values (the language id, the specific index and the -** level within the index) are encoded in 64-bit integer values stored -** in the %_segdir table on disk. This function is used to convert three -** separate component values into the single 64-bit integer value that -** can be used to query the %_segdir table. -** -** Specifically, each language-id/index combination is allocated 1024 -** 64-bit integer level values ("absolute levels"). The main terms index -** for language-id 0 is allocate values 0-1023. The first prefix index -** (if any) for language-id 0 is allocated values 1024-2047. And so on. -** Language 1 indexes are allocated immediately following language 0. -** -** So, for a system with nPrefix prefix indexes configured, the block of -** absolute levels that corresponds to language-id iLangid and index -** iIndex starts at absolute level ((iLangid * (nPrefix+1) + iIndex) * 1024). -*/ -static sqlite3_int64 getAbsoluteLevel( - Fts3Table *p, /* FTS3 table handle */ - int iLangid, /* Language id */ - int iIndex, /* Index in p->aIndex[] */ - int iLevel /* Level of segments */ -){ - sqlite3_int64 iBase; /* First absolute level for iLangid/iIndex */ - assert( iLangid>=0 ); - assert( p->nIndex>0 ); - assert( iIndex>=0 && iIndex<p->nIndex ); - - iBase = ((sqlite3_int64)iLangid * p->nIndex + iIndex) * FTS3_SEGDIR_MAXLEVEL; - return iBase + iLevel; -} - -/* -** Set *ppStmt to a statement handle that may be used to iterate through -** all rows in the %_segdir table, from oldest to newest. If successful, -** return SQLITE_OK. If an error occurs while preparing the statement, -** return an SQLite error code. -** -** There is only ever one instance of this SQL statement compiled for -** each FTS3 table. -** -** The statement returns the following columns from the %_segdir table: -** -** 0: idx -** 1: start_block -** 2: leaves_end_block -** 3: end_block -** 4: root -*/ -int sqlite3Fts3AllSegdirs( - Fts3Table *p, /* FTS3 table */ - int iLangid, /* Language being queried */ - int iIndex, /* Index for p->aIndex[] */ - int iLevel, /* Level to select (relative level) */ - sqlite3_stmt **ppStmt /* OUT: Compiled statement */ -){ - int rc; - sqlite3_stmt *pStmt = 0; - - assert( iLevel==FTS3_SEGCURSOR_ALL || iLevel>=0 ); - assert( iLevel<FTS3_SEGDIR_MAXLEVEL ); - assert( iIndex>=0 && iIndex<p->nIndex ); - - if( iLevel<0 ){ - /* "SELECT * FROM %_segdir WHERE level BETWEEN ? AND ? ORDER BY ..." */ - rc = fts3SqlStmt(p, SQL_SELECT_LEVEL_RANGE, &pStmt, 0); - if( rc==SQLITE_OK ){ - sqlite3_bind_int64(pStmt, 1, getAbsoluteLevel(p, iLangid, iIndex, 0)); - sqlite3_bind_int64(pStmt, 2, - getAbsoluteLevel(p, iLangid, iIndex, FTS3_SEGDIR_MAXLEVEL-1) - ); - } - }else{ - /* "SELECT * FROM %_segdir WHERE level = ? ORDER BY ..." */ - rc = fts3SqlStmt(p, SQL_SELECT_LEVEL, &pStmt, 0); - if( rc==SQLITE_OK ){ - sqlite3_bind_int64(pStmt, 1, getAbsoluteLevel(p, iLangid, iIndex,iLevel)); - } - } - *ppStmt = pStmt; - return rc; -} - - -/* -** Append a single varint to a PendingList buffer. SQLITE_OK is returned -** if successful, or an SQLite error code otherwise. -** -** This function also serves to allocate the PendingList structure itself. -** For example, to create a new PendingList structure containing two -** varints: -** -** PendingList *p = 0; -** fts3PendingListAppendVarint(&p, 1); -** fts3PendingListAppendVarint(&p, 2); -*/ -static int fts3PendingListAppendVarint( - PendingList **pp, /* IN/OUT: Pointer to PendingList struct */ - sqlite3_int64 i /* Value to append to data */ -){ - PendingList *p = *pp; - - /* Allocate or grow the PendingList as required. */ - if( !p ){ - p = sqlite3_malloc(sizeof(*p) + 100); - if( !p ){ - return SQLITE_NOMEM; - } - p->nSpace = 100; - p->aData = (char *)&p[1]; - p->nData = 0; - } - else if( p->nData+FTS3_VARINT_MAX+1>p->nSpace ){ - int nNew = p->nSpace * 2; - p = sqlite3_realloc(p, sizeof(*p) + nNew); - if( !p ){ - sqlite3_free(*pp); - *pp = 0; - return SQLITE_NOMEM; - } - p->nSpace = nNew; - p->aData = (char *)&p[1]; - } - - /* Append the new serialized varint to the end of the list. */ - p->nData += sqlite3Fts3PutVarint(&p->aData[p->nData], i); - p->aData[p->nData] = '\0'; - *pp = p; - return SQLITE_OK; -} - -/* -** Add a docid/column/position entry to a PendingList structure. Non-zero -** is returned if the structure is sqlite3_realloced as part of adding -** the entry. Otherwise, zero. -** -** If an OOM error occurs, *pRc is set to SQLITE_NOMEM before returning. -** Zero is always returned in this case. Otherwise, if no OOM error occurs, -** it is set to SQLITE_OK. -*/ -static int fts3PendingListAppend( - PendingList **pp, /* IN/OUT: PendingList structure */ - sqlite3_int64 iDocid, /* Docid for entry to add */ - sqlite3_int64 iCol, /* Column for entry to add */ - sqlite3_int64 iPos, /* Position of term for entry to add */ - int *pRc /* OUT: Return code */ -){ - PendingList *p = *pp; - int rc = SQLITE_OK; - - assert( !p || p->iLastDocid<=iDocid ); - - if( !p || p->iLastDocid!=iDocid ){ - sqlite3_int64 iDelta = iDocid - (p ? p->iLastDocid : 0); - if( p ){ - assert( p->nData<p->nSpace ); - assert( p->aData[p->nData]==0 ); - p->nData++; - } - if( SQLITE_OK!=(rc = fts3PendingListAppendVarint(&p, iDelta)) ){ - goto pendinglistappend_out; - } - p->iLastCol = -1; - p->iLastPos = 0; - p->iLastDocid = iDocid; - } - if( iCol>0 && p->iLastCol!=iCol ){ - if( SQLITE_OK!=(rc = fts3PendingListAppendVarint(&p, 1)) - || SQLITE_OK!=(rc = fts3PendingListAppendVarint(&p, iCol)) - ){ - goto pendinglistappend_out; - } - p->iLastCol = iCol; - p->iLastPos = 0; - } - if( iCol>=0 ){ - assert( iPos>p->iLastPos || (iPos==0 && p->iLastPos==0) ); - rc = fts3PendingListAppendVarint(&p, 2+iPos-p->iLastPos); - if( rc==SQLITE_OK ){ - p->iLastPos = iPos; - } - } - - pendinglistappend_out: - *pRc = rc; - if( p!=*pp ){ - *pp = p; - return 1; - } - return 0; -} - -/* -** Free a PendingList object allocated by fts3PendingListAppend(). -*/ -static void fts3PendingListDelete(PendingList *pList){ - sqlite3_free(pList); -} - -/* -** Add an entry to one of the pending-terms hash tables. -*/ -static int fts3PendingTermsAddOne( - Fts3Table *p, - int iCol, - int iPos, - Fts3Hash *pHash, /* Pending terms hash table to add entry to */ - const char *zToken, - int nToken -){ - PendingList *pList; - int rc = SQLITE_OK; - - pList = (PendingList *)fts3HashFind(pHash, zToken, nToken); - if( pList ){ - p->nPendingData -= (pList->nData + nToken + sizeof(Fts3HashElem)); - } - if( fts3PendingListAppend(&pList, p->iPrevDocid, iCol, iPos, &rc) ){ - if( pList==fts3HashInsert(pHash, zToken, nToken, pList) ){ - /* Malloc failed while inserting the new entry. This can only - ** happen if there was no previous entry for this token. - */ - assert( 0==fts3HashFind(pHash, zToken, nToken) ); - sqlite3_free(pList); - rc = SQLITE_NOMEM; - } - } - if( rc==SQLITE_OK ){ - p->nPendingData += (pList->nData + nToken + sizeof(Fts3HashElem)); - } - return rc; -} - -/* -** Tokenize the nul-terminated string zText and add all tokens to the -** pending-terms hash-table. The docid used is that currently stored in -** p->iPrevDocid, and the column is specified by argument iCol. -** -** If successful, SQLITE_OK is returned. Otherwise, an SQLite error code. -*/ -static int fts3PendingTermsAdd( - Fts3Table *p, /* Table into which text will be inserted */ - int iLangid, /* Language id to use */ - const char *zText, /* Text of document to be inserted */ - int iCol, /* Column into which text is being inserted */ - u32 *pnWord /* IN/OUT: Incr. by number tokens inserted */ -){ - int rc; - int iStart = 0; - int iEnd = 0; - int iPos = 0; - int nWord = 0; - - char const *zToken; - int nToken = 0; - - sqlite3_tokenizer *pTokenizer = p->pTokenizer; - sqlite3_tokenizer_module const *pModule = pTokenizer->pModule; - sqlite3_tokenizer_cursor *pCsr; - int (*xNext)(sqlite3_tokenizer_cursor *pCursor, - const char**,int*,int*,int*,int*); - - assert( pTokenizer && pModule ); - - /* If the user has inserted a NULL value, this function may be called with - ** zText==0. In this case, add zero token entries to the hash table and - ** return early. */ - if( zText==0 ){ - *pnWord = 0; - return SQLITE_OK; - } - - rc = sqlite3Fts3OpenTokenizer(pTokenizer, iLangid, zText, -1, &pCsr); - if( rc!=SQLITE_OK ){ - return rc; - } - - xNext = pModule->xNext; - while( SQLITE_OK==rc - && SQLITE_OK==(rc = xNext(pCsr, &zToken, &nToken, &iStart, &iEnd, &iPos)) - ){ - int i; - if( iPos>=nWord ) nWord = iPos+1; - - /* Positions cannot be negative; we use -1 as a terminator internally. - ** Tokens must have a non-zero length. - */ - if( iPos<0 || !zToken || nToken<=0 ){ - rc = SQLITE_ERROR; - break; - } - - /* Add the term to the terms index */ - rc = fts3PendingTermsAddOne( - p, iCol, iPos, &p->aIndex[0].hPending, zToken, nToken - ); - - /* Add the term to each of the prefix indexes that it is not too - ** short for. */ - for(i=1; rc==SQLITE_OK && i<p->nIndex; i++){ - struct Fts3Index *pIndex = &p->aIndex[i]; - if( nToken<pIndex->nPrefix ) continue; - rc = fts3PendingTermsAddOne( - p, iCol, iPos, &pIndex->hPending, zToken, pIndex->nPrefix - ); - } - } - - pModule->xClose(pCsr); - *pnWord += nWord; - return (rc==SQLITE_DONE ? SQLITE_OK : rc); -} - -/* -** Calling this function indicates that subsequent calls to -** fts3PendingTermsAdd() are to add term/position-list pairs for the -** contents of the document with docid iDocid. -*/ -static int fts3PendingTermsDocid( - Fts3Table *p, /* Full-text table handle */ - int iLangid, /* Language id of row being written */ - sqlite_int64 iDocid /* Docid of row being written */ -){ - assert( iLangid>=0 ); - - /* TODO(shess) Explore whether partially flushing the buffer on - ** forced-flush would provide better performance. I suspect that if - ** we ordered the doclists by size and flushed the largest until the - ** buffer was half empty, that would let the less frequent terms - ** generate longer doclists. - */ - if( iDocid<=p->iPrevDocid - || p->iPrevLangid!=iLangid - || p->nPendingData>p->nMaxPendingData - ){ - int rc = sqlite3Fts3PendingTermsFlush(p); - if( rc!=SQLITE_OK ) return rc; - } - p->iPrevDocid = iDocid; - p->iPrevLangid = iLangid; - return SQLITE_OK; -} - -/* -** Discard the contents of the pending-terms hash tables. -*/ -void sqlite3Fts3PendingTermsClear(Fts3Table *p){ - int i; - for(i=0; i<p->nIndex; i++){ - Fts3HashElem *pElem; - Fts3Hash *pHash = &p->aIndex[i].hPending; - for(pElem=fts3HashFirst(pHash); pElem; pElem=fts3HashNext(pElem)){ - PendingList *pList = (PendingList *)fts3HashData(pElem); - fts3PendingListDelete(pList); - } - fts3HashClear(pHash); - } - p->nPendingData = 0; -} - -/* -** This function is called by the xUpdate() method as part of an INSERT -** operation. It adds entries for each term in the new record to the -** pendingTerms hash table. -** -** Argument apVal is the same as the similarly named argument passed to -** fts3InsertData(). Parameter iDocid is the docid of the new row. -*/ -static int fts3InsertTerms( - Fts3Table *p, - int iLangid, - sqlite3_value **apVal, - u32 *aSz -){ - int i; /* Iterator variable */ - for(i=2; i<p->nColumn+2; i++){ - const char *zText = (const char *)sqlite3_value_text(apVal[i]); - int rc = fts3PendingTermsAdd(p, iLangid, zText, i-2, &aSz[i-2]); - if( rc!=SQLITE_OK ){ - return rc; - } - aSz[p->nColumn] += sqlite3_value_bytes(apVal[i]); - } - return SQLITE_OK; -} - -/* -** This function is called by the xUpdate() method for an INSERT operation. -** The apVal parameter is passed a copy of the apVal argument passed by -** SQLite to the xUpdate() method. i.e: -** -** apVal[0] Not used for INSERT. -** apVal[1] rowid -** apVal[2] Left-most user-defined column -** ... -** apVal[p->nColumn+1] Right-most user-defined column -** apVal[p->nColumn+2] Hidden column with same name as table -** apVal[p->nColumn+3] Hidden "docid" column (alias for rowid) -** apVal[p->nColumn+4] Hidden languageid column -*/ -static int fts3InsertData( - Fts3Table *p, /* Full-text table */ - sqlite3_value **apVal, /* Array of values to insert */ - sqlite3_int64 *piDocid /* OUT: Docid for row just inserted */ -){ - int rc; /* Return code */ - sqlite3_stmt *pContentInsert; /* INSERT INTO %_content VALUES(...) */ - - if( p->zContentTbl ){ - sqlite3_value *pRowid = apVal[p->nColumn+3]; - if( sqlite3_value_type(pRowid)==SQLITE_NULL ){ - pRowid = apVal[1]; - } - if( sqlite3_value_type(pRowid)!=SQLITE_INTEGER ){ - return SQLITE_CONSTRAINT; - } - *piDocid = sqlite3_value_int64(pRowid); - return SQLITE_OK; - } - - /* Locate the statement handle used to insert data into the %_content - ** table. The SQL for this statement is: - ** - ** INSERT INTO %_content VALUES(?, ?, ?, ...) - ** - ** The statement features N '?' variables, where N is the number of user - ** defined columns in the FTS3 table, plus one for the docid field. - */ - rc = fts3SqlStmt(p, SQL_CONTENT_INSERT, &pContentInsert, &apVal[1]); - if( rc==SQLITE_OK && p->zLanguageid ){ - rc = sqlite3_bind_int( - pContentInsert, p->nColumn+2, - sqlite3_value_int(apVal[p->nColumn+4]) - ); - } - if( rc!=SQLITE_OK ) return rc; - - /* There is a quirk here. The users INSERT statement may have specified - ** a value for the "rowid" field, for the "docid" field, or for both. - ** Which is a problem, since "rowid" and "docid" are aliases for the - ** same value. For example: - ** - ** INSERT INTO fts3tbl(rowid, docid) VALUES(1, 2); - ** - ** In FTS3, this is an error. It is an error to specify non-NULL values - ** for both docid and some other rowid alias. - */ - if( SQLITE_NULL!=sqlite3_value_type(apVal[3+p->nColumn]) ){ - if( SQLITE_NULL==sqlite3_value_type(apVal[0]) - && SQLITE_NULL!=sqlite3_value_type(apVal[1]) - ){ - /* A rowid/docid conflict. */ - return SQLITE_ERROR; - } - rc = sqlite3_bind_value(pContentInsert, 1, apVal[3+p->nColumn]); - if( rc!=SQLITE_OK ) return rc; - } - - /* Execute the statement to insert the record. Set *piDocid to the - ** new docid value. - */ - sqlite3_step(pContentInsert); - rc = sqlite3_reset(pContentInsert); - - *piDocid = sqlite3_last_insert_rowid(p->db); - return rc; -} - - - -/* -** Remove all data from the FTS3 table. Clear the hash table containing -** pending terms. -*/ -static int fts3DeleteAll(Fts3Table *p, int bContent){ - int rc = SQLITE_OK; /* Return code */ - - /* Discard the contents of the pending-terms hash table. */ - sqlite3Fts3PendingTermsClear(p); - - /* Delete everything from the shadow tables. Except, leave %_content as - ** is if bContent is false. */ - assert( p->zContentTbl==0 || bContent==0 ); - if( bContent ) fts3SqlExec(&rc, p, SQL_DELETE_ALL_CONTENT, 0); - fts3SqlExec(&rc, p, SQL_DELETE_ALL_SEGMENTS, 0); - fts3SqlExec(&rc, p, SQL_DELETE_ALL_SEGDIR, 0); - if( p->bHasDocsize ){ - fts3SqlExec(&rc, p, SQL_DELETE_ALL_DOCSIZE, 0); - } - if( p->bHasStat ){ - fts3SqlExec(&rc, p, SQL_DELETE_ALL_STAT, 0); - } - return rc; -} - -/* -** -*/ -static int langidFromSelect(Fts3Table *p, sqlite3_stmt *pSelect){ - int iLangid = 0; - if( p->zLanguageid ) iLangid = sqlite3_column_int(pSelect, p->nColumn+1); - return iLangid; -} - -/* -** The first element in the apVal[] array is assumed to contain the docid -** (an integer) of a row about to be deleted. Remove all terms from the -** full-text index. -*/ -static void fts3DeleteTerms( - int *pRC, /* Result code */ - Fts3Table *p, /* The FTS table to delete from */ - sqlite3_value *pRowid, /* The docid to be deleted */ - u32 *aSz, /* Sizes of deleted document written here */ - int *pbFound /* OUT: Set to true if row really does exist */ -){ - int rc; - sqlite3_stmt *pSelect; - - assert( *pbFound==0 ); - if( *pRC ) return; - rc = fts3SqlStmt(p, SQL_SELECT_CONTENT_BY_ROWID, &pSelect, &pRowid); - if( rc==SQLITE_OK ){ - if( SQLITE_ROW==sqlite3_step(pSelect) ){ - int i; - int iLangid = langidFromSelect(p, pSelect); - rc = fts3PendingTermsDocid(p, iLangid, sqlite3_column_int64(pSelect, 0)); - for(i=1; rc==SQLITE_OK && i<=p->nColumn; i++){ - const char *zText = (const char *)sqlite3_column_text(pSelect, i); - rc = fts3PendingTermsAdd(p, iLangid, zText, -1, &aSz[i-1]); - aSz[p->nColumn] += sqlite3_column_bytes(pSelect, i); - } - if( rc!=SQLITE_OK ){ - sqlite3_reset(pSelect); - *pRC = rc; - return; - } - *pbFound = 1; - } - rc = sqlite3_reset(pSelect); - }else{ - sqlite3_reset(pSelect); - } - *pRC = rc; -} - -/* -** Forward declaration to account for the circular dependency between -** functions fts3SegmentMerge() and fts3AllocateSegdirIdx(). -*/ -static int fts3SegmentMerge(Fts3Table *, int, int, int); - -/* -** This function allocates a new level iLevel index in the segdir table. -** Usually, indexes are allocated within a level sequentially starting -** with 0, so the allocated index is one greater than the value returned -** by: -** -** SELECT max(idx) FROM %_segdir WHERE level = :iLevel -** -** However, if there are already FTS3_MERGE_COUNT indexes at the requested -** level, they are merged into a single level (iLevel+1) segment and the -** allocated index is 0. -** -** If successful, *piIdx is set to the allocated index slot and SQLITE_OK -** returned. Otherwise, an SQLite error code is returned. -*/ -static int fts3AllocateSegdirIdx( - Fts3Table *p, - int iLangid, /* Language id */ - int iIndex, /* Index for p->aIndex */ - int iLevel, - int *piIdx -){ - int rc; /* Return Code */ - sqlite3_stmt *pNextIdx; /* Query for next idx at level iLevel */ - int iNext = 0; /* Result of query pNextIdx */ - - assert( iLangid>=0 ); - assert( p->nIndex>=1 ); - - /* Set variable iNext to the next available segdir index at level iLevel. */ - rc = fts3SqlStmt(p, SQL_NEXT_SEGMENT_INDEX, &pNextIdx, 0); - if( rc==SQLITE_OK ){ - sqlite3_bind_int64( - pNextIdx, 1, getAbsoluteLevel(p, iLangid, iIndex, iLevel) - ); - if( SQLITE_ROW==sqlite3_step(pNextIdx) ){ - iNext = sqlite3_column_int(pNextIdx, 0); - } - rc = sqlite3_reset(pNextIdx); - } - - if( rc==SQLITE_OK ){ - /* If iNext is FTS3_MERGE_COUNT, indicating that level iLevel is already - ** full, merge all segments in level iLevel into a single iLevel+1 - ** segment and allocate (newly freed) index 0 at level iLevel. Otherwise, - ** if iNext is less than FTS3_MERGE_COUNT, allocate index iNext. - */ - if( iNext>=FTS3_MERGE_COUNT ){ - fts3LogMerge(16, getAbsoluteLevel(p, iLangid, iIndex, iLevel)); - rc = fts3SegmentMerge(p, iLangid, iIndex, iLevel); - *piIdx = 0; - }else{ - *piIdx = iNext; - } - } - - return rc; -} - -/* -** The %_segments table is declared as follows: -** -** CREATE TABLE %_segments(blockid INTEGER PRIMARY KEY, block BLOB) -** -** This function reads data from a single row of the %_segments table. The -** specific row is identified by the iBlockid parameter. If paBlob is not -** NULL, then a buffer is allocated using sqlite3_malloc() and populated -** with the contents of the blob stored in the "block" column of the -** identified table row is. Whether or not paBlob is NULL, *pnBlob is set -** to the size of the blob in bytes before returning. -** -** If an error occurs, or the table does not contain the specified row, -** an SQLite error code is returned. Otherwise, SQLITE_OK is returned. If -** paBlob is non-NULL, then it is the responsibility of the caller to -** eventually free the returned buffer. -** -** This function may leave an open sqlite3_blob* handle in the -** Fts3Table.pSegments variable. This handle is reused by subsequent calls -** to this function. The handle may be closed by calling the -** sqlite3Fts3SegmentsClose() function. Reusing a blob handle is a handy -** performance improvement, but the blob handle should always be closed -** before control is returned to the user (to prevent a lock being held -** on the database file for longer than necessary). Thus, any virtual table -** method (xFilter etc.) that may directly or indirectly call this function -** must call sqlite3Fts3SegmentsClose() before returning. -*/ -int sqlite3Fts3ReadBlock( - Fts3Table *p, /* FTS3 table handle */ - sqlite3_int64 iBlockid, /* Access the row with blockid=$iBlockid */ - char **paBlob, /* OUT: Blob data in malloc'd buffer */ - int *pnBlob, /* OUT: Size of blob data */ - int *pnLoad /* OUT: Bytes actually loaded */ -){ - int rc; /* Return code */ - - /* pnBlob must be non-NULL. paBlob may be NULL or non-NULL. */ - assert( pnBlob ); - - if( p->pSegments ){ - rc = sqlite3_blob_reopen(p->pSegments, iBlockid); - }else{ - if( 0==p->zSegmentsTbl ){ - p->zSegmentsTbl = sqlite3_mprintf("%s_segments", p->zName); - if( 0==p->zSegmentsTbl ) return SQLITE_NOMEM; - } - rc = sqlite3_blob_open( - p->db, p->zDb, p->zSegmentsTbl, "block", iBlockid, 0, &p->pSegments - ); - } - - if( rc==SQLITE_OK ){ - int nByte = sqlite3_blob_bytes(p->pSegments); - *pnBlob = nByte; - if( paBlob ){ - char *aByte = sqlite3_malloc(nByte + FTS3_NODE_PADDING); - if( !aByte ){ - rc = SQLITE_NOMEM; - }else{ - if( pnLoad && nByte>(FTS3_NODE_CHUNK_THRESHOLD) ){ - nByte = FTS3_NODE_CHUNKSIZE; - *pnLoad = nByte; - } - rc = sqlite3_blob_read(p->pSegments, aByte, nByte, 0); - memset(&aByte[nByte], 0, FTS3_NODE_PADDING); - if( rc!=SQLITE_OK ){ - sqlite3_free(aByte); - aByte = 0; - } - } - *paBlob = aByte; - } - } - - return rc; -} - -/* -** Close the blob handle at p->pSegments, if it is open. See comments above -** the sqlite3Fts3ReadBlock() function for details. -*/ -void sqlite3Fts3SegmentsClose(Fts3Table *p){ - sqlite3_blob_close(p->pSegments); - p->pSegments = 0; -} - -static int fts3SegReaderIncrRead(Fts3SegReader *pReader){ - int nRead; /* Number of bytes to read */ - int rc; /* Return code */ - - nRead = MIN(pReader->nNode - pReader->nPopulate, FTS3_NODE_CHUNKSIZE); - rc = sqlite3_blob_read( - pReader->pBlob, - &pReader->aNode[pReader->nPopulate], - nRead, - pReader->nPopulate - ); - - if( rc==SQLITE_OK ){ - pReader->nPopulate += nRead; - memset(&pReader->aNode[pReader->nPopulate], 0, FTS3_NODE_PADDING); - if( pReader->nPopulate==pReader->nNode ){ - sqlite3_blob_close(pReader->pBlob); - pReader->pBlob = 0; - pReader->nPopulate = 0; - } - } - return rc; -} - -static int fts3SegReaderRequire(Fts3SegReader *pReader, char *pFrom, int nByte){ - int rc = SQLITE_OK; - assert( !pReader->pBlob - || (pFrom>=pReader->aNode && pFrom<&pReader->aNode[pReader->nNode]) - ); - while( pReader->pBlob && rc==SQLITE_OK - && (pFrom - pReader->aNode + nByte)>pReader->nPopulate - ){ - rc = fts3SegReaderIncrRead(pReader); - } - return rc; -} - -/* -** Set an Fts3SegReader cursor to point at EOF. -*/ -static void fts3SegReaderSetEof(Fts3SegReader *pSeg){ - if( !fts3SegReaderIsRootOnly(pSeg) ){ - sqlite3_free(pSeg->aNode); - sqlite3_blob_close(pSeg->pBlob); - pSeg->pBlob = 0; - } - pSeg->aNode = 0; -} - -/* -** Move the iterator passed as the first argument to the next term in the -** segment. If successful, SQLITE_OK is returned. If there is no next term, -** SQLITE_DONE. Otherwise, an SQLite error code. -*/ -static int fts3SegReaderNext( - Fts3Table *p, - Fts3SegReader *pReader, - int bIncr -){ - int rc; /* Return code of various sub-routines */ - char *pNext; /* Cursor variable */ - int nPrefix; /* Number of bytes in term prefix */ - int nSuffix; /* Number of bytes in term suffix */ - - if( !pReader->aDoclist ){ - pNext = pReader->aNode; - }else{ - pNext = &pReader->aDoclist[pReader->nDoclist]; - } - - if( !pNext || pNext>=&pReader->aNode[pReader->nNode] ){ - - if( fts3SegReaderIsPending(pReader) ){ - Fts3HashElem *pElem = *(pReader->ppNextElem); - if( pElem==0 ){ - pReader->aNode = 0; - }else{ - PendingList *pList = (PendingList *)fts3HashData(pElem); - pReader->zTerm = (char *)fts3HashKey(pElem); - pReader->nTerm = fts3HashKeysize(pElem); - pReader->nNode = pReader->nDoclist = pList->nData + 1; - pReader->aNode = pReader->aDoclist = pList->aData; - pReader->ppNextElem++; - assert( pReader->aNode ); - } - return SQLITE_OK; - } - - fts3SegReaderSetEof(pReader); - - /* If iCurrentBlock>=iLeafEndBlock, this is an EOF condition. All leaf - ** blocks have already been traversed. */ - assert( pReader->iCurrentBlock<=pReader->iLeafEndBlock ); - if( pReader->iCurrentBlock>=pReader->iLeafEndBlock ){ - return SQLITE_OK; - } - - rc = sqlite3Fts3ReadBlock( - p, ++pReader->iCurrentBlock, &pReader->aNode, &pReader->nNode, - (bIncr ? &pReader->nPopulate : 0) - ); - if( rc!=SQLITE_OK ) return rc; - assert( pReader->pBlob==0 ); - if( bIncr && pReader->nPopulate<pReader->nNode ){ - pReader->pBlob = p->pSegments; - p->pSegments = 0; - } - pNext = pReader->aNode; - } - - assert( !fts3SegReaderIsPending(pReader) ); - - rc = fts3SegReaderRequire(pReader, pNext, FTS3_VARINT_MAX*2); - if( rc!=SQLITE_OK ) return rc; - - /* Because of the FTS3_NODE_PADDING bytes of padding, the following is - ** safe (no risk of overread) even if the node data is corrupted. */ - pNext += sqlite3Fts3GetVarint32(pNext, &nPrefix); - pNext += sqlite3Fts3GetVarint32(pNext, &nSuffix); - if( nPrefix<0 || nSuffix<=0 - || &pNext[nSuffix]>&pReader->aNode[pReader->nNode] - ){ - return FTS_CORRUPT_VTAB; - } - - if( nPrefix+nSuffix>pReader->nTermAlloc ){ - int nNew = (nPrefix+nSuffix)*2; - char *zNew = sqlite3_realloc(pReader->zTerm, nNew); - if( !zNew ){ - return SQLITE_NOMEM; - } - pReader->zTerm = zNew; - pReader->nTermAlloc = nNew; - } - - rc = fts3SegReaderRequire(pReader, pNext, nSuffix+FTS3_VARINT_MAX); - if( rc!=SQLITE_OK ) return rc; - - memcpy(&pReader->zTerm[nPrefix], pNext, nSuffix); - pReader->nTerm = nPrefix+nSuffix; - pNext += nSuffix; - pNext += sqlite3Fts3GetVarint32(pNext, &pReader->nDoclist); - pReader->aDoclist = pNext; - pReader->pOffsetList = 0; - - /* Check that the doclist does not appear to extend past the end of the - ** b-tree node. And that the final byte of the doclist is 0x00. If either - ** of these statements is untrue, then the data structure is corrupt. - */ - if( &pReader->aDoclist[pReader->nDoclist]>&pReader->aNode[pReader->nNode] - || (pReader->nPopulate==0 && pReader->aDoclist[pReader->nDoclist-1]) - ){ - return FTS_CORRUPT_VTAB; - } - return SQLITE_OK; -} - -/* -** Set the SegReader to point to the first docid in the doclist associated -** with the current term. -*/ -static int fts3SegReaderFirstDocid(Fts3Table *pTab, Fts3SegReader *pReader){ - int rc = SQLITE_OK; - assert( pReader->aDoclist ); - assert( !pReader->pOffsetList ); - if( pTab->bDescIdx && fts3SegReaderIsPending(pReader) ){ - u8 bEof = 0; - pReader->iDocid = 0; - pReader->nOffsetList = 0; - sqlite3Fts3DoclistPrev(0, - pReader->aDoclist, pReader->nDoclist, &pReader->pOffsetList, - &pReader->iDocid, &pReader->nOffsetList, &bEof - ); - }else{ - rc = fts3SegReaderRequire(pReader, pReader->aDoclist, FTS3_VARINT_MAX); - if( rc==SQLITE_OK ){ - int n = sqlite3Fts3GetVarint(pReader->aDoclist, &pReader->iDocid); - pReader->pOffsetList = &pReader->aDoclist[n]; - } - } - return rc; -} - -/* -** Advance the SegReader to point to the next docid in the doclist -** associated with the current term. -** -** If arguments ppOffsetList and pnOffsetList are not NULL, then -** *ppOffsetList is set to point to the first column-offset list -** in the doclist entry (i.e. immediately past the docid varint). -** *pnOffsetList is set to the length of the set of column-offset -** lists, not including the nul-terminator byte. For example: -*/ -static int fts3SegReaderNextDocid( - Fts3Table *pTab, - Fts3SegReader *pReader, /* Reader to advance to next docid */ - char **ppOffsetList, /* OUT: Pointer to current position-list */ - int *pnOffsetList /* OUT: Length of *ppOffsetList in bytes */ -){ - int rc = SQLITE_OK; - char *p = pReader->pOffsetList; - char c = 0; - - assert( p ); - - if( pTab->bDescIdx && fts3SegReaderIsPending(pReader) ){ - /* A pending-terms seg-reader for an FTS4 table that uses order=desc. - ** Pending-terms doclists are always built up in ascending order, so - ** we have to iterate through them backwards here. */ - u8 bEof = 0; - if( ppOffsetList ){ - *ppOffsetList = pReader->pOffsetList; - *pnOffsetList = pReader->nOffsetList - 1; - } - sqlite3Fts3DoclistPrev(0, - pReader->aDoclist, pReader->nDoclist, &p, &pReader->iDocid, - &pReader->nOffsetList, &bEof - ); - if( bEof ){ - pReader->pOffsetList = 0; - }else{ - pReader->pOffsetList = p; - } - }else{ - char *pEnd = &pReader->aDoclist[pReader->nDoclist]; - - /* Pointer p currently points at the first byte of an offset list. The - ** following block advances it to point one byte past the end of - ** the same offset list. */ - while( 1 ){ - - /* The following line of code (and the "p++" below the while() loop) is - ** normally all that is required to move pointer p to the desired - ** position. The exception is if this node is being loaded from disk - ** incrementally and pointer "p" now points to the first byte passed - ** the populated part of pReader->aNode[]. - */ - while( *p | c ) c = *p++ & 0x80; - assert( *p==0 ); - - if( pReader->pBlob==0 || p<&pReader->aNode[pReader->nPopulate] ) break; - rc = fts3SegReaderIncrRead(pReader); - if( rc!=SQLITE_OK ) return rc; - } - p++; - - /* If required, populate the output variables with a pointer to and the - ** size of the previous offset-list. - */ - if( ppOffsetList ){ - *ppOffsetList = pReader->pOffsetList; - *pnOffsetList = (int)(p - pReader->pOffsetList - 1); - } - - while( p<pEnd && *p==0 ) p++; - - /* If there are no more entries in the doclist, set pOffsetList to - ** NULL. Otherwise, set Fts3SegReader.iDocid to the next docid and - ** Fts3SegReader.pOffsetList to point to the next offset list before - ** returning. - */ - if( p>=pEnd ){ - pReader->pOffsetList = 0; - }else{ - rc = fts3SegReaderRequire(pReader, p, FTS3_VARINT_MAX); - if( rc==SQLITE_OK ){ - sqlite3_int64 iDelta; - pReader->pOffsetList = p + sqlite3Fts3GetVarint(p, &iDelta); - if( pTab->bDescIdx ){ - pReader->iDocid -= iDelta; - }else{ - pReader->iDocid += iDelta; - } - } - } - } - - return SQLITE_OK; -} - - -int sqlite3Fts3MsrOvfl( - Fts3Cursor *pCsr, - Fts3MultiSegReader *pMsr, - int *pnOvfl -){ - Fts3Table *p = (Fts3Table*)pCsr->base.pVtab; - int nOvfl = 0; - int ii; - int rc = SQLITE_OK; - int pgsz = p->nPgsz; - - assert( p->bFts4 ); - assert( pgsz>0 ); - - for(ii=0; rc==SQLITE_OK && ii<pMsr->nSegment; ii++){ - Fts3SegReader *pReader = pMsr->apSegment[ii]; - if( !fts3SegReaderIsPending(pReader) - && !fts3SegReaderIsRootOnly(pReader) - ){ - sqlite3_int64 jj; - for(jj=pReader->iStartBlock; jj<=pReader->iLeafEndBlock; jj++){ - int nBlob; - rc = sqlite3Fts3ReadBlock(p, jj, 0, &nBlob, 0); - if( rc!=SQLITE_OK ) break; - if( (nBlob+35)>pgsz ){ - nOvfl += (nBlob + 34)/pgsz; - } - } - } - } - *pnOvfl = nOvfl; - return rc; -} - -/* -** Free all allocations associated with the iterator passed as the -** second argument. -*/ -void sqlite3Fts3SegReaderFree(Fts3SegReader *pReader){ - if( pReader && !fts3SegReaderIsPending(pReader) ){ - sqlite3_free(pReader->zTerm); - if( !fts3SegReaderIsRootOnly(pReader) ){ - sqlite3_free(pReader->aNode); - sqlite3_blob_close(pReader->pBlob); - } - } - sqlite3_free(pReader); -} - -/* -** Allocate a new SegReader object. -*/ -int sqlite3Fts3SegReaderNew( - int iAge, /* Segment "age". */ - int bLookup, /* True for a lookup only */ - sqlite3_int64 iStartLeaf, /* First leaf to traverse */ - sqlite3_int64 iEndLeaf, /* Final leaf to traverse */ - sqlite3_int64 iEndBlock, /* Final block of segment */ - const char *zRoot, /* Buffer containing root node */ - int nRoot, /* Size of buffer containing root node */ - Fts3SegReader **ppReader /* OUT: Allocated Fts3SegReader */ -){ - Fts3SegReader *pReader; /* Newly allocated SegReader object */ - int nExtra = 0; /* Bytes to allocate segment root node */ - - assert( iStartLeaf<=iEndLeaf ); - if( iStartLeaf==0 ){ - nExtra = nRoot + FTS3_NODE_PADDING; - } - - pReader = (Fts3SegReader *)sqlite3_malloc(sizeof(Fts3SegReader) + nExtra); - if( !pReader ){ - return SQLITE_NOMEM; - } - memset(pReader, 0, sizeof(Fts3SegReader)); - pReader->iIdx = iAge; - pReader->bLookup = bLookup!=0; - pReader->iStartBlock = iStartLeaf; - pReader->iLeafEndBlock = iEndLeaf; - pReader->iEndBlock = iEndBlock; - - if( nExtra ){ - /* The entire segment is stored in the root node. */ - pReader->aNode = (char *)&pReader[1]; - pReader->rootOnly = 1; - pReader->nNode = nRoot; - memcpy(pReader->aNode, zRoot, nRoot); - memset(&pReader->aNode[nRoot], 0, FTS3_NODE_PADDING); - }else{ - pReader->iCurrentBlock = iStartLeaf-1; - } - *ppReader = pReader; - return SQLITE_OK; -} - -/* -** This is a comparison function used as a qsort() callback when sorting -** an array of pending terms by term. This occurs as part of flushing -** the contents of the pending-terms hash table to the database. -*/ -static int fts3CompareElemByTerm(const void *lhs, const void *rhs){ - char *z1 = fts3HashKey(*(Fts3HashElem **)lhs); - char *z2 = fts3HashKey(*(Fts3HashElem **)rhs); - int n1 = fts3HashKeysize(*(Fts3HashElem **)lhs); - int n2 = fts3HashKeysize(*(Fts3HashElem **)rhs); - - int n = (n1<n2 ? n1 : n2); - int c = memcmp(z1, z2, n); - if( c==0 ){ - c = n1 - n2; - } - return c; -} - -/* -** This function is used to allocate an Fts3SegReader that iterates through -** a subset of the terms stored in the Fts3Table.pendingTerms array. -** -** If the isPrefixIter parameter is zero, then the returned SegReader iterates -** through each term in the pending-terms table. Or, if isPrefixIter is -** non-zero, it iterates through each term and its prefixes. For example, if -** the pending terms hash table contains the terms "sqlite", "mysql" and -** "firebird", then the iterator visits the following 'terms' (in the order -** shown): -** -** f fi fir fire fireb firebi firebir firebird -** m my mys mysq mysql -** s sq sql sqli sqlit sqlite -** -** Whereas if isPrefixIter is zero, the terms visited are: -** -** firebird mysql sqlite -*/ -int sqlite3Fts3SegReaderPending( - Fts3Table *p, /* Virtual table handle */ - int iIndex, /* Index for p->aIndex */ - const char *zTerm, /* Term to search for */ - int nTerm, /* Size of buffer zTerm */ - int bPrefix, /* True for a prefix iterator */ - Fts3SegReader **ppReader /* OUT: SegReader for pending-terms */ -){ - Fts3SegReader *pReader = 0; /* Fts3SegReader object to return */ - Fts3HashElem *pE; /* Iterator variable */ - Fts3HashElem **aElem = 0; /* Array of term hash entries to scan */ - int nElem = 0; /* Size of array at aElem */ - int rc = SQLITE_OK; /* Return Code */ - Fts3Hash *pHash; - - pHash = &p->aIndex[iIndex].hPending; - if( bPrefix ){ - int nAlloc = 0; /* Size of allocated array at aElem */ - - for(pE=fts3HashFirst(pHash); pE; pE=fts3HashNext(pE)){ - char *zKey = (char *)fts3HashKey(pE); - int nKey = fts3HashKeysize(pE); - if( nTerm==0 || (nKey>=nTerm && 0==memcmp(zKey, zTerm, nTerm)) ){ - if( nElem==nAlloc ){ - Fts3HashElem **aElem2; - nAlloc += 16; - aElem2 = (Fts3HashElem **)sqlite3_realloc( - aElem, nAlloc*sizeof(Fts3HashElem *) - ); - if( !aElem2 ){ - rc = SQLITE_NOMEM; - nElem = 0; - break; - } - aElem = aElem2; - } - - aElem[nElem++] = pE; - } - } - - /* If more than one term matches the prefix, sort the Fts3HashElem - ** objects in term order using qsort(). This uses the same comparison - ** callback as is used when flushing terms to disk. - */ - if( nElem>1 ){ - qsort(aElem, nElem, sizeof(Fts3HashElem *), fts3CompareElemByTerm); - } - - }else{ - /* The query is a simple term lookup that matches at most one term in - ** the index. All that is required is a straight hash-lookup. - ** - ** Because the stack address of pE may be accessed via the aElem pointer - ** below, the "Fts3HashElem *pE" must be declared so that it is valid - ** within this entire function, not just this "else{...}" block. - */ - pE = fts3HashFindElem(pHash, zTerm, nTerm); - if( pE ){ - aElem = &pE; - nElem = 1; - } - } - - if( nElem>0 ){ - int nByte = sizeof(Fts3SegReader) + (nElem+1)*sizeof(Fts3HashElem *); - pReader = (Fts3SegReader *)sqlite3_malloc(nByte); - if( !pReader ){ - rc = SQLITE_NOMEM; - }else{ - memset(pReader, 0, nByte); - pReader->iIdx = 0x7FFFFFFF; - pReader->ppNextElem = (Fts3HashElem **)&pReader[1]; - memcpy(pReader->ppNextElem, aElem, nElem*sizeof(Fts3HashElem *)); - } - } - - if( bPrefix ){ - sqlite3_free(aElem); - } - *ppReader = pReader; - return rc; -} - -/* -** Compare the entries pointed to by two Fts3SegReader structures. -** Comparison is as follows: -** -** 1) EOF is greater than not EOF. -** -** 2) The current terms (if any) are compared using memcmp(). If one -** term is a prefix of another, the longer term is considered the -** larger. -** -** 3) By segment age. An older segment is considered larger. -*/ -static int fts3SegReaderCmp(Fts3SegReader *pLhs, Fts3SegReader *pRhs){ - int rc; - if( pLhs->aNode && pRhs->aNode ){ - int rc2 = pLhs->nTerm - pRhs->nTerm; - if( rc2<0 ){ - rc = memcmp(pLhs->zTerm, pRhs->zTerm, pLhs->nTerm); - }else{ - rc = memcmp(pLhs->zTerm, pRhs->zTerm, pRhs->nTerm); - } - if( rc==0 ){ - rc = rc2; - } - }else{ - rc = (pLhs->aNode==0) - (pRhs->aNode==0); - } - if( rc==0 ){ - rc = pRhs->iIdx - pLhs->iIdx; - } - assert( rc!=0 ); - return rc; -} - -/* -** A different comparison function for SegReader structures. In this -** version, it is assumed that each SegReader points to an entry in -** a doclist for identical terms. Comparison is made as follows: -** -** 1) EOF (end of doclist in this case) is greater than not EOF. -** -** 2) By current docid. -** -** 3) By segment age. An older segment is considered larger. -*/ -static int fts3SegReaderDoclistCmp(Fts3SegReader *pLhs, Fts3SegReader *pRhs){ - int rc = (pLhs->pOffsetList==0)-(pRhs->pOffsetList==0); - if( rc==0 ){ - if( pLhs->iDocid==pRhs->iDocid ){ - rc = pRhs->iIdx - pLhs->iIdx; - }else{ - rc = (pLhs->iDocid > pRhs->iDocid) ? 1 : -1; - } - } - assert( pLhs->aNode && pRhs->aNode ); - return rc; -} -static int fts3SegReaderDoclistCmpRev(Fts3SegReader *pLhs, Fts3SegReader *pRhs){ - int rc = (pLhs->pOffsetList==0)-(pRhs->pOffsetList==0); - if( rc==0 ){ - if( pLhs->iDocid==pRhs->iDocid ){ - rc = pRhs->iIdx - pLhs->iIdx; - }else{ - rc = (pLhs->iDocid < pRhs->iDocid) ? 1 : -1; - } - } - assert( pLhs->aNode && pRhs->aNode ); - return rc; -} - -/* -** Compare the term that the Fts3SegReader object passed as the first argument -** points to with the term specified by arguments zTerm and nTerm. -** -** If the pSeg iterator is already at EOF, return 0. Otherwise, return -** -ve if the pSeg term is less than zTerm/nTerm, 0 if the two terms are -** equal, or +ve if the pSeg term is greater than zTerm/nTerm. -*/ -static int fts3SegReaderTermCmp( - Fts3SegReader *pSeg, /* Segment reader object */ - const char *zTerm, /* Term to compare to */ - int nTerm /* Size of term zTerm in bytes */ -){ - int res = 0; - if( pSeg->aNode ){ - if( pSeg->nTerm>nTerm ){ - res = memcmp(pSeg->zTerm, zTerm, nTerm); - }else{ - res = memcmp(pSeg->zTerm, zTerm, pSeg->nTerm); - } - if( res==0 ){ - res = pSeg->nTerm-nTerm; - } - } - return res; -} - -/* -** Argument apSegment is an array of nSegment elements. It is known that -** the final (nSegment-nSuspect) members are already in sorted order -** (according to the comparison function provided). This function shuffles -** the array around until all entries are in sorted order. -*/ -static void fts3SegReaderSort( - Fts3SegReader **apSegment, /* Array to sort entries of */ - int nSegment, /* Size of apSegment array */ - int nSuspect, /* Unsorted entry count */ - int (*xCmp)(Fts3SegReader *, Fts3SegReader *) /* Comparison function */ -){ - int i; /* Iterator variable */ - - assert( nSuspect<=nSegment ); - - if( nSuspect==nSegment ) nSuspect--; - for(i=nSuspect-1; i>=0; i--){ - int j; - for(j=i; j<(nSegment-1); j++){ - Fts3SegReader *pTmp; - if( xCmp(apSegment[j], apSegment[j+1])<0 ) break; - pTmp = apSegment[j+1]; - apSegment[j+1] = apSegment[j]; - apSegment[j] = pTmp; - } - } - -#ifndef NDEBUG - /* Check that the list really is sorted now. */ - for(i=0; i<(nSuspect-1); i++){ - assert( xCmp(apSegment[i], apSegment[i+1])<0 ); - } -#endif -} - -/* -** Insert a record into the %_segments table. -*/ -static int fts3WriteSegment( - Fts3Table *p, /* Virtual table handle */ - sqlite3_int64 iBlock, /* Block id for new block */ - char *z, /* Pointer to buffer containing block data */ - int n /* Size of buffer z in bytes */ -){ - sqlite3_stmt *pStmt; - int rc = fts3SqlStmt(p, SQL_INSERT_SEGMENTS, &pStmt, 0); - if( rc==SQLITE_OK ){ - sqlite3_bind_int64(pStmt, 1, iBlock); - sqlite3_bind_blob(pStmt, 2, z, n, SQLITE_STATIC); - sqlite3_step(pStmt); - rc = sqlite3_reset(pStmt); - } - return rc; -} - -/* -** Find the largest relative level number in the table. If successful, set -** *pnMax to this value and return SQLITE_OK. Otherwise, if an error occurs, -** set *pnMax to zero and return an SQLite error code. -*/ -int sqlite3Fts3MaxLevel(Fts3Table *p, int *pnMax){ - int rc; - int mxLevel = 0; - sqlite3_stmt *pStmt = 0; - - rc = fts3SqlStmt(p, SQL_SELECT_MXLEVEL, &pStmt, 0); - if( rc==SQLITE_OK ){ - if( SQLITE_ROW==sqlite3_step(pStmt) ){ - mxLevel = sqlite3_column_int(pStmt, 0); - } - rc = sqlite3_reset(pStmt); - } - *pnMax = mxLevel; - return rc; -} - -/* -** Insert a record into the %_segdir table. -*/ -static int fts3WriteSegdir( - Fts3Table *p, /* Virtual table handle */ - sqlite3_int64 iLevel, /* Value for "level" field (absolute level) */ - int iIdx, /* Value for "idx" field */ - sqlite3_int64 iStartBlock, /* Value for "start_block" field */ - sqlite3_int64 iLeafEndBlock, /* Value for "leaves_end_block" field */ - sqlite3_int64 iEndBlock, /* Value for "end_block" field */ - char *zRoot, /* Blob value for "root" field */ - int nRoot /* Number of bytes in buffer zRoot */ -){ - sqlite3_stmt *pStmt; - int rc = fts3SqlStmt(p, SQL_INSERT_SEGDIR, &pStmt, 0); - if( rc==SQLITE_OK ){ - sqlite3_bind_int64(pStmt, 1, iLevel); - sqlite3_bind_int(pStmt, 2, iIdx); - sqlite3_bind_int64(pStmt, 3, iStartBlock); - sqlite3_bind_int64(pStmt, 4, iLeafEndBlock); - sqlite3_bind_int64(pStmt, 5, iEndBlock); - sqlite3_bind_blob(pStmt, 6, zRoot, nRoot, SQLITE_STATIC); - sqlite3_step(pStmt); - rc = sqlite3_reset(pStmt); - } - return rc; -} - -/* -** Return the size of the common prefix (if any) shared by zPrev and -** zNext, in bytes. For example, -** -** fts3PrefixCompress("abc", 3, "abcdef", 6) // returns 3 -** fts3PrefixCompress("abX", 3, "abcdef", 6) // returns 2 -** fts3PrefixCompress("abX", 3, "Xbcdef", 6) // returns 0 -*/ -static int fts3PrefixCompress( - const char *zPrev, /* Buffer containing previous term */ - int nPrev, /* Size of buffer zPrev in bytes */ - const char *zNext, /* Buffer containing next term */ - int nNext /* Size of buffer zNext in bytes */ -){ - int n; - UNUSED_PARAMETER(nNext); - for(n=0; n<nPrev && zPrev[n]==zNext[n]; n++); - return n; -} - -/* -** Add term zTerm to the SegmentNode. It is guaranteed that zTerm is larger -** (according to memcmp) than the previous term. -*/ -static int fts3NodeAddTerm( - Fts3Table *p, /* Virtual table handle */ - SegmentNode **ppTree, /* IN/OUT: SegmentNode handle */ - int isCopyTerm, /* True if zTerm/nTerm is transient */ - const char *zTerm, /* Pointer to buffer containing term */ - int nTerm /* Size of term in bytes */ -){ - SegmentNode *pTree = *ppTree; - int rc; - SegmentNode *pNew; - - /* First try to append the term to the current node. Return early if - ** this is possible. - */ - if( pTree ){ - int nData = pTree->nData; /* Current size of node in bytes */ - int nReq = nData; /* Required space after adding zTerm */ - int nPrefix; /* Number of bytes of prefix compression */ - int nSuffix; /* Suffix length */ - - nPrefix = fts3PrefixCompress(pTree->zTerm, pTree->nTerm, zTerm, nTerm); - nSuffix = nTerm-nPrefix; - - nReq += sqlite3Fts3VarintLen(nPrefix)+sqlite3Fts3VarintLen(nSuffix)+nSuffix; - if( nReq<=p->nNodeSize || !pTree->zTerm ){ - - if( nReq>p->nNodeSize ){ - /* An unusual case: this is the first term to be added to the node - ** and the static node buffer (p->nNodeSize bytes) is not large - ** enough. Use a separately malloced buffer instead This wastes - ** p->nNodeSize bytes, but since this scenario only comes about when - ** the database contain two terms that share a prefix of almost 2KB, - ** this is not expected to be a serious problem. - */ - assert( pTree->aData==(char *)&pTree[1] ); - pTree->aData = (char *)sqlite3_malloc(nReq); - if( !pTree->aData ){ - return SQLITE_NOMEM; - } - } - - if( pTree->zTerm ){ - /* There is no prefix-length field for first term in a node */ - nData += sqlite3Fts3PutVarint(&pTree->aData[nData], nPrefix); - } - - nData += sqlite3Fts3PutVarint(&pTree->aData[nData], nSuffix); - memcpy(&pTree->aData[nData], &zTerm[nPrefix], nSuffix); - pTree->nData = nData + nSuffix; - pTree->nEntry++; - - if( isCopyTerm ){ - if( pTree->nMalloc<nTerm ){ - char *zNew = sqlite3_realloc(pTree->zMalloc, nTerm*2); - if( !zNew ){ - return SQLITE_NOMEM; - } - pTree->nMalloc = nTerm*2; - pTree->zMalloc = zNew; - } - pTree->zTerm = pTree->zMalloc; - memcpy(pTree->zTerm, zTerm, nTerm); - pTree->nTerm = nTerm; - }else{ - pTree->zTerm = (char *)zTerm; - pTree->nTerm = nTerm; - } - return SQLITE_OK; - } - } - - /* If control flows to here, it was not possible to append zTerm to the - ** current node. Create a new node (a right-sibling of the current node). - ** If this is the first node in the tree, the term is added to it. - ** - ** Otherwise, the term is not added to the new node, it is left empty for - ** now. Instead, the term is inserted into the parent of pTree. If pTree - ** has no parent, one is created here. - */ - pNew = (SegmentNode *)sqlite3_malloc(sizeof(SegmentNode) + p->nNodeSize); - if( !pNew ){ - return SQLITE_NOMEM; - } - memset(pNew, 0, sizeof(SegmentNode)); - pNew->nData = 1 + FTS3_VARINT_MAX; - pNew->aData = (char *)&pNew[1]; - - if( pTree ){ - SegmentNode *pParent = pTree->pParent; - rc = fts3NodeAddTerm(p, &pParent, isCopyTerm, zTerm, nTerm); - if( pTree->pParent==0 ){ - pTree->pParent = pParent; - } - pTree->pRight = pNew; - pNew->pLeftmost = pTree->pLeftmost; - pNew->pParent = pParent; - pNew->zMalloc = pTree->zMalloc; - pNew->nMalloc = pTree->nMalloc; - pTree->zMalloc = 0; - }else{ - pNew->pLeftmost = pNew; - rc = fts3NodeAddTerm(p, &pNew, isCopyTerm, zTerm, nTerm); - } - - *ppTree = pNew; - return rc; -} - -/* -** Helper function for fts3NodeWrite(). -*/ -static int fts3TreeFinishNode( - SegmentNode *pTree, - int iHeight, - sqlite3_int64 iLeftChild -){ - int nStart; - assert( iHeight>=1 && iHeight<128 ); - nStart = FTS3_VARINT_MAX - sqlite3Fts3VarintLen(iLeftChild); - pTree->aData[nStart] = (char)iHeight; - sqlite3Fts3PutVarint(&pTree->aData[nStart+1], iLeftChild); - return nStart; -} - -/* -** Write the buffer for the segment node pTree and all of its peers to the -** database. Then call this function recursively to write the parent of -** pTree and its peers to the database. -** -** Except, if pTree is a root node, do not write it to the database. Instead, -** set output variables *paRoot and *pnRoot to contain the root node. -** -** If successful, SQLITE_OK is returned and output variable *piLast is -** set to the largest blockid written to the database (or zero if no -** blocks were written to the db). Otherwise, an SQLite error code is -** returned. -*/ -static int fts3NodeWrite( - Fts3Table *p, /* Virtual table handle */ - SegmentNode *pTree, /* SegmentNode handle */ - int iHeight, /* Height of this node in tree */ - sqlite3_int64 iLeaf, /* Block id of first leaf node */ - sqlite3_int64 iFree, /* Block id of next free slot in %_segments */ - sqlite3_int64 *piLast, /* OUT: Block id of last entry written */ - char **paRoot, /* OUT: Data for root node */ - int *pnRoot /* OUT: Size of root node in bytes */ -){ - int rc = SQLITE_OK; - - if( !pTree->pParent ){ - /* Root node of the tree. */ - int nStart = fts3TreeFinishNode(pTree, iHeight, iLeaf); - *piLast = iFree-1; - *pnRoot = pTree->nData - nStart; - *paRoot = &pTree->aData[nStart]; - }else{ - SegmentNode *pIter; - sqlite3_int64 iNextFree = iFree; - sqlite3_int64 iNextLeaf = iLeaf; - for(pIter=pTree->pLeftmost; pIter && rc==SQLITE_OK; pIter=pIter->pRight){ - int nStart = fts3TreeFinishNode(pIter, iHeight, iNextLeaf); - int nWrite = pIter->nData - nStart; - - rc = fts3WriteSegment(p, iNextFree, &pIter->aData[nStart], nWrite); - iNextFree++; - iNextLeaf += (pIter->nEntry+1); - } - if( rc==SQLITE_OK ){ - assert( iNextLeaf==iFree ); - rc = fts3NodeWrite( - p, pTree->pParent, iHeight+1, iFree, iNextFree, piLast, paRoot, pnRoot - ); - } - } - - return rc; -} - -/* -** Free all memory allocations associated with the tree pTree. -*/ -static void fts3NodeFree(SegmentNode *pTree){ - if( pTree ){ - SegmentNode *p = pTree->pLeftmost; - fts3NodeFree(p->pParent); - while( p ){ - SegmentNode *pRight = p->pRight; - if( p->aData!=(char *)&p[1] ){ - sqlite3_free(p->aData); - } - assert( pRight==0 || p->zMalloc==0 ); - sqlite3_free(p->zMalloc); - sqlite3_free(p); - p = pRight; - } - } -} - -/* -** Add a term to the segment being constructed by the SegmentWriter object -** *ppWriter. When adding the first term to a segment, *ppWriter should -** be passed NULL. This function will allocate a new SegmentWriter object -** and return it via the input/output variable *ppWriter in this case. -** -** If successful, SQLITE_OK is returned. Otherwise, an SQLite error code. -*/ -static int fts3SegWriterAdd( - Fts3Table *p, /* Virtual table handle */ - SegmentWriter **ppWriter, /* IN/OUT: SegmentWriter handle */ - int isCopyTerm, /* True if buffer zTerm must be copied */ - const char *zTerm, /* Pointer to buffer containing term */ - int nTerm, /* Size of term in bytes */ - const char *aDoclist, /* Pointer to buffer containing doclist */ - int nDoclist /* Size of doclist in bytes */ -){ - int nPrefix; /* Size of term prefix in bytes */ - int nSuffix; /* Size of term suffix in bytes */ - int nReq; /* Number of bytes required on leaf page */ - int nData; - SegmentWriter *pWriter = *ppWriter; - - if( !pWriter ){ - int rc; - sqlite3_stmt *pStmt; - - /* Allocate the SegmentWriter structure */ - pWriter = (SegmentWriter *)sqlite3_malloc(sizeof(SegmentWriter)); - if( !pWriter ) return SQLITE_NOMEM; - memset(pWriter, 0, sizeof(SegmentWriter)); - *ppWriter = pWriter; - - /* Allocate a buffer in which to accumulate data */ - pWriter->aData = (char *)sqlite3_malloc(p->nNodeSize); - if( !pWriter->aData ) return SQLITE_NOMEM; - pWriter->nSize = p->nNodeSize; - - /* Find the next free blockid in the %_segments table */ - rc = fts3SqlStmt(p, SQL_NEXT_SEGMENTS_ID, &pStmt, 0); - if( rc!=SQLITE_OK ) return rc; - if( SQLITE_ROW==sqlite3_step(pStmt) ){ - pWriter->iFree = sqlite3_column_int64(pStmt, 0); - pWriter->iFirst = pWriter->iFree; - } - rc = sqlite3_reset(pStmt); - if( rc!=SQLITE_OK ) return rc; - } - nData = pWriter->nData; - - nPrefix = fts3PrefixCompress(pWriter->zTerm, pWriter->nTerm, zTerm, nTerm); - nSuffix = nTerm-nPrefix; - - /* Figure out how many bytes are required by this new entry */ - nReq = sqlite3Fts3VarintLen(nPrefix) + /* varint containing prefix size */ - sqlite3Fts3VarintLen(nSuffix) + /* varint containing suffix size */ - nSuffix + /* Term suffix */ - sqlite3Fts3VarintLen(nDoclist) + /* Size of doclist */ - nDoclist; /* Doclist data */ - - if( nData>0 && nData+nReq>p->nNodeSize ){ - int rc; - - /* The current leaf node is full. Write it out to the database. */ - rc = fts3WriteSegment(p, pWriter->iFree++, pWriter->aData, nData); - if( rc!=SQLITE_OK ) return rc; - p->nLeafAdd++; - - /* Add the current term to the interior node tree. The term added to - ** the interior tree must: - ** - ** a) be greater than the largest term on the leaf node just written - ** to the database (still available in pWriter->zTerm), and - ** - ** b) be less than or equal to the term about to be added to the new - ** leaf node (zTerm/nTerm). - ** - ** In other words, it must be the prefix of zTerm 1 byte longer than - ** the common prefix (if any) of zTerm and pWriter->zTerm. - */ - assert( nPrefix<nTerm ); - rc = fts3NodeAddTerm(p, &pWriter->pTree, isCopyTerm, zTerm, nPrefix+1); - if( rc!=SQLITE_OK ) return rc; - - nData = 0; - pWriter->nTerm = 0; - - nPrefix = 0; - nSuffix = nTerm; - nReq = 1 + /* varint containing prefix size */ - sqlite3Fts3VarintLen(nTerm) + /* varint containing suffix size */ - nTerm + /* Term suffix */ - sqlite3Fts3VarintLen(nDoclist) + /* Size of doclist */ - nDoclist; /* Doclist data */ - } - - /* If the buffer currently allocated is too small for this entry, realloc - ** the buffer to make it large enough. - */ - if( nReq>pWriter->nSize ){ - char *aNew = sqlite3_realloc(pWriter->aData, nReq); - if( !aNew ) return SQLITE_NOMEM; - pWriter->aData = aNew; - pWriter->nSize = nReq; - } - assert( nData+nReq<=pWriter->nSize ); - - /* Append the prefix-compressed term and doclist to the buffer. */ - nData += sqlite3Fts3PutVarint(&pWriter->aData[nData], nPrefix); - nData += sqlite3Fts3PutVarint(&pWriter->aData[nData], nSuffix); - memcpy(&pWriter->aData[nData], &zTerm[nPrefix], nSuffix); - nData += nSuffix; - nData += sqlite3Fts3PutVarint(&pWriter->aData[nData], nDoclist); - memcpy(&pWriter->aData[nData], aDoclist, nDoclist); - pWriter->nData = nData + nDoclist; - - /* Save the current term so that it can be used to prefix-compress the next. - ** If the isCopyTerm parameter is true, then the buffer pointed to by - ** zTerm is transient, so take a copy of the term data. Otherwise, just - ** store a copy of the pointer. - */ - if( isCopyTerm ){ - if( nTerm>pWriter->nMalloc ){ - char *zNew = sqlite3_realloc(pWriter->zMalloc, nTerm*2); - if( !zNew ){ - return SQLITE_NOMEM; - } - pWriter->nMalloc = nTerm*2; - pWriter->zMalloc = zNew; - pWriter->zTerm = zNew; - } - assert( pWriter->zTerm==pWriter->zMalloc ); - memcpy(pWriter->zTerm, zTerm, nTerm); - }else{ - pWriter->zTerm = (char *)zTerm; - } - pWriter->nTerm = nTerm; - - return SQLITE_OK; -} - -/* -** Flush all data associated with the SegmentWriter object pWriter to the -** database. This function must be called after all terms have been added -** to the segment using fts3SegWriterAdd(). If successful, SQLITE_OK is -** returned. Otherwise, an SQLite error code. -*/ -static int fts3SegWriterFlush( - Fts3Table *p, /* Virtual table handle */ - SegmentWriter *pWriter, /* SegmentWriter to flush to the db */ - sqlite3_int64 iLevel, /* Value for 'level' column of %_segdir */ - int iIdx /* Value for 'idx' column of %_segdir */ -){ - int rc; /* Return code */ - if( pWriter->pTree ){ - sqlite3_int64 iLast = 0; /* Largest block id written to database */ - sqlite3_int64 iLastLeaf; /* Largest leaf block id written to db */ - char *zRoot = NULL; /* Pointer to buffer containing root node */ - int nRoot = 0; /* Size of buffer zRoot */ - - iLastLeaf = pWriter->iFree; - rc = fts3WriteSegment(p, pWriter->iFree++, pWriter->aData, pWriter->nData); - if( rc==SQLITE_OK ){ - rc = fts3NodeWrite(p, pWriter->pTree, 1, - pWriter->iFirst, pWriter->iFree, &iLast, &zRoot, &nRoot); - } - if( rc==SQLITE_OK ){ - rc = fts3WriteSegdir( - p, iLevel, iIdx, pWriter->iFirst, iLastLeaf, iLast, zRoot, nRoot); - } - }else{ - /* The entire tree fits on the root node. Write it to the segdir table. */ - rc = fts3WriteSegdir( - p, iLevel, iIdx, 0, 0, 0, pWriter->aData, pWriter->nData); - } - p->nLeafAdd++; - return rc; -} - -/* -** Release all memory held by the SegmentWriter object passed as the -** first argument. -*/ -static void fts3SegWriterFree(SegmentWriter *pWriter){ - if( pWriter ){ - sqlite3_free(pWriter->aData); - sqlite3_free(pWriter->zMalloc); - fts3NodeFree(pWriter->pTree); - sqlite3_free(pWriter); - } -} - -/* -** The first value in the apVal[] array is assumed to contain an integer. -** This function tests if there exist any documents with docid values that -** are different from that integer. i.e. if deleting the document with docid -** pRowid would mean the FTS3 table were empty. -** -** If successful, *pisEmpty is set to true if the table is empty except for -** document pRowid, or false otherwise, and SQLITE_OK is returned. If an -** error occurs, an SQLite error code is returned. -*/ -static int fts3IsEmpty(Fts3Table *p, sqlite3_value *pRowid, int *pisEmpty){ - sqlite3_stmt *pStmt; - int rc; - if( p->zContentTbl ){ - /* If using the content=xxx option, assume the table is never empty */ - *pisEmpty = 0; - rc = SQLITE_OK; - }else{ - rc = fts3SqlStmt(p, SQL_IS_EMPTY, &pStmt, &pRowid); - if( rc==SQLITE_OK ){ - if( SQLITE_ROW==sqlite3_step(pStmt) ){ - *pisEmpty = sqlite3_column_int(pStmt, 0); - } - rc = sqlite3_reset(pStmt); - } - } - return rc; -} - -/* -** Set *pnMax to the largest segment level in the database for the index -** iIndex. -** -** Segment levels are stored in the 'level' column of the %_segdir table. -** -** Return SQLITE_OK if successful, or an SQLite error code if not. -*/ -static int fts3SegmentMaxLevel( - Fts3Table *p, - int iLangid, - int iIndex, - sqlite3_int64 *pnMax -){ - sqlite3_stmt *pStmt; - int rc; - assert( iIndex>=0 && iIndex<p->nIndex ); - - /* Set pStmt to the compiled version of: - ** - ** SELECT max(level) FROM %Q.'%q_segdir' WHERE level BETWEEN ? AND ? - ** - ** (1024 is actually the value of macro FTS3_SEGDIR_PREFIXLEVEL_STR). - */ - rc = fts3SqlStmt(p, SQL_SELECT_SEGDIR_MAX_LEVEL, &pStmt, 0); - if( rc!=SQLITE_OK ) return rc; - sqlite3_bind_int64(pStmt, 1, getAbsoluteLevel(p, iLangid, iIndex, 0)); - sqlite3_bind_int64(pStmt, 2, - getAbsoluteLevel(p, iLangid, iIndex, FTS3_SEGDIR_MAXLEVEL-1) - ); - if( SQLITE_ROW==sqlite3_step(pStmt) ){ - *pnMax = sqlite3_column_int64(pStmt, 0); - } - return sqlite3_reset(pStmt); -} - -/* -** Delete all entries in the %_segments table associated with the segment -** opened with seg-reader pSeg. This function does not affect the contents -** of the %_segdir table. -*/ -static int fts3DeleteSegment( - Fts3Table *p, /* FTS table handle */ - Fts3SegReader *pSeg /* Segment to delete */ -){ - int rc = SQLITE_OK; /* Return code */ - if( pSeg->iStartBlock ){ - sqlite3_stmt *pDelete; /* SQL statement to delete rows */ - rc = fts3SqlStmt(p, SQL_DELETE_SEGMENTS_RANGE, &pDelete, 0); - if( rc==SQLITE_OK ){ - sqlite3_bind_int64(pDelete, 1, pSeg->iStartBlock); - sqlite3_bind_int64(pDelete, 2, pSeg->iEndBlock); - sqlite3_step(pDelete); - rc = sqlite3_reset(pDelete); - } - } - return rc; -} - -/* -** This function is used after merging multiple segments into a single large -** segment to delete the old, now redundant, segment b-trees. Specifically, -** it: -** -** 1) Deletes all %_segments entries for the segments associated with -** each of the SegReader objects in the array passed as the third -** argument, and -** -** 2) deletes all %_segdir entries with level iLevel, or all %_segdir -** entries regardless of level if (iLevel<0). -** -** SQLITE_OK is returned if successful, otherwise an SQLite error code. -*/ -static int fts3DeleteSegdir( - Fts3Table *p, /* Virtual table handle */ - int iLangid, /* Language id */ - int iIndex, /* Index for p->aIndex */ - int iLevel, /* Level of %_segdir entries to delete */ - Fts3SegReader **apSegment, /* Array of SegReader objects */ - int nReader /* Size of array apSegment */ -){ - int rc = SQLITE_OK; /* Return Code */ - int i; /* Iterator variable */ - sqlite3_stmt *pDelete = 0; /* SQL statement to delete rows */ - - for(i=0; rc==SQLITE_OK && i<nReader; i++){ - rc = fts3DeleteSegment(p, apSegment[i]); - } - if( rc!=SQLITE_OK ){ - return rc; - } - - assert( iLevel>=0 || iLevel==FTS3_SEGCURSOR_ALL ); - if( iLevel==FTS3_SEGCURSOR_ALL ){ - rc = fts3SqlStmt(p, SQL_DELETE_SEGDIR_RANGE, &pDelete, 0); - if( rc==SQLITE_OK ){ - sqlite3_bind_int64(pDelete, 1, getAbsoluteLevel(p, iLangid, iIndex, 0)); - sqlite3_bind_int64(pDelete, 2, - getAbsoluteLevel(p, iLangid, iIndex, FTS3_SEGDIR_MAXLEVEL-1) - ); - } - }else{ - rc = fts3SqlStmt(p, SQL_DELETE_SEGDIR_LEVEL, &pDelete, 0); - if( rc==SQLITE_OK ){ - sqlite3_bind_int64( - pDelete, 1, getAbsoluteLevel(p, iLangid, iIndex, iLevel) - ); - } - } - - if( rc==SQLITE_OK ){ - sqlite3_step(pDelete); - rc = sqlite3_reset(pDelete); - } - - return rc; -} - -/* -** When this function is called, buffer *ppList (size *pnList bytes) contains -** a position list that may (or may not) feature multiple columns. This -** function adjusts the pointer *ppList and the length *pnList so that they -** identify the subset of the position list that corresponds to column iCol. -** -** If there are no entries in the input position list for column iCol, then -** *pnList is set to zero before returning. -*/ -static void fts3ColumnFilter( - int iCol, /* Column to filter on */ - char **ppList, /* IN/OUT: Pointer to position list */ - int *pnList /* IN/OUT: Size of buffer *ppList in bytes */ -){ - char *pList = *ppList; - int nList = *pnList; - char *pEnd = &pList[nList]; - int iCurrent = 0; - char *p = pList; - - assert( iCol>=0 ); - while( 1 ){ - char c = 0; - while( p<pEnd && (c | *p)&0xFE ) c = *p++ & 0x80; - - if( iCol==iCurrent ){ - nList = (int)(p - pList); - break; - } - - nList -= (int)(p - pList); - pList = p; - if( nList==0 ){ - break; - } - p = &pList[1]; - p += sqlite3Fts3GetVarint32(p, &iCurrent); - } - - *ppList = pList; - *pnList = nList; -} - -/* -** Cache data in the Fts3MultiSegReader.aBuffer[] buffer (overwriting any -** existing data). Grow the buffer if required. -** -** If successful, return SQLITE_OK. Otherwise, if an OOM error is encountered -** trying to resize the buffer, return SQLITE_NOMEM. -*/ -static int fts3MsrBufferData( - Fts3MultiSegReader *pMsr, /* Multi-segment-reader handle */ - char *pList, - int nList -){ - if( nList>pMsr->nBuffer ){ - char *pNew; - pMsr->nBuffer = nList*2; - pNew = (char *)sqlite3_realloc(pMsr->aBuffer, pMsr->nBuffer); - if( !pNew ) return SQLITE_NOMEM; - pMsr->aBuffer = pNew; - } - - memcpy(pMsr->aBuffer, pList, nList); - return SQLITE_OK; -} - -int sqlite3Fts3MsrIncrNext( - Fts3Table *p, /* Virtual table handle */ - Fts3MultiSegReader *pMsr, /* Multi-segment-reader handle */ - sqlite3_int64 *piDocid, /* OUT: Docid value */ - char **paPoslist, /* OUT: Pointer to position list */ - int *pnPoslist /* OUT: Size of position list in bytes */ -){ - int nMerge = pMsr->nAdvance; - Fts3SegReader **apSegment = pMsr->apSegment; - int (*xCmp)(Fts3SegReader *, Fts3SegReader *) = ( - p->bDescIdx ? fts3SegReaderDoclistCmpRev : fts3SegReaderDoclistCmp - ); - - if( nMerge==0 ){ - *paPoslist = 0; - return SQLITE_OK; - } - - while( 1 ){ - Fts3SegReader *pSeg; - pSeg = pMsr->apSegment[0]; - - if( pSeg->pOffsetList==0 ){ - *paPoslist = 0; - break; - }else{ - int rc; - char *pList; - int nList; - int j; - sqlite3_int64 iDocid = apSegment[0]->iDocid; - - rc = fts3SegReaderNextDocid(p, apSegment[0], &pList, &nList); - j = 1; - while( rc==SQLITE_OK - && j<nMerge - && apSegment[j]->pOffsetList - && apSegment[j]->iDocid==iDocid - ){ - rc = fts3SegReaderNextDocid(p, apSegment[j], 0, 0); - j++; - } - if( rc!=SQLITE_OK ) return rc; - fts3SegReaderSort(pMsr->apSegment, nMerge, j, xCmp); - - if( pMsr->iColFilter>=0 ){ - fts3ColumnFilter(pMsr->iColFilter, &pList, &nList); - } - - if( nList>0 ){ - if( fts3SegReaderIsPending(apSegment[0]) ){ - rc = fts3MsrBufferData(pMsr, pList, nList+1); - if( rc!=SQLITE_OK ) return rc; - *paPoslist = pMsr->aBuffer; - assert( (pMsr->aBuffer[nList] & 0xFE)==0x00 ); - }else{ - *paPoslist = pList; - } - *piDocid = iDocid; - *pnPoslist = nList; - break; - } - } - } - - return SQLITE_OK; -} - -static int fts3SegReaderStart( - Fts3Table *p, /* Virtual table handle */ - Fts3MultiSegReader *pCsr, /* Cursor object */ - const char *zTerm, /* Term searched for (or NULL) */ - int nTerm /* Length of zTerm in bytes */ -){ - int i; - int nSeg = pCsr->nSegment; - - /* If the Fts3SegFilter defines a specific term (or term prefix) to search - ** for, then advance each segment iterator until it points to a term of - ** equal or greater value than the specified term. This prevents many - ** unnecessary merge/sort operations for the case where single segment - ** b-tree leaf nodes contain more than one term. - */ - for(i=0; pCsr->bRestart==0 && i<pCsr->nSegment; i++){ - int res = 0; - Fts3SegReader *pSeg = pCsr->apSegment[i]; - do { - int rc = fts3SegReaderNext(p, pSeg, 0); - if( rc!=SQLITE_OK ) return rc; - }while( zTerm && (res = fts3SegReaderTermCmp(pSeg, zTerm, nTerm))<0 ); - - if( pSeg->bLookup && res!=0 ){ - fts3SegReaderSetEof(pSeg); - } - } - fts3SegReaderSort(pCsr->apSegment, nSeg, nSeg, fts3SegReaderCmp); - - return SQLITE_OK; -} - -int sqlite3Fts3SegReaderStart( - Fts3Table *p, /* Virtual table handle */ - Fts3MultiSegReader *pCsr, /* Cursor object */ - Fts3SegFilter *pFilter /* Restrictions on range of iteration */ -){ - pCsr->pFilter = pFilter; - return fts3SegReaderStart(p, pCsr, pFilter->zTerm, pFilter->nTerm); -} - -int sqlite3Fts3MsrIncrStart( - Fts3Table *p, /* Virtual table handle */ - Fts3MultiSegReader *pCsr, /* Cursor object */ - int iCol, /* Column to match on. */ - const char *zTerm, /* Term to iterate through a doclist for */ - int nTerm /* Number of bytes in zTerm */ -){ - int i; - int rc; - int nSegment = pCsr->nSegment; - int (*xCmp)(Fts3SegReader *, Fts3SegReader *) = ( - p->bDescIdx ? fts3SegReaderDoclistCmpRev : fts3SegReaderDoclistCmp - ); - - assert( pCsr->pFilter==0 ); - assert( zTerm && nTerm>0 ); - - /* Advance each segment iterator until it points to the term zTerm/nTerm. */ - rc = fts3SegReaderStart(p, pCsr, zTerm, nTerm); - if( rc!=SQLITE_OK ) return rc; - - /* Determine how many of the segments actually point to zTerm/nTerm. */ - for(i=0; i<nSegment; i++){ - Fts3SegReader *pSeg = pCsr->apSegment[i]; - if( !pSeg->aNode || fts3SegReaderTermCmp(pSeg, zTerm, nTerm) ){ - break; - } - } - pCsr->nAdvance = i; - - /* Advance each of the segments to point to the first docid. */ - for(i=0; i<pCsr->nAdvance; i++){ - rc = fts3SegReaderFirstDocid(p, pCsr->apSegment[i]); - if( rc!=SQLITE_OK ) return rc; - } - fts3SegReaderSort(pCsr->apSegment, i, i, xCmp); - - assert( iCol<0 || iCol<p->nColumn ); - pCsr->iColFilter = iCol; - - return SQLITE_OK; -} - -/* -** This function is called on a MultiSegReader that has been started using -** sqlite3Fts3MsrIncrStart(). One or more calls to MsrIncrNext() may also -** have been made. Calling this function puts the MultiSegReader in such -** a state that if the next two calls are: -** -** sqlite3Fts3SegReaderStart() -** sqlite3Fts3SegReaderStep() -** -** then the entire doclist for the term is available in -** MultiSegReader.aDoclist/nDoclist. -*/ -int sqlite3Fts3MsrIncrRestart(Fts3MultiSegReader *pCsr){ - int i; /* Used to iterate through segment-readers */ - - assert( pCsr->zTerm==0 ); - assert( pCsr->nTerm==0 ); - assert( pCsr->aDoclist==0 ); - assert( pCsr->nDoclist==0 ); - - pCsr->nAdvance = 0; - pCsr->bRestart = 1; - for(i=0; i<pCsr->nSegment; i++){ - pCsr->apSegment[i]->pOffsetList = 0; - pCsr->apSegment[i]->nOffsetList = 0; - pCsr->apSegment[i]->iDocid = 0; - } - - return SQLITE_OK; -} - - -int sqlite3Fts3SegReaderStep( - Fts3Table *p, /* Virtual table handle */ - Fts3MultiSegReader *pCsr /* Cursor object */ -){ - int rc = SQLITE_OK; - - int isIgnoreEmpty = (pCsr->pFilter->flags & FTS3_SEGMENT_IGNORE_EMPTY); - int isRequirePos = (pCsr->pFilter->flags & FTS3_SEGMENT_REQUIRE_POS); - int isColFilter = (pCsr->pFilter->flags & FTS3_SEGMENT_COLUMN_FILTER); - int isPrefix = (pCsr->pFilter->flags & FTS3_SEGMENT_PREFIX); - int isScan = (pCsr->pFilter->flags & FTS3_SEGMENT_SCAN); - int isFirst = (pCsr->pFilter->flags & FTS3_SEGMENT_FIRST); - - Fts3SegReader **apSegment = pCsr->apSegment; - int nSegment = pCsr->nSegment; - Fts3SegFilter *pFilter = pCsr->pFilter; - int (*xCmp)(Fts3SegReader *, Fts3SegReader *) = ( - p->bDescIdx ? fts3SegReaderDoclistCmpRev : fts3SegReaderDoclistCmp - ); - - if( pCsr->nSegment==0 ) return SQLITE_OK; - - do { - int nMerge; - int i; - - /* Advance the first pCsr->nAdvance entries in the apSegment[] array - ** forward. Then sort the list in order of current term again. - */ - for(i=0; i<pCsr->nAdvance; i++){ - Fts3SegReader *pSeg = apSegment[i]; - if( pSeg->bLookup ){ - fts3SegReaderSetEof(pSeg); - }else{ - rc = fts3SegReaderNext(p, pSeg, 0); - } - if( rc!=SQLITE_OK ) return rc; - } - fts3SegReaderSort(apSegment, nSegment, pCsr->nAdvance, fts3SegReaderCmp); - pCsr->nAdvance = 0; - - /* If all the seg-readers are at EOF, we're finished. return SQLITE_OK. */ - assert( rc==SQLITE_OK ); - if( apSegment[0]->aNode==0 ) break; - - pCsr->nTerm = apSegment[0]->nTerm; - pCsr->zTerm = apSegment[0]->zTerm; - - /* If this is a prefix-search, and if the term that apSegment[0] points - ** to does not share a suffix with pFilter->zTerm/nTerm, then all - ** required callbacks have been made. In this case exit early. - ** - ** Similarly, if this is a search for an exact match, and the first term - ** of segment apSegment[0] is not a match, exit early. - */ - if( pFilter->zTerm && !isScan ){ - if( pCsr->nTerm<pFilter->nTerm - || (!isPrefix && pCsr->nTerm>pFilter->nTerm) - || memcmp(pCsr->zTerm, pFilter->zTerm, pFilter->nTerm) - ){ - break; - } - } - - nMerge = 1; - while( nMerge<nSegment - && apSegment[nMerge]->aNode - && apSegment[nMerge]->nTerm==pCsr->nTerm - && 0==memcmp(pCsr->zTerm, apSegment[nMerge]->zTerm, pCsr->nTerm) - ){ - nMerge++; - } - - assert( isIgnoreEmpty || (isRequirePos && !isColFilter) ); - if( nMerge==1 - && !isIgnoreEmpty - && !isFirst - && (p->bDescIdx==0 || fts3SegReaderIsPending(apSegment[0])==0) - ){ - pCsr->nDoclist = apSegment[0]->nDoclist; - if( fts3SegReaderIsPending(apSegment[0]) ){ - rc = fts3MsrBufferData(pCsr, apSegment[0]->aDoclist, pCsr->nDoclist); - pCsr->aDoclist = pCsr->aBuffer; - }else{ - pCsr->aDoclist = apSegment[0]->aDoclist; - } - if( rc==SQLITE_OK ) rc = SQLITE_ROW; - }else{ - int nDoclist = 0; /* Size of doclist */ - sqlite3_int64 iPrev = 0; /* Previous docid stored in doclist */ - - /* The current term of the first nMerge entries in the array - ** of Fts3SegReader objects is the same. The doclists must be merged - ** and a single term returned with the merged doclist. - */ - for(i=0; i<nMerge; i++){ - fts3SegReaderFirstDocid(p, apSegment[i]); - } - fts3SegReaderSort(apSegment, nMerge, nMerge, xCmp); - while( apSegment[0]->pOffsetList ){ - int j; /* Number of segments that share a docid */ - char *pList; - int nList; - int nByte; - sqlite3_int64 iDocid = apSegment[0]->iDocid; - fts3SegReaderNextDocid(p, apSegment[0], &pList, &nList); - j = 1; - while( j<nMerge - && apSegment[j]->pOffsetList - && apSegment[j]->iDocid==iDocid - ){ - fts3SegReaderNextDocid(p, apSegment[j], 0, 0); - j++; - } - - if( isColFilter ){ - fts3ColumnFilter(pFilter->iCol, &pList, &nList); - } - - if( !isIgnoreEmpty || nList>0 ){ - - /* Calculate the 'docid' delta value to write into the merged - ** doclist. */ - sqlite3_int64 iDelta; - if( p->bDescIdx && nDoclist>0 ){ - iDelta = iPrev - iDocid; - }else{ - iDelta = iDocid - iPrev; - } - assert( iDelta>0 || (nDoclist==0 && iDelta==iDocid) ); - assert( nDoclist>0 || iDelta==iDocid ); - - nByte = sqlite3Fts3VarintLen(iDelta) + (isRequirePos?nList+1:0); - if( nDoclist+nByte>pCsr->nBuffer ){ - char *aNew; - pCsr->nBuffer = (nDoclist+nByte)*2; - aNew = sqlite3_realloc(pCsr->aBuffer, pCsr->nBuffer); - if( !aNew ){ - return SQLITE_NOMEM; - } - pCsr->aBuffer = aNew; - } - - if( isFirst ){ - char *a = &pCsr->aBuffer[nDoclist]; - int nWrite; - - nWrite = sqlite3Fts3FirstFilter(iDelta, pList, nList, a); - if( nWrite ){ - iPrev = iDocid; - nDoclist += nWrite; - } - }else{ - nDoclist += sqlite3Fts3PutVarint(&pCsr->aBuffer[nDoclist], iDelta); - iPrev = iDocid; - if( isRequirePos ){ - memcpy(&pCsr->aBuffer[nDoclist], pList, nList); - nDoclist += nList; - pCsr->aBuffer[nDoclist++] = '\0'; - } - } - } - - fts3SegReaderSort(apSegment, nMerge, j, xCmp); - } - if( nDoclist>0 ){ - pCsr->aDoclist = pCsr->aBuffer; - pCsr->nDoclist = nDoclist; - rc = SQLITE_ROW; - } - } - pCsr->nAdvance = nMerge; - }while( rc==SQLITE_OK ); - - return rc; -} - - -void sqlite3Fts3SegReaderFinish( - Fts3MultiSegReader *pCsr /* Cursor object */ -){ - if( pCsr ){ - int i; - for(i=0; i<pCsr->nSegment; i++){ - sqlite3Fts3SegReaderFree(pCsr->apSegment[i]); - } - sqlite3_free(pCsr->apSegment); - sqlite3_free(pCsr->aBuffer); - - pCsr->nSegment = 0; - pCsr->apSegment = 0; - pCsr->aBuffer = 0; - } -} - -/* -** Merge all level iLevel segments in the database into a single -** iLevel+1 segment. Or, if iLevel<0, merge all segments into a -** single segment with a level equal to the numerically largest level -** currently present in the database. -** -** If this function is called with iLevel<0, but there is only one -** segment in the database, SQLITE_DONE is returned immediately. -** Otherwise, if successful, SQLITE_OK is returned. If an error occurs, -** an SQLite error code is returned. -*/ -static int fts3SegmentMerge( - Fts3Table *p, - int iLangid, /* Language id to merge */ - int iIndex, /* Index in p->aIndex[] to merge */ - int iLevel /* Level to merge */ -){ - int rc; /* Return code */ - int iIdx = 0; /* Index of new segment */ - sqlite3_int64 iNewLevel = 0; /* Level/index to create new segment at */ - SegmentWriter *pWriter = 0; /* Used to write the new, merged, segment */ - Fts3SegFilter filter; /* Segment term filter condition */ - Fts3MultiSegReader csr; /* Cursor to iterate through level(s) */ - int bIgnoreEmpty = 0; /* True to ignore empty segments */ - - assert( iLevel==FTS3_SEGCURSOR_ALL - || iLevel==FTS3_SEGCURSOR_PENDING - || iLevel>=0 - ); - assert( iLevel<FTS3_SEGDIR_MAXLEVEL ); - assert( iIndex>=0 && iIndex<p->nIndex ); - - rc = sqlite3Fts3SegReaderCursor(p, iLangid, iIndex, iLevel, 0, 0, 1, 0, &csr); - if( rc!=SQLITE_OK || csr.nSegment==0 ) goto finished; - - if( iLevel==FTS3_SEGCURSOR_ALL ){ - /* This call is to merge all segments in the database to a single - ** segment. The level of the new segment is equal to the numerically - ** greatest segment level currently present in the database for this - ** index. The idx of the new segment is always 0. */ - if( csr.nSegment==1 ){ - rc = SQLITE_DONE; - goto finished; - } - rc = fts3SegmentMaxLevel(p, iLangid, iIndex, &iNewLevel); - bIgnoreEmpty = 1; - - }else if( iLevel==FTS3_SEGCURSOR_PENDING ){ - iNewLevel = getAbsoluteLevel(p, iLangid, iIndex, 0); - rc = fts3AllocateSegdirIdx(p, iLangid, iIndex, 0, &iIdx); - }else{ - /* This call is to merge all segments at level iLevel. find the next - ** available segment index at level iLevel+1. The call to - ** fts3AllocateSegdirIdx() will merge the segments at level iLevel+1 to - ** a single iLevel+2 segment if necessary. */ - rc = fts3AllocateSegdirIdx(p, iLangid, iIndex, iLevel+1, &iIdx); - iNewLevel = getAbsoluteLevel(p, iLangid, iIndex, iLevel+1); - } - if( rc!=SQLITE_OK ) goto finished; - assert( csr.nSegment>0 ); - assert( iNewLevel>=getAbsoluteLevel(p, iLangid, iIndex, 0) ); - assert( iNewLevel<getAbsoluteLevel(p, iLangid, iIndex,FTS3_SEGDIR_MAXLEVEL) ); - - memset(&filter, 0, sizeof(Fts3SegFilter)); - filter.flags = FTS3_SEGMENT_REQUIRE_POS; - filter.flags |= (bIgnoreEmpty ? FTS3_SEGMENT_IGNORE_EMPTY : 0); - - rc = sqlite3Fts3SegReaderStart(p, &csr, &filter); - while( SQLITE_OK==rc ){ - rc = sqlite3Fts3SegReaderStep(p, &csr); - if( rc!=SQLITE_ROW ) break; - rc = fts3SegWriterAdd(p, &pWriter, 1, - csr.zTerm, csr.nTerm, csr.aDoclist, csr.nDoclist); - } - if( rc!=SQLITE_OK ) goto finished; - assert( pWriter ); - - if( iLevel!=FTS3_SEGCURSOR_PENDING ){ - rc = fts3DeleteSegdir( - p, iLangid, iIndex, iLevel, csr.apSegment, csr.nSegment - ); - if( rc!=SQLITE_OK ) goto finished; - } - rc = fts3SegWriterFlush(p, pWriter, iNewLevel, iIdx); - - finished: - fts3SegWriterFree(pWriter); - sqlite3Fts3SegReaderFinish(&csr); - return rc; -} - - -/* -** Flush the contents of pendingTerms to level 0 segments. -*/ -int sqlite3Fts3PendingTermsFlush(Fts3Table *p){ - int rc = SQLITE_OK; - int i; - - for(i=0; rc==SQLITE_OK && i<p->nIndex; i++){ - rc = fts3SegmentMerge(p, p->iPrevLangid, i, FTS3_SEGCURSOR_PENDING); - if( rc==SQLITE_DONE ) rc = SQLITE_OK; - } - sqlite3Fts3PendingTermsClear(p); - - /* Determine the auto-incr-merge setting if unknown. If enabled, - ** estimate the number of leaf blocks of content to be written - */ - if( rc==SQLITE_OK && p->bHasStat - && p->bAutoincrmerge==0xff && p->nLeafAdd>0 - ){ - sqlite3_stmt *pStmt = 0; - rc = fts3SqlStmt(p, SQL_SELECT_STAT, &pStmt, 0); - if( rc==SQLITE_OK ){ - sqlite3_bind_int(pStmt, 1, FTS_STAT_AUTOINCRMERGE); - rc = sqlite3_step(pStmt); - p->bAutoincrmerge = (rc==SQLITE_ROW && sqlite3_column_int(pStmt, 0)); - rc = sqlite3_reset(pStmt); - } - } - return rc; -} - -/* -** Encode N integers as varints into a blob. -*/ -static void fts3EncodeIntArray( - int N, /* The number of integers to encode */ - u32 *a, /* The integer values */ - char *zBuf, /* Write the BLOB here */ - int *pNBuf /* Write number of bytes if zBuf[] used here */ -){ - int i, j; - for(i=j=0; i<N; i++){ - j += sqlite3Fts3PutVarint(&zBuf[j], (sqlite3_int64)a[i]); - } - *pNBuf = j; -} - -/* -** Decode a blob of varints into N integers -*/ -static void fts3DecodeIntArray( - int N, /* The number of integers to decode */ - u32 *a, /* Write the integer values */ - const char *zBuf, /* The BLOB containing the varints */ - int nBuf /* size of the BLOB */ -){ - int i, j; - UNUSED_PARAMETER(nBuf); - for(i=j=0; i<N; i++){ - sqlite3_int64 x; - j += sqlite3Fts3GetVarint(&zBuf[j], &x); - assert(j<=nBuf); - a[i] = (u32)(x & 0xffffffff); - } -} - -/* -** Insert the sizes (in tokens) for each column of the document -** with docid equal to p->iPrevDocid. The sizes are encoded as -** a blob of varints. -*/ -static void fts3InsertDocsize( - int *pRC, /* Result code */ - Fts3Table *p, /* Table into which to insert */ - u32 *aSz /* Sizes of each column, in tokens */ -){ - char *pBlob; /* The BLOB encoding of the document size */ - int nBlob; /* Number of bytes in the BLOB */ - sqlite3_stmt *pStmt; /* Statement used to insert the encoding */ - int rc; /* Result code from subfunctions */ - - if( *pRC ) return; - pBlob = sqlite3_malloc( 10*p->nColumn ); - if( pBlob==0 ){ - *pRC = SQLITE_NOMEM; - return; - } - fts3EncodeIntArray(p->nColumn, aSz, pBlob, &nBlob); - rc = fts3SqlStmt(p, SQL_REPLACE_DOCSIZE, &pStmt, 0); - if( rc ){ - sqlite3_free(pBlob); - *pRC = rc; - return; - } - sqlite3_bind_int64(pStmt, 1, p->iPrevDocid); - sqlite3_bind_blob(pStmt, 2, pBlob, nBlob, sqlite3_free); - sqlite3_step(pStmt); - *pRC = sqlite3_reset(pStmt); -} - -/* -** Record 0 of the %_stat table contains a blob consisting of N varints, -** where N is the number of user defined columns in the fts3 table plus -** two. If nCol is the number of user defined columns, then values of the -** varints are set as follows: -** -** Varint 0: Total number of rows in the table. -** -** Varint 1..nCol: For each column, the total number of tokens stored in -** the column for all rows of the table. -** -** Varint 1+nCol: The total size, in bytes, of all text values in all -** columns of all rows of the table. -** -*/ -static void fts3UpdateDocTotals( - int *pRC, /* The result code */ - Fts3Table *p, /* Table being updated */ - u32 *aSzIns, /* Size increases */ - u32 *aSzDel, /* Size decreases */ - int nChng /* Change in the number of documents */ -){ - char *pBlob; /* Storage for BLOB written into %_stat */ - int nBlob; /* Size of BLOB written into %_stat */ - u32 *a; /* Array of integers that becomes the BLOB */ - sqlite3_stmt *pStmt; /* Statement for reading and writing */ - int i; /* Loop counter */ - int rc; /* Result code from subfunctions */ - - const int nStat = p->nColumn+2; - - if( *pRC ) return; - a = sqlite3_malloc( (sizeof(u32)+10)*nStat ); - if( a==0 ){ - *pRC = SQLITE_NOMEM; - return; - } - pBlob = (char*)&a[nStat]; - rc = fts3SqlStmt(p, SQL_SELECT_STAT, &pStmt, 0); - if( rc ){ - sqlite3_free(a); - *pRC = rc; - return; - } - sqlite3_bind_int(pStmt, 1, FTS_STAT_DOCTOTAL); - if( sqlite3_step(pStmt)==SQLITE_ROW ){ - fts3DecodeIntArray(nStat, a, - sqlite3_column_blob(pStmt, 0), - sqlite3_column_bytes(pStmt, 0)); - }else{ - memset(a, 0, sizeof(u32)*(nStat) ); - } - rc = sqlite3_reset(pStmt); - if( rc!=SQLITE_OK ){ - sqlite3_free(a); - *pRC = rc; - return; - } - if( nChng<0 && a[0]<(u32)(-nChng) ){ - a[0] = 0; - }else{ - a[0] += nChng; - } - for(i=0; i<p->nColumn+1; i++){ - u32 x = a[i+1]; - if( x+aSzIns[i] < aSzDel[i] ){ - x = 0; - }else{ - x = x + aSzIns[i] - aSzDel[i]; - } - a[i+1] = x; - } - fts3EncodeIntArray(nStat, a, pBlob, &nBlob); - rc = fts3SqlStmt(p, SQL_REPLACE_STAT, &pStmt, 0); - if( rc ){ - sqlite3_free(a); - *pRC = rc; - return; - } - sqlite3_bind_int(pStmt, 1, FTS_STAT_DOCTOTAL); - sqlite3_bind_blob(pStmt, 2, pBlob, nBlob, SQLITE_STATIC); - sqlite3_step(pStmt); - *pRC = sqlite3_reset(pStmt); - sqlite3_free(a); -} - -/* -** Merge the entire database so that there is one segment for each -** iIndex/iLangid combination. -*/ -static int fts3DoOptimize(Fts3Table *p, int bReturnDone){ - int bSeenDone = 0; - int rc; - sqlite3_stmt *pAllLangid = 0; - - rc = fts3SqlStmt(p, SQL_SELECT_ALL_LANGID, &pAllLangid, 0); - if( rc==SQLITE_OK ){ - int rc2; - sqlite3_bind_int(pAllLangid, 1, p->nIndex); - while( sqlite3_step(pAllLangid)==SQLITE_ROW ){ - int i; - int iLangid = sqlite3_column_int(pAllLangid, 0); - for(i=0; rc==SQLITE_OK && i<p->nIndex; i++){ - rc = fts3SegmentMerge(p, iLangid, i, FTS3_SEGCURSOR_ALL); - if( rc==SQLITE_DONE ){ - bSeenDone = 1; - rc = SQLITE_OK; - } - } - } - rc2 = sqlite3_reset(pAllLangid); - if( rc==SQLITE_OK ) rc = rc2; - } - - sqlite3Fts3SegmentsClose(p); - sqlite3Fts3PendingTermsClear(p); - - return (rc==SQLITE_OK && bReturnDone && bSeenDone) ? SQLITE_DONE : rc; -} - -/* -** This function is called when the user executes the following statement: -** -** INSERT INTO <tbl>(<tbl>) VALUES('rebuild'); -** -** The entire FTS index is discarded and rebuilt. If the table is one -** created using the content=xxx option, then the new index is based on -** the current contents of the xxx table. Otherwise, it is rebuilt based -** on the contents of the %_content table. -*/ -static int fts3DoRebuild(Fts3Table *p){ - int rc; /* Return Code */ - - rc = fts3DeleteAll(p, 0); - if( rc==SQLITE_OK ){ - u32 *aSz = 0; - u32 *aSzIns = 0; - u32 *aSzDel = 0; - sqlite3_stmt *pStmt = 0; - int nEntry = 0; - - /* Compose and prepare an SQL statement to loop through the content table */ - char *zSql = sqlite3_mprintf("SELECT %s" , p->zReadExprlist); - if( !zSql ){ - rc = SQLITE_NOMEM; - }else{ - rc = sqlite3_prepare_v2(p->db, zSql, -1, &pStmt, 0); - sqlite3_free(zSql); - } - - if( rc==SQLITE_OK ){ - int nByte = sizeof(u32) * (p->nColumn+1)*3; - aSz = (u32 *)sqlite3_malloc(nByte); - if( aSz==0 ){ - rc = SQLITE_NOMEM; - }else{ - memset(aSz, 0, nByte); - aSzIns = &aSz[p->nColumn+1]; - aSzDel = &aSzIns[p->nColumn+1]; - } - } - - while( rc==SQLITE_OK && SQLITE_ROW==sqlite3_step(pStmt) ){ - int iCol; - int iLangid = langidFromSelect(p, pStmt); - rc = fts3PendingTermsDocid(p, iLangid, sqlite3_column_int64(pStmt, 0)); - memset(aSz, 0, sizeof(aSz[0]) * (p->nColumn+1)); - for(iCol=0; rc==SQLITE_OK && iCol<p->nColumn; iCol++){ - const char *z = (const char *) sqlite3_column_text(pStmt, iCol+1); - rc = fts3PendingTermsAdd(p, iLangid, z, iCol, &aSz[iCol]); - aSz[p->nColumn] += sqlite3_column_bytes(pStmt, iCol+1); - } - if( p->bHasDocsize ){ - fts3InsertDocsize(&rc, p, aSz); - } - if( rc!=SQLITE_OK ){ - sqlite3_finalize(pStmt); - pStmt = 0; - }else{ - nEntry++; - for(iCol=0; iCol<=p->nColumn; iCol++){ - aSzIns[iCol] += aSz[iCol]; - } - } - } - if( p->bFts4 ){ - fts3UpdateDocTotals(&rc, p, aSzIns, aSzDel, nEntry); - } - sqlite3_free(aSz); - - if( pStmt ){ - int rc2 = sqlite3_finalize(pStmt); - if( rc==SQLITE_OK ){ - rc = rc2; - } - } - } - - return rc; -} - - -/* -** This function opens a cursor used to read the input data for an -** incremental merge operation. Specifically, it opens a cursor to scan -** the oldest nSeg segments (idx=0 through idx=(nSeg-1)) in absolute -** level iAbsLevel. -*/ -static int fts3IncrmergeCsr( - Fts3Table *p, /* FTS3 table handle */ - sqlite3_int64 iAbsLevel, /* Absolute level to open */ - int nSeg, /* Number of segments to merge */ - Fts3MultiSegReader *pCsr /* Cursor object to populate */ -){ - int rc; /* Return Code */ - sqlite3_stmt *pStmt = 0; /* Statement used to read %_segdir entry */ - int nByte; /* Bytes allocated at pCsr->apSegment[] */ - - /* Allocate space for the Fts3MultiSegReader.aCsr[] array */ - memset(pCsr, 0, sizeof(*pCsr)); - nByte = sizeof(Fts3SegReader *) * nSeg; - pCsr->apSegment = (Fts3SegReader **)sqlite3_malloc(nByte); - - if( pCsr->apSegment==0 ){ - rc = SQLITE_NOMEM; - }else{ - memset(pCsr->apSegment, 0, nByte); - rc = fts3SqlStmt(p, SQL_SELECT_LEVEL, &pStmt, 0); - } - if( rc==SQLITE_OK ){ - int i; - int rc2; - sqlite3_bind_int64(pStmt, 1, iAbsLevel); - assert( pCsr->nSegment==0 ); - for(i=0; rc==SQLITE_OK && sqlite3_step(pStmt)==SQLITE_ROW && i<nSeg; i++){ - rc = sqlite3Fts3SegReaderNew(i, 0, - sqlite3_column_int64(pStmt, 1), /* segdir.start_block */ - sqlite3_column_int64(pStmt, 2), /* segdir.leaves_end_block */ - sqlite3_column_int64(pStmt, 3), /* segdir.end_block */ - sqlite3_column_blob(pStmt, 4), /* segdir.root */ - sqlite3_column_bytes(pStmt, 4), /* segdir.root */ - &pCsr->apSegment[i] - ); - pCsr->nSegment++; - } - rc2 = sqlite3_reset(pStmt); - if( rc==SQLITE_OK ) rc = rc2; - } - - return rc; -} - -typedef struct IncrmergeWriter IncrmergeWriter; -typedef struct NodeWriter NodeWriter; -typedef struct Blob Blob; -typedef struct NodeReader NodeReader; - -/* -** An instance of the following structure is used as a dynamic buffer -** to build up nodes or other blobs of data in. -** -** The function blobGrowBuffer() is used to extend the allocation. -*/ -struct Blob { - char *a; /* Pointer to allocation */ - int n; /* Number of valid bytes of data in a[] */ - int nAlloc; /* Allocated size of a[] (nAlloc>=n) */ -}; - -/* -** This structure is used to build up buffers containing segment b-tree -** nodes (blocks). -*/ -struct NodeWriter { - sqlite3_int64 iBlock; /* Current block id */ - Blob key; /* Last key written to the current block */ - Blob block; /* Current block image */ -}; - -/* -** An object of this type contains the state required to create or append -** to an appendable b-tree segment. -*/ -struct IncrmergeWriter { - int nLeafEst; /* Space allocated for leaf blocks */ - int nWork; /* Number of leaf pages flushed */ - sqlite3_int64 iAbsLevel; /* Absolute level of input segments */ - int iIdx; /* Index of *output* segment in iAbsLevel+1 */ - sqlite3_int64 iStart; /* Block number of first allocated block */ - sqlite3_int64 iEnd; /* Block number of last allocated block */ - NodeWriter aNodeWriter[FTS_MAX_APPENDABLE_HEIGHT]; -}; - -/* -** An object of the following type is used to read data from a single -** FTS segment node. See the following functions: -** -** nodeReaderInit() -** nodeReaderNext() -** nodeReaderRelease() -*/ -struct NodeReader { - const char *aNode; - int nNode; - int iOff; /* Current offset within aNode[] */ - - /* Output variables. Containing the current node entry. */ - sqlite3_int64 iChild; /* Pointer to child node */ - Blob term; /* Current term */ - const char *aDoclist; /* Pointer to doclist */ - int nDoclist; /* Size of doclist in bytes */ -}; - -/* -** If *pRc is not SQLITE_OK when this function is called, it is a no-op. -** Otherwise, if the allocation at pBlob->a is not already at least nMin -** bytes in size, extend (realloc) it to be so. -** -** If an OOM error occurs, set *pRc to SQLITE_NOMEM and leave pBlob->a -** unmodified. Otherwise, if the allocation succeeds, update pBlob->nAlloc -** to reflect the new size of the pBlob->a[] buffer. -*/ -static void blobGrowBuffer(Blob *pBlob, int nMin, int *pRc){ - if( *pRc==SQLITE_OK && nMin>pBlob->nAlloc ){ - int nAlloc = nMin; - char *a = (char *)sqlite3_realloc(pBlob->a, nAlloc); - if( a ){ - pBlob->nAlloc = nAlloc; - pBlob->a = a; - }else{ - *pRc = SQLITE_NOMEM; - } - } -} - -/* -** Attempt to advance the node-reader object passed as the first argument to -** the next entry on the node. -** -** Return an error code if an error occurs (SQLITE_NOMEM is possible). -** Otherwise return SQLITE_OK. If there is no next entry on the node -** (e.g. because the current entry is the last) set NodeReader->aNode to -** NULL to indicate EOF. Otherwise, populate the NodeReader structure output -** variables for the new entry. -*/ -static int nodeReaderNext(NodeReader *p){ - int bFirst = (p->term.n==0); /* True for first term on the node */ - int nPrefix = 0; /* Bytes to copy from previous term */ - int nSuffix = 0; /* Bytes to append to the prefix */ - int rc = SQLITE_OK; /* Return code */ - - assert( p->aNode ); - if( p->iChild && bFirst==0 ) p->iChild++; - if( p->iOff>=p->nNode ){ - /* EOF */ - p->aNode = 0; - }else{ - if( bFirst==0 ){ - p->iOff += sqlite3Fts3GetVarint32(&p->aNode[p->iOff], &nPrefix); - } - p->iOff += sqlite3Fts3GetVarint32(&p->aNode[p->iOff], &nSuffix); - - blobGrowBuffer(&p->term, nPrefix+nSuffix, &rc); - if( rc==SQLITE_OK ){ - memcpy(&p->term.a[nPrefix], &p->aNode[p->iOff], nSuffix); - p->term.n = nPrefix+nSuffix; - p->iOff += nSuffix; - if( p->iChild==0 ){ - p->iOff += sqlite3Fts3GetVarint32(&p->aNode[p->iOff], &p->nDoclist); - p->aDoclist = &p->aNode[p->iOff]; - p->iOff += p->nDoclist; - } - } - } - - assert( p->iOff<=p->nNode ); - - return rc; -} - -/* -** Release all dynamic resources held by node-reader object *p. -*/ -static void nodeReaderRelease(NodeReader *p){ - sqlite3_free(p->term.a); -} - -/* -** Initialize a node-reader object to read the node in buffer aNode/nNode. -** -** If successful, SQLITE_OK is returned and the NodeReader object set to -** point to the first entry on the node (if any). Otherwise, an SQLite -** error code is returned. -*/ -static int nodeReaderInit(NodeReader *p, const char *aNode, int nNode){ - memset(p, 0, sizeof(NodeReader)); - p->aNode = aNode; - p->nNode = nNode; - - /* Figure out if this is a leaf or an internal node. */ - if( p->aNode[0] ){ - /* An internal node. */ - p->iOff = 1 + sqlite3Fts3GetVarint(&p->aNode[1], &p->iChild); - }else{ - p->iOff = 1; - } - - return nodeReaderNext(p); -} - -/* -** This function is called while writing an FTS segment each time a leaf o -** node is finished and written to disk. The key (zTerm/nTerm) is guaranteed -** to be greater than the largest key on the node just written, but smaller -** than or equal to the first key that will be written to the next leaf -** node. -** -** The block id of the leaf node just written to disk may be found in -** (pWriter->aNodeWriter[0].iBlock) when this function is called. -*/ -static int fts3IncrmergePush( - Fts3Table *p, /* Fts3 table handle */ - IncrmergeWriter *pWriter, /* Writer object */ - const char *zTerm, /* Term to write to internal node */ - int nTerm /* Bytes at zTerm */ -){ - sqlite3_int64 iPtr = pWriter->aNodeWriter[0].iBlock; - int iLayer; - - assert( nTerm>0 ); - for(iLayer=1; ALWAYS(iLayer<FTS_MAX_APPENDABLE_HEIGHT); iLayer++){ - sqlite3_int64 iNextPtr = 0; - NodeWriter *pNode = &pWriter->aNodeWriter[iLayer]; - int rc = SQLITE_OK; - int nPrefix; - int nSuffix; - int nSpace; - - /* Figure out how much space the key will consume if it is written to - ** the current node of layer iLayer. Due to the prefix compression, - ** the space required changes depending on which node the key is to - ** be added to. */ - nPrefix = fts3PrefixCompress(pNode->key.a, pNode->key.n, zTerm, nTerm); - nSuffix = nTerm - nPrefix; - nSpace = sqlite3Fts3VarintLen(nPrefix); - nSpace += sqlite3Fts3VarintLen(nSuffix) + nSuffix; - - if( pNode->key.n==0 || (pNode->block.n + nSpace)<=p->nNodeSize ){ - /* If the current node of layer iLayer contains zero keys, or if adding - ** the key to it will not cause it to grow to larger than nNodeSize - ** bytes in size, write the key here. */ - - Blob *pBlk = &pNode->block; - if( pBlk->n==0 ){ - blobGrowBuffer(pBlk, p->nNodeSize, &rc); - if( rc==SQLITE_OK ){ - pBlk->a[0] = (char)iLayer; - pBlk->n = 1 + sqlite3Fts3PutVarint(&pBlk->a[1], iPtr); - } - } - blobGrowBuffer(pBlk, pBlk->n + nSpace, &rc); - blobGrowBuffer(&pNode->key, nTerm, &rc); - - if( rc==SQLITE_OK ){ - if( pNode->key.n ){ - pBlk->n += sqlite3Fts3PutVarint(&pBlk->a[pBlk->n], nPrefix); - } - pBlk->n += sqlite3Fts3PutVarint(&pBlk->a[pBlk->n], nSuffix); - memcpy(&pBlk->a[pBlk->n], &zTerm[nPrefix], nSuffix); - pBlk->n += nSuffix; - - memcpy(pNode->key.a, zTerm, nTerm); - pNode->key.n = nTerm; - } - }else{ - /* Otherwise, flush the current node of layer iLayer to disk. - ** Then allocate a new, empty sibling node. The key will be written - ** into the parent of this node. */ - rc = fts3WriteSegment(p, pNode->iBlock, pNode->block.a, pNode->block.n); - - assert( pNode->block.nAlloc>=p->nNodeSize ); - pNode->block.a[0] = (char)iLayer; - pNode->block.n = 1 + sqlite3Fts3PutVarint(&pNode->block.a[1], iPtr+1); - - iNextPtr = pNode->iBlock; - pNode->iBlock++; - pNode->key.n = 0; - } - - if( rc!=SQLITE_OK || iNextPtr==0 ) return rc; - iPtr = iNextPtr; - } - - assert( 0 ); - return 0; -} - -/* -** Append a term and (optionally) doclist to the FTS segment node currently -** stored in blob *pNode. The node need not contain any terms, but the -** header must be written before this function is called. -** -** A node header is a single 0x00 byte for a leaf node, or a height varint -** followed by the left-hand-child varint for an internal node. -** -** The term to be appended is passed via arguments zTerm/nTerm. For a -** leaf node, the doclist is passed as aDoclist/nDoclist. For an internal -** node, both aDoclist and nDoclist must be passed 0. -** -** If the size of the value in blob pPrev is zero, then this is the first -** term written to the node. Otherwise, pPrev contains a copy of the -** previous term. Before this function returns, it is updated to contain a -** copy of zTerm/nTerm. -** -** It is assumed that the buffer associated with pNode is already large -** enough to accommodate the new entry. The buffer associated with pPrev -** is extended by this function if requrired. -** -** If an error (i.e. OOM condition) occurs, an SQLite error code is -** returned. Otherwise, SQLITE_OK. -*/ -static int fts3AppendToNode( - Blob *pNode, /* Current node image to append to */ - Blob *pPrev, /* Buffer containing previous term written */ - const char *zTerm, /* New term to write */ - int nTerm, /* Size of zTerm in bytes */ - const char *aDoclist, /* Doclist (or NULL) to write */ - int nDoclist /* Size of aDoclist in bytes */ -){ - int rc = SQLITE_OK; /* Return code */ - int bFirst = (pPrev->n==0); /* True if this is the first term written */ - int nPrefix; /* Size of term prefix in bytes */ - int nSuffix; /* Size of term suffix in bytes */ - - /* Node must have already been started. There must be a doclist for a - ** leaf node, and there must not be a doclist for an internal node. */ - assert( pNode->n>0 ); - assert( (pNode->a[0]=='\0')==(aDoclist!=0) ); - - blobGrowBuffer(pPrev, nTerm, &rc); - if( rc!=SQLITE_OK ) return rc; - - nPrefix = fts3PrefixCompress(pPrev->a, pPrev->n, zTerm, nTerm); - nSuffix = nTerm - nPrefix; - memcpy(pPrev->a, zTerm, nTerm); - pPrev->n = nTerm; - - if( bFirst==0 ){ - pNode->n += sqlite3Fts3PutVarint(&pNode->a[pNode->n], nPrefix); - } - pNode->n += sqlite3Fts3PutVarint(&pNode->a[pNode->n], nSuffix); - memcpy(&pNode->a[pNode->n], &zTerm[nPrefix], nSuffix); - pNode->n += nSuffix; - - if( aDoclist ){ - pNode->n += sqlite3Fts3PutVarint(&pNode->a[pNode->n], nDoclist); - memcpy(&pNode->a[pNode->n], aDoclist, nDoclist); - pNode->n += nDoclist; - } - - assert( pNode->n<=pNode->nAlloc ); - - return SQLITE_OK; -} - -/* -** Append the current term and doclist pointed to by cursor pCsr to the -** appendable b-tree segment opened for writing by pWriter. -** -** Return SQLITE_OK if successful, or an SQLite error code otherwise. -*/ -static int fts3IncrmergeAppend( - Fts3Table *p, /* Fts3 table handle */ - IncrmergeWriter *pWriter, /* Writer object */ - Fts3MultiSegReader *pCsr /* Cursor containing term and doclist */ -){ - const char *zTerm = pCsr->zTerm; - int nTerm = pCsr->nTerm; - const char *aDoclist = pCsr->aDoclist; - int nDoclist = pCsr->nDoclist; - int rc = SQLITE_OK; /* Return code */ - int nSpace; /* Total space in bytes required on leaf */ - int nPrefix; /* Size of prefix shared with previous term */ - int nSuffix; /* Size of suffix (nTerm - nPrefix) */ - NodeWriter *pLeaf; /* Object used to write leaf nodes */ - - pLeaf = &pWriter->aNodeWriter[0]; - nPrefix = fts3PrefixCompress(pLeaf->key.a, pLeaf->key.n, zTerm, nTerm); - nSuffix = nTerm - nPrefix; - - nSpace = sqlite3Fts3VarintLen(nPrefix); - nSpace += sqlite3Fts3VarintLen(nSuffix) + nSuffix; - nSpace += sqlite3Fts3VarintLen(nDoclist) + nDoclist; - - /* If the current block is not empty, and if adding this term/doclist - ** to the current block would make it larger than Fts3Table.nNodeSize - ** bytes, write this block out to the database. */ - if( pLeaf->block.n>0 && (pLeaf->block.n + nSpace)>p->nNodeSize ){ - rc = fts3WriteSegment(p, pLeaf->iBlock, pLeaf->block.a, pLeaf->block.n); - pWriter->nWork++; - - /* Add the current term to the parent node. The term added to the - ** parent must: - ** - ** a) be greater than the largest term on the leaf node just written - ** to the database (still available in pLeaf->key), and - ** - ** b) be less than or equal to the term about to be added to the new - ** leaf node (zTerm/nTerm). - ** - ** In other words, it must be the prefix of zTerm 1 byte longer than - ** the common prefix (if any) of zTerm and pWriter->zTerm. - */ - if( rc==SQLITE_OK ){ - rc = fts3IncrmergePush(p, pWriter, zTerm, nPrefix+1); - } - - /* Advance to the next output block */ - pLeaf->iBlock++; - pLeaf->key.n = 0; - pLeaf->block.n = 0; - - nSuffix = nTerm; - nSpace = 1; - nSpace += sqlite3Fts3VarintLen(nSuffix) + nSuffix; - nSpace += sqlite3Fts3VarintLen(nDoclist) + nDoclist; - } - - blobGrowBuffer(&pLeaf->block, pLeaf->block.n + nSpace, &rc); - - if( rc==SQLITE_OK ){ - if( pLeaf->block.n==0 ){ - pLeaf->block.n = 1; - pLeaf->block.a[0] = '\0'; - } - rc = fts3AppendToNode( - &pLeaf->block, &pLeaf->key, zTerm, nTerm, aDoclist, nDoclist - ); - } - - return rc; -} - -/* -** This function is called to release all dynamic resources held by the -** merge-writer object pWriter, and if no error has occurred, to flush -** all outstanding node buffers held by pWriter to disk. -** -** If *pRc is not SQLITE_OK when this function is called, then no attempt -** is made to write any data to disk. Instead, this function serves only -** to release outstanding resources. -** -** Otherwise, if *pRc is initially SQLITE_OK and an error occurs while -** flushing buffers to disk, *pRc is set to an SQLite error code before -** returning. -*/ -static void fts3IncrmergeRelease( - Fts3Table *p, /* FTS3 table handle */ - IncrmergeWriter *pWriter, /* Merge-writer object */ - int *pRc /* IN/OUT: Error code */ -){ - int i; /* Used to iterate through non-root layers */ - int iRoot; /* Index of root in pWriter->aNodeWriter */ - NodeWriter *pRoot; /* NodeWriter for root node */ - int rc = *pRc; /* Error code */ - - /* Set iRoot to the index in pWriter->aNodeWriter[] of the output segment - ** root node. If the segment fits entirely on a single leaf node, iRoot - ** will be set to 0. If the root node is the parent of the leaves, iRoot - ** will be 1. And so on. */ - for(iRoot=FTS_MAX_APPENDABLE_HEIGHT-1; iRoot>=0; iRoot--){ - NodeWriter *pNode = &pWriter->aNodeWriter[iRoot]; - if( pNode->block.n>0 ) break; - assert( *pRc || pNode->block.nAlloc==0 ); - assert( *pRc || pNode->key.nAlloc==0 ); - sqlite3_free(pNode->block.a); - sqlite3_free(pNode->key.a); - } - - /* Empty output segment. This is a no-op. */ - if( iRoot<0 ) return; - - /* The entire output segment fits on a single node. Normally, this means - ** the node would be stored as a blob in the "root" column of the %_segdir - ** table. However, this is not permitted in this case. The problem is that - ** space has already been reserved in the %_segments table, and so the - ** start_block and end_block fields of the %_segdir table must be populated. - ** And, by design or by accident, released versions of FTS cannot handle - ** segments that fit entirely on the root node with start_block!=0. - ** - ** Instead, create a synthetic root node that contains nothing but a - ** pointer to the single content node. So that the segment consists of a - ** single leaf and a single interior (root) node. - ** - ** Todo: Better might be to defer allocating space in the %_segments - ** table until we are sure it is needed. - */ - if( iRoot==0 ){ - Blob *pBlock = &pWriter->aNodeWriter[1].block; - blobGrowBuffer(pBlock, 1 + FTS3_VARINT_MAX, &rc); - if( rc==SQLITE_OK ){ - pBlock->a[0] = 0x01; - pBlock->n = 1 + sqlite3Fts3PutVarint( - &pBlock->a[1], pWriter->aNodeWriter[0].iBlock - ); - } - iRoot = 1; - } - pRoot = &pWriter->aNodeWriter[iRoot]; - - /* Flush all currently outstanding nodes to disk. */ - for(i=0; i<iRoot; i++){ - NodeWriter *pNode = &pWriter->aNodeWriter[i]; - if( pNode->block.n>0 && rc==SQLITE_OK ){ - rc = fts3WriteSegment(p, pNode->iBlock, pNode->block.a, pNode->block.n); - } - sqlite3_free(pNode->block.a); - sqlite3_free(pNode->key.a); - } - - /* Write the %_segdir record. */ - if( rc==SQLITE_OK ){ - rc = fts3WriteSegdir(p, - pWriter->iAbsLevel+1, /* level */ - pWriter->iIdx, /* idx */ - pWriter->iStart, /* start_block */ - pWriter->aNodeWriter[0].iBlock, /* leaves_end_block */ - pWriter->iEnd, /* end_block */ - pRoot->block.a, pRoot->block.n /* root */ - ); - } - sqlite3_free(pRoot->block.a); - sqlite3_free(pRoot->key.a); - - *pRc = rc; -} - -/* -** Compare the term in buffer zLhs (size in bytes nLhs) with that in -** zRhs (size in bytes nRhs) using memcmp. If one term is a prefix of -** the other, it is considered to be smaller than the other. -** -** Return -ve if zLhs is smaller than zRhs, 0 if it is equal, or +ve -** if it is greater. -*/ -static int fts3TermCmp( - const char *zLhs, int nLhs, /* LHS of comparison */ - const char *zRhs, int nRhs /* RHS of comparison */ -){ - int nCmp = MIN(nLhs, nRhs); - int res; - - res = memcmp(zLhs, zRhs, nCmp); - if( res==0 ) res = nLhs - nRhs; - - return res; -} - - -/* -** Query to see if the entry in the %_segments table with blockid iEnd is -** NULL. If no error occurs and the entry is NULL, set *pbRes 1 before -** returning. Otherwise, set *pbRes to 0. -** -** Or, if an error occurs while querying the database, return an SQLite -** error code. The final value of *pbRes is undefined in this case. -** -** This is used to test if a segment is an "appendable" segment. If it -** is, then a NULL entry has been inserted into the %_segments table -** with blockid %_segdir.end_block. -*/ -static int fts3IsAppendable(Fts3Table *p, sqlite3_int64 iEnd, int *pbRes){ - int bRes = 0; /* Result to set *pbRes to */ - sqlite3_stmt *pCheck = 0; /* Statement to query database with */ - int rc; /* Return code */ - - rc = fts3SqlStmt(p, SQL_SEGMENT_IS_APPENDABLE, &pCheck, 0); - if( rc==SQLITE_OK ){ - sqlite3_bind_int64(pCheck, 1, iEnd); - if( SQLITE_ROW==sqlite3_step(pCheck) ) bRes = 1; - rc = sqlite3_reset(pCheck); - } - - *pbRes = bRes; - return rc; -} - -/* -** This function is called when initializing an incremental-merge operation. -** It checks if the existing segment with index value iIdx at absolute level -** (iAbsLevel+1) can be appended to by the incremental merge. If it can, the -** merge-writer object *pWriter is initialized to write to it. -** -** An existing segment can be appended to by an incremental merge if: -** -** * It was initially created as an appendable segment (with all required -** space pre-allocated), and -** -** * The first key read from the input (arguments zKey and nKey) is -** greater than the largest key currently stored in the potential -** output segment. -*/ -static int fts3IncrmergeLoad( - Fts3Table *p, /* Fts3 table handle */ - sqlite3_int64 iAbsLevel, /* Absolute level of input segments */ - int iIdx, /* Index of candidate output segment */ - const char *zKey, /* First key to write */ - int nKey, /* Number of bytes in nKey */ - IncrmergeWriter *pWriter /* Populate this object */ -){ - int rc; /* Return code */ - sqlite3_stmt *pSelect = 0; /* SELECT to read %_segdir entry */ - - rc = fts3SqlStmt(p, SQL_SELECT_SEGDIR, &pSelect, 0); - if( rc==SQLITE_OK ){ - sqlite3_int64 iStart = 0; /* Value of %_segdir.start_block */ - sqlite3_int64 iLeafEnd = 0; /* Value of %_segdir.leaves_end_block */ - sqlite3_int64 iEnd = 0; /* Value of %_segdir.end_block */ - const char *aRoot = 0; /* Pointer to %_segdir.root buffer */ - int nRoot = 0; /* Size of aRoot[] in bytes */ - int rc2; /* Return code from sqlite3_reset() */ - int bAppendable = 0; /* Set to true if segment is appendable */ - - /* Read the %_segdir entry for index iIdx absolute level (iAbsLevel+1) */ - sqlite3_bind_int64(pSelect, 1, iAbsLevel+1); - sqlite3_bind_int(pSelect, 2, iIdx); - if( sqlite3_step(pSelect)==SQLITE_ROW ){ - iStart = sqlite3_column_int64(pSelect, 1); - iLeafEnd = sqlite3_column_int64(pSelect, 2); - iEnd = sqlite3_column_int64(pSelect, 3); - nRoot = sqlite3_column_bytes(pSelect, 4); - aRoot = sqlite3_column_blob(pSelect, 4); - }else{ - return sqlite3_reset(pSelect); - } - - /* Check for the zero-length marker in the %_segments table */ - rc = fts3IsAppendable(p, iEnd, &bAppendable); - - /* Check that zKey/nKey is larger than the largest key the candidate */ - if( rc==SQLITE_OK && bAppendable ){ - char *aLeaf = 0; - int nLeaf = 0; - - rc = sqlite3Fts3ReadBlock(p, iLeafEnd, &aLeaf, &nLeaf, 0); - if( rc==SQLITE_OK ){ - NodeReader reader; - for(rc = nodeReaderInit(&reader, aLeaf, nLeaf); - rc==SQLITE_OK && reader.aNode; - rc = nodeReaderNext(&reader) - ){ - assert( reader.aNode ); - } - if( fts3TermCmp(zKey, nKey, reader.term.a, reader.term.n)<=0 ){ - bAppendable = 0; - } - nodeReaderRelease(&reader); - } - sqlite3_free(aLeaf); - } - - if( rc==SQLITE_OK && bAppendable ){ - /* It is possible to append to this segment. Set up the IncrmergeWriter - ** object to do so. */ - int i; - int nHeight = (int)aRoot[0]; - NodeWriter *pNode; - - pWriter->nLeafEst = (int)((iEnd - iStart) + 1)/FTS_MAX_APPENDABLE_HEIGHT; - pWriter->iStart = iStart; - pWriter->iEnd = iEnd; - pWriter->iAbsLevel = iAbsLevel; - pWriter->iIdx = iIdx; - - for(i=nHeight+1; i<FTS_MAX_APPENDABLE_HEIGHT; i++){ - pWriter->aNodeWriter[i].iBlock = pWriter->iStart + i*pWriter->nLeafEst; - } - - pNode = &pWriter->aNodeWriter[nHeight]; - pNode->iBlock = pWriter->iStart + pWriter->nLeafEst*nHeight; - blobGrowBuffer(&pNode->block, MAX(nRoot, p->nNodeSize), &rc); - if( rc==SQLITE_OK ){ - memcpy(pNode->block.a, aRoot, nRoot); - pNode->block.n = nRoot; - } - - for(i=nHeight; i>=0 && rc==SQLITE_OK; i--){ - NodeReader reader; - pNode = &pWriter->aNodeWriter[i]; - - rc = nodeReaderInit(&reader, pNode->block.a, pNode->block.n); - while( reader.aNode && rc==SQLITE_OK ) rc = nodeReaderNext(&reader); - blobGrowBuffer(&pNode->key, reader.term.n, &rc); - if( rc==SQLITE_OK ){ - memcpy(pNode->key.a, reader.term.a, reader.term.n); - pNode->key.n = reader.term.n; - if( i>0 ){ - char *aBlock = 0; - int nBlock = 0; - pNode = &pWriter->aNodeWriter[i-1]; - pNode->iBlock = reader.iChild; - rc = sqlite3Fts3ReadBlock(p, reader.iChild, &aBlock, &nBlock, 0); - blobGrowBuffer(&pNode->block, MAX(nBlock, p->nNodeSize), &rc); - if( rc==SQLITE_OK ){ - memcpy(pNode->block.a, aBlock, nBlock); - pNode->block.n = nBlock; - } - sqlite3_free(aBlock); - } - } - nodeReaderRelease(&reader); - } - } - - rc2 = sqlite3_reset(pSelect); - if( rc==SQLITE_OK ) rc = rc2; - } - - return rc; -} - -/* -** Determine the largest segment index value that exists within absolute -** level iAbsLevel+1. If no error occurs, set *piIdx to this value plus -** one before returning SQLITE_OK. Or, if there are no segments at all -** within level iAbsLevel, set *piIdx to zero. -** -** If an error occurs, return an SQLite error code. The final value of -** *piIdx is undefined in this case. -*/ -static int fts3IncrmergeOutputIdx( - Fts3Table *p, /* FTS Table handle */ - sqlite3_int64 iAbsLevel, /* Absolute index of input segments */ - int *piIdx /* OUT: Next free index at iAbsLevel+1 */ -){ - int rc; - sqlite3_stmt *pOutputIdx = 0; /* SQL used to find output index */ - - rc = fts3SqlStmt(p, SQL_NEXT_SEGMENT_INDEX, &pOutputIdx, 0); - if( rc==SQLITE_OK ){ - sqlite3_bind_int64(pOutputIdx, 1, iAbsLevel+1); - sqlite3_step(pOutputIdx); - *piIdx = sqlite3_column_int(pOutputIdx, 0); - rc = sqlite3_reset(pOutputIdx); - } - - return rc; -} - -/* -** Allocate an appendable output segment on absolute level iAbsLevel+1 -** with idx value iIdx. -** -** In the %_segdir table, a segment is defined by the values in three -** columns: -** -** start_block -** leaves_end_block -** end_block -** -** When an appendable segment is allocated, it is estimated that the -** maximum number of leaf blocks that may be required is the sum of the -** number of leaf blocks consumed by the input segments, plus the number -** of input segments, multiplied by two. This value is stored in stack -** variable nLeafEst. -** -** A total of 16*nLeafEst blocks are allocated when an appendable segment -** is created ((1 + end_block - start_block)==16*nLeafEst). The contiguous -** array of leaf nodes starts at the first block allocated. The array -** of interior nodes that are parents of the leaf nodes start at block -** (start_block + (1 + end_block - start_block) / 16). And so on. -** -** In the actual code below, the value "16" is replaced with the -** pre-processor macro FTS_MAX_APPENDABLE_HEIGHT. -*/ -static int fts3IncrmergeWriter( - Fts3Table *p, /* Fts3 table handle */ - sqlite3_int64 iAbsLevel, /* Absolute level of input segments */ - int iIdx, /* Index of new output segment */ - Fts3MultiSegReader *pCsr, /* Cursor that data will be read from */ - IncrmergeWriter *pWriter /* Populate this object */ -){ - int rc; /* Return Code */ - int i; /* Iterator variable */ - int nLeafEst = 0; /* Blocks allocated for leaf nodes */ - sqlite3_stmt *pLeafEst = 0; /* SQL used to determine nLeafEst */ - sqlite3_stmt *pFirstBlock = 0; /* SQL used to determine first block */ - - /* Calculate nLeafEst. */ - rc = fts3SqlStmt(p, SQL_MAX_LEAF_NODE_ESTIMATE, &pLeafEst, 0); - if( rc==SQLITE_OK ){ - sqlite3_bind_int64(pLeafEst, 1, iAbsLevel); - sqlite3_bind_int64(pLeafEst, 2, pCsr->nSegment); - if( SQLITE_ROW==sqlite3_step(pLeafEst) ){ - nLeafEst = sqlite3_column_int(pLeafEst, 0); - } - rc = sqlite3_reset(pLeafEst); - } - if( rc!=SQLITE_OK ) return rc; - - /* Calculate the first block to use in the output segment */ - rc = fts3SqlStmt(p, SQL_NEXT_SEGMENTS_ID, &pFirstBlock, 0); - if( rc==SQLITE_OK ){ - if( SQLITE_ROW==sqlite3_step(pFirstBlock) ){ - pWriter->iStart = sqlite3_column_int64(pFirstBlock, 0); - pWriter->iEnd = pWriter->iStart - 1; - pWriter->iEnd += nLeafEst * FTS_MAX_APPENDABLE_HEIGHT; - } - rc = sqlite3_reset(pFirstBlock); - } - if( rc!=SQLITE_OK ) return rc; - - /* Insert the marker in the %_segments table to make sure nobody tries - ** to steal the space just allocated. This is also used to identify - ** appendable segments. */ - rc = fts3WriteSegment(p, pWriter->iEnd, 0, 0); - if( rc!=SQLITE_OK ) return rc; - - pWriter->iAbsLevel = iAbsLevel; - pWriter->nLeafEst = nLeafEst; - pWriter->iIdx = iIdx; - - /* Set up the array of NodeWriter objects */ - for(i=0; i<FTS_MAX_APPENDABLE_HEIGHT; i++){ - pWriter->aNodeWriter[i].iBlock = pWriter->iStart + i*pWriter->nLeafEst; - } - return SQLITE_OK; -} - -/* -** Remove an entry from the %_segdir table. This involves running the -** following two statements: -** -** DELETE FROM %_segdir WHERE level = :iAbsLevel AND idx = :iIdx -** UPDATE %_segdir SET idx = idx - 1 WHERE level = :iAbsLevel AND idx > :iIdx -** -** The DELETE statement removes the specific %_segdir level. The UPDATE -** statement ensures that the remaining segments have contiguously allocated -** idx values. -*/ -static int fts3RemoveSegdirEntry( - Fts3Table *p, /* FTS3 table handle */ - sqlite3_int64 iAbsLevel, /* Absolute level to delete from */ - int iIdx /* Index of %_segdir entry to delete */ -){ - int rc; /* Return code */ - sqlite3_stmt *pDelete = 0; /* DELETE statement */ - - rc = fts3SqlStmt(p, SQL_DELETE_SEGDIR_ENTRY, &pDelete, 0); - if( rc==SQLITE_OK ){ - sqlite3_bind_int64(pDelete, 1, iAbsLevel); - sqlite3_bind_int(pDelete, 2, iIdx); - sqlite3_step(pDelete); - rc = sqlite3_reset(pDelete); - } - - return rc; -} - -/* -** One or more segments have just been removed from absolute level iAbsLevel. -** Update the 'idx' values of the remaining segments in the level so that -** the idx values are a contiguous sequence starting from 0. -*/ -static int fts3RepackSegdirLevel( - Fts3Table *p, /* FTS3 table handle */ - sqlite3_int64 iAbsLevel /* Absolute level to repack */ -){ - int rc; /* Return code */ - int *aIdx = 0; /* Array of remaining idx values */ - int nIdx = 0; /* Valid entries in aIdx[] */ - int nAlloc = 0; /* Allocated size of aIdx[] */ - int i; /* Iterator variable */ - sqlite3_stmt *pSelect = 0; /* Select statement to read idx values */ - sqlite3_stmt *pUpdate = 0; /* Update statement to modify idx values */ - - rc = fts3SqlStmt(p, SQL_SELECT_INDEXES, &pSelect, 0); - if( rc==SQLITE_OK ){ - int rc2; - sqlite3_bind_int64(pSelect, 1, iAbsLevel); - while( SQLITE_ROW==sqlite3_step(pSelect) ){ - if( nIdx>=nAlloc ){ - int *aNew; - nAlloc += 16; - aNew = sqlite3_realloc(aIdx, nAlloc*sizeof(int)); - if( !aNew ){ - rc = SQLITE_NOMEM; - break; - } - aIdx = aNew; - } - aIdx[nIdx++] = sqlite3_column_int(pSelect, 0); - } - rc2 = sqlite3_reset(pSelect); - if( rc==SQLITE_OK ) rc = rc2; - } - - if( rc==SQLITE_OK ){ - rc = fts3SqlStmt(p, SQL_SHIFT_SEGDIR_ENTRY, &pUpdate, 0); - } - if( rc==SQLITE_OK ){ - sqlite3_bind_int64(pUpdate, 2, iAbsLevel); - } - - assert( p->bIgnoreSavepoint==0 ); - p->bIgnoreSavepoint = 1; - for(i=0; rc==SQLITE_OK && i<nIdx; i++){ - if( aIdx[i]!=i ){ - sqlite3_bind_int(pUpdate, 3, aIdx[i]); - sqlite3_bind_int(pUpdate, 1, i); - sqlite3_step(pUpdate); - rc = sqlite3_reset(pUpdate); - } - } - p->bIgnoreSavepoint = 0; - - sqlite3_free(aIdx); - return rc; -} - -static void fts3StartNode(Blob *pNode, int iHeight, sqlite3_int64 iChild){ - pNode->a[0] = (char)iHeight; - if( iChild ){ - assert( pNode->nAlloc>=1+sqlite3Fts3VarintLen(iChild) ); - pNode->n = 1 + sqlite3Fts3PutVarint(&pNode->a[1], iChild); - }else{ - assert( pNode->nAlloc>=1 ); - pNode->n = 1; - } -} - -/* -** The first two arguments are a pointer to and the size of a segment b-tree -** node. The node may be a leaf or an internal node. -** -** This function creates a new node image in blob object *pNew by copying -** all terms that are greater than or equal to zTerm/nTerm (for leaf nodes) -** or greater than zTerm/nTerm (for internal nodes) from aNode/nNode. -*/ -static int fts3TruncateNode( - const char *aNode, /* Current node image */ - int nNode, /* Size of aNode in bytes */ - Blob *pNew, /* OUT: Write new node image here */ - const char *zTerm, /* Omit all terms smaller than this */ - int nTerm, /* Size of zTerm in bytes */ - sqlite3_int64 *piBlock /* OUT: Block number in next layer down */ -){ - NodeReader reader; /* Reader object */ - Blob prev = {0, 0, 0}; /* Previous term written to new node */ - int rc = SQLITE_OK; /* Return code */ - int bLeaf = aNode[0]=='\0'; /* True for a leaf node */ - - /* Allocate required output space */ - blobGrowBuffer(pNew, nNode, &rc); - if( rc!=SQLITE_OK ) return rc; - pNew->n = 0; - - /* Populate new node buffer */ - for(rc = nodeReaderInit(&reader, aNode, nNode); - rc==SQLITE_OK && reader.aNode; - rc = nodeReaderNext(&reader) - ){ - if( pNew->n==0 ){ - int res = fts3TermCmp(reader.term.a, reader.term.n, zTerm, nTerm); - if( res<0 || (bLeaf==0 && res==0) ) continue; - fts3StartNode(pNew, (int)aNode[0], reader.iChild); - *piBlock = reader.iChild; - } - rc = fts3AppendToNode( - pNew, &prev, reader.term.a, reader.term.n, - reader.aDoclist, reader.nDoclist - ); - if( rc!=SQLITE_OK ) break; - } - if( pNew->n==0 ){ - fts3StartNode(pNew, (int)aNode[0], reader.iChild); - *piBlock = reader.iChild; - } - assert( pNew->n<=pNew->nAlloc ); - - nodeReaderRelease(&reader); - sqlite3_free(prev.a); - return rc; -} - -/* -** Remove all terms smaller than zTerm/nTerm from segment iIdx in absolute -** level iAbsLevel. This may involve deleting entries from the %_segments -** table, and modifying existing entries in both the %_segments and %_segdir -** tables. -** -** SQLITE_OK is returned if the segment is updated successfully. Or an -** SQLite error code otherwise. -*/ -static int fts3TruncateSegment( - Fts3Table *p, /* FTS3 table handle */ - sqlite3_int64 iAbsLevel, /* Absolute level of segment to modify */ - int iIdx, /* Index within level of segment to modify */ - const char *zTerm, /* Remove terms smaller than this */ - int nTerm /* Number of bytes in buffer zTerm */ -){ - int rc = SQLITE_OK; /* Return code */ - Blob root = {0,0,0}; /* New root page image */ - Blob block = {0,0,0}; /* Buffer used for any other block */ - sqlite3_int64 iBlock = 0; /* Block id */ - sqlite3_int64 iNewStart = 0; /* New value for iStartBlock */ - sqlite3_int64 iOldStart = 0; /* Old value for iStartBlock */ - sqlite3_stmt *pFetch = 0; /* Statement used to fetch segdir */ - - rc = fts3SqlStmt(p, SQL_SELECT_SEGDIR, &pFetch, 0); - if( rc==SQLITE_OK ){ - int rc2; /* sqlite3_reset() return code */ - sqlite3_bind_int64(pFetch, 1, iAbsLevel); - sqlite3_bind_int(pFetch, 2, iIdx); - if( SQLITE_ROW==sqlite3_step(pFetch) ){ - const char *aRoot = sqlite3_column_blob(pFetch, 4); - int nRoot = sqlite3_column_bytes(pFetch, 4); - iOldStart = sqlite3_column_int64(pFetch, 1); - rc = fts3TruncateNode(aRoot, nRoot, &root, zTerm, nTerm, &iBlock); - } - rc2 = sqlite3_reset(pFetch); - if( rc==SQLITE_OK ) rc = rc2; - } - - while( rc==SQLITE_OK && iBlock ){ - char *aBlock = 0; - int nBlock = 0; - iNewStart = iBlock; - - rc = sqlite3Fts3ReadBlock(p, iBlock, &aBlock, &nBlock, 0); - if( rc==SQLITE_OK ){ - rc = fts3TruncateNode(aBlock, nBlock, &block, zTerm, nTerm, &iBlock); - } - if( rc==SQLITE_OK ){ - rc = fts3WriteSegment(p, iNewStart, block.a, block.n); - } - sqlite3_free(aBlock); - } - - /* Variable iNewStart now contains the first valid leaf node. */ - if( rc==SQLITE_OK && iNewStart ){ - sqlite3_stmt *pDel = 0; - rc = fts3SqlStmt(p, SQL_DELETE_SEGMENTS_RANGE, &pDel, 0); - if( rc==SQLITE_OK ){ - sqlite3_bind_int64(pDel, 1, iOldStart); - sqlite3_bind_int64(pDel, 2, iNewStart-1); - sqlite3_step(pDel); - rc = sqlite3_reset(pDel); - } - } - - if( rc==SQLITE_OK ){ - sqlite3_stmt *pChomp = 0; - rc = fts3SqlStmt(p, SQL_CHOMP_SEGDIR, &pChomp, 0); - if( rc==SQLITE_OK ){ - sqlite3_bind_int64(pChomp, 1, iNewStart); - sqlite3_bind_blob(pChomp, 2, root.a, root.n, SQLITE_STATIC); - sqlite3_bind_int64(pChomp, 3, iAbsLevel); - sqlite3_bind_int(pChomp, 4, iIdx); - sqlite3_step(pChomp); - rc = sqlite3_reset(pChomp); - } - } - - sqlite3_free(root.a); - sqlite3_free(block.a); - return rc; -} - -/* -** This function is called after an incrmental-merge operation has run to -** merge (or partially merge) two or more segments from absolute level -** iAbsLevel. -** -** Each input segment is either removed from the db completely (if all of -** its data was copied to the output segment by the incrmerge operation) -** or modified in place so that it no longer contains those entries that -** have been duplicated in the output segment. -*/ -static int fts3IncrmergeChomp( - Fts3Table *p, /* FTS table handle */ - sqlite3_int64 iAbsLevel, /* Absolute level containing segments */ - Fts3MultiSegReader *pCsr, /* Chomp all segments opened by this cursor */ - int *pnRem /* Number of segments not deleted */ -){ - int i; - int nRem = 0; - int rc = SQLITE_OK; - - for(i=pCsr->nSegment-1; i>=0 && rc==SQLITE_OK; i--){ - Fts3SegReader *pSeg = 0; - int j; - - /* Find the Fts3SegReader object with Fts3SegReader.iIdx==i. It is hiding - ** somewhere in the pCsr->apSegment[] array. */ - for(j=0; ALWAYS(j<pCsr->nSegment); j++){ - pSeg = pCsr->apSegment[j]; - if( pSeg->iIdx==i ) break; - } - assert( j<pCsr->nSegment && pSeg->iIdx==i ); - - if( pSeg->aNode==0 ){ - /* Seg-reader is at EOF. Remove the entire input segment. */ - rc = fts3DeleteSegment(p, pSeg); - if( rc==SQLITE_OK ){ - rc = fts3RemoveSegdirEntry(p, iAbsLevel, pSeg->iIdx); - } - *pnRem = 0; - }else{ - /* The incremental merge did not copy all the data from this - ** segment to the upper level. The segment is modified in place - ** so that it contains no keys smaller than zTerm/nTerm. */ - const char *zTerm = pSeg->zTerm; - int nTerm = pSeg->nTerm; - rc = fts3TruncateSegment(p, iAbsLevel, pSeg->iIdx, zTerm, nTerm); - nRem++; - } - } - - if( rc==SQLITE_OK && nRem!=pCsr->nSegment ){ - rc = fts3RepackSegdirLevel(p, iAbsLevel); - } - - *pnRem = nRem; - return rc; -} - -/* -** Store an incr-merge hint in the database. -*/ -static int fts3IncrmergeHintStore(Fts3Table *p, Blob *pHint){ - sqlite3_stmt *pReplace = 0; - int rc; /* Return code */ - - rc = fts3SqlStmt(p, SQL_REPLACE_STAT, &pReplace, 0); - if( rc==SQLITE_OK ){ - sqlite3_bind_int(pReplace, 1, FTS_STAT_INCRMERGEHINT); - sqlite3_bind_blob(pReplace, 2, pHint->a, pHint->n, SQLITE_STATIC); - sqlite3_step(pReplace); - rc = sqlite3_reset(pReplace); - } - - return rc; -} - -/* -** Load an incr-merge hint from the database. The incr-merge hint, if one -** exists, is stored in the rowid==1 row of the %_stat table. -** -** If successful, populate blob *pHint with the value read from the %_stat -** table and return SQLITE_OK. Otherwise, if an error occurs, return an -** SQLite error code. -*/ -static int fts3IncrmergeHintLoad(Fts3Table *p, Blob *pHint){ - sqlite3_stmt *pSelect = 0; - int rc; - - pHint->n = 0; - rc = fts3SqlStmt(p, SQL_SELECT_STAT, &pSelect, 0); - if( rc==SQLITE_OK ){ - int rc2; - sqlite3_bind_int(pSelect, 1, FTS_STAT_INCRMERGEHINT); - if( SQLITE_ROW==sqlite3_step(pSelect) ){ - const char *aHint = sqlite3_column_blob(pSelect, 0); - int nHint = sqlite3_column_bytes(pSelect, 0); - if( aHint ){ - blobGrowBuffer(pHint, nHint, &rc); - if( rc==SQLITE_OK ){ - memcpy(pHint->a, aHint, nHint); - pHint->n = nHint; - } - } - } - rc2 = sqlite3_reset(pSelect); - if( rc==SQLITE_OK ) rc = rc2; - } - - return rc; -} - -/* -** If *pRc is not SQLITE_OK when this function is called, it is a no-op. -** Otherwise, append an entry to the hint stored in blob *pHint. Each entry -** consists of two varints, the absolute level number of the input segments -** and the number of input segments. -** -** If successful, leave *pRc set to SQLITE_OK and return. If an error occurs, -** set *pRc to an SQLite error code before returning. -*/ -static void fts3IncrmergeHintPush( - Blob *pHint, /* Hint blob to append to */ - i64 iAbsLevel, /* First varint to store in hint */ - int nInput, /* Second varint to store in hint */ - int *pRc /* IN/OUT: Error code */ -){ - blobGrowBuffer(pHint, pHint->n + 2*FTS3_VARINT_MAX, pRc); - if( *pRc==SQLITE_OK ){ - pHint->n += sqlite3Fts3PutVarint(&pHint->a[pHint->n], iAbsLevel); - pHint->n += sqlite3Fts3PutVarint(&pHint->a[pHint->n], (i64)nInput); - } -} - -/* -** Read the last entry (most recently pushed) from the hint blob *pHint -** and then remove the entry. Write the two values read to *piAbsLevel and -** *pnInput before returning. -** -** If no error occurs, return SQLITE_OK. If the hint blob in *pHint does -** not contain at least two valid varints, return SQLITE_CORRUPT_VTAB. -*/ -static int fts3IncrmergeHintPop(Blob *pHint, i64 *piAbsLevel, int *pnInput){ - const int nHint = pHint->n; - int i; - - i = pHint->n-2; - while( i>0 && (pHint->a[i-1] & 0x80) ) i--; - while( i>0 && (pHint->a[i-1] & 0x80) ) i--; - - pHint->n = i; - i += sqlite3Fts3GetVarint(&pHint->a[i], piAbsLevel); - i += sqlite3Fts3GetVarint32(&pHint->a[i], pnInput); - if( i!=nHint ) return SQLITE_CORRUPT_VTAB; - - return SQLITE_OK; -} - - -/* -** Attempt an incremental merge that writes nMerge leaf blocks. -** -** Incremental merges happen nMin segments at a time. The two -** segments to be merged are the nMin oldest segments (the ones with -** the smallest indexes) in the highest level that contains at least -** nMin segments. Multiple merges might occur in an attempt to write the -** quota of nMerge leaf blocks. -*/ -int sqlite3Fts3Incrmerge(Fts3Table *p, int nMerge, int nMin){ - int rc; /* Return code */ - int nRem = nMerge; /* Number of leaf pages yet to be written */ - Fts3MultiSegReader *pCsr; /* Cursor used to read input data */ - Fts3SegFilter *pFilter; /* Filter used with cursor pCsr */ - IncrmergeWriter *pWriter; /* Writer object */ - int nSeg = 0; /* Number of input segments */ - sqlite3_int64 iAbsLevel = 0; /* Absolute level number to work on */ - Blob hint = {0, 0, 0}; /* Hint read from %_stat table */ - int bDirtyHint = 0; /* True if blob 'hint' has been modified */ - - /* Allocate space for the cursor, filter and writer objects */ - const int nAlloc = sizeof(*pCsr) + sizeof(*pFilter) + sizeof(*pWriter); - pWriter = (IncrmergeWriter *)sqlite3_malloc(nAlloc); - if( !pWriter ) return SQLITE_NOMEM; - pFilter = (Fts3SegFilter *)&pWriter[1]; - pCsr = (Fts3MultiSegReader *)&pFilter[1]; - - rc = fts3IncrmergeHintLoad(p, &hint); - while( rc==SQLITE_OK && nRem>0 ){ - const i64 nMod = FTS3_SEGDIR_MAXLEVEL * p->nIndex; - sqlite3_stmt *pFindLevel = 0; /* SQL used to determine iAbsLevel */ - int bUseHint = 0; /* True if attempting to append */ - - /* Search the %_segdir table for the absolute level with the smallest - ** relative level number that contains at least nMin segments, if any. - ** If one is found, set iAbsLevel to the absolute level number and - ** nSeg to nMin. If no level with at least nMin segments can be found, - ** set nSeg to -1. - */ - rc = fts3SqlStmt(p, SQL_FIND_MERGE_LEVEL, &pFindLevel, 0); - sqlite3_bind_int(pFindLevel, 1, nMin); - if( sqlite3_step(pFindLevel)==SQLITE_ROW ){ - iAbsLevel = sqlite3_column_int64(pFindLevel, 0); - nSeg = nMin; - }else{ - nSeg = -1; - } - rc = sqlite3_reset(pFindLevel); - - /* If the hint read from the %_stat table is not empty, check if the - ** last entry in it specifies a relative level smaller than or equal - ** to the level identified by the block above (if any). If so, this - ** iteration of the loop will work on merging at the hinted level. - */ - if( rc==SQLITE_OK && hint.n ){ - int nHint = hint.n; - sqlite3_int64 iHintAbsLevel = 0; /* Hint level */ - int nHintSeg = 0; /* Hint number of segments */ - - rc = fts3IncrmergeHintPop(&hint, &iHintAbsLevel, &nHintSeg); - if( nSeg<0 || (iAbsLevel % nMod) >= (iHintAbsLevel % nMod) ){ - iAbsLevel = iHintAbsLevel; - nSeg = nHintSeg; - bUseHint = 1; - bDirtyHint = 1; - }else{ - /* This undoes the effect of the HintPop() above - so that no entry - ** is removed from the hint blob. */ - hint.n = nHint; - } - } - - /* If nSeg is less that zero, then there is no level with at least - ** nMin segments and no hint in the %_stat table. No work to do. - ** Exit early in this case. */ - if( nSeg<0 ) break; - - /* Open a cursor to iterate through the contents of the oldest nSeg - ** indexes of absolute level iAbsLevel. If this cursor is opened using - ** the 'hint' parameters, it is possible that there are less than nSeg - ** segments available in level iAbsLevel. In this case, no work is - ** done on iAbsLevel - fall through to the next iteration of the loop - ** to start work on some other level. */ - memset(pWriter, 0, nAlloc); - pFilter->flags = FTS3_SEGMENT_REQUIRE_POS; - if( rc==SQLITE_OK ){ - rc = fts3IncrmergeCsr(p, iAbsLevel, nSeg, pCsr); - } - if( SQLITE_OK==rc && pCsr->nSegment==nSeg - && SQLITE_OK==(rc = sqlite3Fts3SegReaderStart(p, pCsr, pFilter)) - && SQLITE_ROW==(rc = sqlite3Fts3SegReaderStep(p, pCsr)) - ){ - int iIdx = 0; /* Largest idx in level (iAbsLevel+1) */ - rc = fts3IncrmergeOutputIdx(p, iAbsLevel, &iIdx); - if( rc==SQLITE_OK ){ - if( bUseHint && iIdx>0 ){ - const char *zKey = pCsr->zTerm; - int nKey = pCsr->nTerm; - rc = fts3IncrmergeLoad(p, iAbsLevel, iIdx-1, zKey, nKey, pWriter); - }else{ - rc = fts3IncrmergeWriter(p, iAbsLevel, iIdx, pCsr, pWriter); - } - } - - if( rc==SQLITE_OK && pWriter->nLeafEst ){ - fts3LogMerge(nSeg, iAbsLevel); - do { - rc = fts3IncrmergeAppend(p, pWriter, pCsr); - if( rc==SQLITE_OK ) rc = sqlite3Fts3SegReaderStep(p, pCsr); - if( pWriter->nWork>=nRem && rc==SQLITE_ROW ) rc = SQLITE_OK; - }while( rc==SQLITE_ROW ); - - /* Update or delete the input segments */ - if( rc==SQLITE_OK ){ - nRem -= (1 + pWriter->nWork); - rc = fts3IncrmergeChomp(p, iAbsLevel, pCsr, &nSeg); - if( nSeg!=0 ){ - bDirtyHint = 1; - fts3IncrmergeHintPush(&hint, iAbsLevel, nSeg, &rc); - } - } - } - - fts3IncrmergeRelease(p, pWriter, &rc); - } - - sqlite3Fts3SegReaderFinish(pCsr); - } - - /* Write the hint values into the %_stat table for the next incr-merger */ - if( bDirtyHint && rc==SQLITE_OK ){ - rc = fts3IncrmergeHintStore(p, &hint); - } - - sqlite3_free(pWriter); - sqlite3_free(hint.a); - return rc; -} - -/* -** Convert the text beginning at *pz into an integer and return -** its value. Advance *pz to point to the first character past -** the integer. -*/ -static int fts3Getint(const char **pz){ - const char *z = *pz; - int i = 0; - while( (*z)>='0' && (*z)<='9' ) i = 10*i + *(z++) - '0'; - *pz = z; - return i; -} - -/* -** Process statements of the form: -** -** INSERT INTO table(table) VALUES('merge=A,B'); -** -** A and B are integers that decode to be the number of leaf pages -** written for the merge, and the minimum number of segments on a level -** before it will be selected for a merge, respectively. -*/ -static int fts3DoIncrmerge( - Fts3Table *p, /* FTS3 table handle */ - const char *zParam /* Nul-terminated string containing "A,B" */ -){ - int rc; - int nMin = (FTS3_MERGE_COUNT / 2); - int nMerge = 0; - const char *z = zParam; - - /* Read the first integer value */ - nMerge = fts3Getint(&z); - - /* If the first integer value is followed by a ',', read the second - ** integer value. */ - if( z[0]==',' && z[1]!='\0' ){ - z++; - nMin = fts3Getint(&z); - } - - if( z[0]!='\0' || nMin<2 ){ - rc = SQLITE_ERROR; - }else{ - rc = SQLITE_OK; - if( !p->bHasStat ){ - assert( p->bFts4==0 ); - sqlite3Fts3CreateStatTable(&rc, p); - } - if( rc==SQLITE_OK ){ - rc = sqlite3Fts3Incrmerge(p, nMerge, nMin); - } - sqlite3Fts3SegmentsClose(p); - } - return rc; -} - -/* -** Process statements of the form: -** -** INSERT INTO table(table) VALUES('automerge=X'); -** -** where X is an integer. X==0 means to turn automerge off. X!=0 means -** turn it on. The setting is persistent. -*/ -static int fts3DoAutoincrmerge( - Fts3Table *p, /* FTS3 table handle */ - const char *zParam /* Nul-terminated string containing boolean */ -){ - int rc = SQLITE_OK; - sqlite3_stmt *pStmt = 0; - p->bAutoincrmerge = fts3Getint(&zParam)!=0; - if( !p->bHasStat ){ - assert( p->bFts4==0 ); - sqlite3Fts3CreateStatTable(&rc, p); - if( rc ) return rc; - } - rc = fts3SqlStmt(p, SQL_REPLACE_STAT, &pStmt, 0); - if( rc ) return rc;; - sqlite3_bind_int(pStmt, 1, FTS_STAT_AUTOINCRMERGE); - sqlite3_bind_int(pStmt, 2, p->bAutoincrmerge); - sqlite3_step(pStmt); - rc = sqlite3_reset(pStmt); - return rc; -} - -/* -** Return a 64-bit checksum for the FTS index entry specified by the -** arguments to this function. -*/ -static u64 fts3ChecksumEntry( - const char *zTerm, /* Pointer to buffer containing term */ - int nTerm, /* Size of zTerm in bytes */ - int iLangid, /* Language id for current row */ - int iIndex, /* Index (0..Fts3Table.nIndex-1) */ - i64 iDocid, /* Docid for current row. */ - int iCol, /* Column number */ - int iPos /* Position */ -){ - int i; - u64 ret = (u64)iDocid; - - ret += (ret<<3) + iLangid; - ret += (ret<<3) + iIndex; - ret += (ret<<3) + iCol; - ret += (ret<<3) + iPos; - for(i=0; i<nTerm; i++) ret += (ret<<3) + zTerm[i]; - - return ret; -} - -/* -** Return a checksum of all entries in the FTS index that correspond to -** language id iLangid. The checksum is calculated by XORing the checksums -** of each individual entry (see fts3ChecksumEntry()) together. -** -** If successful, the checksum value is returned and *pRc set to SQLITE_OK. -** Otherwise, if an error occurs, *pRc is set to an SQLite error code. The -** return value is undefined in this case. -*/ -static u64 fts3ChecksumIndex( - Fts3Table *p, /* FTS3 table handle */ - int iLangid, /* Language id to return cksum for */ - int iIndex, /* Index to cksum (0..p->nIndex-1) */ - int *pRc /* OUT: Return code */ -){ - Fts3SegFilter filter; - Fts3MultiSegReader csr; - int rc; - u64 cksum = 0; - - assert( *pRc==SQLITE_OK ); - - memset(&filter, 0, sizeof(filter)); - memset(&csr, 0, sizeof(csr)); - filter.flags = FTS3_SEGMENT_REQUIRE_POS|FTS3_SEGMENT_IGNORE_EMPTY; - filter.flags |= FTS3_SEGMENT_SCAN; - - rc = sqlite3Fts3SegReaderCursor( - p, iLangid, iIndex, FTS3_SEGCURSOR_ALL, 0, 0, 0, 1,&csr - ); - if( rc==SQLITE_OK ){ - rc = sqlite3Fts3SegReaderStart(p, &csr, &filter); - } - - if( rc==SQLITE_OK ){ - while( SQLITE_ROW==(rc = sqlite3Fts3SegReaderStep(p, &csr)) ){ - char *pCsr = csr.aDoclist; - char *pEnd = &pCsr[csr.nDoclist]; - - i64 iDocid = 0; - i64 iCol = 0; - i64 iPos = 0; - - pCsr += sqlite3Fts3GetVarint(pCsr, &iDocid); - while( pCsr<pEnd ){ - i64 iVal = 0; - pCsr += sqlite3Fts3GetVarint(pCsr, &iVal); - if( pCsr<pEnd ){ - if( iVal==0 || iVal==1 ){ - iCol = 0; - iPos = 0; - if( iVal ){ - pCsr += sqlite3Fts3GetVarint(pCsr, &iCol); - }else{ - pCsr += sqlite3Fts3GetVarint(pCsr, &iVal); - iDocid += iVal; - } - }else{ - iPos += (iVal - 2); - cksum = cksum ^ fts3ChecksumEntry( - csr.zTerm, csr.nTerm, iLangid, iIndex, iDocid, - (int)iCol, (int)iPos - ); - } - } - } - } - } - sqlite3Fts3SegReaderFinish(&csr); - - *pRc = rc; - return cksum; -} - -/* -** Check if the contents of the FTS index match the current contents of the -** content table. If no error occurs and the contents do match, set *pbOk -** to true and return SQLITE_OK. Or if the contents do not match, set *pbOk -** to false before returning. -** -** If an error occurs (e.g. an OOM or IO error), return an SQLite error -** code. The final value of *pbOk is undefined in this case. -*/ -static int fts3IntegrityCheck(Fts3Table *p, int *pbOk){ - int rc = SQLITE_OK; /* Return code */ - u64 cksum1 = 0; /* Checksum based on FTS index contents */ - u64 cksum2 = 0; /* Checksum based on %_content contents */ - sqlite3_stmt *pAllLangid = 0; /* Statement to return all language-ids */ - - /* This block calculates the checksum according to the FTS index. */ - rc = fts3SqlStmt(p, SQL_SELECT_ALL_LANGID, &pAllLangid, 0); - if( rc==SQLITE_OK ){ - int rc2; - sqlite3_bind_int(pAllLangid, 1, p->nIndex); - while( rc==SQLITE_OK && sqlite3_step(pAllLangid)==SQLITE_ROW ){ - int iLangid = sqlite3_column_int(pAllLangid, 0); - int i; - for(i=0; i<p->nIndex; i++){ - cksum1 = cksum1 ^ fts3ChecksumIndex(p, iLangid, i, &rc); - } - } - rc2 = sqlite3_reset(pAllLangid); - if( rc==SQLITE_OK ) rc = rc2; - } - - /* This block calculates the checksum according to the %_content table */ - rc = fts3SqlStmt(p, SQL_SELECT_ALL_LANGID, &pAllLangid, 0); - if( rc==SQLITE_OK ){ - sqlite3_tokenizer_module const *pModule = p->pTokenizer->pModule; - sqlite3_stmt *pStmt = 0; - char *zSql; - - zSql = sqlite3_mprintf("SELECT %s" , p->zReadExprlist); - if( !zSql ){ - rc = SQLITE_NOMEM; - }else{ - rc = sqlite3_prepare_v2(p->db, zSql, -1, &pStmt, 0); - sqlite3_free(zSql); - } - - while( rc==SQLITE_OK && SQLITE_ROW==sqlite3_step(pStmt) ){ - i64 iDocid = sqlite3_column_int64(pStmt, 0); - int iLang = langidFromSelect(p, pStmt); - int iCol; - - for(iCol=0; rc==SQLITE_OK && iCol<p->nColumn; iCol++){ - const char *zText = (const char *)sqlite3_column_text(pStmt, iCol+1); - int nText = sqlite3_column_bytes(pStmt, iCol+1); - sqlite3_tokenizer_cursor *pT = 0; - - rc = sqlite3Fts3OpenTokenizer(p->pTokenizer, iLang, zText, nText, &pT); - while( rc==SQLITE_OK ){ - char const *zToken; /* Buffer containing token */ - int nToken = 0; /* Number of bytes in token */ - int iDum1 = 0, iDum2 = 0; /* Dummy variables */ - int iPos = 0; /* Position of token in zText */ - - rc = pModule->xNext(pT, &zToken, &nToken, &iDum1, &iDum2, &iPos); - if( rc==SQLITE_OK ){ - int i; - cksum2 = cksum2 ^ fts3ChecksumEntry( - zToken, nToken, iLang, 0, iDocid, iCol, iPos - ); - for(i=1; i<p->nIndex; i++){ - if( p->aIndex[i].nPrefix<=nToken ){ - cksum2 = cksum2 ^ fts3ChecksumEntry( - zToken, p->aIndex[i].nPrefix, iLang, i, iDocid, iCol, iPos - ); - } - } - } - } - if( pT ) pModule->xClose(pT); - if( rc==SQLITE_DONE ) rc = SQLITE_OK; - } - } - - sqlite3_finalize(pStmt); - } - - *pbOk = (cksum1==cksum2); - return rc; -} - -/* -** Run the integrity-check. If no error occurs and the current contents of -** the FTS index are correct, return SQLITE_OK. Or, if the contents of the -** FTS index are incorrect, return SQLITE_CORRUPT_VTAB. -** -** Or, if an error (e.g. an OOM or IO error) occurs, return an SQLite -** error code. -** -** The integrity-check works as follows. For each token and indexed token -** prefix in the document set, a 64-bit checksum is calculated (by code -** in fts3ChecksumEntry()) based on the following: -** -** + The index number (0 for the main index, 1 for the first prefix -** index etc.), -** + The token (or token prefix) text itself, -** + The language-id of the row it appears in, -** + The docid of the row it appears in, -** + The column it appears in, and -** + The tokens position within that column. -** -** The checksums for all entries in the index are XORed together to create -** a single checksum for the entire index. -** -** The integrity-check code calculates the same checksum in two ways: -** -** 1. By scanning the contents of the FTS index, and -** 2. By scanning and tokenizing the content table. -** -** If the two checksums are identical, the integrity-check is deemed to have -** passed. -*/ -static int fts3DoIntegrityCheck( - Fts3Table *p /* FTS3 table handle */ -){ - int rc; - int bOk = 0; - rc = fts3IntegrityCheck(p, &bOk); - if( rc==SQLITE_OK && bOk==0 ) rc = SQLITE_CORRUPT_VTAB; - return rc; -} - -/* -** Handle a 'special' INSERT of the form: -** -** "INSERT INTO tbl(tbl) VALUES(<expr>)" -** -** Argument pVal contains the result of <expr>. Currently the only -** meaningful value to insert is the text 'optimize'. -*/ -static int fts3SpecialInsert(Fts3Table *p, sqlite3_value *pVal){ - int rc; /* Return Code */ - const char *zVal = (const char *)sqlite3_value_text(pVal); - int nVal = sqlite3_value_bytes(pVal); - - if( !zVal ){ - return SQLITE_NOMEM; - }else if( nVal==8 && 0==sqlite3_strnicmp(zVal, "optimize", 8) ){ - rc = fts3DoOptimize(p, 0); - }else if( nVal==7 && 0==sqlite3_strnicmp(zVal, "rebuild", 7) ){ - rc = fts3DoRebuild(p); - }else if( nVal==15 && 0==sqlite3_strnicmp(zVal, "integrity-check", 15) ){ - rc = fts3DoIntegrityCheck(p); - }else if( nVal>6 && 0==sqlite3_strnicmp(zVal, "merge=", 6) ){ - rc = fts3DoIncrmerge(p, &zVal[6]); - }else if( nVal>10 && 0==sqlite3_strnicmp(zVal, "automerge=", 10) ){ - rc = fts3DoAutoincrmerge(p, &zVal[10]); -#ifdef SQLITE_TEST - }else if( nVal>9 && 0==sqlite3_strnicmp(zVal, "nodesize=", 9) ){ - p->nNodeSize = atoi(&zVal[9]); - rc = SQLITE_OK; - }else if( nVal>11 && 0==sqlite3_strnicmp(zVal, "maxpending=", 9) ){ - p->nMaxPendingData = atoi(&zVal[11]); - rc = SQLITE_OK; -#endif - }else{ - rc = SQLITE_ERROR; - } - - return rc; -} - -#ifndef SQLITE_DISABLE_FTS4_DEFERRED -/* -** Delete all cached deferred doclists. Deferred doclists are cached -** (allocated) by the sqlite3Fts3CacheDeferredDoclists() function. -*/ -void sqlite3Fts3FreeDeferredDoclists(Fts3Cursor *pCsr){ - Fts3DeferredToken *pDef; - for(pDef=pCsr->pDeferred; pDef; pDef=pDef->pNext){ - fts3PendingListDelete(pDef->pList); - pDef->pList = 0; - } -} - -/* -** Free all entries in the pCsr->pDeffered list. Entries are added to -** this list using sqlite3Fts3DeferToken(). -*/ -void sqlite3Fts3FreeDeferredTokens(Fts3Cursor *pCsr){ - Fts3DeferredToken *pDef; - Fts3DeferredToken *pNext; - for(pDef=pCsr->pDeferred; pDef; pDef=pNext){ - pNext = pDef->pNext; - fts3PendingListDelete(pDef->pList); - sqlite3_free(pDef); - } - pCsr->pDeferred = 0; -} - -/* -** Generate deferred-doclists for all tokens in the pCsr->pDeferred list -** based on the row that pCsr currently points to. -** -** A deferred-doclist is like any other doclist with position information -** included, except that it only contains entries for a single row of the -** table, not for all rows. -*/ -int sqlite3Fts3CacheDeferredDoclists(Fts3Cursor *pCsr){ - int rc = SQLITE_OK; /* Return code */ - if( pCsr->pDeferred ){ - int i; /* Used to iterate through table columns */ - sqlite3_int64 iDocid; /* Docid of the row pCsr points to */ - Fts3DeferredToken *pDef; /* Used to iterate through deferred tokens */ - - Fts3Table *p = (Fts3Table *)pCsr->base.pVtab; - sqlite3_tokenizer *pT = p->pTokenizer; - sqlite3_tokenizer_module const *pModule = pT->pModule; - - assert( pCsr->isRequireSeek==0 ); - iDocid = sqlite3_column_int64(pCsr->pStmt, 0); - - for(i=0; i<p->nColumn && rc==SQLITE_OK; i++){ - const char *zText = (const char *)sqlite3_column_text(pCsr->pStmt, i+1); - sqlite3_tokenizer_cursor *pTC = 0; - - rc = sqlite3Fts3OpenTokenizer(pT, pCsr->iLangid, zText, -1, &pTC); - while( rc==SQLITE_OK ){ - char const *zToken; /* Buffer containing token */ - int nToken = 0; /* Number of bytes in token */ - int iDum1 = 0, iDum2 = 0; /* Dummy variables */ - int iPos = 0; /* Position of token in zText */ - - rc = pModule->xNext(pTC, &zToken, &nToken, &iDum1, &iDum2, &iPos); - for(pDef=pCsr->pDeferred; pDef && rc==SQLITE_OK; pDef=pDef->pNext){ - Fts3PhraseToken *pPT = pDef->pToken; - if( (pDef->iCol>=p->nColumn || pDef->iCol==i) - && (pPT->bFirst==0 || iPos==0) - && (pPT->n==nToken || (pPT->isPrefix && pPT->n<nToken)) - && (0==memcmp(zToken, pPT->z, pPT->n)) - ){ - fts3PendingListAppend(&pDef->pList, iDocid, i, iPos, &rc); - } - } - } - if( pTC ) pModule->xClose(pTC); - if( rc==SQLITE_DONE ) rc = SQLITE_OK; - } - - for(pDef=pCsr->pDeferred; pDef && rc==SQLITE_OK; pDef=pDef->pNext){ - if( pDef->pList ){ - rc = fts3PendingListAppendVarint(&pDef->pList, 0); - } - } - } - - return rc; -} - -int sqlite3Fts3DeferredTokenList( - Fts3DeferredToken *p, - char **ppData, - int *pnData -){ - char *pRet; - int nSkip; - sqlite3_int64 dummy; - - *ppData = 0; - *pnData = 0; - - if( p->pList==0 ){ - return SQLITE_OK; - } - - pRet = (char *)sqlite3_malloc(p->pList->nData); - if( !pRet ) return SQLITE_NOMEM; - - nSkip = sqlite3Fts3GetVarint(p->pList->aData, &dummy); - *pnData = p->pList->nData - nSkip; - *ppData = pRet; - - memcpy(pRet, &p->pList->aData[nSkip], *pnData); - return SQLITE_OK; -} - -/* -** Add an entry for token pToken to the pCsr->pDeferred list. -*/ -int sqlite3Fts3DeferToken( - Fts3Cursor *pCsr, /* Fts3 table cursor */ - Fts3PhraseToken *pToken, /* Token to defer */ - int iCol /* Column that token must appear in (or -1) */ -){ - Fts3DeferredToken *pDeferred; - pDeferred = sqlite3_malloc(sizeof(*pDeferred)); - if( !pDeferred ){ - return SQLITE_NOMEM; - } - memset(pDeferred, 0, sizeof(*pDeferred)); - pDeferred->pToken = pToken; - pDeferred->pNext = pCsr->pDeferred; - pDeferred->iCol = iCol; - pCsr->pDeferred = pDeferred; - - assert( pToken->pDeferred==0 ); - pToken->pDeferred = pDeferred; - - return SQLITE_OK; -} -#endif - -/* -** SQLite value pRowid contains the rowid of a row that may or may not be -** present in the FTS3 table. If it is, delete it and adjust the contents -** of subsiduary data structures accordingly. -*/ -static int fts3DeleteByRowid( - Fts3Table *p, - sqlite3_value *pRowid, - int *pnChng, /* IN/OUT: Decrement if row is deleted */ - u32 *aSzDel -){ - int rc = SQLITE_OK; /* Return code */ - int bFound = 0; /* True if *pRowid really is in the table */ - - fts3DeleteTerms(&rc, p, pRowid, aSzDel, &bFound); - if( bFound && rc==SQLITE_OK ){ - int isEmpty = 0; /* Deleting *pRowid leaves the table empty */ - rc = fts3IsEmpty(p, pRowid, &isEmpty); - if( rc==SQLITE_OK ){ - if( isEmpty ){ - /* Deleting this row means the whole table is empty. In this case - ** delete the contents of all three tables and throw away any - ** data in the pendingTerms hash table. */ - rc = fts3DeleteAll(p, 1); - *pnChng = 0; - memset(aSzDel, 0, sizeof(u32) * (p->nColumn+1) * 2); - }else{ - *pnChng = *pnChng - 1; - if( p->zContentTbl==0 ){ - fts3SqlExec(&rc, p, SQL_DELETE_CONTENT, &pRowid); - } - if( p->bHasDocsize ){ - fts3SqlExec(&rc, p, SQL_DELETE_DOCSIZE, &pRowid); - } - } - } - } - - return rc; -} - -/* -** This function does the work for the xUpdate method of FTS3 virtual -** tables. The schema of the virtual table being: -** -** CREATE TABLE <table name>( -** <user columns>, -** <table name> HIDDEN, -** docid HIDDEN, -** <langid> HIDDEN -** ); -** -** -*/ -int sqlite3Fts3UpdateMethod( - sqlite3_vtab *pVtab, /* FTS3 vtab object */ - int nArg, /* Size of argument array */ - sqlite3_value **apVal, /* Array of arguments */ - sqlite_int64 *pRowid /* OUT: The affected (or effected) rowid */ -){ - Fts3Table *p = (Fts3Table *)pVtab; - int rc = SQLITE_OK; /* Return Code */ - int isRemove = 0; /* True for an UPDATE or DELETE */ - u32 *aSzIns = 0; /* Sizes of inserted documents */ - u32 *aSzDel = 0; /* Sizes of deleted documents */ - int nChng = 0; /* Net change in number of documents */ - int bInsertDone = 0; - - assert( p->pSegments==0 ); - assert( - nArg==1 /* DELETE operations */ - || nArg==(2 + p->nColumn + 3) /* INSERT or UPDATE operations */ - ); - - /* Check for a "special" INSERT operation. One of the form: - ** - ** INSERT INTO xyz(xyz) VALUES('command'); - */ - if( nArg>1 - && sqlite3_value_type(apVal[0])==SQLITE_NULL - && sqlite3_value_type(apVal[p->nColumn+2])!=SQLITE_NULL - ){ - rc = fts3SpecialInsert(p, apVal[p->nColumn+2]); - goto update_out; - } - - if( nArg>1 && sqlite3_value_int(apVal[2 + p->nColumn + 2])<0 ){ - rc = SQLITE_CONSTRAINT; - goto update_out; - } - - /* Allocate space to hold the change in document sizes */ - aSzDel = sqlite3_malloc( sizeof(aSzDel[0])*(p->nColumn+1)*2 ); - if( aSzDel==0 ){ - rc = SQLITE_NOMEM; - goto update_out; - } - aSzIns = &aSzDel[p->nColumn+1]; - memset(aSzDel, 0, sizeof(aSzDel[0])*(p->nColumn+1)*2); - - /* If this is an INSERT operation, or an UPDATE that modifies the rowid - ** value, then this operation requires constraint handling. - ** - ** If the on-conflict mode is REPLACE, this means that the existing row - ** should be deleted from the database before inserting the new row. Or, - ** if the on-conflict mode is other than REPLACE, then this method must - ** detect the conflict and return SQLITE_CONSTRAINT before beginning to - ** modify the database file. - */ - if( nArg>1 && p->zContentTbl==0 ){ - /* Find the value object that holds the new rowid value. */ - sqlite3_value *pNewRowid = apVal[3+p->nColumn]; - if( sqlite3_value_type(pNewRowid)==SQLITE_NULL ){ - pNewRowid = apVal[1]; - } - - if( sqlite3_value_type(pNewRowid)!=SQLITE_NULL && ( - sqlite3_value_type(apVal[0])==SQLITE_NULL - || sqlite3_value_int64(apVal[0])!=sqlite3_value_int64(pNewRowid) - )){ - /* The new rowid is not NULL (in this case the rowid will be - ** automatically assigned and there is no chance of a conflict), and - ** the statement is either an INSERT or an UPDATE that modifies the - ** rowid column. So if the conflict mode is REPLACE, then delete any - ** existing row with rowid=pNewRowid. - ** - ** Or, if the conflict mode is not REPLACE, insert the new record into - ** the %_content table. If we hit the duplicate rowid constraint (or any - ** other error) while doing so, return immediately. - ** - ** This branch may also run if pNewRowid contains a value that cannot - ** be losslessly converted to an integer. In this case, the eventual - ** call to fts3InsertData() (either just below or further on in this - ** function) will return SQLITE_MISMATCH. If fts3DeleteByRowid is - ** invoked, it will delete zero rows (since no row will have - ** docid=$pNewRowid if $pNewRowid is not an integer value). - */ - if( sqlite3_vtab_on_conflict(p->db)==SQLITE_REPLACE ){ - rc = fts3DeleteByRowid(p, pNewRowid, &nChng, aSzDel); - }else{ - rc = fts3InsertData(p, apVal, pRowid); - bInsertDone = 1; - } - } - } - if( rc!=SQLITE_OK ){ - goto update_out; - } - - /* If this is a DELETE or UPDATE operation, remove the old record. */ - if( sqlite3_value_type(apVal[0])!=SQLITE_NULL ){ - assert( sqlite3_value_type(apVal[0])==SQLITE_INTEGER ); - rc = fts3DeleteByRowid(p, apVal[0], &nChng, aSzDel); - isRemove = 1; - } - - /* If this is an INSERT or UPDATE operation, insert the new record. */ - if( nArg>1 && rc==SQLITE_OK ){ - int iLangid = sqlite3_value_int(apVal[2 + p->nColumn + 2]); - if( bInsertDone==0 ){ - rc = fts3InsertData(p, apVal, pRowid); - if( rc==SQLITE_CONSTRAINT && p->zContentTbl==0 ){ - rc = FTS_CORRUPT_VTAB; - } - } - if( rc==SQLITE_OK && (!isRemove || *pRowid!=p->iPrevDocid ) ){ - rc = fts3PendingTermsDocid(p, iLangid, *pRowid); - } - if( rc==SQLITE_OK ){ - assert( p->iPrevDocid==*pRowid ); - rc = fts3InsertTerms(p, iLangid, apVal, aSzIns); - } - if( p->bHasDocsize ){ - fts3InsertDocsize(&rc, p, aSzIns); - } - nChng++; - } - - if( p->bFts4 ){ - fts3UpdateDocTotals(&rc, p, aSzIns, aSzDel, nChng); - } - - update_out: - sqlite3_free(aSzDel); - sqlite3Fts3SegmentsClose(p); - return rc; -} - -/* -** Flush any data in the pending-terms hash table to disk. If successful, -** merge all segments in the database (including the new segment, if -** there was any data to flush) into a single segment. -*/ -int sqlite3Fts3Optimize(Fts3Table *p){ - int rc; - rc = sqlite3_exec(p->db, "SAVEPOINT fts3", 0, 0, 0); - if( rc==SQLITE_OK ){ - rc = fts3DoOptimize(p, 1); - if( rc==SQLITE_OK || rc==SQLITE_DONE ){ - int rc2 = sqlite3_exec(p->db, "RELEASE fts3", 0, 0, 0); - if( rc2!=SQLITE_OK ) rc = rc2; - }else{ - sqlite3_exec(p->db, "ROLLBACK TO fts3", 0, 0, 0); - sqlite3_exec(p->db, "RELEASE fts3", 0, 0, 0); - } - } - sqlite3Fts3SegmentsClose(p); - return rc; -} - -#endif diff --git a/src/libtracker-fts/fts5.c b/src/libtracker-fts/fts5.c new file mode 100644 index 000000000..5e2fb1471 --- /dev/null +++ b/src/libtracker-fts/fts5.c @@ -0,0 +1,20402 @@ + + +#if !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS5) + +#if !defined(NDEBUG) && !defined(SQLITE_DEBUG) +# define NDEBUG 1 +#endif +#if defined(NDEBUG) && defined(SQLITE_DEBUG) +# undef NDEBUG +#endif + +#line 1 "fts5.h" +/* +** 2014 May 31 +** +** The author disclaims copyright to this source code. In place of +** a legal notice, here is a blessing: +** +** May you do good and not evil. +** May you find forgiveness for yourself and forgive others. +** May you share freely, never taking more than you give. +** +****************************************************************************** +** +** Interfaces to extend FTS5. Using the interfaces defined in this file, +** FTS5 may be extended with: +** +** * custom tokenizers, and +** * custom auxiliary functions. +*/ + + +#ifndef _FTS5_H +#define _FTS5_H + +#include "sqlite3.h" + +#ifdef __cplusplus +extern "C" { +#endif + +/************************************************************************* +** CUSTOM AUXILIARY FUNCTIONS +** +** Virtual table implementations may overload SQL functions by implementing +** the sqlite3_module.xFindFunction() method. +*/ + +typedef struct Fts5ExtensionApi Fts5ExtensionApi; +typedef struct Fts5Context Fts5Context; +typedef struct Fts5PhraseIter Fts5PhraseIter; + +typedef void (*fts5_extension_function)( + const Fts5ExtensionApi *pApi, /* API offered by current FTS version */ + Fts5Context *pFts, /* First arg to pass to pApi functions */ + sqlite3_context *pCtx, /* Context for returning result/error */ + int nVal, /* Number of values in apVal[] array */ + sqlite3_value **apVal /* Array of trailing arguments */ +); + +struct Fts5PhraseIter { + const unsigned char *a; + const unsigned char *b; +}; + +/* +** EXTENSION API FUNCTIONS +** +** xUserData(pFts): +** Return a copy of the context pointer the extension function was +** registered with. +** +** xColumnTotalSize(pFts, iCol, pnToken): +** If parameter iCol is less than zero, set output variable *pnToken +** to the total number of tokens in the FTS5 table. Or, if iCol is +** non-negative but less than the number of columns in the table, return +** the total number of tokens in column iCol, considering all rows in +** the FTS5 table. +** +** If parameter iCol is greater than or equal to the number of columns +** in the table, SQLITE_RANGE is returned. Or, if an error occurs (e.g. +** an OOM condition or IO error), an appropriate SQLite error code is +** returned. +** +** xColumnCount(pFts): +** Return the number of columns in the table. +** +** xColumnSize(pFts, iCol, pnToken): +** If parameter iCol is less than zero, set output variable *pnToken +** to the total number of tokens in the current row. Or, if iCol is +** non-negative but less than the number of columns in the table, set +** *pnToken to the number of tokens in column iCol of the current row. +** +** If parameter iCol is greater than or equal to the number of columns +** in the table, SQLITE_RANGE is returned. Or, if an error occurs (e.g. +** an OOM condition or IO error), an appropriate SQLite error code is +** returned. +** +** This function may be quite inefficient if used with an FTS5 table +** created with the "columnsize=0" option. +** +** xColumnText: +** This function attempts to retrieve the text of column iCol of the +** current document. If successful, (*pz) is set to point to a buffer +** containing the text in utf-8 encoding, (*pn) is set to the size in bytes +** (not characters) of the buffer and SQLITE_OK is returned. Otherwise, +** if an error occurs, an SQLite error code is returned and the final values +** of (*pz) and (*pn) are undefined. +** +** xPhraseCount: +** Returns the number of phrases in the current query expression. +** +** xPhraseSize: +** Returns the number of tokens in phrase iPhrase of the query. Phrases +** are numbered starting from zero. +** +** xInstCount: +** Set *pnInst to the total number of occurrences of all phrases within +** the query within the current row. Return SQLITE_OK if successful, or +** an error code (i.e. SQLITE_NOMEM) if an error occurs. +** +** This API can be quite slow if used with an FTS5 table created with the +** "detail=none" or "detail=column" option. If the FTS5 table is created +** with either "detail=none" or "detail=column" and "content=" option +** (i.e. if it is a contentless table), then this API always returns 0. +** +** xInst: +** Query for the details of phrase match iIdx within the current row. +** Phrase matches are numbered starting from zero, so the iIdx argument +** should be greater than or equal to zero and smaller than the value +** output by xInstCount(). +** +** Usually, output parameter *piPhrase is set to the phrase number, *piCol +** to the column in which it occurs and *piOff the token offset of the +** first token of the phrase. The exception is if the table was created +** with the offsets=0 option specified. In this case *piOff is always +** set to -1. +** +** Returns SQLITE_OK if successful, or an error code (i.e. SQLITE_NOMEM) +** if an error occurs. +** +** This API can be quite slow if used with an FTS5 table created with the +** "detail=none" or "detail=column" option. +** +** xRowid: +** Returns the rowid of the current row. +** +** xTokenize: +** Tokenize text using the tokenizer belonging to the FTS5 table. +** +** xQueryPhrase(pFts5, iPhrase, pUserData, xCallback): +** This API function is used to query the FTS table for phrase iPhrase +** of the current query. Specifically, a query equivalent to: +** +** ... FROM ftstable WHERE ftstable MATCH $p ORDER BY rowid +** +** with $p set to a phrase equivalent to the phrase iPhrase of the +** current query is executed. For each row visited, the callback function +** passed as the fourth argument is invoked. The context and API objects +** passed to the callback function may be used to access the properties of +** each matched row. Invoking Api.xUserData() returns a copy of the pointer +** passed as the third argument to pUserData. +** +** If the callback function returns any value other than SQLITE_OK, the +** query is abandoned and the xQueryPhrase function returns immediately. +** If the returned value is SQLITE_DONE, xQueryPhrase returns SQLITE_OK. +** Otherwise, the error code is propagated upwards. +** +** If the query runs to completion without incident, SQLITE_OK is returned. +** Or, if some error occurs before the query completes or is aborted by +** the callback, an SQLite error code is returned. +** +** +** xSetAuxdata(pFts5, pAux, xDelete) +** +** Save the pointer passed as the second argument as the extension functions +** "auxiliary data". The pointer may then be retrieved by the current or any +** future invocation of the same fts5 extension function made as part of +** of the same MATCH query using the xGetAuxdata() API. +** +** Each extension function is allocated a single auxiliary data slot for +** each FTS query (MATCH expression). If the extension function is invoked +** more than once for a single FTS query, then all invocations share a +** single auxiliary data context. +** +** If there is already an auxiliary data pointer when this function is +** invoked, then it is replaced by the new pointer. If an xDelete callback +** was specified along with the original pointer, it is invoked at this +** point. +** +** The xDelete callback, if one is specified, is also invoked on the +** auxiliary data pointer after the FTS5 query has finished. +** +** If an error (e.g. an OOM condition) occurs within this function, an +** the auxiliary data is set to NULL and an error code returned. If the +** xDelete parameter was not NULL, it is invoked on the auxiliary data +** pointer before returning. +** +** +** xGetAuxdata(pFts5, bClear) +** +** Returns the current auxiliary data pointer for the fts5 extension +** function. See the xSetAuxdata() method for details. +** +** If the bClear argument is non-zero, then the auxiliary data is cleared +** (set to NULL) before this function returns. In this case the xDelete, +** if any, is not invoked. +** +** +** xRowCount(pFts5, pnRow) +** +** This function is used to retrieve the total number of rows in the table. +** In other words, the same value that would be returned by: +** +** SELECT count(*) FROM ftstable; +** +** xPhraseFirst() +** This function is used, along with type Fts5PhraseIter and the xPhraseNext +** method, to iterate through all instances of a single query phrase within +** the current row. This is the same information as is accessible via the +** xInstCount/xInst APIs. While the xInstCount/xInst APIs are more convenient +** to use, this API may be faster under some circumstances. To iterate +** through instances of phrase iPhrase, use the following code: +** +** Fts5PhraseIter iter; +** int iCol, iOff; +** for(pApi->xPhraseFirst(pFts, iPhrase, &iter, &iCol, &iOff); +** iCol>=0; +** pApi->xPhraseNext(pFts, &iter, &iCol, &iOff) +** ){ +** // An instance of phrase iPhrase at offset iOff of column iCol +** } +** +** The Fts5PhraseIter structure is defined above. Applications should not +** modify this structure directly - it should only be used as shown above +** with the xPhraseFirst() and xPhraseNext() API methods (and by +** xPhraseFirstColumn() and xPhraseNextColumn() as illustrated below). +** +** This API can be quite slow if used with an FTS5 table created with the +** "detail=none" or "detail=column" option. If the FTS5 table is created +** with either "detail=none" or "detail=column" and "content=" option +** (i.e. if it is a contentless table), then this API always iterates +** through an empty set (all calls to xPhraseFirst() set iCol to -1). +** +** xPhraseNext() +** See xPhraseFirst above. +** +** xPhraseFirstColumn() +** This function and xPhraseNextColumn() are similar to the xPhraseFirst() +** and xPhraseNext() APIs described above. The difference is that instead +** of iterating through all instances of a phrase in the current row, these +** APIs are used to iterate through the set of columns in the current row +** that contain one or more instances of a specified phrase. For example: +** +** Fts5PhraseIter iter; +** int iCol; +** for(pApi->xPhraseFirstColumn(pFts, iPhrase, &iter, &iCol); +** iCol>=0; +** pApi->xPhraseNextColumn(pFts, &iter, &iCol) +** ){ +** // Column iCol contains at least one instance of phrase iPhrase +** } +** +** This API can be quite slow if used with an FTS5 table created with the +** "detail=none" option. If the FTS5 table is created with either +** "detail=none" "content=" option (i.e. if it is a contentless table), +** then this API always iterates through an empty set (all calls to +** xPhraseFirstColumn() set iCol to -1). +** +** The information accessed using this API and its companion +** xPhraseFirstColumn() may also be obtained using xPhraseFirst/xPhraseNext +** (or xInst/xInstCount). The chief advantage of this API is that it is +** significantly more efficient than those alternatives when used with +** "detail=column" tables. +** +** xPhraseNextColumn() +** See xPhraseFirstColumn above. +*/ +struct Fts5ExtensionApi { + int iVersion; /* Currently always set to 3 */ + + void *(*xUserData)(Fts5Context*); + + int (*xColumnCount)(Fts5Context*); + int (*xRowCount)(Fts5Context*, sqlite3_int64 *pnRow); + int (*xColumnTotalSize)(Fts5Context*, int iCol, sqlite3_int64 *pnToken); + + int (*xTokenize)(Fts5Context*, + const char *pText, int nText, /* Text to tokenize */ + void *pCtx, /* Context passed to xToken() */ + int (*xToken)(void*, int, const char*, int, int, int) /* Callback */ + ); + + int (*xPhraseCount)(Fts5Context*); + int (*xPhraseSize)(Fts5Context*, int iPhrase); + + int (*xInstCount)(Fts5Context*, int *pnInst); + int (*xInst)(Fts5Context*, int iIdx, int *piPhrase, int *piCol, int *piOff); + + sqlite3_int64 (*xRowid)(Fts5Context*); + int (*xColumnText)(Fts5Context*, int iCol, const char **pz, int *pn); + int (*xColumnSize)(Fts5Context*, int iCol, int *pnToken); + + int (*xQueryPhrase)(Fts5Context*, int iPhrase, void *pUserData, + int(*)(const Fts5ExtensionApi*,Fts5Context*,void*) + ); + int (*xSetAuxdata)(Fts5Context*, void *pAux, void(*xDelete)(void*)); + void *(*xGetAuxdata)(Fts5Context*, int bClear); + + int (*xPhraseFirst)(Fts5Context*, int iPhrase, Fts5PhraseIter*, int*, int*); + void (*xPhraseNext)(Fts5Context*, Fts5PhraseIter*, int *piCol, int *piOff); + + int (*xPhraseFirstColumn)(Fts5Context*, int iPhrase, Fts5PhraseIter*, int*); + void (*xPhraseNextColumn)(Fts5Context*, Fts5PhraseIter*, int *piCol); +}; + +/* +** CUSTOM AUXILIARY FUNCTIONS +*************************************************************************/ + +/************************************************************************* +** CUSTOM TOKENIZERS +** +** Applications may also register custom tokenizer types. A tokenizer +** is registered by providing fts5 with a populated instance of the +** following structure. All structure methods must be defined, setting +** any member of the fts5_tokenizer struct to NULL leads to undefined +** behaviour. The structure methods are expected to function as follows: +** +** xCreate: +** This function is used to allocate and inititalize a tokenizer instance. +** A tokenizer instance is required to actually tokenize text. +** +** The first argument passed to this function is a copy of the (void*) +** pointer provided by the application when the fts5_tokenizer object +** was registered with FTS5 (the third argument to xCreateTokenizer()). +** The second and third arguments are an array of nul-terminated strings +** containing the tokenizer arguments, if any, specified following the +** tokenizer name as part of the CREATE VIRTUAL TABLE statement used +** to create the FTS5 table. +** +** The final argument is an output variable. If successful, (*ppOut) +** should be set to point to the new tokenizer handle and SQLITE_OK +** returned. If an error occurs, some value other than SQLITE_OK should +** be returned. In this case, fts5 assumes that the final value of *ppOut +** is undefined. +** +** xDelete: +** This function is invoked to delete a tokenizer handle previously +** allocated using xCreate(). Fts5 guarantees that this function will +** be invoked exactly once for each successful call to xCreate(). +** +** xTokenize: +** This function is expected to tokenize the nText byte string indicated +** by argument pText. pText may or may not be nul-terminated. The first +** argument passed to this function is a pointer to an Fts5Tokenizer object +** returned by an earlier call to xCreate(). +** +** The second argument indicates the reason that FTS5 is requesting +** tokenization of the supplied text. This is always one of the following +** four values: +** +** <ul><li> <b>FTS5_TOKENIZE_DOCUMENT</b> - A document is being inserted into +** or removed from the FTS table. The tokenizer is being invoked to +** determine the set of tokens to add to (or delete from) the +** FTS index. +** +** <li> <b>FTS5_TOKENIZE_QUERY</b> - A MATCH query is being executed +** against the FTS index. The tokenizer is being called to tokenize +** a bareword or quoted string specified as part of the query. +** +** <li> <b>(FTS5_TOKENIZE_QUERY | FTS5_TOKENIZE_PREFIX)</b> - Same as +** FTS5_TOKENIZE_QUERY, except that the bareword or quoted string is +** followed by a "*" character, indicating that the last token +** returned by the tokenizer will be treated as a token prefix. +** +** <li> <b>FTS5_TOKENIZE_AUX</b> - The tokenizer is being invoked to +** satisfy an fts5_api.xTokenize() request made by an auxiliary +** function. Or an fts5_api.xColumnSize() request made by the same +** on a columnsize=0 database. +** </ul> +** +** For each token in the input string, the supplied callback xToken() must +** be invoked. The first argument to it should be a copy of the pointer +** passed as the second argument to xTokenize(). The third and fourth +** arguments are a pointer to a buffer containing the token text, and the +** size of the token in bytes. The 4th and 5th arguments are the byte offsets +** of the first byte of and first byte immediately following the text from +** which the token is derived within the input. +** +** The second argument passed to the xToken() callback ("tflags") should +** normally be set to 0. The exception is if the tokenizer supports +** synonyms. In this case see the discussion below for details. +** +** FTS5 assumes the xToken() callback is invoked for each token in the +** order that they occur within the input text. +** +** If an xToken() callback returns any value other than SQLITE_OK, then +** the tokenization should be abandoned and the xTokenize() method should +** immediately return a copy of the xToken() return value. Or, if the +** input buffer is exhausted, xTokenize() should return SQLITE_OK. Finally, +** if an error occurs with the xTokenize() implementation itself, it +** may abandon the tokenization and return any error code other than +** SQLITE_OK or SQLITE_DONE. +** +** SYNONYM SUPPORT +** +** Custom tokenizers may also support synonyms. Consider a case in which a +** user wishes to query for a phrase such as "first place". Using the +** built-in tokenizers, the FTS5 query 'first + place' will match instances +** of "first place" within the document set, but not alternative forms +** such as "1st place". In some applications, it would be better to match +** all instances of "first place" or "1st place" regardless of which form +** the user specified in the MATCH query text. +** +** There are several ways to approach this in FTS5: +** +** <ol><li> By mapping all synonyms to a single token. In this case, the +** In the above example, this means that the tokenizer returns the +** same token for inputs "first" and "1st". Say that token is in +** fact "first", so that when the user inserts the document "I won +** 1st place" entries are added to the index for tokens "i", "won", +** "first" and "place". If the user then queries for '1st + place', +** the tokenizer substitutes "first" for "1st" and the query works +** as expected. +** +** <li> By adding multiple synonyms for a single term to the FTS index. +** In this case, when tokenizing query text, the tokenizer may +** provide multiple synonyms for a single term within the document. +** FTS5 then queries the index for each synonym individually. For +** example, faced with the query: +** +** <codeblock> +** ... MATCH 'first place'</codeblock> +** +** the tokenizer offers both "1st" and "first" as synonyms for the +** first token in the MATCH query and FTS5 effectively runs a query +** similar to: +** +** <codeblock> +** ... MATCH '(first OR 1st) place'</codeblock> +** +** except that, for the purposes of auxiliary functions, the query +** still appears to contain just two phrases - "(first OR 1st)" +** being treated as a single phrase. +** +** <li> By adding multiple synonyms for a single term to the FTS index. +** Using this method, when tokenizing document text, the tokenizer +** provides multiple synonyms for each token. So that when a +** document such as "I won first place" is tokenized, entries are +** added to the FTS index for "i", "won", "first", "1st" and +** "place". +** +** This way, even if the tokenizer does not provide synonyms +** when tokenizing query text (it should not - to do would be +** inefficient), it doesn't matter if the user queries for +** 'first + place' or '1st + place', as there are entires in the +** FTS index corresponding to both forms of the first token. +** </ol> +** +** Whether it is parsing document or query text, any call to xToken that +** specifies a <i>tflags</i> argument with the FTS5_TOKEN_COLOCATED bit +** is considered to supply a synonym for the previous token. For example, +** when parsing the document "I won first place", a tokenizer that supports +** synonyms would call xToken() 5 times, as follows: +** +** <codeblock> +** xToken(pCtx, 0, "i", 1, 0, 1); +** xToken(pCtx, 0, "won", 3, 2, 5); +** xToken(pCtx, 0, "first", 5, 6, 11); +** xToken(pCtx, FTS5_TOKEN_COLOCATED, "1st", 3, 6, 11); +** xToken(pCtx, 0, "place", 5, 12, 17); +**</codeblock> +** +** It is an error to specify the FTS5_TOKEN_COLOCATED flag the first time +** xToken() is called. Multiple synonyms may be specified for a single token +** by making multiple calls to xToken(FTS5_TOKEN_COLOCATED) in sequence. +** There is no limit to the number of synonyms that may be provided for a +** single token. +** +** In many cases, method (1) above is the best approach. It does not add +** extra data to the FTS index or require FTS5 to query for multiple terms, +** so it is efficient in terms of disk space and query speed. However, it +** does not support prefix queries very well. If, as suggested above, the +** token "first" is subsituted for "1st" by the tokenizer, then the query: +** +** <codeblock> +** ... MATCH '1s*'</codeblock> +** +** will not match documents that contain the token "1st" (as the tokenizer +** will probably not map "1s" to any prefix of "first"). +** +** For full prefix support, method (3) may be preferred. In this case, +** because the index contains entries for both "first" and "1st", prefix +** queries such as 'fi*' or '1s*' will match correctly. However, because +** extra entries are added to the FTS index, this method uses more space +** within the database. +** +** Method (2) offers a midpoint between (1) and (3). Using this method, +** a query such as '1s*' will match documents that contain the literal +** token "1st", but not "first" (assuming the tokenizer is not able to +** provide synonyms for prefixes). However, a non-prefix query like '1st' +** will match against "1st" and "first". This method does not require +** extra disk space, as no extra entries are added to the FTS index. +** On the other hand, it may require more CPU cycles to run MATCH queries, +** as separate queries of the FTS index are required for each synonym. +** +** When using methods (2) or (3), it is important that the tokenizer only +** provide synonyms when tokenizing document text (method (2)) or query +** text (method (3)), not both. Doing so will not cause any errors, but is +** inefficient. +*/ +typedef struct Fts5Tokenizer Fts5Tokenizer; +typedef struct fts5_tokenizer fts5_tokenizer; +struct fts5_tokenizer { + int (*xCreate)(void*, const char **azArg, int nArg, Fts5Tokenizer **ppOut); + void (*xDelete)(Fts5Tokenizer*); + int (*xTokenize)(Fts5Tokenizer*, + void *pCtx, + int flags, /* Mask of FTS5_TOKENIZE_* flags */ + const char *pText, int nText, + int (*xToken)( + void *pCtx, /* Copy of 2nd argument to xTokenize() */ + int tflags, /* Mask of FTS5_TOKEN_* flags */ + const char *pToken, /* Pointer to buffer containing token */ + int nToken, /* Size of token in bytes */ + int iStart, /* Byte offset of token within input text */ + int iEnd /* Byte offset of end of token within input text */ + ) + ); +}; + +/* Flags that may be passed as the third argument to xTokenize() */ +#define FTS5_TOKENIZE_QUERY 0x0001 +#define FTS5_TOKENIZE_PREFIX 0x0002 +#define FTS5_TOKENIZE_DOCUMENT 0x0004 +#define FTS5_TOKENIZE_AUX 0x0008 + +/* Flags that may be passed by the tokenizer implementation back to FTS5 +** as the third argument to the supplied xToken callback. */ +#define FTS5_TOKEN_COLOCATED 0x0001 /* Same position as prev. token */ + +/* +** END OF CUSTOM TOKENIZERS +*************************************************************************/ + +/************************************************************************* +** FTS5 EXTENSION REGISTRATION API +*/ +typedef struct fts5_api fts5_api; +struct fts5_api { + int iVersion; /* Currently always set to 2 */ + + /* Create a new tokenizer */ + int (*xCreateTokenizer)( + fts5_api *pApi, + const char *zName, + void *pContext, + fts5_tokenizer *pTokenizer, + void (*xDestroy)(void*) + ); + + /* Find an existing tokenizer */ + int (*xFindTokenizer)( + fts5_api *pApi, + const char *zName, + void **ppContext, + fts5_tokenizer *pTokenizer + ); + + /* Create a new auxiliary function */ + int (*xCreateFunction)( + fts5_api *pApi, + const char *zName, + void *pContext, + fts5_extension_function xFunction, + void (*xDestroy)(void*) + ); +}; + +/* +** END OF REGISTRATION API +*************************************************************************/ + +#ifdef __cplusplus +} /* end of the 'extern "C"' block */ +#endif + +#endif /* _FTS5_H */ + + +#line 1 "fts5Int.h" +/* +** 2014 May 31 +** +** The author disclaims copyright to this source code. In place of +** a legal notice, here is a blessing: +** +** May you do good and not evil. +** May you find forgiveness for yourself and forgive others. +** May you share freely, never taking more than you give. +** +****************************************************************************** +** +*/ +#ifndef _FTS5INT_H +#define _FTS5INT_H + +/* #include "fts5.h" */ +#include "sqlite3ext.h" +SQLITE_EXTENSION_INIT1 + +#include <string.h> +#include <assert.h> + +#ifndef SQLITE_AMALGAMATION + +typedef unsigned char u8; +typedef unsigned int u32; +typedef unsigned short u16; +typedef short i16; +typedef sqlite3_int64 i64; +typedef sqlite3_uint64 u64; + +#define ArraySize(x) ((int)(sizeof(x) / sizeof(x[0]))) + +#define testcase(x) +#define ALWAYS(x) 1 +#define NEVER(x) 0 + +#define MIN(x,y) (((x) < (y)) ? (x) : (y)) +#define MAX(x,y) (((x) > (y)) ? (x) : (y)) + +/* +** Constants for the largest and smallest possible 64-bit signed integers. +*/ +# define LARGEST_INT64 (0xffffffff|(((i64)0x7fffffff)<<32)) +# define SMALLEST_INT64 (((i64)-1) - LARGEST_INT64) + +#endif + + +/* +** Maximum number of prefix indexes on single FTS5 table. This must be +** less than 32. If it is set to anything large than that, an #error +** directive in fts5_index.c will cause the build to fail. +*/ +#define FTS5_MAX_PREFIX_INDEXES 31 + +#define FTS5_DEFAULT_NEARDIST 10 +#define FTS5_DEFAULT_RANK "bm25" + +/* Name of rank and rowid columns */ +#define FTS5_RANK_NAME "rank" +#define FTS5_ROWID_NAME "rowid" + +#ifdef SQLITE_DEBUG +# define FTS5_CORRUPT sqlite3Fts5Corrupt() +static int sqlite3Fts5Corrupt(void); +#else +# define FTS5_CORRUPT SQLITE_CORRUPT_VTAB +#endif + +/* +** The assert_nc() macro is similar to the assert() macro, except that it +** is used for assert() conditions that are true only if it can be +** guranteed that the database is not corrupt. +*/ +#ifdef SQLITE_DEBUG +extern int sqlite3_fts5_may_be_corrupt; +# define assert_nc(x) assert(sqlite3_fts5_may_be_corrupt || (x)) +#else +# define assert_nc(x) assert(x) +#endif + +/* Mark a function parameter as unused, to suppress nuisance compiler +** warnings. */ +#ifndef UNUSED_PARAM +# define UNUSED_PARAM(X) (void)(X) +#endif + +#ifndef UNUSED_PARAM2 +# define UNUSED_PARAM2(X, Y) (void)(X), (void)(Y) +#endif + +typedef struct Fts5Global Fts5Global; +typedef struct Fts5Colset Fts5Colset; + +/* If a NEAR() clump or phrase may only match a specific set of columns, +** then an object of the following type is used to record the set of columns. +** Each entry in the aiCol[] array is a column that may be matched. +** +** This object is used by fts5_expr.c and fts5_index.c. +*/ +struct Fts5Colset { + int nCol; + int aiCol[1]; +}; + + + +/************************************************************************** +** Interface to code in fts5_config.c. fts5_config.c contains contains code +** to parse the arguments passed to the CREATE VIRTUAL TABLE statement. +*/ + +typedef struct Fts5Config Fts5Config; + +/* +** An instance of the following structure encodes all information that can +** be gleaned from the CREATE VIRTUAL TABLE statement. +** +** And all information loaded from the %_config table. +** +** nAutomerge: +** The minimum number of segments that an auto-merge operation should +** attempt to merge together. A value of 1 sets the object to use the +** compile time default. Zero disables auto-merge altogether. +** +** zContent: +** +** zContentRowid: +** The value of the content_rowid= option, if one was specified. Or +** the string "rowid" otherwise. This text is not quoted - if it is +** used as part of an SQL statement it needs to be quoted appropriately. +** +** zContentExprlist: +** +** pzErrmsg: +** This exists in order to allow the fts5_index.c module to return a +** decent error message if it encounters a file-format version it does +** not understand. +** +** bColumnsize: +** True if the %_docsize table is created. +** +** bPrefixIndex: +** This is only used for debugging. If set to false, any prefix indexes +** are ignored. This value is configured using: +** +** INSERT INTO tbl(tbl, rank) VALUES('prefix-index', $bPrefixIndex); +** +*/ +struct Fts5Config { + sqlite3 *db; /* Database handle */ + char *zDb; /* Database holding FTS index (e.g. "main") */ + char *zName; /* Name of FTS index */ + int nCol; /* Number of columns */ + char **azCol; /* Column names */ + u8 *abUnindexed; /* True for unindexed columns */ + int nPrefix; /* Number of prefix indexes */ + int *aPrefix; /* Sizes in bytes of nPrefix prefix indexes */ + int eContent; /* An FTS5_CONTENT value */ + char *zContent; /* content table */ + char *zContentRowid; /* "content_rowid=" option value */ + int bColumnsize; /* "columnsize=" option value (dflt==1) */ + int eDetail; /* FTS5_DETAIL_XXX value */ + char *zContentExprlist; + Fts5Tokenizer *pTok; + fts5_tokenizer *pTokApi; + + /* Values loaded from the %_config table */ + int iCookie; /* Incremented when %_config is modified */ + int pgsz; /* Approximate page size used in %_data */ + int nAutomerge; /* 'automerge' setting */ + int nCrisisMerge; /* Maximum allowed segments per level */ + int nHashSize; /* Bytes of memory for in-memory hash */ + char *zRank; /* Name of rank function */ + char *zRankArgs; /* Arguments to rank function */ + + /* If non-NULL, points to sqlite3_vtab.base.zErrmsg. Often NULL. */ + char **pzErrmsg; + +#ifdef SQLITE_DEBUG + int bPrefixIndex; /* True to use prefix-indexes */ +#endif +}; + +/* Current expected value of %_config table 'version' field */ +#define FTS5_CURRENT_VERSION 4 + +#define FTS5_CONTENT_NORMAL 0 +#define FTS5_CONTENT_NONE 1 +#define FTS5_CONTENT_EXTERNAL 2 + +#define FTS5_DETAIL_FULL 0 +#define FTS5_DETAIL_NONE 1 +#define FTS5_DETAIL_COLUMNS 2 + + + +static int sqlite3Fts5ConfigParse( + Fts5Global*, sqlite3*, int, const char **, Fts5Config**, char** +); +static void sqlite3Fts5ConfigFree(Fts5Config*); + +static int sqlite3Fts5ConfigDeclareVtab(Fts5Config *pConfig); + +static int sqlite3Fts5Tokenize( + Fts5Config *pConfig, /* FTS5 Configuration object */ + int flags, /* FTS5_TOKENIZE_* flags */ + const char *pText, int nText, /* Text to tokenize */ + void *pCtx, /* Context passed to xToken() */ + int (*xToken)(void*, int, const char*, int, int, int) /* Callback */ +); + +static void sqlite3Fts5Dequote(char *z); + +/* Load the contents of the %_config table */ +static int sqlite3Fts5ConfigLoad(Fts5Config*, int); + +/* Set the value of a single config attribute */ +static int sqlite3Fts5ConfigSetValue(Fts5Config*, const char*, sqlite3_value*, int*); + +static int sqlite3Fts5ConfigParseRank(const char*, char**, char**); + +/* +** End of interface to code in fts5_config.c. +**************************************************************************/ + +/************************************************************************** +** Interface to code in fts5_buffer.c. +*/ + +/* +** Buffer object for the incremental building of string data. +*/ +typedef struct Fts5Buffer Fts5Buffer; +struct Fts5Buffer { + u8 *p; + int n; + int nSpace; +}; + +static int sqlite3Fts5BufferSize(int*, Fts5Buffer*, u32); +static void sqlite3Fts5BufferAppendVarint(int*, Fts5Buffer*, i64); +static void sqlite3Fts5BufferAppendBlob(int*, Fts5Buffer*, u32, const u8*); +static void sqlite3Fts5BufferAppendString(int *, Fts5Buffer*, const char*); +static void sqlite3Fts5BufferFree(Fts5Buffer*); +static void sqlite3Fts5BufferZero(Fts5Buffer*); +static void sqlite3Fts5BufferSet(int*, Fts5Buffer*, int, const u8*); +static void sqlite3Fts5BufferAppendPrintf(int *, Fts5Buffer*, char *zFmt, ...); + +static char *sqlite3Fts5Mprintf(int *pRc, const char *zFmt, ...); + +#define fts5BufferZero(x) sqlite3Fts5BufferZero(x) +#define fts5BufferAppendVarint(a,b,c) sqlite3Fts5BufferAppendVarint(a,b,c) +#define fts5BufferFree(a) sqlite3Fts5BufferFree(a) +#define fts5BufferAppendBlob(a,b,c,d) sqlite3Fts5BufferAppendBlob(a,b,c,d) +#define fts5BufferSet(a,b,c,d) sqlite3Fts5BufferSet(a,b,c,d) + +#define fts5BufferGrow(pRc,pBuf,nn) ( \ + (u32)((pBuf)->n) + (u32)(nn) <= (u32)((pBuf)->nSpace) ? 0 : \ + sqlite3Fts5BufferSize((pRc),(pBuf),(nn)+(pBuf)->n) \ +) + +/* Write and decode big-endian 32-bit integer values */ +static void sqlite3Fts5Put32(u8*, int); +static int sqlite3Fts5Get32(const u8*); + +#define FTS5_POS2COLUMN(iPos) (int)(iPos >> 32) +#define FTS5_POS2OFFSET(iPos) (int)(iPos & 0xFFFFFFFF) + +typedef struct Fts5PoslistReader Fts5PoslistReader; +struct Fts5PoslistReader { + /* Variables used only by sqlite3Fts5PoslistIterXXX() functions. */ + const u8 *a; /* Position list to iterate through */ + int n; /* Size of buffer at a[] in bytes */ + int i; /* Current offset in a[] */ + + u8 bFlag; /* For client use (any custom purpose) */ + + /* Output variables */ + u8 bEof; /* Set to true at EOF */ + i64 iPos; /* (iCol<<32) + iPos */ +}; +static int sqlite3Fts5PoslistReaderInit( + const u8 *a, int n, /* Poslist buffer to iterate through */ + Fts5PoslistReader *pIter /* Iterator object to initialize */ +); +static int sqlite3Fts5PoslistReaderNext(Fts5PoslistReader*); + +typedef struct Fts5PoslistWriter Fts5PoslistWriter; +struct Fts5PoslistWriter { + i64 iPrev; +}; +static int sqlite3Fts5PoslistWriterAppend(Fts5Buffer*, Fts5PoslistWriter*, i64); +static void sqlite3Fts5PoslistSafeAppend(Fts5Buffer*, i64*, i64); + +static int sqlite3Fts5PoslistNext64( + const u8 *a, int n, /* Buffer containing poslist */ + int *pi, /* IN/OUT: Offset within a[] */ + i64 *piOff /* IN/OUT: Current offset */ +); + +/* Malloc utility */ +static void *sqlite3Fts5MallocZero(int *pRc, int nByte); +static char *sqlite3Fts5Strndup(int *pRc, const char *pIn, int nIn); + +/* Character set tests (like isspace(), isalpha() etc.) */ +static int sqlite3Fts5IsBareword(char t); + + +/* Bucket of terms object used by the integrity-check in offsets=0 mode. */ +typedef struct Fts5Termset Fts5Termset; +static int sqlite3Fts5TermsetNew(Fts5Termset**); +static int sqlite3Fts5TermsetAdd(Fts5Termset*, int, const char*, int, int *pbPresent); +static void sqlite3Fts5TermsetFree(Fts5Termset*); + +/* +** End of interface to code in fts5_buffer.c. +**************************************************************************/ + +/************************************************************************** +** Interface to code in fts5_index.c. fts5_index.c contains contains code +** to access the data stored in the %_data table. +*/ + +typedef struct Fts5Index Fts5Index; +typedef struct Fts5IndexIter Fts5IndexIter; + +struct Fts5IndexIter { + i64 iRowid; + const u8 *pData; + int nData; + u8 bEof; +}; + +#define sqlite3Fts5IterEof(x) ((x)->bEof) + +/* +** Values used as part of the flags argument passed to IndexQuery(). +*/ +#define FTS5INDEX_QUERY_PREFIX 0x0001 /* Prefix query */ +#define FTS5INDEX_QUERY_DESC 0x0002 /* Docs in descending rowid order */ +#define FTS5INDEX_QUERY_TEST_NOIDX 0x0004 /* Do not use prefix index */ +#define FTS5INDEX_QUERY_SCAN 0x0008 /* Scan query (fts5vocab) */ + +/* The following are used internally by the fts5_index.c module. They are +** defined here only to make it easier to avoid clashes with the flags +** above. */ +#define FTS5INDEX_QUERY_SKIPEMPTY 0x0010 +#define FTS5INDEX_QUERY_NOOUTPUT 0x0020 + +/* +** Create/destroy an Fts5Index object. +*/ +static int sqlite3Fts5IndexOpen(Fts5Config *pConfig, int bCreate, Fts5Index**, char**); +static int sqlite3Fts5IndexClose(Fts5Index *p); + +/* +** Return a simple checksum value based on the arguments. +*/ +static u64 sqlite3Fts5IndexEntryCksum( + i64 iRowid, + int iCol, + int iPos, + int iIdx, + const char *pTerm, + int nTerm +); + +/* +** Argument p points to a buffer containing utf-8 text that is n bytes in +** size. Return the number of bytes in the nChar character prefix of the +** buffer, or 0 if there are less than nChar characters in total. +*/ +static int sqlite3Fts5IndexCharlenToBytelen( + const char *p, + int nByte, + int nChar +); + +/* +** Open a new iterator to iterate though all rowids that match the +** specified token or token prefix. +*/ +static int sqlite3Fts5IndexQuery( + Fts5Index *p, /* FTS index to query */ + const char *pToken, int nToken, /* Token (or prefix) to query for */ + int flags, /* Mask of FTS5INDEX_QUERY_X flags */ + Fts5Colset *pColset, /* Match these columns only */ + Fts5IndexIter **ppIter /* OUT: New iterator object */ +); + +/* +** The various operations on open token or token prefix iterators opened +** using sqlite3Fts5IndexQuery(). +*/ +static int sqlite3Fts5IterNext(Fts5IndexIter*); +static int sqlite3Fts5IterNextFrom(Fts5IndexIter*, i64 iMatch); + +/* +** Close an iterator opened by sqlite3Fts5IndexQuery(). +*/ +static void sqlite3Fts5IterClose(Fts5IndexIter*); + +/* +** This interface is used by the fts5vocab module. +*/ +static const char *sqlite3Fts5IterTerm(Fts5IndexIter*, int*); +static int sqlite3Fts5IterNextScan(Fts5IndexIter*); + + +/* +** Insert or remove data to or from the index. Each time a document is +** added to or removed from the index, this function is called one or more +** times. +** +** For an insert, it must be called once for each token in the new document. +** If the operation is a delete, it must be called (at least) once for each +** unique token in the document with an iCol value less than zero. The iPos +** argument is ignored for a delete. +*/ +static int sqlite3Fts5IndexWrite( + Fts5Index *p, /* Index to write to */ + int iCol, /* Column token appears in (-ve -> delete) */ + int iPos, /* Position of token within column */ + const char *pToken, int nToken /* Token to add or remove to or from index */ +); + +/* +** Indicate that subsequent calls to sqlite3Fts5IndexWrite() pertain to +** document iDocid. +*/ +static int sqlite3Fts5IndexBeginWrite( + Fts5Index *p, /* Index to write to */ + int bDelete, /* True if current operation is a delete */ + i64 iDocid /* Docid to add or remove data from */ +); + +/* +** Flush any data stored in the in-memory hash tables to the database. +** If the bCommit flag is true, also close any open blob handles. +*/ +static int sqlite3Fts5IndexSync(Fts5Index *p, int bCommit); + +/* +** Discard any data stored in the in-memory hash tables. Do not write it +** to the database. Additionally, assume that the contents of the %_data +** table may have changed on disk. So any in-memory caches of %_data +** records must be invalidated. +*/ +static int sqlite3Fts5IndexRollback(Fts5Index *p); + +/* +** Get or set the "averages" values. +*/ +static int sqlite3Fts5IndexGetAverages(Fts5Index *p, i64 *pnRow, i64 *anSize); +static int sqlite3Fts5IndexSetAverages(Fts5Index *p, const u8*, int); + +/* +** Functions called by the storage module as part of integrity-check. +*/ +static int sqlite3Fts5IndexIntegrityCheck(Fts5Index*, u64 cksum); + +/* +** Called during virtual module initialization to register UDF +** fts5_decode() with SQLite +*/ +static int sqlite3Fts5IndexInit(sqlite3*); + +static int sqlite3Fts5IndexSetCookie(Fts5Index*, int); + +/* +** Return the total number of entries read from the %_data table by +** this connection since it was created. +*/ +static int sqlite3Fts5IndexReads(Fts5Index *p); + +static int sqlite3Fts5IndexReinit(Fts5Index *p); +static int sqlite3Fts5IndexOptimize(Fts5Index *p); +static int sqlite3Fts5IndexMerge(Fts5Index *p, int nMerge); + +static int sqlite3Fts5IndexLoadConfig(Fts5Index *p); + +/* +** End of interface to code in fts5_index.c. +**************************************************************************/ + +/************************************************************************** +** Interface to code in fts5_varint.c. +*/ +static int sqlite3Fts5GetVarint32(const unsigned char *p, u32 *v); +static int sqlite3Fts5GetVarintLen(u32 iVal); +static u8 sqlite3Fts5GetVarint(const unsigned char*, u64*); +static int sqlite3Fts5PutVarint(unsigned char *p, u64 v); + +#define fts5GetVarint32(a,b) sqlite3Fts5GetVarint32(a,(u32*)&b) +#define fts5GetVarint sqlite3Fts5GetVarint + +#define fts5FastGetVarint32(a, iOff, nVal) { \ + nVal = (a)[iOff++]; \ + if( nVal & 0x80 ){ \ + iOff--; \ + iOff += fts5GetVarint32(&(a)[iOff], nVal); \ + } \ +} + + +/* +** End of interface to code in fts5_varint.c. +**************************************************************************/ + + +/************************************************************************** +** Interface to code in fts5.c. +*/ + +static int sqlite3Fts5GetTokenizer( + Fts5Global*, + const char **azArg, + int nArg, + Fts5Tokenizer**, + fts5_tokenizer**, + char **pzErr +); + +static Fts5Index *sqlite3Fts5IndexFromCsrid(Fts5Global*, i64, Fts5Config **); + +/* +** End of interface to code in fts5.c. +**************************************************************************/ + +/************************************************************************** +** Interface to code in fts5_hash.c. +*/ +typedef struct Fts5Hash Fts5Hash; + +/* +** Create a hash table, free a hash table. +*/ +static int sqlite3Fts5HashNew(Fts5Config*, Fts5Hash**, int *pnSize); +static void sqlite3Fts5HashFree(Fts5Hash*); + +static int sqlite3Fts5HashWrite( + Fts5Hash*, + i64 iRowid, /* Rowid for this entry */ + int iCol, /* Column token appears in (-ve -> delete) */ + int iPos, /* Position of token within column */ + char bByte, + const char *pToken, int nToken /* Token to add or remove to or from index */ +); + +/* +** Empty (but do not delete) a hash table. +*/ +static void sqlite3Fts5HashClear(Fts5Hash*); + +static int sqlite3Fts5HashQuery( + Fts5Hash*, /* Hash table to query */ + const char *pTerm, int nTerm, /* Query term */ + const u8 **ppDoclist, /* OUT: Pointer to doclist for pTerm */ + int *pnDoclist /* OUT: Size of doclist in bytes */ +); + +static int sqlite3Fts5HashScanInit( + Fts5Hash*, /* Hash table to query */ + const char *pTerm, int nTerm /* Query prefix */ +); +static void sqlite3Fts5HashScanNext(Fts5Hash*); +static int sqlite3Fts5HashScanEof(Fts5Hash*); +static void sqlite3Fts5HashScanEntry(Fts5Hash *, + const char **pzTerm, /* OUT: term (nul-terminated) */ + const u8 **ppDoclist, /* OUT: pointer to doclist */ + int *pnDoclist /* OUT: size of doclist in bytes */ +); + + +/* +** End of interface to code in fts5_hash.c. +**************************************************************************/ + +/************************************************************************** +** Interface to code in fts5_storage.c. fts5_storage.c contains contains +** code to access the data stored in the %_content and %_docsize tables. +*/ + +#define FTS5_STMT_SCAN_ASC 0 /* SELECT rowid, * FROM ... ORDER BY 1 ASC */ +#define FTS5_STMT_SCAN_DESC 1 /* SELECT rowid, * FROM ... ORDER BY 1 DESC */ +#define FTS5_STMT_LOOKUP 2 /* SELECT rowid, * FROM ... WHERE rowid=? */ + +typedef struct Fts5Storage Fts5Storage; + +static int sqlite3Fts5StorageOpen(Fts5Config*, Fts5Index*, int, Fts5Storage**, char**); +static int sqlite3Fts5StorageClose(Fts5Storage *p); +static int sqlite3Fts5StorageRename(Fts5Storage*, const char *zName); + +static int sqlite3Fts5DropAll(Fts5Config*); +static int sqlite3Fts5CreateTable(Fts5Config*, const char*, const char*, int, char **); + +static int sqlite3Fts5StorageDelete(Fts5Storage *p, i64, sqlite3_value**); +static int sqlite3Fts5StorageContentInsert(Fts5Storage *p, sqlite3_value**, i64*); +static int sqlite3Fts5StorageIndexInsert(Fts5Storage *p, sqlite3_value**, i64); + +static int sqlite3Fts5StorageIntegrity(Fts5Storage *p); + +static int sqlite3Fts5StorageStmt(Fts5Storage *p, int eStmt, sqlite3_stmt**, char**); +static void sqlite3Fts5StorageStmtRelease(Fts5Storage *p, int eStmt, sqlite3_stmt*); + +static int sqlite3Fts5StorageDocsize(Fts5Storage *p, i64 iRowid, int *aCol); +static int sqlite3Fts5StorageSize(Fts5Storage *p, int iCol, i64 *pnAvg); +static int sqlite3Fts5StorageRowCount(Fts5Storage *p, i64 *pnRow); + +static int sqlite3Fts5StorageSync(Fts5Storage *p, int bCommit); +static int sqlite3Fts5StorageRollback(Fts5Storage *p); + +static int sqlite3Fts5StorageConfigValue( + Fts5Storage *p, const char*, sqlite3_value*, int +); + +static int sqlite3Fts5StorageDeleteAll(Fts5Storage *p); +static int sqlite3Fts5StorageRebuild(Fts5Storage *p); +static int sqlite3Fts5StorageOptimize(Fts5Storage *p); +static int sqlite3Fts5StorageMerge(Fts5Storage *p, int nMerge); + +/* +** End of interface to code in fts5_storage.c. +**************************************************************************/ + + +/************************************************************************** +** Interface to code in fts5_expr.c. +*/ +typedef struct Fts5Expr Fts5Expr; +typedef struct Fts5ExprNode Fts5ExprNode; +typedef struct Fts5Parse Fts5Parse; +typedef struct Fts5Token Fts5Token; +typedef struct Fts5ExprPhrase Fts5ExprPhrase; +typedef struct Fts5ExprNearset Fts5ExprNearset; + +struct Fts5Token { + const char *p; /* Token text (not NULL terminated) */ + int n; /* Size of buffer p in bytes */ +}; + +/* Parse a MATCH expression. */ +static int sqlite3Fts5ExprNew( + Fts5Config *pConfig, + const char *zExpr, + Fts5Expr **ppNew, + char **pzErr +); + +/* +** for(rc = sqlite3Fts5ExprFirst(pExpr, pIdx, bDesc); +** rc==SQLITE_OK && 0==sqlite3Fts5ExprEof(pExpr); +** rc = sqlite3Fts5ExprNext(pExpr) +** ){ +** // The document with rowid iRowid matches the expression! +** i64 iRowid = sqlite3Fts5ExprRowid(pExpr); +** } +*/ +static int sqlite3Fts5ExprFirst(Fts5Expr*, Fts5Index *pIdx, i64 iMin, int bDesc); +static int sqlite3Fts5ExprNext(Fts5Expr*, i64 iMax); +static int sqlite3Fts5ExprEof(Fts5Expr*); +static i64 sqlite3Fts5ExprRowid(Fts5Expr*); + +static void sqlite3Fts5ExprFree(Fts5Expr*); + +/* Called during startup to register a UDF with SQLite */ +static int sqlite3Fts5ExprInit(Fts5Global*, sqlite3*); + +static int sqlite3Fts5ExprPhraseCount(Fts5Expr*); +static int sqlite3Fts5ExprPhraseSize(Fts5Expr*, int iPhrase); +static int sqlite3Fts5ExprPoslist(Fts5Expr*, int, const u8 **); + +typedef struct Fts5PoslistPopulator Fts5PoslistPopulator; +static Fts5PoslistPopulator *sqlite3Fts5ExprClearPoslists(Fts5Expr*, int); +static int sqlite3Fts5ExprPopulatePoslists( + Fts5Config*, Fts5Expr*, Fts5PoslistPopulator*, int, const char*, int +); +static void sqlite3Fts5ExprCheckPoslists(Fts5Expr*, i64); +static void sqlite3Fts5ExprClearEof(Fts5Expr*); + +static int sqlite3Fts5ExprClonePhrase(Fts5Expr*, int, Fts5Expr**); + +static int sqlite3Fts5ExprPhraseCollist(Fts5Expr *, int, const u8 **, int *); + +/******************************************* +** The fts5_expr.c API above this point is used by the other hand-written +** C code in this module. The interfaces below this point are called by +** the parser code in fts5parse.y. */ + +static void sqlite3Fts5ParseError(Fts5Parse *pParse, const char *zFmt, ...); + +static Fts5ExprNode *sqlite3Fts5ParseNode( + Fts5Parse *pParse, + int eType, + Fts5ExprNode *pLeft, + Fts5ExprNode *pRight, + Fts5ExprNearset *pNear +); + +static Fts5ExprPhrase *sqlite3Fts5ParseTerm( + Fts5Parse *pParse, + Fts5ExprPhrase *pPhrase, + Fts5Token *pToken, + int bPrefix +); + +static Fts5ExprNearset *sqlite3Fts5ParseNearset( + Fts5Parse*, + Fts5ExprNearset*, + Fts5ExprPhrase* +); + +static Fts5Colset *sqlite3Fts5ParseColset( + Fts5Parse*, + Fts5Colset*, + Fts5Token * +); + +static void sqlite3Fts5ParsePhraseFree(Fts5ExprPhrase*); +static void sqlite3Fts5ParseNearsetFree(Fts5ExprNearset*); +static void sqlite3Fts5ParseNodeFree(Fts5ExprNode*); + +static void sqlite3Fts5ParseSetDistance(Fts5Parse*, Fts5ExprNearset*, Fts5Token*); +static void sqlite3Fts5ParseSetColset(Fts5Parse*, Fts5ExprNearset*, Fts5Colset*); +static void sqlite3Fts5ParseFinished(Fts5Parse *pParse, Fts5ExprNode *p); +static void sqlite3Fts5ParseNear(Fts5Parse *pParse, Fts5Token*); + +/* +** End of interface to code in fts5_expr.c. +**************************************************************************/ + + + +/************************************************************************** +** Interface to code in fts5_aux.c. +*/ + +static int sqlite3Fts5AuxInit(fts5_api*); +/* +** End of interface to code in fts5_aux.c. +**************************************************************************/ + +/************************************************************************** +** Interface to code in fts5_tokenizer.c. +*/ + +static int sqlite3Fts5TokenizerInit(fts5_api*); +/* +** End of interface to code in fts5_tokenizer.c. +**************************************************************************/ + +/************************************************************************** +** Interface to code in fts5_vocab.c. +*/ + +static int sqlite3Fts5VocabInit(Fts5Global*, sqlite3*); + +/* +** End of interface to code in fts5_vocab.c. +**************************************************************************/ + + +/************************************************************************** +** Interface to automatically generated code in fts5_unicode2.c. +*/ +static int sqlite3Fts5UnicodeIsalnum(int c); +static int sqlite3Fts5UnicodeIsdiacritic(int c); +static int sqlite3Fts5UnicodeFold(int c, int bRemoveDiacritic); +/* +** End of interface to code in fts5_unicode2.c. +**************************************************************************/ + +#endif + +#line 1 "fts5parse.h" +#define FTS5_OR 1 +#define FTS5_AND 2 +#define FTS5_NOT 3 +#define FTS5_TERM 4 +#define FTS5_COLON 5 +#define FTS5_LP 6 +#define FTS5_RP 7 +#define FTS5_LCP 8 +#define FTS5_RCP 9 +#define FTS5_STRING 10 +#define FTS5_COMMA 11 +#define FTS5_PLUS 12 +#define FTS5_STAR 13 + +#line 1 "fts5parse.c" +/* +** 2000-05-29 +** +** The author disclaims copyright to this source code. In place of +** a legal notice, here is a blessing: +** +** May you do good and not evil. +** May you find forgiveness for yourself and forgive others. +** May you share freely, never taking more than you give. +** +************************************************************************* +** Driver template for the LEMON parser generator. +** +** The "lemon" program processes an LALR(1) input grammar file, then uses +** this template to construct a parser. The "lemon" program inserts text +** at each "%%" line. Also, any "P-a-r-s-e" identifer prefix (without the +** interstitial "-" characters) contained in this template is changed into +** the value of the %name directive from the grammar. Otherwise, the content +** of this template is copied straight through into the generate parser +** source file. +** +** The following is the concatenation of all %include directives from the +** input grammar file: +*/ +#include <stdio.h> +/************ Begin %include sections from the grammar ************************/ +#line 47 "fts5parse.y" + +/* #include "fts5Int.h" */ +/* #include "fts5parse.h" */ + +/* +** Disable all error recovery processing in the parser push-down +** automaton. +*/ +#define fts5YYNOERRORRECOVERY 1 + +/* +** Make fts5yytestcase() the same as testcase() +*/ +#define fts5yytestcase(X) testcase(X) + +/* +** Indicate that sqlite3ParserFree() will never be called with a null +** pointer. +*/ +#define fts5YYPARSEFREENOTNULL 1 + +/* +** Alternative datatype for the argument to the malloc() routine passed +** into sqlite3ParserAlloc(). The default is size_t. +*/ +#define fts5YYMALLOCARGTYPE u64 + +#line 56 "fts5parse.c" +/**************** End of %include directives **********************************/ +/* These constants specify the various numeric values for terminal symbols +** in a format understandable to "makeheaders". This section is blank unless +** "lemon" is run with the "-m" command-line option. +***************** Begin makeheaders token definitions *************************/ +/**************** End makeheaders token definitions ***************************/ + +/* The next sections is a series of control #defines. +** various aspects of the generated parser. +** fts5YYCODETYPE is the data type used to store the integer codes +** that represent terminal and non-terminal symbols. +** "unsigned char" is used if there are fewer than +** 256 symbols. Larger types otherwise. +** fts5YYNOCODE is a number of type fts5YYCODETYPE that is not used for +** any terminal or nonterminal symbol. +** fts5YYFALLBACK If defined, this indicates that one or more tokens +** (also known as: "terminal symbols") have fall-back +** values which should be used if the original symbol +** would not parse. This permits keywords to sometimes +** be used as identifiers, for example. +** fts5YYACTIONTYPE is the data type used for "action codes" - numbers +** that indicate what to do in response to the next +** token. +** sqlite3Fts5ParserFTS5TOKENTYPE is the data type used for minor type for terminal +** symbols. Background: A "minor type" is a semantic +** value associated with a terminal or non-terminal +** symbols. For example, for an "ID" terminal symbol, +** the minor type might be the name of the identifier. +** Each non-terminal can have a different minor type. +** Terminal symbols all have the same minor type, though. +** This macros defines the minor type for terminal +** symbols. +** fts5YYMINORTYPE is the data type used for all minor types. +** This is typically a union of many types, one of +** which is sqlite3Fts5ParserFTS5TOKENTYPE. The entry in the union +** for terminal symbols is called "fts5yy0". +** fts5YYSTACKDEPTH is the maximum depth of the parser's stack. If +** zero the stack is dynamically sized using realloc() +** sqlite3Fts5ParserARG_SDECL A static variable declaration for the %extra_argument +** sqlite3Fts5ParserARG_PDECL A parameter declaration for the %extra_argument +** sqlite3Fts5ParserARG_STORE Code to store %extra_argument into fts5yypParser +** sqlite3Fts5ParserARG_FETCH Code to extract %extra_argument from fts5yypParser +** fts5YYERRORSYMBOL is the code number of the error symbol. If not +** defined, then do no error processing. +** fts5YYNSTATE the combined number of states. +** fts5YYNRULE the number of rules in the grammar +** fts5YY_MAX_SHIFT Maximum value for shift actions +** fts5YY_MIN_SHIFTREDUCE Minimum value for shift-reduce actions +** fts5YY_MAX_SHIFTREDUCE Maximum value for shift-reduce actions +** fts5YY_MIN_REDUCE Maximum value for reduce actions +** fts5YY_ERROR_ACTION The fts5yy_action[] code for syntax error +** fts5YY_ACCEPT_ACTION The fts5yy_action[] code for accept +** fts5YY_NO_ACTION The fts5yy_action[] code for no-op +*/ +#ifndef INTERFACE +# define INTERFACE 1 +#endif +/************* Begin control #defines *****************************************/ +#define fts5YYCODETYPE unsigned char +#define fts5YYNOCODE 27 +#define fts5YYACTIONTYPE unsigned char +#define sqlite3Fts5ParserFTS5TOKENTYPE Fts5Token +typedef union { + int fts5yyinit; + sqlite3Fts5ParserFTS5TOKENTYPE fts5yy0; + Fts5Colset* fts5yy3; + Fts5ExprPhrase* fts5yy11; + Fts5ExprNode* fts5yy18; + int fts5yy20; + Fts5ExprNearset* fts5yy26; +} fts5YYMINORTYPE; +#ifndef fts5YYSTACKDEPTH +#define fts5YYSTACKDEPTH 100 +#endif +#define sqlite3Fts5ParserARG_SDECL Fts5Parse *pParse; +#define sqlite3Fts5ParserARG_PDECL ,Fts5Parse *pParse +#define sqlite3Fts5ParserARG_FETCH Fts5Parse *pParse = fts5yypParser->pParse +#define sqlite3Fts5ParserARG_STORE fts5yypParser->pParse = pParse +#define fts5YYNSTATE 26 +#define fts5YYNRULE 24 +#define fts5YY_MAX_SHIFT 25 +#define fts5YY_MIN_SHIFTREDUCE 40 +#define fts5YY_MAX_SHIFTREDUCE 63 +#define fts5YY_MIN_REDUCE 64 +#define fts5YY_MAX_REDUCE 87 +#define fts5YY_ERROR_ACTION 88 +#define fts5YY_ACCEPT_ACTION 89 +#define fts5YY_NO_ACTION 90 +/************* End control #defines *******************************************/ + +/* Define the fts5yytestcase() macro to be a no-op if is not already defined +** otherwise. +** +** Applications can choose to define fts5yytestcase() in the %include section +** to a macro that can assist in verifying code coverage. For production +** code the fts5yytestcase() macro should be turned off. But it is useful +** for testing. +*/ +#ifndef fts5yytestcase +# define fts5yytestcase(X) +#endif + + +/* Next are the tables used to determine what action to take based on the +** current state and lookahead token. These tables are used to implement +** functions that take a state number and lookahead value and return an +** action integer. +** +** Suppose the action integer is N. Then the action is determined as +** follows +** +** 0 <= N <= fts5YY_MAX_SHIFT Shift N. That is, push the lookahead +** token onto the stack and goto state N. +** +** N between fts5YY_MIN_SHIFTREDUCE Shift to an arbitrary state then +** and fts5YY_MAX_SHIFTREDUCE reduce by rule N-fts5YY_MIN_SHIFTREDUCE. +** +** N between fts5YY_MIN_REDUCE Reduce by rule N-fts5YY_MIN_REDUCE +** and fts5YY_MAX_REDUCE + +** N == fts5YY_ERROR_ACTION A syntax error has occurred. +** +** N == fts5YY_ACCEPT_ACTION The parser accepts its input. +** +** N == fts5YY_NO_ACTION No such action. Denotes unused +** slots in the fts5yy_action[] table. +** +** The action table is constructed as a single large table named fts5yy_action[]. +** Given state S and lookahead X, the action is computed as +** +** fts5yy_action[ fts5yy_shift_ofst[S] + X ] +** +** If the index value fts5yy_shift_ofst[S]+X is out of range or if the value +** fts5yy_lookahead[fts5yy_shift_ofst[S]+X] is not equal to X or if fts5yy_shift_ofst[S] +** is equal to fts5YY_SHIFT_USE_DFLT, it means that the action is not in the table +** and that fts5yy_default[S] should be used instead. +** +** The formula above is for computing the action when the lookahead is +** a terminal symbol. If the lookahead is a non-terminal (as occurs after +** a reduce action) then the fts5yy_reduce_ofst[] array is used in place of +** the fts5yy_shift_ofst[] array and fts5YY_REDUCE_USE_DFLT is used in place of +** fts5YY_SHIFT_USE_DFLT. +** +** The following are the tables generated in this section: +** +** fts5yy_action[] A single table containing all actions. +** fts5yy_lookahead[] A table containing the lookahead for each entry in +** fts5yy_action. Used to detect hash collisions. +** fts5yy_shift_ofst[] For each state, the offset into fts5yy_action for +** shifting terminals. +** fts5yy_reduce_ofst[] For each state, the offset into fts5yy_action for +** shifting non-terminals after a reduce. +** fts5yy_default[] Default action for each state. +** +*********** Begin parsing tables **********************************************/ +#define fts5YY_ACTTAB_COUNT (78) +static const fts5YYACTIONTYPE fts5yy_action[] = { + /* 0 */ 89, 15, 46, 5, 48, 24, 12, 19, 23, 14, + /* 10 */ 46, 5, 48, 24, 20, 21, 23, 43, 46, 5, + /* 20 */ 48, 24, 6, 18, 23, 17, 46, 5, 48, 24, + /* 30 */ 75, 7, 23, 25, 46, 5, 48, 24, 62, 47, + /* 40 */ 23, 48, 24, 7, 11, 23, 9, 3, 4, 2, + /* 50 */ 62, 50, 52, 44, 64, 3, 4, 2, 49, 4, + /* 60 */ 2, 1, 23, 11, 16, 9, 12, 2, 10, 61, + /* 70 */ 53, 59, 62, 60, 22, 13, 55, 8, +}; +static const fts5YYCODETYPE fts5yy_lookahead[] = { + /* 0 */ 15, 16, 17, 18, 19, 20, 10, 11, 23, 16, + /* 10 */ 17, 18, 19, 20, 23, 24, 23, 16, 17, 18, + /* 20 */ 19, 20, 22, 23, 23, 16, 17, 18, 19, 20, + /* 30 */ 5, 6, 23, 16, 17, 18, 19, 20, 13, 17, + /* 40 */ 23, 19, 20, 6, 8, 23, 10, 1, 2, 3, + /* 50 */ 13, 9, 10, 7, 0, 1, 2, 3, 19, 2, + /* 60 */ 3, 6, 23, 8, 21, 10, 10, 3, 10, 25, + /* 70 */ 10, 10, 13, 25, 12, 10, 7, 5, +}; +#define fts5YY_SHIFT_USE_DFLT (-5) +#define fts5YY_SHIFT_COUNT (25) +#define fts5YY_SHIFT_MIN (-4) +#define fts5YY_SHIFT_MAX (72) +static const signed char fts5yy_shift_ofst[] = { + /* 0 */ 55, 55, 55, 55, 55, 36, -4, 56, 58, 25, + /* 10 */ 37, 60, 59, 59, 46, 54, 42, 57, 62, 61, + /* 20 */ 62, 69, 65, 62, 72, 64, +}; +#define fts5YY_REDUCE_USE_DFLT (-16) +#define fts5YY_REDUCE_COUNT (13) +#define fts5YY_REDUCE_MIN (-15) +#define fts5YY_REDUCE_MAX (48) +static const signed char fts5yy_reduce_ofst[] = { + /* 0 */ -15, -7, 1, 9, 17, 22, -9, 0, 39, 44, + /* 10 */ 44, 43, 44, 48, +}; +static const fts5YYACTIONTYPE fts5yy_default[] = { + /* 0 */ 88, 88, 88, 88, 88, 69, 82, 88, 88, 87, + /* 10 */ 87, 88, 87, 87, 88, 88, 88, 66, 80, 88, + /* 20 */ 81, 88, 88, 78, 88, 65, +}; +/********** End of lemon-generated parsing tables *****************************/ + +/* The next table maps tokens (terminal symbols) into fallback tokens. +** If a construct like the following: +** +** %fallback ID X Y Z. +** +** appears in the grammar, then ID becomes a fallback token for X, Y, +** and Z. Whenever one of the tokens X, Y, or Z is input to the parser +** but it does not parse, the type of the token is changed to ID and +** the parse is retried before an error is thrown. +** +** This feature can be used, for example, to cause some keywords in a language +** to revert to identifiers if they keyword does not apply in the context where +** it appears. +*/ +#ifdef fts5YYFALLBACK +static const fts5YYCODETYPE fts5yyFallback[] = { +}; +#endif /* fts5YYFALLBACK */ + +/* The following structure represents a single element of the +** parser's stack. Information stored includes: +** +** + The state number for the parser at this level of the stack. +** +** + The value of the token stored at this level of the stack. +** (In other words, the "major" token.) +** +** + The semantic value stored at this level of the stack. This is +** the information used by the action routines in the grammar. +** It is sometimes called the "minor" token. +** +** After the "shift" half of a SHIFTREDUCE action, the stateno field +** actually contains the reduce action for the second half of the +** SHIFTREDUCE. +*/ +struct fts5yyStackEntry { + fts5YYACTIONTYPE stateno; /* The state-number, or reduce action in SHIFTREDUCE */ + fts5YYCODETYPE major; /* The major token value. This is the code + ** number for the token at this stack level */ + fts5YYMINORTYPE minor; /* The user-supplied minor token value. This + ** is the value of the token */ +}; +typedef struct fts5yyStackEntry fts5yyStackEntry; + +/* The state of the parser is completely contained in an instance of +** the following structure */ +struct fts5yyParser { + int fts5yyidx; /* Index of top element in stack */ +#ifdef fts5YYTRACKMAXSTACKDEPTH + int fts5yyidxMax; /* Maximum value of fts5yyidx */ +#endif +#ifndef fts5YYNOERRORRECOVERY + int fts5yyerrcnt; /* Shifts left before out of the error */ +#endif + sqlite3Fts5ParserARG_SDECL /* A place to hold %extra_argument */ +#if fts5YYSTACKDEPTH<=0 + int fts5yystksz; /* Current side of the stack */ + fts5yyStackEntry *fts5yystack; /* The parser's stack */ +#else + fts5yyStackEntry fts5yystack[fts5YYSTACKDEPTH]; /* The parser's stack */ +#endif +}; +typedef struct fts5yyParser fts5yyParser; + +#ifndef NDEBUG +#include <stdio.h> +static FILE *fts5yyTraceFILE = 0; +static char *fts5yyTracePrompt = 0; +#endif /* NDEBUG */ + +#ifndef NDEBUG +/* +** Turn parser tracing on by giving a stream to which to write the trace +** and a prompt to preface each trace message. Tracing is turned off +** by making either argument NULL +** +** Inputs: +** <ul> +** <li> A FILE* to which trace output should be written. +** If NULL, then tracing is turned off. +** <li> A prefix string written at the beginning of every +** line of trace output. If NULL, then tracing is +** turned off. +** </ul> +** +** Outputs: +** None. +*/ +static void sqlite3Fts5ParserTrace(FILE *TraceFILE, char *zTracePrompt){ + fts5yyTraceFILE = TraceFILE; + fts5yyTracePrompt = zTracePrompt; + if( fts5yyTraceFILE==0 ) fts5yyTracePrompt = 0; + else if( fts5yyTracePrompt==0 ) fts5yyTraceFILE = 0; +} +#endif /* NDEBUG */ + +#ifndef NDEBUG +/* For tracing shifts, the names of all terminals and nonterminals +** are required. The following table supplies these names */ +static const char *const fts5yyTokenName[] = { + "$", "OR", "AND", "NOT", + "TERM", "COLON", "LP", "RP", + "LCP", "RCP", "STRING", "COMMA", + "PLUS", "STAR", "error", "input", + "expr", "cnearset", "exprlist", "nearset", + "colset", "colsetlist", "nearphrases", "phrase", + "neardist_opt", "star_opt", +}; +#endif /* NDEBUG */ + +#ifndef NDEBUG +/* For tracing reduce actions, the names of all rules are required. +*/ +static const char *const fts5yyRuleName[] = { + /* 0 */ "input ::= expr", + /* 1 */ "expr ::= expr AND expr", + /* 2 */ "expr ::= expr OR expr", + /* 3 */ "expr ::= expr NOT expr", + /* 4 */ "expr ::= LP expr RP", + /* 5 */ "expr ::= exprlist", + /* 6 */ "exprlist ::= cnearset", + /* 7 */ "exprlist ::= exprlist cnearset", + /* 8 */ "cnearset ::= nearset", + /* 9 */ "cnearset ::= colset COLON nearset", + /* 10 */ "colset ::= LCP colsetlist RCP", + /* 11 */ "colset ::= STRING", + /* 12 */ "colsetlist ::= colsetlist STRING", + /* 13 */ "colsetlist ::= STRING", + /* 14 */ "nearset ::= phrase", + /* 15 */ "nearset ::= STRING LP nearphrases neardist_opt RP", + /* 16 */ "nearphrases ::= phrase", + /* 17 */ "nearphrases ::= nearphrases phrase", + /* 18 */ "neardist_opt ::=", + /* 19 */ "neardist_opt ::= COMMA STRING", + /* 20 */ "phrase ::= phrase PLUS STRING star_opt", + /* 21 */ "phrase ::= STRING star_opt", + /* 22 */ "star_opt ::= STAR", + /* 23 */ "star_opt ::=", +}; +#endif /* NDEBUG */ + + +#if fts5YYSTACKDEPTH<=0 +/* +** Try to increase the size of the parser stack. +*/ +static void fts5yyGrowStack(fts5yyParser *p){ + int newSize; + fts5yyStackEntry *pNew; + + newSize = p->fts5yystksz*2 + 100; + pNew = realloc(p->fts5yystack, newSize*sizeof(pNew[0])); + if( pNew ){ + p->fts5yystack = pNew; + p->fts5yystksz = newSize; +#ifndef NDEBUG + if( fts5yyTraceFILE ){ + fprintf(fts5yyTraceFILE,"%sStack grows to %d entries!\n", + fts5yyTracePrompt, p->fts5yystksz); + } +#endif + } +} +#endif + +/* Datatype of the argument to the memory allocated passed as the +** second argument to sqlite3Fts5ParserAlloc() below. This can be changed by +** putting an appropriate #define in the %include section of the input +** grammar. +*/ +#ifndef fts5YYMALLOCARGTYPE +# define fts5YYMALLOCARGTYPE size_t +#endif + +/* +** This function allocates a new parser. +** The only argument is a pointer to a function which works like +** malloc. +** +** Inputs: +** A pointer to the function used to allocate memory. +** +** Outputs: +** A pointer to a parser. This pointer is used in subsequent calls +** to sqlite3Fts5Parser and sqlite3Fts5ParserFree. +*/ +static void *sqlite3Fts5ParserAlloc(void *(*mallocProc)(fts5YYMALLOCARGTYPE)){ + fts5yyParser *pParser; + pParser = (fts5yyParser*)(*mallocProc)( (fts5YYMALLOCARGTYPE)sizeof(fts5yyParser) ); + if( pParser ){ + pParser->fts5yyidx = -1; +#ifdef fts5YYTRACKMAXSTACKDEPTH + pParser->fts5yyidxMax = 0; +#endif +#if fts5YYSTACKDEPTH<=0 + pParser->fts5yystack = NULL; + pParser->fts5yystksz = 0; + fts5yyGrowStack(pParser); +#endif + } + return pParser; +} + +/* The following function deletes the "minor type" or semantic value +** associated with a symbol. The symbol can be either a terminal +** or nonterminal. "fts5yymajor" is the symbol code, and "fts5yypminor" is +** a pointer to the value to be deleted. The code used to do the +** deletions is derived from the %destructor and/or %token_destructor +** directives of the input grammar. +*/ +static void fts5yy_destructor( + fts5yyParser *fts5yypParser, /* The parser */ + fts5YYCODETYPE fts5yymajor, /* Type code for object to destroy */ + fts5YYMINORTYPE *fts5yypminor /* The object to be destroyed */ +){ + sqlite3Fts5ParserARG_FETCH; + switch( fts5yymajor ){ + /* Here is inserted the actions which take place when a + ** terminal or non-terminal is destroyed. This can happen + ** when the symbol is popped from the stack during a + ** reduce or during error processing or when a parser is + ** being destroyed before it is finished parsing. + ** + ** Note: during a reduce, the only symbols destroyed are those + ** which appear on the RHS of the rule, but which are *not* used + ** inside the C code. + */ +/********* Begin destructor definitions ***************************************/ + case 15: /* input */ +{ +#line 83 "fts5parse.y" + (void)pParse; +#line 489 "fts5parse.c" +} + break; + case 16: /* expr */ + case 17: /* cnearset */ + case 18: /* exprlist */ +{ +#line 89 "fts5parse.y" + sqlite3Fts5ParseNodeFree((fts5yypminor->fts5yy18)); +#line 498 "fts5parse.c" +} + break; + case 19: /* nearset */ + case 22: /* nearphrases */ +{ +#line 137 "fts5parse.y" + sqlite3Fts5ParseNearsetFree((fts5yypminor->fts5yy26)); +#line 506 "fts5parse.c" +} + break; + case 20: /* colset */ + case 21: /* colsetlist */ +{ +#line 119 "fts5parse.y" + sqlite3_free((fts5yypminor->fts5yy3)); +#line 514 "fts5parse.c" +} + break; + case 23: /* phrase */ +{ +#line 168 "fts5parse.y" + sqlite3Fts5ParsePhraseFree((fts5yypminor->fts5yy11)); +#line 521 "fts5parse.c" +} + break; +/********* End destructor definitions *****************************************/ + default: break; /* If no destructor action specified: do nothing */ + } +} + +/* +** Pop the parser's stack once. +** +** If there is a destructor routine associated with the token which +** is popped from the stack, then call it. +*/ +static void fts5yy_pop_parser_stack(fts5yyParser *pParser){ + fts5yyStackEntry *fts5yytos; + assert( pParser->fts5yyidx>=0 ); + fts5yytos = &pParser->fts5yystack[pParser->fts5yyidx--]; +#ifndef NDEBUG + if( fts5yyTraceFILE ){ + fprintf(fts5yyTraceFILE,"%sPopping %s\n", + fts5yyTracePrompt, + fts5yyTokenName[fts5yytos->major]); + } +#endif + fts5yy_destructor(pParser, fts5yytos->major, &fts5yytos->minor); +} + +/* +** Deallocate and destroy a parser. Destructors are called for +** all stack elements before shutting the parser down. +** +** If the fts5YYPARSEFREENEVERNULL macro exists (for example because it +** is defined in a %include section of the input grammar) then it is +** assumed that the input pointer is never NULL. +*/ +static void sqlite3Fts5ParserFree( + void *p, /* The parser to be deleted */ + void (*freeProc)(void*) /* Function used to reclaim memory */ +){ + fts5yyParser *pParser = (fts5yyParser*)p; +#ifndef fts5YYPARSEFREENEVERNULL + if( pParser==0 ) return; +#endif + while( pParser->fts5yyidx>=0 ) fts5yy_pop_parser_stack(pParser); +#if fts5YYSTACKDEPTH<=0 + free(pParser->fts5yystack); +#endif + (*freeProc)((void*)pParser); +} + +/* +** Return the peak depth of the stack for a parser. +*/ +#ifdef fts5YYTRACKMAXSTACKDEPTH +static int sqlite3Fts5ParserStackPeak(void *p){ + fts5yyParser *pParser = (fts5yyParser*)p; + return pParser->fts5yyidxMax; +} +#endif + +/* +** Find the appropriate action for a parser given the terminal +** look-ahead token iLookAhead. +*/ +static int fts5yy_find_shift_action( + fts5yyParser *pParser, /* The parser */ + fts5YYCODETYPE iLookAhead /* The look-ahead token */ +){ + int i; + int stateno = pParser->fts5yystack[pParser->fts5yyidx].stateno; + + if( stateno>=fts5YY_MIN_REDUCE ) return stateno; + assert( stateno <= fts5YY_SHIFT_COUNT ); + do{ + i = fts5yy_shift_ofst[stateno]; + if( i==fts5YY_SHIFT_USE_DFLT ) return fts5yy_default[stateno]; + assert( iLookAhead!=fts5YYNOCODE ); + i += iLookAhead; + if( i<0 || i>=fts5YY_ACTTAB_COUNT || fts5yy_lookahead[i]!=iLookAhead ){ + if( iLookAhead>0 ){ +#ifdef fts5YYFALLBACK + fts5YYCODETYPE iFallback; /* Fallback token */ + if( iLookAhead<sizeof(fts5yyFallback)/sizeof(fts5yyFallback[0]) + && (iFallback = fts5yyFallback[iLookAhead])!=0 ){ +#ifndef NDEBUG + if( fts5yyTraceFILE ){ + fprintf(fts5yyTraceFILE, "%sFALLBACK %s => %s\n", + fts5yyTracePrompt, fts5yyTokenName[iLookAhead], fts5yyTokenName[iFallback]); + } +#endif + assert( fts5yyFallback[iFallback]==0 ); /* Fallback loop must terminate */ + iLookAhead = iFallback; + continue; + } +#endif +#ifdef fts5YYWILDCARD + { + int j = i - iLookAhead + fts5YYWILDCARD; + if( +#if fts5YY_SHIFT_MIN+fts5YYWILDCARD<0 + j>=0 && +#endif +#if fts5YY_SHIFT_MAX+fts5YYWILDCARD>=fts5YY_ACTTAB_COUNT + j<fts5YY_ACTTAB_COUNT && +#endif + fts5yy_lookahead[j]==fts5YYWILDCARD + ){ +#ifndef NDEBUG + if( fts5yyTraceFILE ){ + fprintf(fts5yyTraceFILE, "%sWILDCARD %s => %s\n", + fts5yyTracePrompt, fts5yyTokenName[iLookAhead], + fts5yyTokenName[fts5YYWILDCARD]); + } +#endif /* NDEBUG */ + return fts5yy_action[j]; + } + } +#endif /* fts5YYWILDCARD */ + } + return fts5yy_default[stateno]; + }else{ + return fts5yy_action[i]; + } + }while(1); +} + +/* +** Find the appropriate action for a parser given the non-terminal +** look-ahead token iLookAhead. +*/ +static int fts5yy_find_reduce_action( + int stateno, /* Current state number */ + fts5YYCODETYPE iLookAhead /* The look-ahead token */ +){ + int i; +#ifdef fts5YYERRORSYMBOL + if( stateno>fts5YY_REDUCE_COUNT ){ + return fts5yy_default[stateno]; + } +#else + assert( stateno<=fts5YY_REDUCE_COUNT ); +#endif + i = fts5yy_reduce_ofst[stateno]; + assert( i!=fts5YY_REDUCE_USE_DFLT ); + assert( iLookAhead!=fts5YYNOCODE ); + i += iLookAhead; +#ifdef fts5YYERRORSYMBOL + if( i<0 || i>=fts5YY_ACTTAB_COUNT || fts5yy_lookahead[i]!=iLookAhead ){ + return fts5yy_default[stateno]; + } +#else + assert( i>=0 && i<fts5YY_ACTTAB_COUNT ); + assert( fts5yy_lookahead[i]==iLookAhead ); +#endif + return fts5yy_action[i]; +} + +/* +** The following routine is called if the stack overflows. +*/ +static void fts5yyStackOverflow(fts5yyParser *fts5yypParser){ + sqlite3Fts5ParserARG_FETCH; + fts5yypParser->fts5yyidx--; +#ifndef NDEBUG + if( fts5yyTraceFILE ){ + fprintf(fts5yyTraceFILE,"%sStack Overflow!\n",fts5yyTracePrompt); + } +#endif + while( fts5yypParser->fts5yyidx>=0 ) fts5yy_pop_parser_stack(fts5yypParser); + /* Here code is inserted which will execute if the parser + ** stack every overflows */ +/******** Begin %stack_overflow code ******************************************/ +#line 36 "fts5parse.y" + + sqlite3Fts5ParseError(pParse, "fts5: parser stack overflow"); +#line 697 "fts5parse.c" +/******** End %stack_overflow code ********************************************/ + sqlite3Fts5ParserARG_STORE; /* Suppress warning about unused %extra_argument var */ +} + +/* +** Print tracing information for a SHIFT action +*/ +#ifndef NDEBUG +static void fts5yyTraceShift(fts5yyParser *fts5yypParser, int fts5yyNewState){ + if( fts5yyTraceFILE ){ + if( fts5yyNewState<fts5YYNSTATE ){ + fprintf(fts5yyTraceFILE,"%sShift '%s', go to state %d\n", + fts5yyTracePrompt,fts5yyTokenName[fts5yypParser->fts5yystack[fts5yypParser->fts5yyidx].major], + fts5yyNewState); + }else{ + fprintf(fts5yyTraceFILE,"%sShift '%s'\n", + fts5yyTracePrompt,fts5yyTokenName[fts5yypParser->fts5yystack[fts5yypParser->fts5yyidx].major]); + } + } +} +#else +# define fts5yyTraceShift(X,Y) +#endif + +/* +** Perform a shift action. +*/ +static void fts5yy_shift( + fts5yyParser *fts5yypParser, /* The parser to be shifted */ + int fts5yyNewState, /* The new state to shift in */ + int fts5yyMajor, /* The major token to shift in */ + sqlite3Fts5ParserFTS5TOKENTYPE fts5yyMinor /* The minor token to shift in */ +){ + fts5yyStackEntry *fts5yytos; + fts5yypParser->fts5yyidx++; +#ifdef fts5YYTRACKMAXSTACKDEPTH + if( fts5yypParser->fts5yyidx>fts5yypParser->fts5yyidxMax ){ + fts5yypParser->fts5yyidxMax = fts5yypParser->fts5yyidx; + } +#endif +#if fts5YYSTACKDEPTH>0 + if( fts5yypParser->fts5yyidx>=fts5YYSTACKDEPTH ){ + fts5yyStackOverflow(fts5yypParser); + return; + } +#else + if( fts5yypParser->fts5yyidx>=fts5yypParser->fts5yystksz ){ + fts5yyGrowStack(fts5yypParser); + if( fts5yypParser->fts5yyidx>=fts5yypParser->fts5yystksz ){ + fts5yyStackOverflow(fts5yypParser); + return; + } + } +#endif + fts5yytos = &fts5yypParser->fts5yystack[fts5yypParser->fts5yyidx]; + fts5yytos->stateno = (fts5YYACTIONTYPE)fts5yyNewState; + fts5yytos->major = (fts5YYCODETYPE)fts5yyMajor; + fts5yytos->minor.fts5yy0 = fts5yyMinor; + fts5yyTraceShift(fts5yypParser, fts5yyNewState); +} + +/* The following table contains information about every rule that +** is used during the reduce. +*/ +static const struct { + fts5YYCODETYPE lhs; /* Symbol on the left-hand side of the rule */ + unsigned char nrhs; /* Number of right-hand side symbols in the rule */ +} fts5yyRuleInfo[] = { + { 15, 1 }, + { 16, 3 }, + { 16, 3 }, + { 16, 3 }, + { 16, 3 }, + { 16, 1 }, + { 18, 1 }, + { 18, 2 }, + { 17, 1 }, + { 17, 3 }, + { 20, 3 }, + { 20, 1 }, + { 21, 2 }, + { 21, 1 }, + { 19, 1 }, + { 19, 5 }, + { 22, 1 }, + { 22, 2 }, + { 24, 0 }, + { 24, 2 }, + { 23, 4 }, + { 23, 2 }, + { 25, 1 }, + { 25, 0 }, +}; + +static void fts5yy_accept(fts5yyParser*); /* Forward Declaration */ + +/* +** Perform a reduce action and the shift that must immediately +** follow the reduce. +*/ +static void fts5yy_reduce( + fts5yyParser *fts5yypParser, /* The parser */ + int fts5yyruleno /* Number of the rule by which to reduce */ +){ + int fts5yygoto; /* The next state */ + int fts5yyact; /* The next action */ + fts5yyStackEntry *fts5yymsp; /* The top of the parser's stack */ + int fts5yysize; /* Amount to pop the stack */ + sqlite3Fts5ParserARG_FETCH; + fts5yymsp = &fts5yypParser->fts5yystack[fts5yypParser->fts5yyidx]; +#ifndef NDEBUG + if( fts5yyTraceFILE && fts5yyruleno>=0 + && fts5yyruleno<(int)(sizeof(fts5yyRuleName)/sizeof(fts5yyRuleName[0])) ){ + fts5yysize = fts5yyRuleInfo[fts5yyruleno].nrhs; + fprintf(fts5yyTraceFILE, "%sReduce [%s], go to state %d.\n", fts5yyTracePrompt, + fts5yyRuleName[fts5yyruleno], fts5yymsp[-fts5yysize].stateno); + } +#endif /* NDEBUG */ + + /* Check that the stack is large enough to grow by a single entry + ** if the RHS of the rule is empty. This ensures that there is room + ** enough on the stack to push the LHS value */ + if( fts5yyRuleInfo[fts5yyruleno].nrhs==0 ){ +#ifdef fts5YYTRACKMAXSTACKDEPTH + if( fts5yypParser->fts5yyidx>fts5yypParser->fts5yyidxMax ){ + fts5yypParser->fts5yyidxMax = fts5yypParser->fts5yyidx; + } +#endif +#if fts5YYSTACKDEPTH>0 + if( fts5yypParser->fts5yyidx>=fts5YYSTACKDEPTH-1 ){ + fts5yyStackOverflow(fts5yypParser); + return; + } +#else + if( fts5yypParser->fts5yyidx>=fts5yypParser->fts5yystksz-1 ){ + fts5yyGrowStack(fts5yypParser); + if( fts5yypParser->fts5yyidx>=fts5yypParser->fts5yystksz-1 ){ + fts5yyStackOverflow(fts5yypParser); + return; + } + } +#endif + } + + switch( fts5yyruleno ){ + /* Beginning here are the reduction cases. A typical example + ** follows: + ** case 0: + ** #line <lineno> <grammarfile> + ** { ... } // User supplied code + ** #line <lineno> <thisfile> + ** break; + */ +/********** Begin reduce actions **********************************************/ + fts5YYMINORTYPE fts5yylhsminor; + case 0: /* input ::= expr */ +#line 82 "fts5parse.y" +{ sqlite3Fts5ParseFinished(pParse, fts5yymsp[0].minor.fts5yy18); } +#line 856 "fts5parse.c" + break; + case 1: /* expr ::= expr AND expr */ +#line 92 "fts5parse.y" +{ + fts5yylhsminor.fts5yy18 = sqlite3Fts5ParseNode(pParse, FTS5_AND, fts5yymsp[-2].minor.fts5yy18, fts5yymsp[0].minor.fts5yy18, 0); +} +#line 863 "fts5parse.c" + fts5yymsp[-2].minor.fts5yy18 = fts5yylhsminor.fts5yy18; + break; + case 2: /* expr ::= expr OR expr */ +#line 95 "fts5parse.y" +{ + fts5yylhsminor.fts5yy18 = sqlite3Fts5ParseNode(pParse, FTS5_OR, fts5yymsp[-2].minor.fts5yy18, fts5yymsp[0].minor.fts5yy18, 0); +} +#line 871 "fts5parse.c" + fts5yymsp[-2].minor.fts5yy18 = fts5yylhsminor.fts5yy18; + break; + case 3: /* expr ::= expr NOT expr */ +#line 98 "fts5parse.y" +{ + fts5yylhsminor.fts5yy18 = sqlite3Fts5ParseNode(pParse, FTS5_NOT, fts5yymsp[-2].minor.fts5yy18, fts5yymsp[0].minor.fts5yy18, 0); +} +#line 879 "fts5parse.c" + fts5yymsp[-2].minor.fts5yy18 = fts5yylhsminor.fts5yy18; + break; + case 4: /* expr ::= LP expr RP */ +#line 102 "fts5parse.y" +{fts5yymsp[-2].minor.fts5yy18 = fts5yymsp[-1].minor.fts5yy18;} +#line 885 "fts5parse.c" + break; + case 5: /* expr ::= exprlist */ + case 6: /* exprlist ::= cnearset */ fts5yytestcase(fts5yyruleno==6); +#line 103 "fts5parse.y" +{fts5yylhsminor.fts5yy18 = fts5yymsp[0].minor.fts5yy18;} +#line 891 "fts5parse.c" + fts5yymsp[0].minor.fts5yy18 = fts5yylhsminor.fts5yy18; + break; + case 7: /* exprlist ::= exprlist cnearset */ +#line 106 "fts5parse.y" +{ + fts5yylhsminor.fts5yy18 = sqlite3Fts5ParseNode(pParse, FTS5_AND, fts5yymsp[-1].minor.fts5yy18, fts5yymsp[0].minor.fts5yy18, 0); +} +#line 899 "fts5parse.c" + fts5yymsp[-1].minor.fts5yy18 = fts5yylhsminor.fts5yy18; + break; + case 8: /* cnearset ::= nearset */ +#line 110 "fts5parse.y" +{ + fts5yylhsminor.fts5yy18 = sqlite3Fts5ParseNode(pParse, FTS5_STRING, 0, 0, fts5yymsp[0].minor.fts5yy26); +} +#line 907 "fts5parse.c" + fts5yymsp[0].minor.fts5yy18 = fts5yylhsminor.fts5yy18; + break; + case 9: /* cnearset ::= colset COLON nearset */ +#line 113 "fts5parse.y" +{ + sqlite3Fts5ParseSetColset(pParse, fts5yymsp[0].minor.fts5yy26, fts5yymsp[-2].minor.fts5yy3); + fts5yylhsminor.fts5yy18 = sqlite3Fts5ParseNode(pParse, FTS5_STRING, 0, 0, fts5yymsp[0].minor.fts5yy26); +} +#line 916 "fts5parse.c" + fts5yymsp[-2].minor.fts5yy18 = fts5yylhsminor.fts5yy18; + break; + case 10: /* colset ::= LCP colsetlist RCP */ +#line 123 "fts5parse.y" +{ fts5yymsp[-2].minor.fts5yy3 = fts5yymsp[-1].minor.fts5yy3; } +#line 922 "fts5parse.c" + break; + case 11: /* colset ::= STRING */ +#line 124 "fts5parse.y" +{ + fts5yylhsminor.fts5yy3 = sqlite3Fts5ParseColset(pParse, 0, &fts5yymsp[0].minor.fts5yy0); +} +#line 929 "fts5parse.c" + fts5yymsp[0].minor.fts5yy3 = fts5yylhsminor.fts5yy3; + break; + case 12: /* colsetlist ::= colsetlist STRING */ +#line 128 "fts5parse.y" +{ + fts5yylhsminor.fts5yy3 = sqlite3Fts5ParseColset(pParse, fts5yymsp[-1].minor.fts5yy3, &fts5yymsp[0].minor.fts5yy0); } +#line 936 "fts5parse.c" + fts5yymsp[-1].minor.fts5yy3 = fts5yylhsminor.fts5yy3; + break; + case 13: /* colsetlist ::= STRING */ +#line 130 "fts5parse.y" +{ + fts5yylhsminor.fts5yy3 = sqlite3Fts5ParseColset(pParse, 0, &fts5yymsp[0].minor.fts5yy0); +} +#line 944 "fts5parse.c" + fts5yymsp[0].minor.fts5yy3 = fts5yylhsminor.fts5yy3; + break; + case 14: /* nearset ::= phrase */ +#line 140 "fts5parse.y" +{ fts5yylhsminor.fts5yy26 = sqlite3Fts5ParseNearset(pParse, 0, fts5yymsp[0].minor.fts5yy11); } +#line 950 "fts5parse.c" + fts5yymsp[0].minor.fts5yy26 = fts5yylhsminor.fts5yy26; + break; + case 15: /* nearset ::= STRING LP nearphrases neardist_opt RP */ +#line 141 "fts5parse.y" +{ + sqlite3Fts5ParseNear(pParse, &fts5yymsp[-4].minor.fts5yy0); + sqlite3Fts5ParseSetDistance(pParse, fts5yymsp[-2].minor.fts5yy26, &fts5yymsp[-1].minor.fts5yy0); + fts5yylhsminor.fts5yy26 = fts5yymsp[-2].minor.fts5yy26; +} +#line 960 "fts5parse.c" + fts5yymsp[-4].minor.fts5yy26 = fts5yylhsminor.fts5yy26; + break; + case 16: /* nearphrases ::= phrase */ +#line 147 "fts5parse.y" +{ + fts5yylhsminor.fts5yy26 = sqlite3Fts5ParseNearset(pParse, 0, fts5yymsp[0].minor.fts5yy11); +} +#line 968 "fts5parse.c" + fts5yymsp[0].minor.fts5yy26 = fts5yylhsminor.fts5yy26; + break; + case 17: /* nearphrases ::= nearphrases phrase */ +#line 150 "fts5parse.y" +{ + fts5yylhsminor.fts5yy26 = sqlite3Fts5ParseNearset(pParse, fts5yymsp[-1].minor.fts5yy26, fts5yymsp[0].minor.fts5yy11); +} +#line 976 "fts5parse.c" + fts5yymsp[-1].minor.fts5yy26 = fts5yylhsminor.fts5yy26; + break; + case 18: /* neardist_opt ::= */ +#line 157 "fts5parse.y" +{ fts5yymsp[1].minor.fts5yy0.p = 0; fts5yymsp[1].minor.fts5yy0.n = 0; } +#line 982 "fts5parse.c" + break; + case 19: /* neardist_opt ::= COMMA STRING */ +#line 158 "fts5parse.y" +{ fts5yymsp[-1].minor.fts5yy0 = fts5yymsp[0].minor.fts5yy0; } +#line 987 "fts5parse.c" + break; + case 20: /* phrase ::= phrase PLUS STRING star_opt */ +#line 170 "fts5parse.y" +{ + fts5yylhsminor.fts5yy11 = sqlite3Fts5ParseTerm(pParse, fts5yymsp[-3].minor.fts5yy11, &fts5yymsp[-1].minor.fts5yy0, fts5yymsp[0].minor.fts5yy20); +} +#line 994 "fts5parse.c" + fts5yymsp[-3].minor.fts5yy11 = fts5yylhsminor.fts5yy11; + break; + case 21: /* phrase ::= STRING star_opt */ +#line 173 "fts5parse.y" +{ + fts5yylhsminor.fts5yy11 = sqlite3Fts5ParseTerm(pParse, 0, &fts5yymsp[-1].minor.fts5yy0, fts5yymsp[0].minor.fts5yy20); +} +#line 1002 "fts5parse.c" + fts5yymsp[-1].minor.fts5yy11 = fts5yylhsminor.fts5yy11; + break; + case 22: /* star_opt ::= STAR */ +#line 182 "fts5parse.y" +{ fts5yymsp[0].minor.fts5yy20 = 1; } +#line 1008 "fts5parse.c" + break; + case 23: /* star_opt ::= */ +#line 183 "fts5parse.y" +{ fts5yymsp[1].minor.fts5yy20 = 0; } +#line 1013 "fts5parse.c" + break; + default: + break; +/********** End reduce actions ************************************************/ + }; + assert( fts5yyruleno>=0 && fts5yyruleno<sizeof(fts5yyRuleInfo)/sizeof(fts5yyRuleInfo[0]) ); + fts5yygoto = fts5yyRuleInfo[fts5yyruleno].lhs; + fts5yysize = fts5yyRuleInfo[fts5yyruleno].nrhs; + fts5yyact = fts5yy_find_reduce_action(fts5yymsp[-fts5yysize].stateno,(fts5YYCODETYPE)fts5yygoto); + if( fts5yyact <= fts5YY_MAX_SHIFTREDUCE ){ + if( fts5yyact>fts5YY_MAX_SHIFT ) fts5yyact += fts5YY_MIN_REDUCE - fts5YY_MIN_SHIFTREDUCE; + fts5yypParser->fts5yyidx -= fts5yysize - 1; + fts5yymsp -= fts5yysize-1; + fts5yymsp->stateno = (fts5YYACTIONTYPE)fts5yyact; + fts5yymsp->major = (fts5YYCODETYPE)fts5yygoto; + fts5yyTraceShift(fts5yypParser, fts5yyact); + }else{ + assert( fts5yyact == fts5YY_ACCEPT_ACTION ); + fts5yypParser->fts5yyidx -= fts5yysize; + fts5yy_accept(fts5yypParser); + } +} + +/* +** The following code executes when the parse fails +*/ +#ifndef fts5YYNOERRORRECOVERY +static void fts5yy_parse_failed( + fts5yyParser *fts5yypParser /* The parser */ +){ + sqlite3Fts5ParserARG_FETCH; +#ifndef NDEBUG + if( fts5yyTraceFILE ){ + fprintf(fts5yyTraceFILE,"%sFail!\n",fts5yyTracePrompt); + } +#endif + while( fts5yypParser->fts5yyidx>=0 ) fts5yy_pop_parser_stack(fts5yypParser); + /* Here code is inserted which will be executed whenever the + ** parser fails */ +/************ Begin %parse_failure code ***************************************/ +/************ End %parse_failure code *****************************************/ + sqlite3Fts5ParserARG_STORE; /* Suppress warning about unused %extra_argument variable */ +} +#endif /* fts5YYNOERRORRECOVERY */ + +/* +** The following code executes when a syntax error first occurs. +*/ +static void fts5yy_syntax_error( + fts5yyParser *fts5yypParser, /* The parser */ + int fts5yymajor, /* The major type of the error token */ + sqlite3Fts5ParserFTS5TOKENTYPE fts5yyminor /* The minor type of the error token */ +){ + sqlite3Fts5ParserARG_FETCH; +#define FTS5TOKEN fts5yyminor +/************ Begin %syntax_error code ****************************************/ +#line 30 "fts5parse.y" + + UNUSED_PARAM(fts5yymajor); /* Silence a compiler warning */ + sqlite3Fts5ParseError( + pParse, "fts5: syntax error near \"%.*s\"",FTS5TOKEN.n,FTS5TOKEN.p + ); +#line 1076 "fts5parse.c" +/************ End %syntax_error code ******************************************/ + sqlite3Fts5ParserARG_STORE; /* Suppress warning about unused %extra_argument variable */ +} + +/* +** The following is executed when the parser accepts +*/ +static void fts5yy_accept( + fts5yyParser *fts5yypParser /* The parser */ +){ + sqlite3Fts5ParserARG_FETCH; +#ifndef NDEBUG + if( fts5yyTraceFILE ){ + fprintf(fts5yyTraceFILE,"%sAccept!\n",fts5yyTracePrompt); + } +#endif + while( fts5yypParser->fts5yyidx>=0 ) fts5yy_pop_parser_stack(fts5yypParser); + /* Here code is inserted which will be executed whenever the + ** parser accepts */ +/*********** Begin %parse_accept code *****************************************/ +/*********** End %parse_accept code *******************************************/ + sqlite3Fts5ParserARG_STORE; /* Suppress warning about unused %extra_argument variable */ +} + +/* The main parser program. +** The first argument is a pointer to a structure obtained from +** "sqlite3Fts5ParserAlloc" which describes the current state of the parser. +** The second argument is the major token number. The third is +** the minor token. The fourth optional argument is whatever the +** user wants (and specified in the grammar) and is available for +** use by the action routines. +** +** Inputs: +** <ul> +** <li> A pointer to the parser (an opaque structure.) +** <li> The major token number. +** <li> The minor token number. +** <li> An option argument of a grammar-specified type. +** </ul> +** +** Outputs: +** None. +*/ +static void sqlite3Fts5Parser( + void *fts5yyp, /* The parser */ + int fts5yymajor, /* The major token code number */ + sqlite3Fts5ParserFTS5TOKENTYPE fts5yyminor /* The value for the token */ + sqlite3Fts5ParserARG_PDECL /* Optional %extra_argument parameter */ +){ + fts5YYMINORTYPE fts5yyminorunion; + int fts5yyact; /* The parser action. */ +#if !defined(fts5YYERRORSYMBOL) && !defined(fts5YYNOERRORRECOVERY) + int fts5yyendofinput; /* True if we are at the end of input */ +#endif +#ifdef fts5YYERRORSYMBOL + int fts5yyerrorhit = 0; /* True if fts5yymajor has invoked an error */ +#endif + fts5yyParser *fts5yypParser; /* The parser */ + + /* (re)initialize the parser, if necessary */ + fts5yypParser = (fts5yyParser*)fts5yyp; + if( fts5yypParser->fts5yyidx<0 ){ +#if fts5YYSTACKDEPTH<=0 + if( fts5yypParser->fts5yystksz <=0 ){ + fts5yyStackOverflow(fts5yypParser); + return; + } +#endif + fts5yypParser->fts5yyidx = 0; +#ifndef fts5YYNOERRORRECOVERY + fts5yypParser->fts5yyerrcnt = -1; +#endif + fts5yypParser->fts5yystack[0].stateno = 0; + fts5yypParser->fts5yystack[0].major = 0; +#ifndef NDEBUG + if( fts5yyTraceFILE ){ + fprintf(fts5yyTraceFILE,"%sInitialize. Empty stack. State 0\n", + fts5yyTracePrompt); + } +#endif + } +#if !defined(fts5YYERRORSYMBOL) && !defined(fts5YYNOERRORRECOVERY) + fts5yyendofinput = (fts5yymajor==0); +#endif + sqlite3Fts5ParserARG_STORE; + +#ifndef NDEBUG + if( fts5yyTraceFILE ){ + fprintf(fts5yyTraceFILE,"%sInput '%s'\n",fts5yyTracePrompt,fts5yyTokenName[fts5yymajor]); + } +#endif + + do{ + fts5yyact = fts5yy_find_shift_action(fts5yypParser,(fts5YYCODETYPE)fts5yymajor); + if( fts5yyact <= fts5YY_MAX_SHIFTREDUCE ){ + if( fts5yyact > fts5YY_MAX_SHIFT ) fts5yyact += fts5YY_MIN_REDUCE - fts5YY_MIN_SHIFTREDUCE; + fts5yy_shift(fts5yypParser,fts5yyact,fts5yymajor,fts5yyminor); +#ifndef fts5YYNOERRORRECOVERY + fts5yypParser->fts5yyerrcnt--; +#endif + fts5yymajor = fts5YYNOCODE; + }else if( fts5yyact <= fts5YY_MAX_REDUCE ){ + fts5yy_reduce(fts5yypParser,fts5yyact-fts5YY_MIN_REDUCE); + }else{ + assert( fts5yyact == fts5YY_ERROR_ACTION ); + fts5yyminorunion.fts5yy0 = fts5yyminor; +#ifdef fts5YYERRORSYMBOL + int fts5yymx; +#endif +#ifndef NDEBUG + if( fts5yyTraceFILE ){ + fprintf(fts5yyTraceFILE,"%sSyntax Error!\n",fts5yyTracePrompt); + } +#endif +#ifdef fts5YYERRORSYMBOL + /* A syntax error has occurred. + ** The response to an error depends upon whether or not the + ** grammar defines an error token "ERROR". + ** + ** This is what we do if the grammar does define ERROR: + ** + ** * Call the %syntax_error function. + ** + ** * Begin popping the stack until we enter a state where + ** it is legal to shift the error symbol, then shift + ** the error symbol. + ** + ** * Set the error count to three. + ** + ** * Begin accepting and shifting new tokens. No new error + ** processing will occur until three tokens have been + ** shifted successfully. + ** + */ + if( fts5yypParser->fts5yyerrcnt<0 ){ + fts5yy_syntax_error(fts5yypParser,fts5yymajor,fts5yyminor); + } + fts5yymx = fts5yypParser->fts5yystack[fts5yypParser->fts5yyidx].major; + if( fts5yymx==fts5YYERRORSYMBOL || fts5yyerrorhit ){ +#ifndef NDEBUG + if( fts5yyTraceFILE ){ + fprintf(fts5yyTraceFILE,"%sDiscard input token %s\n", + fts5yyTracePrompt,fts5yyTokenName[fts5yymajor]); + } +#endif + fts5yy_destructor(fts5yypParser, (fts5YYCODETYPE)fts5yymajor, &fts5yyminorunion); + fts5yymajor = fts5YYNOCODE; + }else{ + while( + fts5yypParser->fts5yyidx >= 0 && + fts5yymx != fts5YYERRORSYMBOL && + (fts5yyact = fts5yy_find_reduce_action( + fts5yypParser->fts5yystack[fts5yypParser->fts5yyidx].stateno, + fts5YYERRORSYMBOL)) >= fts5YY_MIN_REDUCE + ){ + fts5yy_pop_parser_stack(fts5yypParser); + } + if( fts5yypParser->fts5yyidx < 0 || fts5yymajor==0 ){ + fts5yy_destructor(fts5yypParser,(fts5YYCODETYPE)fts5yymajor,&fts5yyminorunion); + fts5yy_parse_failed(fts5yypParser); + fts5yymajor = fts5YYNOCODE; + }else if( fts5yymx!=fts5YYERRORSYMBOL ){ + fts5yy_shift(fts5yypParser,fts5yyact,fts5YYERRORSYMBOL,fts5yyminor); + } + } + fts5yypParser->fts5yyerrcnt = 3; + fts5yyerrorhit = 1; +#elif defined(fts5YYNOERRORRECOVERY) + /* If the fts5YYNOERRORRECOVERY macro is defined, then do not attempt to + ** do any kind of error recovery. Instead, simply invoke the syntax + ** error routine and continue going as if nothing had happened. + ** + ** Applications can set this macro (for example inside %include) if + ** they intend to abandon the parse upon the first syntax error seen. + */ + fts5yy_syntax_error(fts5yypParser,fts5yymajor, fts5yyminor); + fts5yy_destructor(fts5yypParser,(fts5YYCODETYPE)fts5yymajor,&fts5yyminorunion); + fts5yymajor = fts5YYNOCODE; + +#else /* fts5YYERRORSYMBOL is not defined */ + /* This is what we do if the grammar does not define ERROR: + ** + ** * Report an error message, and throw away the input token. + ** + ** * If the input token is $, then fail the parse. + ** + ** As before, subsequent error messages are suppressed until + ** three input tokens have been successfully shifted. + */ + if( fts5yypParser->fts5yyerrcnt<=0 ){ + fts5yy_syntax_error(fts5yypParser,fts5yymajor, fts5yyminor); + } + fts5yypParser->fts5yyerrcnt = 3; + fts5yy_destructor(fts5yypParser,(fts5YYCODETYPE)fts5yymajor,&fts5yyminorunion); + if( fts5yyendofinput ){ + fts5yy_parse_failed(fts5yypParser); + } + fts5yymajor = fts5YYNOCODE; +#endif + } + }while( fts5yymajor!=fts5YYNOCODE && fts5yypParser->fts5yyidx>=0 ); +#ifndef NDEBUG + if( fts5yyTraceFILE ){ + int i; + fprintf(fts5yyTraceFILE,"%sReturn. Stack=",fts5yyTracePrompt); + for(i=1; i<=fts5yypParser->fts5yyidx; i++) + fprintf(fts5yyTraceFILE,"%c%s", i==1 ? '[' : ' ', + fts5yyTokenName[fts5yypParser->fts5yystack[i].major]); + fprintf(fts5yyTraceFILE,"]\n"); + } +#endif + return; +} + +#line 1 "fts5_aux.c" +/* +** 2014 May 31 +** +** The author disclaims copyright to this source code. In place of +** a legal notice, here is a blessing: +** +** May you do good and not evil. +** May you find forgiveness for yourself and forgive others. +** May you share freely, never taking more than you give. +** +****************************************************************************** +*/ + + +/* #include "fts5Int.h" */ +#include <math.h> /* amalgamator: keep */ + +/* +** Object used to iterate through all "coalesced phrase instances" in +** a single column of the current row. If the phrase instances in the +** column being considered do not overlap, this object simply iterates +** through them. Or, if they do overlap (share one or more tokens in +** common), each set of overlapping instances is treated as a single +** match. See documentation for the highlight() auxiliary function for +** details. +** +** Usage is: +** +** for(rc = fts5CInstIterNext(pApi, pFts, iCol, &iter); +** (rc==SQLITE_OK && 0==fts5CInstIterEof(&iter); +** rc = fts5CInstIterNext(&iter) +** ){ +** printf("instance starts at %d, ends at %d\n", iter.iStart, iter.iEnd); +** } +** +*/ +typedef struct CInstIter CInstIter; +struct CInstIter { + const Fts5ExtensionApi *pApi; /* API offered by current FTS version */ + Fts5Context *pFts; /* First arg to pass to pApi functions */ + int iCol; /* Column to search */ + int iInst; /* Next phrase instance index */ + int nInst; /* Total number of phrase instances */ + + /* Output variables */ + int iStart; /* First token in coalesced phrase instance */ + int iEnd; /* Last token in coalesced phrase instance */ +}; + +/* +** Advance the iterator to the next coalesced phrase instance. Return +** an SQLite error code if an error occurs, or SQLITE_OK otherwise. +*/ +static int fts5CInstIterNext(CInstIter *pIter){ + int rc = SQLITE_OK; + pIter->iStart = -1; + pIter->iEnd = -1; + + while( rc==SQLITE_OK && pIter->iInst<pIter->nInst ){ + int ip; int ic; int io; + rc = pIter->pApi->xInst(pIter->pFts, pIter->iInst, &ip, &ic, &io); + if( rc==SQLITE_OK ){ + if( ic==pIter->iCol ){ + int iEnd = io - 1 + pIter->pApi->xPhraseSize(pIter->pFts, ip); + if( pIter->iStart<0 ){ + pIter->iStart = io; + pIter->iEnd = iEnd; + }else if( io<=pIter->iEnd ){ + if( iEnd>pIter->iEnd ) pIter->iEnd = iEnd; + }else{ + break; + } + } + pIter->iInst++; + } + } + + return rc; +} + +/* +** Initialize the iterator object indicated by the final parameter to +** iterate through coalesced phrase instances in column iCol. +*/ +static int fts5CInstIterInit( + const Fts5ExtensionApi *pApi, + Fts5Context *pFts, + int iCol, + CInstIter *pIter +){ + int rc; + + memset(pIter, 0, sizeof(CInstIter)); + pIter->pApi = pApi; + pIter->pFts = pFts; + pIter->iCol = iCol; + rc = pApi->xInstCount(pFts, &pIter->nInst); + + if( rc==SQLITE_OK ){ + rc = fts5CInstIterNext(pIter); + } + + return rc; +} + + + +/************************************************************************* +** Start of highlight() implementation. +*/ +typedef struct HighlightContext HighlightContext; +struct HighlightContext { + CInstIter iter; /* Coalesced Instance Iterator */ + int iPos; /* Current token offset in zIn[] */ + int iRangeStart; /* First token to include */ + int iRangeEnd; /* If non-zero, last token to include */ + const char *zOpen; /* Opening highlight */ + const char *zClose; /* Closing highlight */ + const char *zIn; /* Input text */ + int nIn; /* Size of input text in bytes */ + int iOff; /* Current offset within zIn[] */ + char *zOut; /* Output value */ +}; + +/* +** Append text to the HighlightContext output string - p->zOut. Argument +** z points to a buffer containing n bytes of text to append. If n is +** negative, everything up until the first '\0' is appended to the output. +** +** If *pRc is set to any value other than SQLITE_OK when this function is +** called, it is a no-op. If an error (i.e. an OOM condition) is encountered, +** *pRc is set to an error code before returning. +*/ +static void fts5HighlightAppend( + int *pRc, + HighlightContext *p, + const char *z, int n +){ + if( *pRc==SQLITE_OK ){ + if( n<0 ) n = (int)strlen(z); + p->zOut = sqlite3_mprintf("%z%.*s", p->zOut, n, z); + if( p->zOut==0 ) *pRc = SQLITE_NOMEM; + } +} + +/* +** Tokenizer callback used by implementation of highlight() function. +*/ +static int fts5HighlightCb( + void *pContext, /* Pointer to HighlightContext object */ + int tflags, /* Mask of FTS5_TOKEN_* flags */ + const char *pToken, /* Buffer containing token */ + int nToken, /* Size of token in bytes */ + int iStartOff, /* Start offset of token */ + int iEndOff /* End offset of token */ +){ + HighlightContext *p = (HighlightContext*)pContext; + int rc = SQLITE_OK; + int iPos; + + UNUSED_PARAM2(pToken, nToken); + + if( tflags & FTS5_TOKEN_COLOCATED ) return SQLITE_OK; + iPos = p->iPos++; + + if( p->iRangeEnd>0 ){ + if( iPos<p->iRangeStart || iPos>p->iRangeEnd ) return SQLITE_OK; + if( p->iRangeStart && iPos==p->iRangeStart ) p->iOff = iStartOff; + } + + if( iPos==p->iter.iStart ){ + fts5HighlightAppend(&rc, p, &p->zIn[p->iOff], iStartOff - p->iOff); + fts5HighlightAppend(&rc, p, p->zOpen, -1); + p->iOff = iStartOff; + } + + if( iPos==p->iter.iEnd ){ + if( p->iRangeEnd && p->iter.iStart<p->iRangeStart ){ + fts5HighlightAppend(&rc, p, p->zOpen, -1); + } + fts5HighlightAppend(&rc, p, &p->zIn[p->iOff], iEndOff - p->iOff); + fts5HighlightAppend(&rc, p, p->zClose, -1); + p->iOff = iEndOff; + if( rc==SQLITE_OK ){ + rc = fts5CInstIterNext(&p->iter); + } + } + + if( p->iRangeEnd>0 && iPos==p->iRangeEnd ){ + fts5HighlightAppend(&rc, p, &p->zIn[p->iOff], iEndOff - p->iOff); + p->iOff = iEndOff; + if( iPos<p->iter.iEnd ){ + fts5HighlightAppend(&rc, p, p->zClose, -1); + } + } + + return rc; +} + +/* +** Implementation of highlight() function. +*/ +static void fts5HighlightFunction( + const Fts5ExtensionApi *pApi, /* API offered by current FTS version */ + Fts5Context *pFts, /* First arg to pass to pApi functions */ + sqlite3_context *pCtx, /* Context for returning result/error */ + int nVal, /* Number of values in apVal[] array */ + sqlite3_value **apVal /* Array of trailing arguments */ +){ + HighlightContext ctx; + int rc; + int iCol; + + if( nVal!=3 ){ + const char *zErr = "wrong number of arguments to function highlight()"; + sqlite3_result_error(pCtx, zErr, -1); + return; + } + + iCol = sqlite3_value_int(apVal[0]); + memset(&ctx, 0, sizeof(HighlightContext)); + ctx.zOpen = (const char*)sqlite3_value_text(apVal[1]); + ctx.zClose = (const char*)sqlite3_value_text(apVal[2]); + rc = pApi->xColumnText(pFts, iCol, &ctx.zIn, &ctx.nIn); + + if( ctx.zIn ){ + if( rc==SQLITE_OK ){ + rc = fts5CInstIterInit(pApi, pFts, iCol, &ctx.iter); + } + + if( rc==SQLITE_OK ){ + rc = pApi->xTokenize(pFts, ctx.zIn, ctx.nIn, (void*)&ctx,fts5HighlightCb); + } + fts5HighlightAppend(&rc, &ctx, &ctx.zIn[ctx.iOff], ctx.nIn - ctx.iOff); + + if( rc==SQLITE_OK ){ + sqlite3_result_text(pCtx, (const char*)ctx.zOut, -1, SQLITE_TRANSIENT); + } + sqlite3_free(ctx.zOut); + } + if( rc!=SQLITE_OK ){ + sqlite3_result_error_code(pCtx, rc); + } +} +/* +** End of highlight() implementation. +**************************************************************************/ + +/* +** Implementation of snippet() function. +*/ +static void fts5SnippetFunction( + const Fts5ExtensionApi *pApi, /* API offered by current FTS version */ + Fts5Context *pFts, /* First arg to pass to pApi functions */ + sqlite3_context *pCtx, /* Context for returning result/error */ + int nVal, /* Number of values in apVal[] array */ + sqlite3_value **apVal /* Array of trailing arguments */ +){ + HighlightContext ctx; + int rc = SQLITE_OK; /* Return code */ + int iCol; /* 1st argument to snippet() */ + const char *zEllips; /* 4th argument to snippet() */ + int nToken; /* 5th argument to snippet() */ + int nInst = 0; /* Number of instance matches this row */ + int i; /* Used to iterate through instances */ + int nPhrase; /* Number of phrases in query */ + unsigned char *aSeen; /* Array of "seen instance" flags */ + int iBestCol; /* Column containing best snippet */ + int iBestStart = 0; /* First token of best snippet */ + int iBestLast; /* Last token of best snippet */ + int nBestScore = 0; /* Score of best snippet */ + int nColSize = 0; /* Total size of iBestCol in tokens */ + + if( nVal!=5 ){ + const char *zErr = "wrong number of arguments to function snippet()"; + sqlite3_result_error(pCtx, zErr, -1); + return; + } + + memset(&ctx, 0, sizeof(HighlightContext)); + iCol = sqlite3_value_int(apVal[0]); + ctx.zOpen = (const char*)sqlite3_value_text(apVal[1]); + ctx.zClose = (const char*)sqlite3_value_text(apVal[2]); + zEllips = (const char*)sqlite3_value_text(apVal[3]); + nToken = sqlite3_value_int(apVal[4]); + iBestLast = nToken-1; + + iBestCol = (iCol>=0 ? iCol : 0); + nPhrase = pApi->xPhraseCount(pFts); + aSeen = sqlite3_malloc(nPhrase); + if( aSeen==0 ){ + rc = SQLITE_NOMEM; + } + + if( rc==SQLITE_OK ){ + rc = pApi->xInstCount(pFts, &nInst); + } + for(i=0; rc==SQLITE_OK && i<nInst; i++){ + int ip, iSnippetCol, iStart; + memset(aSeen, 0, nPhrase); + rc = pApi->xInst(pFts, i, &ip, &iSnippetCol, &iStart); + if( rc==SQLITE_OK && (iCol<0 || iSnippetCol==iCol) ){ + int nScore = 1000; + int iLast = iStart - 1 + pApi->xPhraseSize(pFts, ip); + int j; + aSeen[ip] = 1; + + for(j=i+1; rc==SQLITE_OK && j<nInst; j++){ + int ic; int io; int iFinal; + rc = pApi->xInst(pFts, j, &ip, &ic, &io); + iFinal = io + pApi->xPhraseSize(pFts, ip) - 1; + if( rc==SQLITE_OK && ic==iSnippetCol && iLast<iStart+nToken ){ + nScore += aSeen[ip] ? 1000 : 1; + aSeen[ip] = 1; + if( iFinal>iLast ) iLast = iFinal; + } + } + + if( rc==SQLITE_OK && nScore>nBestScore ){ + iBestCol = iSnippetCol; + iBestStart = iStart; + iBestLast = iLast; + nBestScore = nScore; + } + } + } + + if( rc==SQLITE_OK ){ + rc = pApi->xColumnSize(pFts, iBestCol, &nColSize); + } + if( rc==SQLITE_OK ){ + rc = pApi->xColumnText(pFts, iBestCol, &ctx.zIn, &ctx.nIn); + } + if( ctx.zIn ){ + if( rc==SQLITE_OK ){ + rc = fts5CInstIterInit(pApi, pFts, iBestCol, &ctx.iter); + } + + if( (iBestStart+nToken-1)>iBestLast ){ + iBestStart -= (iBestStart+nToken-1-iBestLast) / 2; + } + if( iBestStart+nToken>nColSize ){ + iBestStart = nColSize - nToken; + } + if( iBestStart<0 ) iBestStart = 0; + + ctx.iRangeStart = iBestStart; + ctx.iRangeEnd = iBestStart + nToken - 1; + + if( iBestStart>0 ){ + fts5HighlightAppend(&rc, &ctx, zEllips, -1); + } + if( rc==SQLITE_OK ){ + rc = pApi->xTokenize(pFts, ctx.zIn, ctx.nIn, (void*)&ctx,fts5HighlightCb); + } + if( ctx.iRangeEnd>=(nColSize-1) ){ + fts5HighlightAppend(&rc, &ctx, &ctx.zIn[ctx.iOff], ctx.nIn - ctx.iOff); + }else{ + fts5HighlightAppend(&rc, &ctx, zEllips, -1); + } + + if( rc==SQLITE_OK ){ + sqlite3_result_text(pCtx, (const char*)ctx.zOut, -1, SQLITE_TRANSIENT); + }else{ + sqlite3_result_error_code(pCtx, rc); + } + sqlite3_free(ctx.zOut); + } + sqlite3_free(aSeen); +} + +/************************************************************************/ + +/* +** The first time the bm25() function is called for a query, an instance +** of the following structure is allocated and populated. +*/ +typedef struct Fts5Bm25Data Fts5Bm25Data; +struct Fts5Bm25Data { + int nPhrase; /* Number of phrases in query */ + double avgdl; /* Average number of tokens in each row */ + double *aIDF; /* IDF for each phrase */ + double *aFreq; /* Array used to calculate phrase freq. */ +}; + +/* +** Callback used by fts5Bm25GetData() to count the number of rows in the +** table matched by each individual phrase within the query. +*/ +static int fts5CountCb( + const Fts5ExtensionApi *pApi, + Fts5Context *pFts, + void *pUserData /* Pointer to sqlite3_int64 variable */ +){ + sqlite3_int64 *pn = (sqlite3_int64*)pUserData; + UNUSED_PARAM2(pApi, pFts); + (*pn)++; + return SQLITE_OK; +} + +/* +** Set *ppData to point to the Fts5Bm25Data object for the current query. +** If the object has not already been allocated, allocate and populate it +** now. +*/ +static int fts5Bm25GetData( + const Fts5ExtensionApi *pApi, + Fts5Context *pFts, + Fts5Bm25Data **ppData /* OUT: bm25-data object for this query */ +){ + int rc = SQLITE_OK; /* Return code */ + Fts5Bm25Data *p; /* Object to return */ + + p = pApi->xGetAuxdata(pFts, 0); + if( p==0 ){ + int nPhrase; /* Number of phrases in query */ + sqlite3_int64 nRow = 0; /* Number of rows in table */ + sqlite3_int64 nToken = 0; /* Number of tokens in table */ + int nByte; /* Bytes of space to allocate */ + int i; + + /* Allocate the Fts5Bm25Data object */ + nPhrase = pApi->xPhraseCount(pFts); + nByte = sizeof(Fts5Bm25Data) + nPhrase*2*sizeof(double); + p = (Fts5Bm25Data*)sqlite3_malloc(nByte); + if( p==0 ){ + rc = SQLITE_NOMEM; + }else{ + memset(p, 0, nByte); + p->nPhrase = nPhrase; + p->aIDF = (double*)&p[1]; + p->aFreq = &p->aIDF[nPhrase]; + } + + /* Calculate the average document length for this FTS5 table */ + if( rc==SQLITE_OK ) rc = pApi->xRowCount(pFts, &nRow); + if( rc==SQLITE_OK ) rc = pApi->xColumnTotalSize(pFts, -1, &nToken); + if( rc==SQLITE_OK ) p->avgdl = (double)nToken / (double)nRow; + + /* Calculate an IDF for each phrase in the query */ + for(i=0; rc==SQLITE_OK && i<nPhrase; i++){ + sqlite3_int64 nHit = 0; + rc = pApi->xQueryPhrase(pFts, i, (void*)&nHit, fts5CountCb); + if( rc==SQLITE_OK ){ + /* Calculate the IDF (Inverse Document Frequency) for phrase i. + ** This is done using the standard BM25 formula as found on wikipedia: + ** + ** IDF = log( (N - nHit + 0.5) / (nHit + 0.5) ) + ** + ** where "N" is the total number of documents in the set and nHit + ** is the number that contain at least one instance of the phrase + ** under consideration. + ** + ** The problem with this is that if (N < 2*nHit), the IDF is + ** negative. Which is undesirable. So the mimimum allowable IDF is + ** (1e-6) - roughly the same as a term that appears in just over + ** half of set of 5,000,000 documents. */ + double idf = log( (nRow - nHit + 0.5) / (nHit + 0.5) ); + if( idf<=0.0 ) idf = 1e-6; + p->aIDF[i] = idf; + } + } + + if( rc!=SQLITE_OK ){ + sqlite3_free(p); + }else{ + rc = pApi->xSetAuxdata(pFts, p, sqlite3_free); + } + if( rc!=SQLITE_OK ) p = 0; + } + *ppData = p; + return rc; +} + +/* +** Implementation of bm25() function. +*/ +static void fts5Bm25Function( + const Fts5ExtensionApi *pApi, /* API offered by current FTS version */ + Fts5Context *pFts, /* First arg to pass to pApi functions */ + sqlite3_context *pCtx, /* Context for returning result/error */ + int nVal, /* Number of values in apVal[] array */ + sqlite3_value **apVal /* Array of trailing arguments */ +){ + const double k1 = 1.2; /* Constant "k1" from BM25 formula */ + const double b = 0.75; /* Constant "b" from BM25 formula */ + int rc = SQLITE_OK; /* Error code */ + double score = 0.0; /* SQL function return value */ + Fts5Bm25Data *pData; /* Values allocated/calculated once only */ + int i; /* Iterator variable */ + int nInst = 0; /* Value returned by xInstCount() */ + double D = 0.0; /* Total number of tokens in row */ + double *aFreq = 0; /* Array of phrase freq. for current row */ + + /* Calculate the phrase frequency (symbol "f(qi,D)" in the documentation) + ** for each phrase in the query for the current row. */ + rc = fts5Bm25GetData(pApi, pFts, &pData); + if( rc==SQLITE_OK ){ + aFreq = pData->aFreq; + memset(aFreq, 0, sizeof(double) * pData->nPhrase); + rc = pApi->xInstCount(pFts, &nInst); + } + for(i=0; rc==SQLITE_OK && i<nInst; i++){ + int ip; int ic; int io; + rc = pApi->xInst(pFts, i, &ip, &ic, &io); + if( rc==SQLITE_OK ){ + double w = (nVal > ic) ? sqlite3_value_double(apVal[ic]) : 1.0; + aFreq[ip] += w; + } + } + + /* Figure out the total size of the current row in tokens. */ + if( rc==SQLITE_OK ){ + int nTok; + rc = pApi->xColumnSize(pFts, -1, &nTok); + D = (double)nTok; + } + + /* Determine the BM25 score for the current row. */ + for(i=0; rc==SQLITE_OK && i<pData->nPhrase; i++){ + score += pData->aIDF[i] * ( + ( aFreq[i] * (k1 + 1.0) ) / + ( aFreq[i] + k1 * (1 - b + b * D / pData->avgdl) ) + ); + } + + /* If no error has occurred, return the calculated score. Otherwise, + ** throw an SQL exception. */ + if( rc==SQLITE_OK ){ + sqlite3_result_double(pCtx, -1.0 * score); + }else{ + sqlite3_result_error_code(pCtx, rc); + } +} + +static int sqlite3Fts5AuxInit(fts5_api *pApi){ + struct Builtin { + const char *zFunc; /* Function name (nul-terminated) */ + void *pUserData; /* User-data pointer */ + fts5_extension_function xFunc;/* Callback function */ + void (*xDestroy)(void*); /* Destructor function */ + } aBuiltin [] = { + { "snippet", 0, fts5SnippetFunction, 0 }, + { "highlight", 0, fts5HighlightFunction, 0 }, + { "bm25", 0, fts5Bm25Function, 0 }, + }; + int rc = SQLITE_OK; /* Return code */ + int i; /* To iterate through builtin functions */ + + for(i=0; rc==SQLITE_OK && i<ArraySize(aBuiltin); i++){ + rc = pApi->xCreateFunction(pApi, + aBuiltin[i].zFunc, + aBuiltin[i].pUserData, + aBuiltin[i].xFunc, + aBuiltin[i].xDestroy + ); + } + + return rc; +} + + + +#line 1 "fts5_buffer.c" +/* +** 2014 May 31 +** +** The author disclaims copyright to this source code. In place of +** a legal notice, here is a blessing: +** +** May you do good and not evil. +** May you find forgiveness for yourself and forgive others. +** May you share freely, never taking more than you give. +** +****************************************************************************** +*/ + + + +/* #include "fts5Int.h" */ + +static int sqlite3Fts5BufferSize(int *pRc, Fts5Buffer *pBuf, u32 nByte){ + if( (u32)pBuf->nSpace<nByte ){ + u32 nNew = pBuf->nSpace ? pBuf->nSpace : 64; + u8 *pNew; + while( nNew<nByte ){ + nNew = nNew * 2; + } + pNew = sqlite3_realloc(pBuf->p, nNew); + if( pNew==0 ){ + *pRc = SQLITE_NOMEM; + return 1; + }else{ + pBuf->nSpace = nNew; + pBuf->p = pNew; + } + } + return 0; +} + + +/* +** Encode value iVal as an SQLite varint and append it to the buffer object +** pBuf. If an OOM error occurs, set the error code in p. +*/ +static void sqlite3Fts5BufferAppendVarint(int *pRc, Fts5Buffer *pBuf, i64 iVal){ + if( fts5BufferGrow(pRc, pBuf, 9) ) return; + pBuf->n += sqlite3Fts5PutVarint(&pBuf->p[pBuf->n], iVal); +} + +static void sqlite3Fts5Put32(u8 *aBuf, int iVal){ + aBuf[0] = (iVal>>24) & 0x00FF; + aBuf[1] = (iVal>>16) & 0x00FF; + aBuf[2] = (iVal>> 8) & 0x00FF; + aBuf[3] = (iVal>> 0) & 0x00FF; +} + +static int sqlite3Fts5Get32(const u8 *aBuf){ + return (aBuf[0] << 24) + (aBuf[1] << 16) + (aBuf[2] << 8) + aBuf[3]; +} + +/* +** Append buffer nData/pData to buffer pBuf. If an OOM error occurs, set +** the error code in p. If an error has already occurred when this function +** is called, it is a no-op. +*/ +static void sqlite3Fts5BufferAppendBlob( + int *pRc, + Fts5Buffer *pBuf, + u32 nData, + const u8 *pData +){ + assert_nc( *pRc || nData>=0 ); + if( fts5BufferGrow(pRc, pBuf, nData) ) return; + memcpy(&pBuf->p[pBuf->n], pData, nData); + pBuf->n += nData; +} + +/* +** Append the nul-terminated string zStr to the buffer pBuf. This function +** ensures that the byte following the buffer data is set to 0x00, even +** though this byte is not included in the pBuf->n count. +*/ +static void sqlite3Fts5BufferAppendString( + int *pRc, + Fts5Buffer *pBuf, + const char *zStr +){ + int nStr = (int)strlen(zStr); + sqlite3Fts5BufferAppendBlob(pRc, pBuf, nStr+1, (const u8*)zStr); + pBuf->n--; +} + +/* +** Argument zFmt is a printf() style format string. This function performs +** the printf() style processing, then appends the results to buffer pBuf. +** +** Like sqlite3Fts5BufferAppendString(), this function ensures that the byte +** following the buffer data is set to 0x00, even though this byte is not +** included in the pBuf->n count. +*/ +static void sqlite3Fts5BufferAppendPrintf( + int *pRc, + Fts5Buffer *pBuf, + char *zFmt, ... +){ + if( *pRc==SQLITE_OK ){ + char *zTmp; + va_list ap; + va_start(ap, zFmt); + zTmp = sqlite3_vmprintf(zFmt, ap); + va_end(ap); + + if( zTmp==0 ){ + *pRc = SQLITE_NOMEM; + }else{ + sqlite3Fts5BufferAppendString(pRc, pBuf, zTmp); + sqlite3_free(zTmp); + } + } +} + +static char *sqlite3Fts5Mprintf(int *pRc, const char *zFmt, ...){ + char *zRet = 0; + if( *pRc==SQLITE_OK ){ + va_list ap; + va_start(ap, zFmt); + zRet = sqlite3_vmprintf(zFmt, ap); + va_end(ap); + if( zRet==0 ){ + *pRc = SQLITE_NOMEM; + } + } + return zRet; +} + + +/* +** Free any buffer allocated by pBuf. Zero the structure before returning. +*/ +static void sqlite3Fts5BufferFree(Fts5Buffer *pBuf){ + sqlite3_free(pBuf->p); + memset(pBuf, 0, sizeof(Fts5Buffer)); +} + +/* +** Zero the contents of the buffer object. But do not free the associated +** memory allocation. +*/ +static void sqlite3Fts5BufferZero(Fts5Buffer *pBuf){ + pBuf->n = 0; +} + +/* +** Set the buffer to contain nData/pData. If an OOM error occurs, leave an +** the error code in p. If an error has already occurred when this function +** is called, it is a no-op. +*/ +static void sqlite3Fts5BufferSet( + int *pRc, + Fts5Buffer *pBuf, + int nData, + const u8 *pData +){ + pBuf->n = 0; + sqlite3Fts5BufferAppendBlob(pRc, pBuf, nData, pData); +} + +static int sqlite3Fts5PoslistNext64( + const u8 *a, int n, /* Buffer containing poslist */ + int *pi, /* IN/OUT: Offset within a[] */ + i64 *piOff /* IN/OUT: Current offset */ +){ + int i = *pi; + if( i>=n ){ + /* EOF */ + *piOff = -1; + return 1; + }else{ + i64 iOff = *piOff; + int iVal; + fts5FastGetVarint32(a, i, iVal); + if( iVal==1 ){ + fts5FastGetVarint32(a, i, iVal); + iOff = ((i64)iVal) << 32; + fts5FastGetVarint32(a, i, iVal); + } + *piOff = iOff + (iVal-2); + *pi = i; + return 0; + } +} + + +/* +** Advance the iterator object passed as the only argument. Return true +** if the iterator reaches EOF, or false otherwise. +*/ +static int sqlite3Fts5PoslistReaderNext(Fts5PoslistReader *pIter){ + if( sqlite3Fts5PoslistNext64(pIter->a, pIter->n, &pIter->i, &pIter->iPos) ){ + pIter->bEof = 1; + } + return pIter->bEof; +} + +static int sqlite3Fts5PoslistReaderInit( + const u8 *a, int n, /* Poslist buffer to iterate through */ + Fts5PoslistReader *pIter /* Iterator object to initialize */ +){ + memset(pIter, 0, sizeof(*pIter)); + pIter->a = a; + pIter->n = n; + sqlite3Fts5PoslistReaderNext(pIter); + return pIter->bEof; +} + +/* +** Append position iPos to the position list being accumulated in buffer +** pBuf, which must be already be large enough to hold the new data. +** The previous position written to this list is *piPrev. *piPrev is set +** to iPos before returning. +*/ +static void sqlite3Fts5PoslistSafeAppend( + Fts5Buffer *pBuf, + i64 *piPrev, + i64 iPos +){ + static const i64 colmask = ((i64)(0x7FFFFFFF)) << 32; + if( (iPos & colmask) != (*piPrev & colmask) ){ + pBuf->p[pBuf->n++] = 1; + pBuf->n += sqlite3Fts5PutVarint(&pBuf->p[pBuf->n], (iPos>>32)); + *piPrev = (iPos & colmask); + } + pBuf->n += sqlite3Fts5PutVarint(&pBuf->p[pBuf->n], (iPos-*piPrev)+2); + *piPrev = iPos; +} + +static int sqlite3Fts5PoslistWriterAppend( + Fts5Buffer *pBuf, + Fts5PoslistWriter *pWriter, + i64 iPos +){ + int rc = 0; /* Initialized only to suppress erroneous warning from Clang */ + if( fts5BufferGrow(&rc, pBuf, 5+5+5) ) return rc; + sqlite3Fts5PoslistSafeAppend(pBuf, &pWriter->iPrev, iPos); + return SQLITE_OK; +} + +static void *sqlite3Fts5MallocZero(int *pRc, int nByte){ + void *pRet = 0; + if( *pRc==SQLITE_OK ){ + pRet = sqlite3_malloc(nByte); + if( pRet==0 && nByte>0 ){ + *pRc = SQLITE_NOMEM; + }else{ + memset(pRet, 0, nByte); + } + } + return pRet; +} + +/* +** Return a nul-terminated copy of the string indicated by pIn. If nIn +** is non-negative, then it is the length of the string in bytes. Otherwise, +** the length of the string is determined using strlen(). +** +** It is the responsibility of the caller to eventually free the returned +** buffer using sqlite3_free(). If an OOM error occurs, NULL is returned. +*/ +static char *sqlite3Fts5Strndup(int *pRc, const char *pIn, int nIn){ + char *zRet = 0; + if( *pRc==SQLITE_OK ){ + if( nIn<0 ){ + nIn = (int)strlen(pIn); + } + zRet = (char*)sqlite3_malloc(nIn+1); + if( zRet ){ + memcpy(zRet, pIn, nIn); + zRet[nIn] = '\0'; + }else{ + *pRc = SQLITE_NOMEM; + } + } + return zRet; +} + + +/* +** Return true if character 't' may be part of an FTS5 bareword, or false +** otherwise. Characters that may be part of barewords: +** +** * All non-ASCII characters, +** * The 52 upper and lower case ASCII characters, and +** * The 10 integer ASCII characters. +** * The underscore character "_" (0x5F). +** * The unicode "subsitute" character (0x1A). +*/ +static int sqlite3Fts5IsBareword(char t){ + u8 aBareword[128] = { + 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 0x00 .. 0x0F */ + 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, /* 0x10 .. 0x1F */ + 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 0x20 .. 0x2F */ + 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, /* 0x30 .. 0x3F */ + 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, /* 0x40 .. 0x4F */ + 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 1, /* 0x50 .. 0x5F */ + 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, /* 0x60 .. 0x6F */ + 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0 /* 0x70 .. 0x7F */ + }; + + return (t & 0x80) || aBareword[(int)t]; +} + + +/************************************************************************* +*/ +typedef struct Fts5TermsetEntry Fts5TermsetEntry; +struct Fts5TermsetEntry { + char *pTerm; + int nTerm; + int iIdx; /* Index (main or aPrefix[] entry) */ + Fts5TermsetEntry *pNext; +}; + +struct Fts5Termset { + Fts5TermsetEntry *apHash[512]; +}; + +static int sqlite3Fts5TermsetNew(Fts5Termset **pp){ + int rc = SQLITE_OK; + *pp = sqlite3Fts5MallocZero(&rc, sizeof(Fts5Termset)); + return rc; +} + +static int sqlite3Fts5TermsetAdd( + Fts5Termset *p, + int iIdx, + const char *pTerm, int nTerm, + int *pbPresent +){ + int rc = SQLITE_OK; + *pbPresent = 0; + if( p ){ + int i; + u32 hash = 13; + Fts5TermsetEntry *pEntry; + + /* Calculate a hash value for this term. This is the same hash checksum + ** used by the fts5_hash.c module. This is not important for correct + ** operation of the module, but is necessary to ensure that some tests + ** designed to produce hash table collisions really do work. */ + for(i=nTerm-1; i>=0; i--){ + hash = (hash << 3) ^ hash ^ pTerm[i]; + } + hash = (hash << 3) ^ hash ^ iIdx; + hash = hash % ArraySize(p->apHash); + + for(pEntry=p->apHash[hash]; pEntry; pEntry=pEntry->pNext){ + if( pEntry->iIdx==iIdx + && pEntry->nTerm==nTerm + && memcmp(pEntry->pTerm, pTerm, nTerm)==0 + ){ + *pbPresent = 1; + break; + } + } + + if( pEntry==0 ){ + pEntry = sqlite3Fts5MallocZero(&rc, sizeof(Fts5TermsetEntry) + nTerm); + if( pEntry ){ + pEntry->pTerm = (char*)&pEntry[1]; + pEntry->nTerm = nTerm; + pEntry->iIdx = iIdx; + memcpy(pEntry->pTerm, pTerm, nTerm); + pEntry->pNext = p->apHash[hash]; + p->apHash[hash] = pEntry; + } + } + } + + return rc; +} + +static void sqlite3Fts5TermsetFree(Fts5Termset *p){ + if( p ){ + u32 i; + for(i=0; i<ArraySize(p->apHash); i++){ + Fts5TermsetEntry *pEntry = p->apHash[i]; + while( pEntry ){ + Fts5TermsetEntry *pDel = pEntry; + pEntry = pEntry->pNext; + sqlite3_free(pDel); + } + } + sqlite3_free(p); + } +} + +#line 1 "fts5_config.c" +/* +** 2014 Jun 09 +** +** The author disclaims copyright to this source code. In place of +** a legal notice, here is a blessing: +** +** May you do good and not evil. +** May you find forgiveness for yourself and forgive others. +** May you share freely, never taking more than you give. +** +****************************************************************************** +** +** This is an SQLite module implementing full-text search. +*/ + + +/* #include "fts5Int.h" */ + +#define FTS5_DEFAULT_PAGE_SIZE 4050 +#define FTS5_DEFAULT_AUTOMERGE 4 +#define FTS5_DEFAULT_CRISISMERGE 16 +#define FTS5_DEFAULT_HASHSIZE (1024*1024) + +/* Maximum allowed page size */ +#define FTS5_MAX_PAGE_SIZE (128*1024) + +static int fts5_iswhitespace(char x){ + return (x==' '); +} + +static int fts5_isopenquote(char x){ + return (x=='"' || x=='\'' || x=='[' || x=='`'); +} + +/* +** Argument pIn points to a character that is part of a nul-terminated +** string. Return a pointer to the first character following *pIn in +** the string that is not a white-space character. +*/ +static const char *fts5ConfigSkipWhitespace(const char *pIn){ + const char *p = pIn; + if( p ){ + while( fts5_iswhitespace(*p) ){ p++; } + } + return p; +} + +/* +** Argument pIn points to a character that is part of a nul-terminated +** string. Return a pointer to the first character following *pIn in +** the string that is not a "bareword" character. +*/ +static const char *fts5ConfigSkipBareword(const char *pIn){ + const char *p = pIn; + while ( sqlite3Fts5IsBareword(*p) ) p++; + if( p==pIn ) p = 0; + return p; +} + +static int fts5_isdigit(char a){ + return (a>='0' && a<='9'); +} + + + +static const char *fts5ConfigSkipLiteral(const char *pIn){ + const char *p = pIn; + switch( *p ){ + case 'n': case 'N': + if( sqlite3_strnicmp("null", p, 4)==0 ){ + p = &p[4]; + }else{ + p = 0; + } + break; + + case 'x': case 'X': + p++; + if( *p=='\'' ){ + p++; + while( (*p>='a' && *p<='f') + || (*p>='A' && *p<='F') + || (*p>='0' && *p<='9') + ){ + p++; + } + if( *p=='\'' && 0==((p-pIn)%2) ){ + p++; + }else{ + p = 0; + } + }else{ + p = 0; + } + break; + + case '\'': + p++; + while( p ){ + if( *p=='\'' ){ + p++; + if( *p!='\'' ) break; + } + p++; + if( *p==0 ) p = 0; + } + break; + + default: + /* maybe a number */ + if( *p=='+' || *p=='-' ) p++; + while( fts5_isdigit(*p) ) p++; + + /* At this point, if the literal was an integer, the parse is + ** finished. Or, if it is a floating point value, it may continue + ** with either a decimal point or an 'E' character. */ + if( *p=='.' && fts5_isdigit(p[1]) ){ + p += 2; + while( fts5_isdigit(*p) ) p++; + } + if( p==pIn ) p = 0; + + break; + } + + return p; +} + +/* +** The first character of the string pointed to by argument z is guaranteed +** to be an open-quote character (see function fts5_isopenquote()). +** +** This function searches for the corresponding close-quote character within +** the string and, if found, dequotes the string in place and adds a new +** nul-terminator byte. +** +** If the close-quote is found, the value returned is the byte offset of +** the character immediately following it. Or, if the close-quote is not +** found, -1 is returned. If -1 is returned, the buffer is left in an +** undefined state. +*/ +static int fts5Dequote(char *z){ + char q; + int iIn = 1; + int iOut = 0; + q = z[0]; + + /* Set stack variable q to the close-quote character */ + assert( q=='[' || q=='\'' || q=='"' || q=='`' ); + if( q=='[' ) q = ']'; + + while( ALWAYS(z[iIn]) ){ + if( z[iIn]==q ){ + if( z[iIn+1]!=q ){ + /* Character iIn was the close quote. */ + iIn++; + break; + }else{ + /* Character iIn and iIn+1 form an escaped quote character. Skip + ** the input cursor past both and copy a single quote character + ** to the output buffer. */ + iIn += 2; + z[iOut++] = q; + } + }else{ + z[iOut++] = z[iIn++]; + } + } + + z[iOut] = '\0'; + return iIn; +} + +/* +** Convert an SQL-style quoted string into a normal string by removing +** the quote characters. The conversion is done in-place. If the +** input does not begin with a quote character, then this routine +** is a no-op. +** +** Examples: +** +** "abc" becomes abc +** 'xyz' becomes xyz +** [pqr] becomes pqr +** `mno` becomes mno +*/ +static void sqlite3Fts5Dequote(char *z){ + char quote; /* Quote character (if any ) */ + + assert( 0==fts5_iswhitespace(z[0]) ); + quote = z[0]; + if( quote=='[' || quote=='\'' || quote=='"' || quote=='`' ){ + fts5Dequote(z); + } +} + + +struct Fts5Enum { + const char *zName; + int eVal; +}; +typedef struct Fts5Enum Fts5Enum; + +static int fts5ConfigSetEnum( + const Fts5Enum *aEnum, + const char *zEnum, + int *peVal +){ + int nEnum = (int)strlen(zEnum); + int i; + int iVal = -1; + + for(i=0; aEnum[i].zName; i++){ + if( sqlite3_strnicmp(aEnum[i].zName, zEnum, nEnum)==0 ){ + if( iVal>=0 ) return SQLITE_ERROR; + iVal = aEnum[i].eVal; + } + } + + *peVal = iVal; + return iVal<0 ? SQLITE_ERROR : SQLITE_OK; +} + +/* +** Parse a "special" CREATE VIRTUAL TABLE directive and update +** configuration object pConfig as appropriate. +** +** If successful, object pConfig is updated and SQLITE_OK returned. If +** an error occurs, an SQLite error code is returned and an error message +** may be left in *pzErr. It is the responsibility of the caller to +** eventually free any such error message using sqlite3_free(). +*/ +static int fts5ConfigParseSpecial( + Fts5Global *pGlobal, + Fts5Config *pConfig, /* Configuration object to update */ + const char *zCmd, /* Special command to parse */ + const char *zArg, /* Argument to parse */ + char **pzErr /* OUT: Error message */ +){ + int rc = SQLITE_OK; + int nCmd = (int)strlen(zCmd); + if( sqlite3_strnicmp("prefix", zCmd, nCmd)==0 ){ + const int nByte = sizeof(int) * FTS5_MAX_PREFIX_INDEXES; + const char *p; + int bFirst = 1; + if( pConfig->aPrefix==0 ){ + pConfig->aPrefix = sqlite3Fts5MallocZero(&rc, nByte); + if( rc ) return rc; + } + + p = zArg; + while( 1 ){ + int nPre = 0; + + while( p[0]==' ' ) p++; + if( bFirst==0 && p[0]==',' ){ + p++; + while( p[0]==' ' ) p++; + }else if( p[0]=='\0' ){ + break; + } + if( p[0]<'0' || p[0]>'9' ){ + *pzErr = sqlite3_mprintf("malformed prefix=... directive"); + rc = SQLITE_ERROR; + break; + } + + if( pConfig->nPrefix==FTS5_MAX_PREFIX_INDEXES ){ + *pzErr = sqlite3_mprintf( + "too many prefix indexes (max %d)", FTS5_MAX_PREFIX_INDEXES + ); + rc = SQLITE_ERROR; + break; + } + + while( p[0]>='0' && p[0]<='9' && nPre<1000 ){ + nPre = nPre*10 + (p[0] - '0'); + p++; + } + + if( nPre<=0 || nPre>=1000 ){ + *pzErr = sqlite3_mprintf("prefix length out of range (max 999)"); + rc = SQLITE_ERROR; + break; + } + + pConfig->aPrefix[pConfig->nPrefix] = nPre; + pConfig->nPrefix++; + bFirst = 0; + } + assert( pConfig->nPrefix<=FTS5_MAX_PREFIX_INDEXES ); + return rc; + } + + if( sqlite3_strnicmp("tokenize", zCmd, nCmd)==0 ){ + const char *p = (const char*)zArg; + int nArg = (int)strlen(zArg) + 1; + char **azArg = sqlite3Fts5MallocZero(&rc, sizeof(char*) * nArg); + char *pDel = sqlite3Fts5MallocZero(&rc, nArg * 2); + char *pSpace = pDel; + + if( azArg && pSpace ){ + if( pConfig->pTok ){ + *pzErr = sqlite3_mprintf("multiple tokenize=... directives"); + rc = SQLITE_ERROR; + }else{ + for(nArg=0; p && *p; nArg++){ + const char *p2 = fts5ConfigSkipWhitespace(p); + if( *p2=='\'' ){ + p = fts5ConfigSkipLiteral(p2); + }else{ + p = fts5ConfigSkipBareword(p2); + } + if( p ){ + memcpy(pSpace, p2, p-p2); + azArg[nArg] = pSpace; + sqlite3Fts5Dequote(pSpace); + pSpace += (p - p2) + 1; + p = fts5ConfigSkipWhitespace(p); + } + } + if( p==0 ){ + *pzErr = sqlite3_mprintf("parse error in tokenize directive"); + rc = SQLITE_ERROR; + }else{ + rc = sqlite3Fts5GetTokenizer(pGlobal, + (const char**)azArg, nArg, &pConfig->pTok, &pConfig->pTokApi, + pzErr + ); + } + } + } + + sqlite3_free(azArg); + sqlite3_free(pDel); + return rc; + } + + if( sqlite3_strnicmp("content", zCmd, nCmd)==0 ){ + if( pConfig->eContent!=FTS5_CONTENT_NORMAL ){ + *pzErr = sqlite3_mprintf("multiple content=... directives"); + rc = SQLITE_ERROR; + }else{ + if( zArg[0] ){ + pConfig->eContent = FTS5_CONTENT_EXTERNAL; + pConfig->zContent = sqlite3Fts5Mprintf(&rc, "%Q.%Q", pConfig->zDb,zArg); + }else{ + pConfig->eContent = FTS5_CONTENT_NONE; + } + } + return rc; + } + + if( sqlite3_strnicmp("content_rowid", zCmd, nCmd)==0 ){ + if( pConfig->zContentRowid ){ + *pzErr = sqlite3_mprintf("multiple content_rowid=... directives"); + rc = SQLITE_ERROR; + }else{ + pConfig->zContentRowid = sqlite3Fts5Strndup(&rc, zArg, -1); + } + return rc; + } + + if( sqlite3_strnicmp("columnsize", zCmd, nCmd)==0 ){ + if( (zArg[0]!='0' && zArg[0]!='1') || zArg[1]!='\0' ){ + *pzErr = sqlite3_mprintf("malformed columnsize=... directive"); + rc = SQLITE_ERROR; + }else{ + pConfig->bColumnsize = (zArg[0]=='1'); + } + return rc; + } + + if( sqlite3_strnicmp("detail", zCmd, nCmd)==0 ){ + const Fts5Enum aDetail[] = { + { "none", FTS5_DETAIL_NONE }, + { "full", FTS5_DETAIL_FULL }, + { "columns", FTS5_DETAIL_COLUMNS }, + { 0, 0 } + }; + + if( (rc = fts5ConfigSetEnum(aDetail, zArg, &pConfig->eDetail)) ){ + *pzErr = sqlite3_mprintf("malformed detail=... directive"); + } + return rc; + } + + *pzErr = sqlite3_mprintf("unrecognized option: \"%.*s\"", nCmd, zCmd); + return SQLITE_ERROR; +} + +/* +** Allocate an instance of the default tokenizer ("simple") at +** Fts5Config.pTokenizer. Return SQLITE_OK if successful, or an SQLite error +** code if an error occurs. +*/ +static int fts5ConfigDefaultTokenizer(Fts5Global *pGlobal, Fts5Config *pConfig){ + assert( pConfig->pTok==0 && pConfig->pTokApi==0 ); + return sqlite3Fts5GetTokenizer( + pGlobal, 0, 0, &pConfig->pTok, &pConfig->pTokApi, 0 + ); +} + +/* +** Gobble up the first bareword or quoted word from the input buffer zIn. +** Return a pointer to the character immediately following the last in +** the gobbled word if successful, or a NULL pointer otherwise (failed +** to find close-quote character). +** +** Before returning, set pzOut to point to a new buffer containing a +** nul-terminated, dequoted copy of the gobbled word. If the word was +** quoted, *pbQuoted is also set to 1 before returning. +** +** If *pRc is other than SQLITE_OK when this function is called, it is +** a no-op (NULL is returned). Otherwise, if an OOM occurs within this +** function, *pRc is set to SQLITE_NOMEM before returning. *pRc is *not* +** set if a parse error (failed to find close quote) occurs. +*/ +static const char *fts5ConfigGobbleWord( + int *pRc, /* IN/OUT: Error code */ + const char *zIn, /* Buffer to gobble string/bareword from */ + char **pzOut, /* OUT: malloc'd buffer containing str/bw */ + int *pbQuoted /* OUT: Set to true if dequoting required */ +){ + const char *zRet = 0; + + int nIn = (int)strlen(zIn); + char *zOut = sqlite3_malloc(nIn+1); + + assert( *pRc==SQLITE_OK ); + *pbQuoted = 0; + *pzOut = 0; + + if( zOut==0 ){ + *pRc = SQLITE_NOMEM; + }else{ + memcpy(zOut, zIn, nIn+1); + if( fts5_isopenquote(zOut[0]) ){ + int ii = fts5Dequote(zOut); + zRet = &zIn[ii]; + *pbQuoted = 1; + }else{ + zRet = fts5ConfigSkipBareword(zIn); + zOut[zRet-zIn] = '\0'; + } + } + + if( zRet==0 ){ + sqlite3_free(zOut); + }else{ + *pzOut = zOut; + } + + return zRet; +} + +static int fts5ConfigParseColumn( + Fts5Config *p, + char *zCol, + char *zArg, + char **pzErr +){ + int rc = SQLITE_OK; + if( 0==sqlite3_stricmp(zCol, FTS5_RANK_NAME) + || 0==sqlite3_stricmp(zCol, FTS5_ROWID_NAME) + ){ + *pzErr = sqlite3_mprintf("reserved fts5 column name: %s", zCol); + rc = SQLITE_ERROR; + }else if( zArg ){ + if( 0==sqlite3_stricmp(zArg, "unindexed") ){ + p->abUnindexed[p->nCol] = 1; + }else{ + *pzErr = sqlite3_mprintf("unrecognized column option: %s", zArg); + rc = SQLITE_ERROR; + } + } + + p->azCol[p->nCol++] = zCol; + return rc; +} + +/* +** Populate the Fts5Config.zContentExprlist string. +*/ +static int fts5ConfigMakeExprlist(Fts5Config *p){ + int i; + int rc = SQLITE_OK; + Fts5Buffer buf = {0, 0, 0}; + + sqlite3Fts5BufferAppendPrintf(&rc, &buf, "T.%Q", p->zContentRowid); + if( p->eContent!=FTS5_CONTENT_NONE ){ + for(i=0; i<p->nCol; i++){ + if( p->eContent==FTS5_CONTENT_EXTERNAL ){ + sqlite3Fts5BufferAppendPrintf(&rc, &buf, ", T.%Q", p->azCol[i]); + }else{ + sqlite3Fts5BufferAppendPrintf(&rc, &buf, ", T.c%d", i); + } + } + } + + assert( p->zContentExprlist==0 ); + p->zContentExprlist = (char*)buf.p; + return rc; +} + +/* +** Arguments nArg/azArg contain the string arguments passed to the xCreate +** or xConnect method of the virtual table. This function attempts to +** allocate an instance of Fts5Config containing the results of parsing +** those arguments. +** +** If successful, SQLITE_OK is returned and *ppOut is set to point to the +** new Fts5Config object. If an error occurs, an SQLite error code is +** returned, *ppOut is set to NULL and an error message may be left in +** *pzErr. It is the responsibility of the caller to eventually free any +** such error message using sqlite3_free(). +*/ +static int sqlite3Fts5ConfigParse( + Fts5Global *pGlobal, + sqlite3 *db, + int nArg, /* Number of arguments */ + const char **azArg, /* Array of nArg CREATE VIRTUAL TABLE args */ + Fts5Config **ppOut, /* OUT: Results of parse */ + char **pzErr /* OUT: Error message */ +){ + int rc = SQLITE_OK; /* Return code */ + Fts5Config *pRet; /* New object to return */ + int i; + int nByte; + + *ppOut = pRet = (Fts5Config*)sqlite3_malloc(sizeof(Fts5Config)); + if( pRet==0 ) return SQLITE_NOMEM; + memset(pRet, 0, sizeof(Fts5Config)); + pRet->db = db; + pRet->iCookie = -1; + + nByte = nArg * (sizeof(char*) + sizeof(u8)); + pRet->azCol = (char**)sqlite3Fts5MallocZero(&rc, nByte); + pRet->abUnindexed = (u8*)&pRet->azCol[nArg]; + pRet->zDb = sqlite3Fts5Strndup(&rc, azArg[1], -1); + pRet->zName = sqlite3Fts5Strndup(&rc, azArg[2], -1); + pRet->bColumnsize = 1; + pRet->eDetail = FTS5_DETAIL_FULL; +#ifdef SQLITE_DEBUG + pRet->bPrefixIndex = 1; +#endif + if( rc==SQLITE_OK && sqlite3_stricmp(pRet->zName, FTS5_RANK_NAME)==0 ){ + *pzErr = sqlite3_mprintf("reserved fts5 table name: %s", pRet->zName); + rc = SQLITE_ERROR; + } + + for(i=3; rc==SQLITE_OK && i<nArg; i++){ + const char *zOrig = azArg[i]; + const char *z; + char *zOne = 0; + char *zTwo = 0; + int bOption = 0; + int bMustBeCol = 0; + + z = fts5ConfigGobbleWord(&rc, zOrig, &zOne, &bMustBeCol); + z = fts5ConfigSkipWhitespace(z); + if( z && *z=='=' ){ + bOption = 1; + z++; + if( bMustBeCol ) z = 0; + } + z = fts5ConfigSkipWhitespace(z); + if( z && z[0] ){ + int bDummy; + z = fts5ConfigGobbleWord(&rc, z, &zTwo, &bDummy); + if( z && z[0] ) z = 0; + } + + if( rc==SQLITE_OK ){ + if( z==0 ){ + *pzErr = sqlite3_mprintf("parse error in \"%s\"", zOrig); + rc = SQLITE_ERROR; + }else{ + if( bOption ){ + rc = fts5ConfigParseSpecial(pGlobal, pRet, zOne, zTwo?zTwo:"", pzErr); + }else{ + rc = fts5ConfigParseColumn(pRet, zOne, zTwo, pzErr); + zOne = 0; + } + } + } + + sqlite3_free(zOne); + sqlite3_free(zTwo); + } + + /* If a tokenizer= option was successfully parsed, the tokenizer has + ** already been allocated. Otherwise, allocate an instance of the default + ** tokenizer (unicode61) now. */ + if( rc==SQLITE_OK && pRet->pTok==0 ){ + rc = fts5ConfigDefaultTokenizer(pGlobal, pRet); + } + + /* If no zContent option was specified, fill in the default values. */ + if( rc==SQLITE_OK && pRet->zContent==0 ){ + const char *zTail = 0; + assert( pRet->eContent==FTS5_CONTENT_NORMAL + || pRet->eContent==FTS5_CONTENT_NONE + ); + if( pRet->eContent==FTS5_CONTENT_NORMAL ){ + zTail = "content"; + }else if( pRet->bColumnsize ){ + zTail = "docsize"; + } + + if( zTail ){ + pRet->zContent = sqlite3Fts5Mprintf( + &rc, "%Q.'%q_%s'", pRet->zDb, pRet->zName, zTail + ); + } + } + + if( rc==SQLITE_OK && pRet->zContentRowid==0 ){ + pRet->zContentRowid = sqlite3Fts5Strndup(&rc, "rowid", -1); + } + + /* Formulate the zContentExprlist text */ + if( rc==SQLITE_OK ){ + rc = fts5ConfigMakeExprlist(pRet); + } + + if( rc!=SQLITE_OK ){ + sqlite3Fts5ConfigFree(pRet); + *ppOut = 0; + } + return rc; +} + +/* +** Free the configuration object passed as the only argument. +*/ +static void sqlite3Fts5ConfigFree(Fts5Config *pConfig){ + if( pConfig ){ + int i; + if( pConfig->pTok ){ + pConfig->pTokApi->xDelete(pConfig->pTok); + } + sqlite3_free(pConfig->zDb); + sqlite3_free(pConfig->zName); + for(i=0; i<pConfig->nCol; i++){ + sqlite3_free(pConfig->azCol[i]); + } + sqlite3_free(pConfig->azCol); + sqlite3_free(pConfig->aPrefix); + sqlite3_free(pConfig->zRank); + sqlite3_free(pConfig->zRankArgs); + sqlite3_free(pConfig->zContent); + sqlite3_free(pConfig->zContentRowid); + sqlite3_free(pConfig->zContentExprlist); + sqlite3_free(pConfig); + } +} + +/* +** Call sqlite3_declare_vtab() based on the contents of the configuration +** object passed as the only argument. Return SQLITE_OK if successful, or +** an SQLite error code if an error occurs. +*/ +static int sqlite3Fts5ConfigDeclareVtab(Fts5Config *pConfig){ + int i; + int rc = SQLITE_OK; + char *zSql; + + zSql = sqlite3Fts5Mprintf(&rc, "CREATE TABLE x("); + for(i=0; zSql && i<pConfig->nCol; i++){ + const char *zSep = (i==0?"":", "); + zSql = sqlite3Fts5Mprintf(&rc, "%z%s%Q", zSql, zSep, pConfig->azCol[i]); + } + zSql = sqlite3Fts5Mprintf(&rc, "%z, %Q HIDDEN, %s HIDDEN)", + zSql, pConfig->zName, FTS5_RANK_NAME + ); + + assert( zSql || rc==SQLITE_NOMEM ); + if( zSql ){ + rc = sqlite3_declare_vtab(pConfig->db, zSql); + sqlite3_free(zSql); + } + + return rc; +} + +/* +** Tokenize the text passed via the second and third arguments. +** +** The callback is invoked once for each token in the input text. The +** arguments passed to it are, in order: +** +** void *pCtx // Copy of 4th argument to sqlite3Fts5Tokenize() +** const char *pToken // Pointer to buffer containing token +** int nToken // Size of token in bytes +** int iStart // Byte offset of start of token within input text +** int iEnd // Byte offset of end of token within input text +** int iPos // Position of token in input (first token is 0) +** +** If the callback returns a non-zero value the tokenization is abandoned +** and no further callbacks are issued. +** +** This function returns SQLITE_OK if successful or an SQLite error code +** if an error occurs. If the tokenization was abandoned early because +** the callback returned SQLITE_DONE, this is not an error and this function +** still returns SQLITE_OK. Or, if the tokenization was abandoned early +** because the callback returned another non-zero value, it is assumed +** to be an SQLite error code and returned to the caller. +*/ +static int sqlite3Fts5Tokenize( + Fts5Config *pConfig, /* FTS5 Configuration object */ + int flags, /* FTS5_TOKENIZE_* flags */ + const char *pText, int nText, /* Text to tokenize */ + void *pCtx, /* Context passed to xToken() */ + int (*xToken)(void*, int, const char*, int, int, int) /* Callback */ +){ + if( pText==0 ) return SQLITE_OK; + return pConfig->pTokApi->xTokenize( + pConfig->pTok, pCtx, flags, pText, nText, xToken + ); +} + +/* +** Argument pIn points to the first character in what is expected to be +** a comma-separated list of SQL literals followed by a ')' character. +** If it actually is this, return a pointer to the ')'. Otherwise, return +** NULL to indicate a parse error. +*/ +static const char *fts5ConfigSkipArgs(const char *pIn){ + const char *p = pIn; + + while( 1 ){ + p = fts5ConfigSkipWhitespace(p); + p = fts5ConfigSkipLiteral(p); + p = fts5ConfigSkipWhitespace(p); + if( p==0 || *p==')' ) break; + if( *p!=',' ){ + p = 0; + break; + } + p++; + } + + return p; +} + +/* +** Parameter zIn contains a rank() function specification. The format of +** this is: +** +** + Bareword (function name) +** + Open parenthesis - "(" +** + Zero or more SQL literals in a comma separated list +** + Close parenthesis - ")" +*/ +static int sqlite3Fts5ConfigParseRank( + const char *zIn, /* Input string */ + char **pzRank, /* OUT: Rank function name */ + char **pzRankArgs /* OUT: Rank function arguments */ +){ + const char *p = zIn; + const char *pRank; + char *zRank = 0; + char *zRankArgs = 0; + int rc = SQLITE_OK; + + *pzRank = 0; + *pzRankArgs = 0; + + if( p==0 ){ + rc = SQLITE_ERROR; + }else{ + p = fts5ConfigSkipWhitespace(p); + pRank = p; + p = fts5ConfigSkipBareword(p); + + if( p ){ + zRank = sqlite3Fts5MallocZero(&rc, 1 + p - pRank); + if( zRank ) memcpy(zRank, pRank, p-pRank); + }else{ + rc = SQLITE_ERROR; + } + + if( rc==SQLITE_OK ){ + p = fts5ConfigSkipWhitespace(p); + if( *p!='(' ) rc = SQLITE_ERROR; + p++; + } + if( rc==SQLITE_OK ){ + const char *pArgs; + p = fts5ConfigSkipWhitespace(p); + pArgs = p; + if( *p!=')' ){ + p = fts5ConfigSkipArgs(p); + if( p==0 ){ + rc = SQLITE_ERROR; + }else{ + zRankArgs = sqlite3Fts5MallocZero(&rc, 1 + p - pArgs); + if( zRankArgs ) memcpy(zRankArgs, pArgs, p-pArgs); + } + } + } + } + + if( rc!=SQLITE_OK ){ + sqlite3_free(zRank); + assert( zRankArgs==0 ); + }else{ + *pzRank = zRank; + *pzRankArgs = zRankArgs; + } + return rc; +} + +static int sqlite3Fts5ConfigSetValue( + Fts5Config *pConfig, + const char *zKey, + sqlite3_value *pVal, + int *pbBadkey +){ + int rc = SQLITE_OK; + + if( 0==sqlite3_stricmp(zKey, "pgsz") ){ + int pgsz = 0; + if( SQLITE_INTEGER==sqlite3_value_numeric_type(pVal) ){ + pgsz = sqlite3_value_int(pVal); + } + if( pgsz<=0 || pgsz>FTS5_MAX_PAGE_SIZE ){ + *pbBadkey = 1; + }else{ + pConfig->pgsz = pgsz; + } + } + + else if( 0==sqlite3_stricmp(zKey, "hashsize") ){ + int nHashSize = -1; + if( SQLITE_INTEGER==sqlite3_value_numeric_type(pVal) ){ + nHashSize = sqlite3_value_int(pVal); + } + if( nHashSize<=0 ){ + *pbBadkey = 1; + }else{ + pConfig->nHashSize = nHashSize; + } + } + + else if( 0==sqlite3_stricmp(zKey, "automerge") ){ + int nAutomerge = -1; + if( SQLITE_INTEGER==sqlite3_value_numeric_type(pVal) ){ + nAutomerge = sqlite3_value_int(pVal); + } + if( nAutomerge<0 || nAutomerge>64 ){ + *pbBadkey = 1; + }else{ + if( nAutomerge==1 ) nAutomerge = FTS5_DEFAULT_AUTOMERGE; + pConfig->nAutomerge = nAutomerge; + } + } + + else if( 0==sqlite3_stricmp(zKey, "crisismerge") ){ + int nCrisisMerge = -1; + if( SQLITE_INTEGER==sqlite3_value_numeric_type(pVal) ){ + nCrisisMerge = sqlite3_value_int(pVal); + } + if( nCrisisMerge<0 ){ + *pbBadkey = 1; + }else{ + if( nCrisisMerge<=1 ) nCrisisMerge = FTS5_DEFAULT_CRISISMERGE; + pConfig->nCrisisMerge = nCrisisMerge; + } + } + + else if( 0==sqlite3_stricmp(zKey, "rank") ){ + const char *zIn = (const char*)sqlite3_value_text(pVal); + char *zRank; + char *zRankArgs; + rc = sqlite3Fts5ConfigParseRank(zIn, &zRank, &zRankArgs); + if( rc==SQLITE_OK ){ + sqlite3_free(pConfig->zRank); + sqlite3_free(pConfig->zRankArgs); + pConfig->zRank = zRank; + pConfig->zRankArgs = zRankArgs; + }else if( rc==SQLITE_ERROR ){ + rc = SQLITE_OK; + *pbBadkey = 1; + } + }else{ + *pbBadkey = 1; + } + return rc; +} + +/* +** Load the contents of the %_config table into memory. +*/ +static int sqlite3Fts5ConfigLoad(Fts5Config *pConfig, int iCookie){ + const char *zSelect = "SELECT k, v FROM %Q.'%q_config'"; + char *zSql; + sqlite3_stmt *p = 0; + int rc = SQLITE_OK; + int iVersion = 0; + + /* Set default values */ + pConfig->pgsz = FTS5_DEFAULT_PAGE_SIZE; + pConfig->nAutomerge = FTS5_DEFAULT_AUTOMERGE; + pConfig->nCrisisMerge = FTS5_DEFAULT_CRISISMERGE; + pConfig->nHashSize = FTS5_DEFAULT_HASHSIZE; + + zSql = sqlite3Fts5Mprintf(&rc, zSelect, pConfig->zDb, pConfig->zName); + if( zSql ){ + rc = sqlite3_prepare_v2(pConfig->db, zSql, -1, &p, 0); + sqlite3_free(zSql); + } + + assert( rc==SQLITE_OK || p==0 ); + if( rc==SQLITE_OK ){ + while( SQLITE_ROW==sqlite3_step(p) ){ + const char *zK = (const char*)sqlite3_column_text(p, 0); + sqlite3_value *pVal = sqlite3_column_value(p, 1); + if( 0==sqlite3_stricmp(zK, "version") ){ + iVersion = sqlite3_value_int(pVal); + }else{ + int bDummy = 0; + sqlite3Fts5ConfigSetValue(pConfig, zK, pVal, &bDummy); + } + } + rc = sqlite3_finalize(p); + } + + if( rc==SQLITE_OK && iVersion!=FTS5_CURRENT_VERSION ){ + rc = SQLITE_ERROR; + if( pConfig->pzErrmsg ){ + assert( 0==*pConfig->pzErrmsg ); + *pConfig->pzErrmsg = sqlite3_mprintf( + "invalid fts5 file format (found %d, expected %d) - run 'rebuild'", + iVersion, FTS5_CURRENT_VERSION + ); + } + } + + if( rc==SQLITE_OK ){ + pConfig->iCookie = iCookie; + } + return rc; +} + +#line 1 "fts5_expr.c" +/* +** 2014 May 31 +** +** The author disclaims copyright to this source code. In place of +** a legal notice, here is a blessing: +** +** May you do good and not evil. +** May you find forgiveness for yourself and forgive others. +** May you share freely, never taking more than you give. +** +****************************************************************************** +** +*/ + + + +/* #include "fts5Int.h" */ +/* #include "fts5parse.h" */ + +/* +** All token types in the generated fts5parse.h file are greater than 0. +*/ +#define FTS5_EOF 0 + +#define FTS5_LARGEST_INT64 (0xffffffff|(((i64)0x7fffffff)<<32)) + +typedef struct Fts5ExprTerm Fts5ExprTerm; + +/* +** Functions generated by lemon from fts5parse.y. +*/ +static void *sqlite3Fts5ParserAlloc(void *(*mallocProc)(u64)); +static void sqlite3Fts5ParserFree(void*, void (*freeProc)(void*)); +static void sqlite3Fts5Parser(void*, int, Fts5Token, Fts5Parse*); +#ifndef NDEBUG +#include <stdio.h> +static void sqlite3Fts5ParserTrace(FILE*, char*); +#endif + + +struct Fts5Expr { + Fts5Index *pIndex; + Fts5Config *pConfig; + Fts5ExprNode *pRoot; + int bDesc; /* Iterate in descending rowid order */ + int nPhrase; /* Number of phrases in expression */ + Fts5ExprPhrase **apExprPhrase; /* Pointers to phrase objects */ +}; + +/* +** eType: +** Expression node type. Always one of: +** +** FTS5_AND (nChild, apChild valid) +** FTS5_OR (nChild, apChild valid) +** FTS5_NOT (nChild, apChild valid) +** FTS5_STRING (pNear valid) +** FTS5_TERM (pNear valid) +*/ +struct Fts5ExprNode { + int eType; /* Node type */ + int bEof; /* True at EOF */ + int bNomatch; /* True if entry is not a match */ + + /* Next method for this node. */ + int (*xNext)(Fts5Expr*, Fts5ExprNode*, int, i64); + + i64 iRowid; /* Current rowid */ + Fts5ExprNearset *pNear; /* For FTS5_STRING - cluster of phrases */ + + /* Child nodes. For a NOT node, this array always contains 2 entries. For + ** AND or OR nodes, it contains 2 or more entries. */ + int nChild; /* Number of child nodes */ + Fts5ExprNode *apChild[1]; /* Array of child nodes */ +}; + +#define Fts5NodeIsString(p) ((p)->eType==FTS5_TERM || (p)->eType==FTS5_STRING) + +/* +** Invoke the xNext method of an Fts5ExprNode object. This macro should be +** used as if it has the same signature as the xNext() methods themselves. +*/ +#define fts5ExprNodeNext(a,b,c,d) (b)->xNext((a), (b), (c), (d)) + +/* +** An instance of the following structure represents a single search term +** or term prefix. +*/ +struct Fts5ExprTerm { + int bPrefix; /* True for a prefix term */ + char *zTerm; /* nul-terminated term */ + Fts5IndexIter *pIter; /* Iterator for this term */ + Fts5ExprTerm *pSynonym; /* Pointer to first in list of synonyms */ +}; + +/* +** A phrase. One or more terms that must appear in a contiguous sequence +** within a document for it to match. +*/ +struct Fts5ExprPhrase { + Fts5ExprNode *pNode; /* FTS5_STRING node this phrase is part of */ + Fts5Buffer poslist; /* Current position list */ + int nTerm; /* Number of entries in aTerm[] */ + Fts5ExprTerm aTerm[1]; /* Terms that make up this phrase */ +}; + +/* +** One or more phrases that must appear within a certain token distance of +** each other within each matching document. +*/ +struct Fts5ExprNearset { + int nNear; /* NEAR parameter */ + Fts5Colset *pColset; /* Columns to search (NULL -> all columns) */ + int nPhrase; /* Number of entries in aPhrase[] array */ + Fts5ExprPhrase *apPhrase[1]; /* Array of phrase pointers */ +}; + + +/* +** Parse context. +*/ +struct Fts5Parse { + Fts5Config *pConfig; + char *zErr; + int rc; + int nPhrase; /* Size of apPhrase array */ + Fts5ExprPhrase **apPhrase; /* Array of all phrases */ + Fts5ExprNode *pExpr; /* Result of a successful parse */ +}; + +static void sqlite3Fts5ParseError(Fts5Parse *pParse, const char *zFmt, ...){ + va_list ap; + va_start(ap, zFmt); + if( pParse->rc==SQLITE_OK ){ + pParse->zErr = sqlite3_vmprintf(zFmt, ap); + pParse->rc = SQLITE_ERROR; + } + va_end(ap); +} + +static int fts5ExprIsspace(char t){ + return t==' ' || t=='\t' || t=='\n' || t=='\r'; +} + +/* +** Read the first token from the nul-terminated string at *pz. +*/ +static int fts5ExprGetToken( + Fts5Parse *pParse, + const char **pz, /* IN/OUT: Pointer into buffer */ + Fts5Token *pToken +){ + const char *z = *pz; + int tok; + + /* Skip past any whitespace */ + while( fts5ExprIsspace(*z) ) z++; + + pToken->p = z; + pToken->n = 1; + switch( *z ){ + case '(': tok = FTS5_LP; break; + case ')': tok = FTS5_RP; break; + case '{': tok = FTS5_LCP; break; + case '}': tok = FTS5_RCP; break; + case ':': tok = FTS5_COLON; break; + case ',': tok = FTS5_COMMA; break; + case '+': tok = FTS5_PLUS; break; + case '*': tok = FTS5_STAR; break; + case '\0': tok = FTS5_EOF; break; + + case '"': { + const char *z2; + tok = FTS5_STRING; + + for(z2=&z[1]; 1; z2++){ + if( z2[0]=='"' ){ + z2++; + if( z2[0]!='"' ) break; + } + if( z2[0]=='\0' ){ + sqlite3Fts5ParseError(pParse, "unterminated string"); + return FTS5_EOF; + } + } + pToken->n = (z2 - z); + break; + } + + default: { + const char *z2; + if( sqlite3Fts5IsBareword(z[0])==0 ){ + sqlite3Fts5ParseError(pParse, "fts5: syntax error near \"%.1s\"", z); + return FTS5_EOF; + } + tok = FTS5_STRING; + for(z2=&z[1]; sqlite3Fts5IsBareword(*z2); z2++); + pToken->n = (z2 - z); + if( pToken->n==2 && memcmp(pToken->p, "OR", 2)==0 ) tok = FTS5_OR; + if( pToken->n==3 && memcmp(pToken->p, "NOT", 3)==0 ) tok = FTS5_NOT; + if( pToken->n==3 && memcmp(pToken->p, "AND", 3)==0 ) tok = FTS5_AND; + break; + } + } + + *pz = &pToken->p[pToken->n]; + return tok; +} + +static void *fts5ParseAlloc(u64 t){ return sqlite3_malloc((int)t); } +static void fts5ParseFree(void *p){ sqlite3_free(p); } + +static int sqlite3Fts5ExprNew( + Fts5Config *pConfig, /* FTS5 Configuration */ + const char *zExpr, /* Expression text */ + Fts5Expr **ppNew, + char **pzErr +){ + Fts5Parse sParse; + Fts5Token token; + const char *z = zExpr; + int t; /* Next token type */ + void *pEngine; + Fts5Expr *pNew; + + *ppNew = 0; + *pzErr = 0; + memset(&sParse, 0, sizeof(sParse)); + pEngine = sqlite3Fts5ParserAlloc(fts5ParseAlloc); + if( pEngine==0 ){ return SQLITE_NOMEM; } + sParse.pConfig = pConfig; + + do { + t = fts5ExprGetToken(&sParse, &z, &token); + sqlite3Fts5Parser(pEngine, t, token, &sParse); + }while( sParse.rc==SQLITE_OK && t!=FTS5_EOF ); + sqlite3Fts5ParserFree(pEngine, fts5ParseFree); + + assert( sParse.rc!=SQLITE_OK || sParse.zErr==0 ); + if( sParse.rc==SQLITE_OK ){ + *ppNew = pNew = sqlite3_malloc(sizeof(Fts5Expr)); + if( pNew==0 ){ + sParse.rc = SQLITE_NOMEM; + sqlite3Fts5ParseNodeFree(sParse.pExpr); + }else{ + if( !sParse.pExpr ){ + const int nByte = sizeof(Fts5ExprNode); + pNew->pRoot = (Fts5ExprNode*)sqlite3Fts5MallocZero(&sParse.rc, nByte); + if( pNew->pRoot ){ + pNew->pRoot->bEof = 1; + } + }else{ + pNew->pRoot = sParse.pExpr; + } + pNew->pIndex = 0; + pNew->pConfig = pConfig; + pNew->apExprPhrase = sParse.apPhrase; + pNew->nPhrase = sParse.nPhrase; + sParse.apPhrase = 0; + } + } + + sqlite3_free(sParse.apPhrase); + *pzErr = sParse.zErr; + return sParse.rc; +} + +/* +** Free the expression node object passed as the only argument. +*/ +static void sqlite3Fts5ParseNodeFree(Fts5ExprNode *p){ + if( p ){ + int i; + for(i=0; i<p->nChild; i++){ + sqlite3Fts5ParseNodeFree(p->apChild[i]); + } + sqlite3Fts5ParseNearsetFree(p->pNear); + sqlite3_free(p); + } +} + +/* +** Free the expression object passed as the only argument. +*/ +static void sqlite3Fts5ExprFree(Fts5Expr *p){ + if( p ){ + sqlite3Fts5ParseNodeFree(p->pRoot); + sqlite3_free(p->apExprPhrase); + sqlite3_free(p); + } +} + +/* +** Argument pTerm must be a synonym iterator. Return the current rowid +** that it points to. +*/ +static i64 fts5ExprSynonymRowid(Fts5ExprTerm *pTerm, int bDesc, int *pbEof){ + i64 iRet = 0; + int bRetValid = 0; + Fts5ExprTerm *p; + + assert( pTerm->pSynonym ); + assert( bDesc==0 || bDesc==1 ); + for(p=pTerm; p; p=p->pSynonym){ + if( 0==sqlite3Fts5IterEof(p->pIter) ){ + i64 iRowid = p->pIter->iRowid; + if( bRetValid==0 || (bDesc!=(iRowid<iRet)) ){ + iRet = iRowid; + bRetValid = 1; + } + } + } + + if( pbEof && bRetValid==0 ) *pbEof = 1; + return iRet; +} + +/* +** Argument pTerm must be a synonym iterator. +*/ +static int fts5ExprSynonymList( + Fts5ExprTerm *pTerm, + i64 iRowid, + Fts5Buffer *pBuf, /* Use this buffer for space if required */ + u8 **pa, int *pn +){ + Fts5PoslistReader aStatic[4]; + Fts5PoslistReader *aIter = aStatic; + int nIter = 0; + int nAlloc = 4; + int rc = SQLITE_OK; + Fts5ExprTerm *p; + + assert( pTerm->pSynonym ); + for(p=pTerm; p; p=p->pSynonym){ + Fts5IndexIter *pIter = p->pIter; + if( sqlite3Fts5IterEof(pIter)==0 && pIter->iRowid==iRowid ){ + if( pIter->nData==0 ) continue; + if( nIter==nAlloc ){ + int nByte = sizeof(Fts5PoslistReader) * nAlloc * 2; + Fts5PoslistReader *aNew = (Fts5PoslistReader*)sqlite3_malloc(nByte); + if( aNew==0 ){ + rc = SQLITE_NOMEM; + goto synonym_poslist_out; + } + memcpy(aNew, aIter, sizeof(Fts5PoslistReader) * nIter); + nAlloc = nAlloc*2; + if( aIter!=aStatic ) sqlite3_free(aIter); + aIter = aNew; + } + sqlite3Fts5PoslistReaderInit(pIter->pData, pIter->nData, &aIter[nIter]); + assert( aIter[nIter].bEof==0 ); + nIter++; + } + } + + if( nIter==1 ){ + *pa = (u8*)aIter[0].a; + *pn = aIter[0].n; + }else{ + Fts5PoslistWriter writer = {0}; + i64 iPrev = -1; + fts5BufferZero(pBuf); + while( 1 ){ + int i; + i64 iMin = FTS5_LARGEST_INT64; + for(i=0; i<nIter; i++){ + if( aIter[i].bEof==0 ){ + if( aIter[i].iPos==iPrev ){ + if( sqlite3Fts5PoslistReaderNext(&aIter[i]) ) continue; + } + if( aIter[i].iPos<iMin ){ + iMin = aIter[i].iPos; + } + } + } + if( iMin==FTS5_LARGEST_INT64 || rc!=SQLITE_OK ) break; + rc = sqlite3Fts5PoslistWriterAppend(pBuf, &writer, iMin); + iPrev = iMin; + } + if( rc==SQLITE_OK ){ + *pa = pBuf->p; + *pn = pBuf->n; + } + } + + synonym_poslist_out: + if( aIter!=aStatic ) sqlite3_free(aIter); + return rc; +} + + +/* +** All individual term iterators in pPhrase are guaranteed to be valid and +** pointing to the same rowid when this function is called. This function +** checks if the current rowid really is a match, and if so populates +** the pPhrase->poslist buffer accordingly. Output parameter *pbMatch +** is set to true if this is really a match, or false otherwise. +** +** SQLITE_OK is returned if an error occurs, or an SQLite error code +** otherwise. It is not considered an error code if the current rowid is +** not a match. +*/ +static int fts5ExprPhraseIsMatch( + Fts5ExprNode *pNode, /* Node pPhrase belongs to */ + Fts5ExprPhrase *pPhrase, /* Phrase object to initialize */ + int *pbMatch /* OUT: Set to true if really a match */ +){ + Fts5PoslistWriter writer = {0}; + Fts5PoslistReader aStatic[4]; + Fts5PoslistReader *aIter = aStatic; + int i; + int rc = SQLITE_OK; + + fts5BufferZero(&pPhrase->poslist); + + /* If the aStatic[] array is not large enough, allocate a large array + ** using sqlite3_malloc(). This approach could be improved upon. */ + if( pPhrase->nTerm>ArraySize(aStatic) ){ + int nByte = sizeof(Fts5PoslistReader) * pPhrase->nTerm; + aIter = (Fts5PoslistReader*)sqlite3_malloc(nByte); + if( !aIter ) return SQLITE_NOMEM; + } + memset(aIter, 0, sizeof(Fts5PoslistReader) * pPhrase->nTerm); + + /* Initialize a term iterator for each term in the phrase */ + for(i=0; i<pPhrase->nTerm; i++){ + Fts5ExprTerm *pTerm = &pPhrase->aTerm[i]; + int n = 0; + int bFlag = 0; + u8 *a = 0; + if( pTerm->pSynonym ){ + Fts5Buffer buf = {0, 0, 0}; + rc = fts5ExprSynonymList(pTerm, pNode->iRowid, &buf, &a, &n); + if( rc ){ + sqlite3_free(a); + goto ismatch_out; + } + if( a==buf.p ) bFlag = 1; + }else{ + a = (u8*)pTerm->pIter->pData; + n = pTerm->pIter->nData; + } + sqlite3Fts5PoslistReaderInit(a, n, &aIter[i]); + aIter[i].bFlag = (u8)bFlag; + if( aIter[i].bEof ) goto ismatch_out; + } + + while( 1 ){ + int bMatch; + i64 iPos = aIter[0].iPos; + do { + bMatch = 1; + for(i=0; i<pPhrase->nTerm; i++){ + Fts5PoslistReader *pPos = &aIter[i]; + i64 iAdj = iPos + i; + if( pPos->iPos!=iAdj ){ + bMatch = 0; + while( pPos->iPos<iAdj ){ + if( sqlite3Fts5PoslistReaderNext(pPos) ) goto ismatch_out; + } + if( pPos->iPos>iAdj ) iPos = pPos->iPos-i; + } + } + }while( bMatch==0 ); + + /* Append position iPos to the output */ + rc = sqlite3Fts5PoslistWriterAppend(&pPhrase->poslist, &writer, iPos); + if( rc!=SQLITE_OK ) goto ismatch_out; + + for(i=0; i<pPhrase->nTerm; i++){ + if( sqlite3Fts5PoslistReaderNext(&aIter[i]) ) goto ismatch_out; + } + } + + ismatch_out: + *pbMatch = (pPhrase->poslist.n>0); + for(i=0; i<pPhrase->nTerm; i++){ + if( aIter[i].bFlag ) sqlite3_free((u8*)aIter[i].a); + } + if( aIter!=aStatic ) sqlite3_free(aIter); + return rc; +} + +typedef struct Fts5LookaheadReader Fts5LookaheadReader; +struct Fts5LookaheadReader { + const u8 *a; /* Buffer containing position list */ + int n; /* Size of buffer a[] in bytes */ + int i; /* Current offset in position list */ + i64 iPos; /* Current position */ + i64 iLookahead; /* Next position */ +}; + +#define FTS5_LOOKAHEAD_EOF (((i64)1) << 62) + +static int fts5LookaheadReaderNext(Fts5LookaheadReader *p){ + p->iPos = p->iLookahead; + if( sqlite3Fts5PoslistNext64(p->a, p->n, &p->i, &p->iLookahead) ){ + p->iLookahead = FTS5_LOOKAHEAD_EOF; + } + return (p->iPos==FTS5_LOOKAHEAD_EOF); +} + +static int fts5LookaheadReaderInit( + const u8 *a, int n, /* Buffer to read position list from */ + Fts5LookaheadReader *p /* Iterator object to initialize */ +){ + memset(p, 0, sizeof(Fts5LookaheadReader)); + p->a = a; + p->n = n; + fts5LookaheadReaderNext(p); + return fts5LookaheadReaderNext(p); +} + +typedef struct Fts5NearTrimmer Fts5NearTrimmer; +struct Fts5NearTrimmer { + Fts5LookaheadReader reader; /* Input iterator */ + Fts5PoslistWriter writer; /* Writer context */ + Fts5Buffer *pOut; /* Output poslist */ +}; + +/* +** The near-set object passed as the first argument contains more than +** one phrase. All phrases currently point to the same row. The +** Fts5ExprPhrase.poslist buffers are populated accordingly. This function +** tests if the current row contains instances of each phrase sufficiently +** close together to meet the NEAR constraint. Non-zero is returned if it +** does, or zero otherwise. +** +** If in/out parameter (*pRc) is set to other than SQLITE_OK when this +** function is called, it is a no-op. Or, if an error (e.g. SQLITE_NOMEM) +** occurs within this function (*pRc) is set accordingly before returning. +** The return value is undefined in both these cases. +** +** If no error occurs and non-zero (a match) is returned, the position-list +** of each phrase object is edited to contain only those entries that +** meet the constraint before returning. +*/ +static int fts5ExprNearIsMatch(int *pRc, Fts5ExprNearset *pNear){ + Fts5NearTrimmer aStatic[4]; + Fts5NearTrimmer *a = aStatic; + Fts5ExprPhrase **apPhrase = pNear->apPhrase; + + int i; + int rc = *pRc; + int bMatch; + + assert( pNear->nPhrase>1 ); + + /* If the aStatic[] array is not large enough, allocate a large array + ** using sqlite3_malloc(). This approach could be improved upon. */ + if( pNear->nPhrase>ArraySize(aStatic) ){ + int nByte = sizeof(Fts5NearTrimmer) * pNear->nPhrase; + a = (Fts5NearTrimmer*)sqlite3Fts5MallocZero(&rc, nByte); + }else{ + memset(aStatic, 0, sizeof(aStatic)); + } + if( rc!=SQLITE_OK ){ + *pRc = rc; + return 0; + } + + /* Initialize a lookahead iterator for each phrase. After passing the + ** buffer and buffer size to the lookaside-reader init function, zero + ** the phrase poslist buffer. The new poslist for the phrase (containing + ** the same entries as the original with some entries removed on account + ** of the NEAR constraint) is written over the original even as it is + ** being read. This is safe as the entries for the new poslist are a + ** subset of the old, so it is not possible for data yet to be read to + ** be overwritten. */ + for(i=0; i<pNear->nPhrase; i++){ + Fts5Buffer *pPoslist = &apPhrase[i]->poslist; + fts5LookaheadReaderInit(pPoslist->p, pPoslist->n, &a[i].reader); + pPoslist->n = 0; + a[i].pOut = pPoslist; + } + + while( 1 ){ + int iAdv; + i64 iMin; + i64 iMax; + + /* This block advances the phrase iterators until they point to a set of + ** entries that together comprise a match. */ + iMax = a[0].reader.iPos; + do { + bMatch = 1; + for(i=0; i<pNear->nPhrase; i++){ + Fts5LookaheadReader *pPos = &a[i].reader; + iMin = iMax - pNear->apPhrase[i]->nTerm - pNear->nNear; + if( pPos->iPos<iMin || pPos->iPos>iMax ){ + bMatch = 0; + while( pPos->iPos<iMin ){ + if( fts5LookaheadReaderNext(pPos) ) goto ismatch_out; + } + if( pPos->iPos>iMax ) iMax = pPos->iPos; + } + } + }while( bMatch==0 ); + + /* Add an entry to each output position list */ + for(i=0; i<pNear->nPhrase; i++){ + i64 iPos = a[i].reader.iPos; + Fts5PoslistWriter *pWriter = &a[i].writer; + if( a[i].pOut->n==0 || iPos!=pWriter->iPrev ){ + sqlite3Fts5PoslistWriterAppend(a[i].pOut, pWriter, iPos); + } + } + + iAdv = 0; + iMin = a[0].reader.iLookahead; + for(i=0; i<pNear->nPhrase; i++){ + if( a[i].reader.iLookahead < iMin ){ + iMin = a[i].reader.iLookahead; + iAdv = i; + } + } + if( fts5LookaheadReaderNext(&a[iAdv].reader) ) goto ismatch_out; + } + + ismatch_out: { + int bRet = a[0].pOut->n>0; + *pRc = rc; + if( a!=aStatic ) sqlite3_free(a); + return bRet; + } +} + +/* +** Advance iterator pIter until it points to a value equal to or laster +** than the initial value of *piLast. If this means the iterator points +** to a value laster than *piLast, update *piLast to the new lastest value. +** +** If the iterator reaches EOF, set *pbEof to true before returning. If +** an error occurs, set *pRc to an error code. If either *pbEof or *pRc +** are set, return a non-zero value. Otherwise, return zero. +*/ +static int fts5ExprAdvanceto( + Fts5IndexIter *pIter, /* Iterator to advance */ + int bDesc, /* True if iterator is "rowid DESC" */ + i64 *piLast, /* IN/OUT: Lastest rowid seen so far */ + int *pRc, /* OUT: Error code */ + int *pbEof /* OUT: Set to true if EOF */ +){ + i64 iLast = *piLast; + i64 iRowid; + + iRowid = pIter->iRowid; + if( (bDesc==0 && iLast>iRowid) || (bDesc && iLast<iRowid) ){ + int rc = sqlite3Fts5IterNextFrom(pIter, iLast); + if( rc || sqlite3Fts5IterEof(pIter) ){ + *pRc = rc; + *pbEof = 1; + return 1; + } + iRowid = pIter->iRowid; + assert( (bDesc==0 && iRowid>=iLast) || (bDesc==1 && iRowid<=iLast) ); + } + *piLast = iRowid; + + return 0; +} + +static int fts5ExprSynonymAdvanceto( + Fts5ExprTerm *pTerm, /* Term iterator to advance */ + int bDesc, /* True if iterator is "rowid DESC" */ + i64 *piLast, /* IN/OUT: Lastest rowid seen so far */ + int *pRc /* OUT: Error code */ +){ + int rc = SQLITE_OK; + i64 iLast = *piLast; + Fts5ExprTerm *p; + int bEof = 0; + + for(p=pTerm; rc==SQLITE_OK && p; p=p->pSynonym){ + if( sqlite3Fts5IterEof(p->pIter)==0 ){ + i64 iRowid = p->pIter->iRowid; + if( (bDesc==0 && iLast>iRowid) || (bDesc && iLast<iRowid) ){ + rc = sqlite3Fts5IterNextFrom(p->pIter, iLast); + } + } + } + + if( rc!=SQLITE_OK ){ + *pRc = rc; + bEof = 1; + }else{ + *piLast = fts5ExprSynonymRowid(pTerm, bDesc, &bEof); + } + return bEof; +} + + +static int fts5ExprNearTest( + int *pRc, + Fts5Expr *pExpr, /* Expression that pNear is a part of */ + Fts5ExprNode *pNode /* The "NEAR" node (FTS5_STRING) */ +){ + Fts5ExprNearset *pNear = pNode->pNear; + int rc = *pRc; + + if( pExpr->pConfig->eDetail!=FTS5_DETAIL_FULL ){ + Fts5ExprTerm *pTerm; + Fts5ExprPhrase *pPhrase = pNear->apPhrase[0]; + pPhrase->poslist.n = 0; + for(pTerm=&pPhrase->aTerm[0]; pTerm; pTerm=pTerm->pSynonym){ + Fts5IndexIter *pIter = pTerm->pIter; + if( sqlite3Fts5IterEof(pIter)==0 ){ + if( pIter->iRowid==pNode->iRowid && pIter->nData>0 ){ + pPhrase->poslist.n = 1; + } + } + } + return pPhrase->poslist.n; + }else{ + int i; + + /* Check that each phrase in the nearset matches the current row. + ** Populate the pPhrase->poslist buffers at the same time. If any + ** phrase is not a match, break out of the loop early. */ + for(i=0; rc==SQLITE_OK && i<pNear->nPhrase; i++){ + Fts5ExprPhrase *pPhrase = pNear->apPhrase[i]; + if( pPhrase->nTerm>1 || pPhrase->aTerm[0].pSynonym || pNear->pColset ){ + int bMatch = 0; + rc = fts5ExprPhraseIsMatch(pNode, pPhrase, &bMatch); + if( bMatch==0 ) break; + }else{ + Fts5IndexIter *pIter = pPhrase->aTerm[0].pIter; + fts5BufferSet(&rc, &pPhrase->poslist, pIter->nData, pIter->pData); + } + } + + *pRc = rc; + if( i==pNear->nPhrase && (i==1 || fts5ExprNearIsMatch(pRc, pNear)) ){ + return 1; + } + return 0; + } +} + + +/* +** Initialize all term iterators in the pNear object. If any term is found +** to match no documents at all, return immediately without initializing any +** further iterators. +*/ +static int fts5ExprNearInitAll( + Fts5Expr *pExpr, + Fts5ExprNode *pNode +){ + Fts5ExprNearset *pNear = pNode->pNear; + int i, j; + int rc = SQLITE_OK; + + assert( pNode->bNomatch==0 ); + for(i=0; rc==SQLITE_OK && i<pNear->nPhrase; i++){ + Fts5ExprPhrase *pPhrase = pNear->apPhrase[i]; + for(j=0; j<pPhrase->nTerm; j++){ + Fts5ExprTerm *pTerm = &pPhrase->aTerm[j]; + Fts5ExprTerm *p; + int bEof = 1; + + for(p=pTerm; p && rc==SQLITE_OK; p=p->pSynonym){ + if( p->pIter ){ + sqlite3Fts5IterClose(p->pIter); + p->pIter = 0; + } + rc = sqlite3Fts5IndexQuery( + pExpr->pIndex, p->zTerm, (int)strlen(p->zTerm), + (pTerm->bPrefix ? FTS5INDEX_QUERY_PREFIX : 0) | + (pExpr->bDesc ? FTS5INDEX_QUERY_DESC : 0), + pNear->pColset, + &p->pIter + ); + assert( rc==SQLITE_OK || p->pIter==0 ); + if( p->pIter && 0==sqlite3Fts5IterEof(p->pIter) ){ + bEof = 0; + } + } + + if( bEof ){ + pNode->bEof = 1; + return rc; + } + } + } + + return rc; +} + +/* +** If pExpr is an ASC iterator, this function returns a value with the +** same sign as: +** +** (iLhs - iRhs) +** +** Otherwise, if this is a DESC iterator, the opposite is returned: +** +** (iRhs - iLhs) +*/ +static int fts5RowidCmp( + Fts5Expr *pExpr, + i64 iLhs, + i64 iRhs +){ + assert( pExpr->bDesc==0 || pExpr->bDesc==1 ); + if( pExpr->bDesc==0 ){ + if( iLhs<iRhs ) return -1; + return (iLhs > iRhs); + }else{ + if( iLhs>iRhs ) return -1; + return (iLhs < iRhs); + } +} + +static void fts5ExprSetEof(Fts5ExprNode *pNode){ + int i; + pNode->bEof = 1; + pNode->bNomatch = 0; + for(i=0; i<pNode->nChild; i++){ + fts5ExprSetEof(pNode->apChild[i]); + } +} + +static void fts5ExprNodeZeroPoslist(Fts5ExprNode *pNode){ + if( pNode->eType==FTS5_STRING || pNode->eType==FTS5_TERM ){ + Fts5ExprNearset *pNear = pNode->pNear; + int i; + for(i=0; i<pNear->nPhrase; i++){ + Fts5ExprPhrase *pPhrase = pNear->apPhrase[i]; + pPhrase->poslist.n = 0; + } + }else{ + int i; + for(i=0; i<pNode->nChild; i++){ + fts5ExprNodeZeroPoslist(pNode->apChild[i]); + } + } +} + + + +/* +** Compare the values currently indicated by the two nodes as follows: +** +** res = (*p1) - (*p2) +** +** Nodes that point to values that come later in the iteration order are +** considered to be larger. Nodes at EOF are the largest of all. +** +** This means that if the iteration order is ASC, then numerically larger +** rowids are considered larger. Or if it is the default DESC, numerically +** smaller rowids are larger. +*/ +static int fts5NodeCompare( + Fts5Expr *pExpr, + Fts5ExprNode *p1, + Fts5ExprNode *p2 +){ + if( p2->bEof ) return -1; + if( p1->bEof ) return +1; + return fts5RowidCmp(pExpr, p1->iRowid, p2->iRowid); +} + +/* +** All individual term iterators in pNear are guaranteed to be valid when +** this function is called. This function checks if all term iterators +** point to the same rowid, and if not, advances them until they do. +** If an EOF is reached before this happens, *pbEof is set to true before +** returning. +** +** SQLITE_OK is returned if an error occurs, or an SQLite error code +** otherwise. It is not considered an error code if an iterator reaches +** EOF. +*/ +static int fts5ExprNodeTest_STRING( + Fts5Expr *pExpr, /* Expression pPhrase belongs to */ + Fts5ExprNode *pNode +){ + Fts5ExprNearset *pNear = pNode->pNear; + Fts5ExprPhrase *pLeft = pNear->apPhrase[0]; + int rc = SQLITE_OK; + i64 iLast; /* Lastest rowid any iterator points to */ + int i, j; /* Phrase and token index, respectively */ + int bMatch; /* True if all terms are at the same rowid */ + const int bDesc = pExpr->bDesc; + + /* Check that this node should not be FTS5_TERM */ + assert( pNear->nPhrase>1 + || pNear->apPhrase[0]->nTerm>1 + || pNear->apPhrase[0]->aTerm[0].pSynonym + ); + + /* Initialize iLast, the "lastest" rowid any iterator points to. If the + ** iterator skips through rowids in the default ascending order, this means + ** the maximum rowid. Or, if the iterator is "ORDER BY rowid DESC", then it + ** means the minimum rowid. */ + if( pLeft->aTerm[0].pSynonym ){ + iLast = fts5ExprSynonymRowid(&pLeft->aTerm[0], bDesc, 0); + }else{ + iLast = pLeft->aTerm[0].pIter->iRowid; + } + + do { + bMatch = 1; + for(i=0; i<pNear->nPhrase; i++){ + Fts5ExprPhrase *pPhrase = pNear->apPhrase[i]; + for(j=0; j<pPhrase->nTerm; j++){ + Fts5ExprTerm *pTerm = &pPhrase->aTerm[j]; + if( pTerm->pSynonym ){ + i64 iRowid = fts5ExprSynonymRowid(pTerm, bDesc, 0); + if( iRowid==iLast ) continue; + bMatch = 0; + if( fts5ExprSynonymAdvanceto(pTerm, bDesc, &iLast, &rc) ){ + pNode->bNomatch = 0; + pNode->bEof = 1; + return rc; + } + }else{ + Fts5IndexIter *pIter = pPhrase->aTerm[j].pIter; + if( pIter->iRowid==iLast ) continue; + bMatch = 0; + if( fts5ExprAdvanceto(pIter, bDesc, &iLast, &rc, &pNode->bEof) ){ + return rc; + } + } + } + } + }while( bMatch==0 ); + + pNode->iRowid = iLast; + pNode->bNomatch = ((0==fts5ExprNearTest(&rc, pExpr, pNode)) && rc==SQLITE_OK); + assert( pNode->bEof==0 || pNode->bNomatch==0 ); + + return rc; +} + +/* +** Advance the first term iterator in the first phrase of pNear. Set output +** variable *pbEof to true if it reaches EOF or if an error occurs. +** +** Return SQLITE_OK if successful, or an SQLite error code if an error +** occurs. +*/ +static int fts5ExprNodeNext_STRING( + Fts5Expr *pExpr, /* Expression pPhrase belongs to */ + Fts5ExprNode *pNode, /* FTS5_STRING or FTS5_TERM node */ + int bFromValid, + i64 iFrom +){ + Fts5ExprTerm *pTerm = &pNode->pNear->apPhrase[0]->aTerm[0]; + int rc = SQLITE_OK; + + pNode->bNomatch = 0; + if( pTerm->pSynonym ){ + int bEof = 1; + Fts5ExprTerm *p; + + /* Find the firstest rowid any synonym points to. */ + i64 iRowid = fts5ExprSynonymRowid(pTerm, pExpr->bDesc, 0); + + /* Advance each iterator that currently points to iRowid. Or, if iFrom + ** is valid - each iterator that points to a rowid before iFrom. */ + for(p=pTerm; p; p=p->pSynonym){ + if( sqlite3Fts5IterEof(p->pIter)==0 ){ + i64 ii = p->pIter->iRowid; + if( ii==iRowid + || (bFromValid && ii!=iFrom && (ii>iFrom)==pExpr->bDesc) + ){ + if( bFromValid ){ + rc = sqlite3Fts5IterNextFrom(p->pIter, iFrom); + }else{ + rc = sqlite3Fts5IterNext(p->pIter); + } + if( rc!=SQLITE_OK ) break; + if( sqlite3Fts5IterEof(p->pIter)==0 ){ + bEof = 0; + } + }else{ + bEof = 0; + } + } + } + + /* Set the EOF flag if either all synonym iterators are at EOF or an + ** error has occurred. */ + pNode->bEof = (rc || bEof); + }else{ + Fts5IndexIter *pIter = pTerm->pIter; + + assert( Fts5NodeIsString(pNode) ); + if( bFromValid ){ + rc = sqlite3Fts5IterNextFrom(pIter, iFrom); + }else{ + rc = sqlite3Fts5IterNext(pIter); + } + + pNode->bEof = (rc || sqlite3Fts5IterEof(pIter)); + } + + if( pNode->bEof==0 ){ + assert( rc==SQLITE_OK ); + rc = fts5ExprNodeTest_STRING(pExpr, pNode); + } + + return rc; +} + + +static int fts5ExprNodeTest_TERM( + Fts5Expr *pExpr, /* Expression that pNear is a part of */ + Fts5ExprNode *pNode /* The "NEAR" node (FTS5_TERM) */ +){ + /* As this "NEAR" object is actually a single phrase that consists + ** of a single term only, grab pointers into the poslist managed by the + ** fts5_index.c iterator object. This is much faster than synthesizing + ** a new poslist the way we have to for more complicated phrase or NEAR + ** expressions. */ + Fts5ExprPhrase *pPhrase = pNode->pNear->apPhrase[0]; + Fts5IndexIter *pIter = pPhrase->aTerm[0].pIter; + + assert( pNode->eType==FTS5_TERM ); + assert( pNode->pNear->nPhrase==1 && pPhrase->nTerm==1 ); + assert( pPhrase->aTerm[0].pSynonym==0 ); + + pPhrase->poslist.n = pIter->nData; + if( pExpr->pConfig->eDetail==FTS5_DETAIL_FULL ){ + pPhrase->poslist.p = (u8*)pIter->pData; + } + pNode->iRowid = pIter->iRowid; + pNode->bNomatch = (pPhrase->poslist.n==0); + return SQLITE_OK; +} + +/* +** xNext() method for a node of type FTS5_TERM. +*/ +static int fts5ExprNodeNext_TERM( + Fts5Expr *pExpr, + Fts5ExprNode *pNode, + int bFromValid, + i64 iFrom +){ + int rc; + Fts5IndexIter *pIter = pNode->pNear->apPhrase[0]->aTerm[0].pIter; + + assert( pNode->bEof==0 ); + if( bFromValid ){ + rc = sqlite3Fts5IterNextFrom(pIter, iFrom); + }else{ + rc = sqlite3Fts5IterNext(pIter); + } + if( rc==SQLITE_OK && sqlite3Fts5IterEof(pIter)==0 ){ + rc = fts5ExprNodeTest_TERM(pExpr, pNode); + }else{ + pNode->bEof = 1; + pNode->bNomatch = 0; + } + return rc; +} + +static void fts5ExprNodeTest_OR( + Fts5Expr *pExpr, /* Expression of which pNode is a part */ + Fts5ExprNode *pNode /* Expression node to test */ +){ + Fts5ExprNode *pNext = pNode->apChild[0]; + int i; + + for(i=1; i<pNode->nChild; i++){ + Fts5ExprNode *pChild = pNode->apChild[i]; + int cmp = fts5NodeCompare(pExpr, pNext, pChild); + if( cmp>0 || (cmp==0 && pChild->bNomatch==0) ){ + pNext = pChild; + } + } + pNode->iRowid = pNext->iRowid; + pNode->bEof = pNext->bEof; + pNode->bNomatch = pNext->bNomatch; +} + +static int fts5ExprNodeNext_OR( + Fts5Expr *pExpr, + Fts5ExprNode *pNode, + int bFromValid, + i64 iFrom +){ + int i; + i64 iLast = pNode->iRowid; + + for(i=0; i<pNode->nChild; i++){ + Fts5ExprNode *p1 = pNode->apChild[i]; + assert( p1->bEof || fts5RowidCmp(pExpr, p1->iRowid, iLast)>=0 ); + if( p1->bEof==0 ){ + if( (p1->iRowid==iLast) + || (bFromValid && fts5RowidCmp(pExpr, p1->iRowid, iFrom)<0) + ){ + int rc = fts5ExprNodeNext(pExpr, p1, bFromValid, iFrom); + if( rc!=SQLITE_OK ) return rc; + } + } + } + + fts5ExprNodeTest_OR(pExpr, pNode); + return SQLITE_OK; +} + +/* +** Argument pNode is an FTS5_AND node. +*/ +static int fts5ExprNodeTest_AND( + Fts5Expr *pExpr, /* Expression pPhrase belongs to */ + Fts5ExprNode *pAnd /* FTS5_AND node to advance */ +){ + int iChild; + i64 iLast = pAnd->iRowid; + int rc = SQLITE_OK; + int bMatch; + + assert( pAnd->bEof==0 ); + do { + pAnd->bNomatch = 0; + bMatch = 1; + for(iChild=0; iChild<pAnd->nChild; iChild++){ + Fts5ExprNode *pChild = pAnd->apChild[iChild]; + int cmp = fts5RowidCmp(pExpr, iLast, pChild->iRowid); + if( cmp>0 ){ + /* Advance pChild until it points to iLast or laster */ + rc = fts5ExprNodeNext(pExpr, pChild, 1, iLast); + if( rc!=SQLITE_OK ) return rc; + } + + /* If the child node is now at EOF, so is the parent AND node. Otherwise, + ** the child node is guaranteed to have advanced at least as far as + ** rowid iLast. So if it is not at exactly iLast, pChild->iRowid is the + ** new lastest rowid seen so far. */ + assert( pChild->bEof || fts5RowidCmp(pExpr, iLast, pChild->iRowid)<=0 ); + if( pChild->bEof ){ + fts5ExprSetEof(pAnd); + bMatch = 1; + break; + }else if( iLast!=pChild->iRowid ){ + bMatch = 0; + iLast = pChild->iRowid; + } + + if( pChild->bNomatch ){ + pAnd->bNomatch = 1; + } + } + }while( bMatch==0 ); + + if( pAnd->bNomatch && pAnd!=pExpr->pRoot ){ + fts5ExprNodeZeroPoslist(pAnd); + } + pAnd->iRowid = iLast; + return SQLITE_OK; +} + +static int fts5ExprNodeNext_AND( + Fts5Expr *pExpr, + Fts5ExprNode *pNode, + int bFromValid, + i64 iFrom +){ + int rc = fts5ExprNodeNext(pExpr, pNode->apChild[0], bFromValid, iFrom); + if( rc==SQLITE_OK ){ + rc = fts5ExprNodeTest_AND(pExpr, pNode); + } + return rc; +} + +static int fts5ExprNodeTest_NOT( + Fts5Expr *pExpr, /* Expression pPhrase belongs to */ + Fts5ExprNode *pNode /* FTS5_NOT node to advance */ +){ + int rc = SQLITE_OK; + Fts5ExprNode *p1 = pNode->apChild[0]; + Fts5ExprNode *p2 = pNode->apChild[1]; + assert( pNode->nChild==2 ); + + while( rc==SQLITE_OK && p1->bEof==0 ){ + int cmp = fts5NodeCompare(pExpr, p1, p2); + if( cmp>0 ){ + rc = fts5ExprNodeNext(pExpr, p2, 1, p1->iRowid); + cmp = fts5NodeCompare(pExpr, p1, p2); + } + assert( rc!=SQLITE_OK || cmp<=0 ); + if( cmp || p2->bNomatch ) break; + rc = fts5ExprNodeNext(pExpr, p1, 0, 0); + } + pNode->bEof = p1->bEof; + pNode->bNomatch = p1->bNomatch; + pNode->iRowid = p1->iRowid; + if( p1->bEof ){ + fts5ExprNodeZeroPoslist(p2); + } + return rc; +} + +static int fts5ExprNodeNext_NOT( + Fts5Expr *pExpr, + Fts5ExprNode *pNode, + int bFromValid, + i64 iFrom +){ + int rc = fts5ExprNodeNext(pExpr, pNode->apChild[0], bFromValid, iFrom); + if( rc==SQLITE_OK ){ + rc = fts5ExprNodeTest_NOT(pExpr, pNode); + } + return rc; +} + +/* +** If pNode currently points to a match, this function returns SQLITE_OK +** without modifying it. Otherwise, pNode is advanced until it does point +** to a match or EOF is reached. +*/ +static int fts5ExprNodeTest( + Fts5Expr *pExpr, /* Expression of which pNode is a part */ + Fts5ExprNode *pNode /* Expression node to test */ +){ + int rc = SQLITE_OK; + if( pNode->bEof==0 ){ + switch( pNode->eType ){ + + case FTS5_STRING: { + rc = fts5ExprNodeTest_STRING(pExpr, pNode); + break; + } + + case FTS5_TERM: { + rc = fts5ExprNodeTest_TERM(pExpr, pNode); + break; + } + + case FTS5_AND: { + rc = fts5ExprNodeTest_AND(pExpr, pNode); + break; + } + + case FTS5_OR: { + fts5ExprNodeTest_OR(pExpr, pNode); + break; + } + + default: assert( pNode->eType==FTS5_NOT ); { + rc = fts5ExprNodeTest_NOT(pExpr, pNode); + break; + } + } + } + return rc; +} + + +/* +** Set node pNode, which is part of expression pExpr, to point to the first +** match. If there are no matches, set the Node.bEof flag to indicate EOF. +** +** Return an SQLite error code if an error occurs, or SQLITE_OK otherwise. +** It is not an error if there are no matches. +*/ +static int fts5ExprNodeFirst(Fts5Expr *pExpr, Fts5ExprNode *pNode){ + int rc = SQLITE_OK; + pNode->bEof = 0; + pNode->bNomatch = 0; + + if( Fts5NodeIsString(pNode) ){ + /* Initialize all term iterators in the NEAR object. */ + rc = fts5ExprNearInitAll(pExpr, pNode); + }else{ + int i; + int nEof = 0; + for(i=0; i<pNode->nChild && rc==SQLITE_OK; i++){ + Fts5ExprNode *pChild = pNode->apChild[i]; + rc = fts5ExprNodeFirst(pExpr, pNode->apChild[i]); + assert( pChild->bEof==0 || pChild->bEof==1 ); + nEof += pChild->bEof; + } + pNode->iRowid = pNode->apChild[0]->iRowid; + + switch( pNode->eType ){ + case FTS5_AND: + if( nEof>0 ) fts5ExprSetEof(pNode); + break; + + case FTS5_OR: + if( pNode->nChild==nEof ) fts5ExprSetEof(pNode); + break; + + default: + assert( pNode->eType==FTS5_NOT ); + pNode->bEof = pNode->apChild[0]->bEof; + break; + } + } + + if( rc==SQLITE_OK ){ + rc = fts5ExprNodeTest(pExpr, pNode); + } + return rc; +} + + +/* +** Begin iterating through the set of documents in index pIdx matched by +** the MATCH expression passed as the first argument. If the "bDesc" +** parameter is passed a non-zero value, iteration is in descending rowid +** order. Or, if it is zero, in ascending order. +** +** If iterating in ascending rowid order (bDesc==0), the first document +** visited is that with the smallest rowid that is larger than or equal +** to parameter iFirst. Or, if iterating in ascending order (bDesc==1), +** then the first document visited must have a rowid smaller than or +** equal to iFirst. +** +** Return SQLITE_OK if successful, or an SQLite error code otherwise. It +** is not considered an error if the query does not match any documents. +*/ +static int sqlite3Fts5ExprFirst(Fts5Expr *p, Fts5Index *pIdx, i64 iFirst, int bDesc){ + Fts5ExprNode *pRoot = p->pRoot; + int rc = SQLITE_OK; + if( pRoot->xNext ){ + p->pIndex = pIdx; + p->bDesc = bDesc; + rc = fts5ExprNodeFirst(p, pRoot); + + /* If not at EOF but the current rowid occurs earlier than iFirst in + ** the iteration order, move to document iFirst or later. */ + if( pRoot->bEof==0 && fts5RowidCmp(p, pRoot->iRowid, iFirst)<0 ){ + rc = fts5ExprNodeNext(p, pRoot, 1, iFirst); + } + + /* If the iterator is not at a real match, skip forward until it is. */ + while( pRoot->bNomatch ){ + assert( pRoot->bEof==0 && rc==SQLITE_OK ); + rc = fts5ExprNodeNext(p, pRoot, 0, 0); + } + } + return rc; +} + +/* +** Move to the next document +** +** Return SQLITE_OK if successful, or an SQLite error code otherwise. It +** is not considered an error if the query does not match any documents. +*/ +static int sqlite3Fts5ExprNext(Fts5Expr *p, i64 iLast){ + int rc; + Fts5ExprNode *pRoot = p->pRoot; + assert( pRoot->bEof==0 && pRoot->bNomatch==0 ); + do { + rc = fts5ExprNodeNext(p, pRoot, 0, 0); + assert( pRoot->bNomatch==0 || (rc==SQLITE_OK && pRoot->bEof==0) ); + }while( pRoot->bNomatch ); + if( fts5RowidCmp(p, pRoot->iRowid, iLast)>0 ){ + pRoot->bEof = 1; + } + return rc; +} + +static int sqlite3Fts5ExprEof(Fts5Expr *p){ + return p->pRoot->bEof; +} + +static i64 sqlite3Fts5ExprRowid(Fts5Expr *p){ + return p->pRoot->iRowid; +} + +static int fts5ParseStringFromToken(Fts5Token *pToken, char **pz){ + int rc = SQLITE_OK; + *pz = sqlite3Fts5Strndup(&rc, pToken->p, pToken->n); + return rc; +} + +/* +** Free the phrase object passed as the only argument. +*/ +static void fts5ExprPhraseFree(Fts5ExprPhrase *pPhrase){ + if( pPhrase ){ + int i; + for(i=0; i<pPhrase->nTerm; i++){ + Fts5ExprTerm *pSyn; + Fts5ExprTerm *pNext; + Fts5ExprTerm *pTerm = &pPhrase->aTerm[i]; + sqlite3_free(pTerm->zTerm); + sqlite3Fts5IterClose(pTerm->pIter); + for(pSyn=pTerm->pSynonym; pSyn; pSyn=pNext){ + pNext = pSyn->pSynonym; + sqlite3Fts5IterClose(pSyn->pIter); + fts5BufferFree((Fts5Buffer*)&pSyn[1]); + sqlite3_free(pSyn); + } + } + if( pPhrase->poslist.nSpace>0 ) fts5BufferFree(&pPhrase->poslist); + sqlite3_free(pPhrase); + } +} + +/* +** If argument pNear is NULL, then a new Fts5ExprNearset object is allocated +** and populated with pPhrase. Or, if pNear is not NULL, phrase pPhrase is +** appended to it and the results returned. +** +** If an OOM error occurs, both the pNear and pPhrase objects are freed and +** NULL returned. +*/ +static Fts5ExprNearset *sqlite3Fts5ParseNearset( + Fts5Parse *pParse, /* Parse context */ + Fts5ExprNearset *pNear, /* Existing nearset, or NULL */ + Fts5ExprPhrase *pPhrase /* Recently parsed phrase */ +){ + const int SZALLOC = 8; + Fts5ExprNearset *pRet = 0; + + if( pParse->rc==SQLITE_OK ){ + if( pPhrase==0 ){ + return pNear; + } + if( pNear==0 ){ + int nByte = sizeof(Fts5ExprNearset) + SZALLOC * sizeof(Fts5ExprPhrase*); + pRet = sqlite3_malloc(nByte); + if( pRet==0 ){ + pParse->rc = SQLITE_NOMEM; + }else{ + memset(pRet, 0, nByte); + } + }else if( (pNear->nPhrase % SZALLOC)==0 ){ + int nNew = pNear->nPhrase + SZALLOC; + int nByte = sizeof(Fts5ExprNearset) + nNew * sizeof(Fts5ExprPhrase*); + + pRet = (Fts5ExprNearset*)sqlite3_realloc(pNear, nByte); + if( pRet==0 ){ + pParse->rc = SQLITE_NOMEM; + } + }else{ + pRet = pNear; + } + } + + if( pRet==0 ){ + assert( pParse->rc!=SQLITE_OK ); + sqlite3Fts5ParseNearsetFree(pNear); + sqlite3Fts5ParsePhraseFree(pPhrase); + }else{ + pRet->apPhrase[pRet->nPhrase++] = pPhrase; + } + return pRet; +} + +typedef struct TokenCtx TokenCtx; +struct TokenCtx { + Fts5ExprPhrase *pPhrase; + int rc; +}; + +/* +** Callback for tokenizing terms used by ParseTerm(). +*/ +static int fts5ParseTokenize( + void *pContext, /* Pointer to Fts5InsertCtx object */ + int tflags, /* Mask of FTS5_TOKEN_* flags */ + const char *pToken, /* Buffer containing token */ + int nToken, /* Size of token in bytes */ + int iUnused1, /* Start offset of token */ + int iUnused2 /* End offset of token */ +){ + int rc = SQLITE_OK; + const int SZALLOC = 8; + TokenCtx *pCtx = (TokenCtx*)pContext; + Fts5ExprPhrase *pPhrase = pCtx->pPhrase; + + UNUSED_PARAM2(iUnused1, iUnused2); + + /* If an error has already occurred, this is a no-op */ + if( pCtx->rc!=SQLITE_OK ) return pCtx->rc; + + assert( pPhrase==0 || pPhrase->nTerm>0 ); + if( pPhrase && (tflags & FTS5_TOKEN_COLOCATED) ){ + Fts5ExprTerm *pSyn; + int nByte = sizeof(Fts5ExprTerm) + sizeof(Fts5Buffer) + nToken+1; + pSyn = (Fts5ExprTerm*)sqlite3_malloc(nByte); + if( pSyn==0 ){ + rc = SQLITE_NOMEM; + }else{ + memset(pSyn, 0, nByte); + pSyn->zTerm = ((char*)pSyn) + sizeof(Fts5ExprTerm) + sizeof(Fts5Buffer); + memcpy(pSyn->zTerm, pToken, nToken); + pSyn->pSynonym = pPhrase->aTerm[pPhrase->nTerm-1].pSynonym; + pPhrase->aTerm[pPhrase->nTerm-1].pSynonym = pSyn; + } + }else{ + Fts5ExprTerm *pTerm; + if( pPhrase==0 || (pPhrase->nTerm % SZALLOC)==0 ){ + Fts5ExprPhrase *pNew; + int nNew = SZALLOC + (pPhrase ? pPhrase->nTerm : 0); + + pNew = (Fts5ExprPhrase*)sqlite3_realloc(pPhrase, + sizeof(Fts5ExprPhrase) + sizeof(Fts5ExprTerm) * nNew + ); + if( pNew==0 ){ + rc = SQLITE_NOMEM; + }else{ + if( pPhrase==0 ) memset(pNew, 0, sizeof(Fts5ExprPhrase)); + pCtx->pPhrase = pPhrase = pNew; + pNew->nTerm = nNew - SZALLOC; + } + } + + if( rc==SQLITE_OK ){ + pTerm = &pPhrase->aTerm[pPhrase->nTerm++]; + memset(pTerm, 0, sizeof(Fts5ExprTerm)); + pTerm->zTerm = sqlite3Fts5Strndup(&rc, pToken, nToken); + } + } + + pCtx->rc = rc; + return rc; +} + + +/* +** Free the phrase object passed as the only argument. +*/ +static void sqlite3Fts5ParsePhraseFree(Fts5ExprPhrase *pPhrase){ + fts5ExprPhraseFree(pPhrase); +} + +/* +** Free the phrase object passed as the second argument. +*/ +static void sqlite3Fts5ParseNearsetFree(Fts5ExprNearset *pNear){ + if( pNear ){ + int i; + for(i=0; i<pNear->nPhrase; i++){ + fts5ExprPhraseFree(pNear->apPhrase[i]); + } + sqlite3_free(pNear->pColset); + sqlite3_free(pNear); + } +} + +static void sqlite3Fts5ParseFinished(Fts5Parse *pParse, Fts5ExprNode *p){ + assert( pParse->pExpr==0 ); + pParse->pExpr = p; +} + +/* +** This function is called by the parser to process a string token. The +** string may or may not be quoted. In any case it is tokenized and a +** phrase object consisting of all tokens returned. +*/ +static Fts5ExprPhrase *sqlite3Fts5ParseTerm( + Fts5Parse *pParse, /* Parse context */ + Fts5ExprPhrase *pAppend, /* Phrase to append to */ + Fts5Token *pToken, /* String to tokenize */ + int bPrefix /* True if there is a trailing "*" */ +){ + Fts5Config *pConfig = pParse->pConfig; + TokenCtx sCtx; /* Context object passed to callback */ + int rc; /* Tokenize return code */ + char *z = 0; + + memset(&sCtx, 0, sizeof(TokenCtx)); + sCtx.pPhrase = pAppend; + + rc = fts5ParseStringFromToken(pToken, &z); + if( rc==SQLITE_OK ){ + int flags = FTS5_TOKENIZE_QUERY | (bPrefix ? FTS5_TOKENIZE_QUERY : 0); + int n; + sqlite3Fts5Dequote(z); + n = (int)strlen(z); + rc = sqlite3Fts5Tokenize(pConfig, flags, z, n, &sCtx, fts5ParseTokenize); + } + sqlite3_free(z); + if( rc || (rc = sCtx.rc) ){ + pParse->rc = rc; + fts5ExprPhraseFree(sCtx.pPhrase); + sCtx.pPhrase = 0; + }else if( sCtx.pPhrase ){ + + if( pAppend==0 ){ + if( (pParse->nPhrase % 8)==0 ){ + int nByte = sizeof(Fts5ExprPhrase*) * (pParse->nPhrase + 8); + Fts5ExprPhrase **apNew; + apNew = (Fts5ExprPhrase**)sqlite3_realloc(pParse->apPhrase, nByte); + if( apNew==0 ){ + pParse->rc = SQLITE_NOMEM; + fts5ExprPhraseFree(sCtx.pPhrase); + return 0; + } + pParse->apPhrase = apNew; + } + pParse->nPhrase++; + } + + pParse->apPhrase[pParse->nPhrase-1] = sCtx.pPhrase; + assert( sCtx.pPhrase->nTerm>0 ); + sCtx.pPhrase->aTerm[sCtx.pPhrase->nTerm-1].bPrefix = bPrefix; + } + + return sCtx.pPhrase; +} + +/* +** Create a new FTS5 expression by cloning phrase iPhrase of the +** expression passed as the second argument. +*/ +static int sqlite3Fts5ExprClonePhrase( + Fts5Expr *pExpr, + int iPhrase, + Fts5Expr **ppNew +){ + int rc = SQLITE_OK; /* Return code */ + Fts5ExprPhrase *pOrig; /* The phrase extracted from pExpr */ + int i; /* Used to iterate through phrase terms */ + Fts5Expr *pNew = 0; /* Expression to return via *ppNew */ + TokenCtx sCtx = {0,0}; /* Context object for fts5ParseTokenize */ + + pOrig = pExpr->apExprPhrase[iPhrase]; + pNew = (Fts5Expr*)sqlite3Fts5MallocZero(&rc, sizeof(Fts5Expr)); + if( rc==SQLITE_OK ){ + pNew->apExprPhrase = (Fts5ExprPhrase**)sqlite3Fts5MallocZero(&rc, + sizeof(Fts5ExprPhrase*)); + } + if( rc==SQLITE_OK ){ + pNew->pRoot = (Fts5ExprNode*)sqlite3Fts5MallocZero(&rc, + sizeof(Fts5ExprNode)); + } + if( rc==SQLITE_OK ){ + pNew->pRoot->pNear = (Fts5ExprNearset*)sqlite3Fts5MallocZero(&rc, + sizeof(Fts5ExprNearset) + sizeof(Fts5ExprPhrase*)); + } + + for(i=0; rc==SQLITE_OK && i<pOrig->nTerm; i++){ + int tflags = 0; + Fts5ExprTerm *p; + for(p=&pOrig->aTerm[i]; p && rc==SQLITE_OK; p=p->pSynonym){ + const char *zTerm = p->zTerm; + rc = fts5ParseTokenize((void*)&sCtx, tflags, zTerm, (int)strlen(zTerm), + 0, 0); + tflags = FTS5_TOKEN_COLOCATED; + } + if( rc==SQLITE_OK ){ + sCtx.pPhrase->aTerm[i].bPrefix = pOrig->aTerm[i].bPrefix; + } + } + + if( rc==SQLITE_OK ){ + /* All the allocations succeeded. Put the expression object together. */ + pNew->pIndex = pExpr->pIndex; + pNew->pConfig = pExpr->pConfig; + pNew->nPhrase = 1; + pNew->apExprPhrase[0] = sCtx.pPhrase; + pNew->pRoot->pNear->apPhrase[0] = sCtx.pPhrase; + pNew->pRoot->pNear->nPhrase = 1; + sCtx.pPhrase->pNode = pNew->pRoot; + + if( pOrig->nTerm==1 && pOrig->aTerm[0].pSynonym==0 ){ + pNew->pRoot->eType = FTS5_TERM; + pNew->pRoot->xNext = fts5ExprNodeNext_TERM; + }else{ + pNew->pRoot->eType = FTS5_STRING; + pNew->pRoot->xNext = fts5ExprNodeNext_STRING; + } + }else{ + sqlite3Fts5ExprFree(pNew); + fts5ExprPhraseFree(sCtx.pPhrase); + pNew = 0; + } + + *ppNew = pNew; + return rc; +} + + +/* +** Token pTok has appeared in a MATCH expression where the NEAR operator +** is expected. If token pTok does not contain "NEAR", store an error +** in the pParse object. +*/ +static void sqlite3Fts5ParseNear(Fts5Parse *pParse, Fts5Token *pTok){ + if( pTok->n!=4 || memcmp("NEAR", pTok->p, 4) ){ + sqlite3Fts5ParseError( + pParse, "fts5: syntax error near \"%.*s\"", pTok->n, pTok->p + ); + } +} + +static void sqlite3Fts5ParseSetDistance( + Fts5Parse *pParse, + Fts5ExprNearset *pNear, + Fts5Token *p +){ + int nNear = 0; + int i; + if( p->n ){ + for(i=0; i<p->n; i++){ + char c = (char)p->p[i]; + if( c<'0' || c>'9' ){ + sqlite3Fts5ParseError( + pParse, "expected integer, got \"%.*s\"", p->n, p->p + ); + return; + } + nNear = nNear * 10 + (p->p[i] - '0'); + } + }else{ + nNear = FTS5_DEFAULT_NEARDIST; + } + pNear->nNear = nNear; +} + +/* +** The second argument passed to this function may be NULL, or it may be +** an existing Fts5Colset object. This function returns a pointer to +** a new colset object containing the contents of (p) with new value column +** number iCol appended. +** +** If an OOM error occurs, store an error code in pParse and return NULL. +** The old colset object (if any) is not freed in this case. +*/ +static Fts5Colset *fts5ParseColset( + Fts5Parse *pParse, /* Store SQLITE_NOMEM here if required */ + Fts5Colset *p, /* Existing colset object */ + int iCol /* New column to add to colset object */ +){ + int nCol = p ? p->nCol : 0; /* Num. columns already in colset object */ + Fts5Colset *pNew; /* New colset object to return */ + + assert( pParse->rc==SQLITE_OK ); + assert( iCol>=0 && iCol<pParse->pConfig->nCol ); + + pNew = sqlite3_realloc(p, sizeof(Fts5Colset) + sizeof(int)*nCol); + if( pNew==0 ){ + pParse->rc = SQLITE_NOMEM; + }else{ + int *aiCol = pNew->aiCol; + int i, j; + for(i=0; i<nCol; i++){ + if( aiCol[i]==iCol ) return pNew; + if( aiCol[i]>iCol ) break; + } + for(j=nCol; j>i; j--){ + aiCol[j] = aiCol[j-1]; + } + aiCol[i] = iCol; + pNew->nCol = nCol+1; + +#ifndef NDEBUG + /* Check that the array is in order and contains no duplicate entries. */ + for(i=1; i<pNew->nCol; i++) assert( pNew->aiCol[i]>pNew->aiCol[i-1] ); +#endif + } + + return pNew; +} + +static Fts5Colset *sqlite3Fts5ParseColset( + Fts5Parse *pParse, /* Store SQLITE_NOMEM here if required */ + Fts5Colset *pColset, /* Existing colset object */ + Fts5Token *p +){ + Fts5Colset *pRet = 0; + int iCol; + char *z; /* Dequoted copy of token p */ + + z = sqlite3Fts5Strndup(&pParse->rc, p->p, p->n); + if( pParse->rc==SQLITE_OK ){ + Fts5Config *pConfig = pParse->pConfig; + sqlite3Fts5Dequote(z); + for(iCol=0; iCol<pConfig->nCol; iCol++){ + if( 0==sqlite3_stricmp(pConfig->azCol[iCol], z) ) break; + } + if( iCol==pConfig->nCol ){ + sqlite3Fts5ParseError(pParse, "no such column: %s", z); + }else{ + pRet = fts5ParseColset(pParse, pColset, iCol); + } + sqlite3_free(z); + } + + if( pRet==0 ){ + assert( pParse->rc!=SQLITE_OK ); + sqlite3_free(pColset); + } + + return pRet; +} + +static void sqlite3Fts5ParseSetColset( + Fts5Parse *pParse, + Fts5ExprNearset *pNear, + Fts5Colset *pColset +){ + if( pParse->pConfig->eDetail==FTS5_DETAIL_NONE ){ + pParse->rc = SQLITE_ERROR; + pParse->zErr = sqlite3_mprintf( + "fts5: column queries are not supported (detail=none)" + ); + sqlite3_free(pColset); + return; + } + + if( pNear ){ + pNear->pColset = pColset; + }else{ + sqlite3_free(pColset); + } +} + +static void fts5ExprAssignXNext(Fts5ExprNode *pNode){ + switch( pNode->eType ){ + case FTS5_STRING: { + Fts5ExprNearset *pNear = pNode->pNear; + if( pNear->nPhrase==1 && pNear->apPhrase[0]->nTerm==1 + && pNear->apPhrase[0]->aTerm[0].pSynonym==0 + ){ + pNode->eType = FTS5_TERM; + pNode->xNext = fts5ExprNodeNext_TERM; + }else{ + pNode->xNext = fts5ExprNodeNext_STRING; + } + break; + }; + + case FTS5_OR: { + pNode->xNext = fts5ExprNodeNext_OR; + break; + }; + + case FTS5_AND: { + pNode->xNext = fts5ExprNodeNext_AND; + break; + }; + + default: assert( pNode->eType==FTS5_NOT ); { + pNode->xNext = fts5ExprNodeNext_NOT; + break; + }; + } +} + +static void fts5ExprAddChildren(Fts5ExprNode *p, Fts5ExprNode *pSub){ + if( p->eType!=FTS5_NOT && pSub->eType==p->eType ){ + int nByte = sizeof(Fts5ExprNode*) * pSub->nChild; + memcpy(&p->apChild[p->nChild], pSub->apChild, nByte); + p->nChild += pSub->nChild; + sqlite3_free(pSub); + }else{ + p->apChild[p->nChild++] = pSub; + } +} + +/* +** Allocate and return a new expression object. If anything goes wrong (i.e. +** OOM error), leave an error code in pParse and return NULL. +*/ +static Fts5ExprNode *sqlite3Fts5ParseNode( + Fts5Parse *pParse, /* Parse context */ + int eType, /* FTS5_STRING, AND, OR or NOT */ + Fts5ExprNode *pLeft, /* Left hand child expression */ + Fts5ExprNode *pRight, /* Right hand child expression */ + Fts5ExprNearset *pNear /* For STRING expressions, the near cluster */ +){ + Fts5ExprNode *pRet = 0; + + if( pParse->rc==SQLITE_OK ){ + int nChild = 0; /* Number of children of returned node */ + int nByte; /* Bytes of space to allocate for this node */ + + assert( (eType!=FTS5_STRING && !pNear) + || (eType==FTS5_STRING && !pLeft && !pRight) + ); + if( eType==FTS5_STRING && pNear==0 ) return 0; + if( eType!=FTS5_STRING && pLeft==0 ) return pRight; + if( eType!=FTS5_STRING && pRight==0 ) return pLeft; + + if( eType==FTS5_NOT ){ + nChild = 2; + }else if( eType==FTS5_AND || eType==FTS5_OR ){ + nChild = 2; + if( pLeft->eType==eType ) nChild += pLeft->nChild-1; + if( pRight->eType==eType ) nChild += pRight->nChild-1; + } + + nByte = sizeof(Fts5ExprNode) + sizeof(Fts5ExprNode*)*(nChild-1); + pRet = (Fts5ExprNode*)sqlite3Fts5MallocZero(&pParse->rc, nByte); + + if( pRet ){ + pRet->eType = eType; + pRet->pNear = pNear; + fts5ExprAssignXNext(pRet); + if( eType==FTS5_STRING ){ + int iPhrase; + for(iPhrase=0; iPhrase<pNear->nPhrase; iPhrase++){ + pNear->apPhrase[iPhrase]->pNode = pRet; + } + + if( pParse->pConfig->eDetail!=FTS5_DETAIL_FULL + && (pNear->nPhrase!=1 || pNear->apPhrase[0]->nTerm!=1) + ){ + assert( pParse->rc==SQLITE_OK ); + pParse->rc = SQLITE_ERROR; + assert( pParse->zErr==0 ); + pParse->zErr = sqlite3_mprintf( + "fts5: %s queries are not supported (detail!=full)", + pNear->nPhrase==1 ? "phrase": "NEAR" + ); + sqlite3_free(pRet); + pRet = 0; + } + + }else{ + fts5ExprAddChildren(pRet, pLeft); + fts5ExprAddChildren(pRet, pRight); + } + } + } + + if( pRet==0 ){ + assert( pParse->rc!=SQLITE_OK ); + sqlite3Fts5ParseNodeFree(pLeft); + sqlite3Fts5ParseNodeFree(pRight); + sqlite3Fts5ParseNearsetFree(pNear); + } + return pRet; +} + +static char *fts5ExprTermPrint(Fts5ExprTerm *pTerm){ + int nByte = 0; + Fts5ExprTerm *p; + char *zQuoted; + + /* Determine the maximum amount of space required. */ + for(p=pTerm; p; p=p->pSynonym){ + nByte += (int)strlen(pTerm->zTerm) * 2 + 3 + 2; + } + zQuoted = sqlite3_malloc(nByte); + + if( zQuoted ){ + int i = 0; + for(p=pTerm; p; p=p->pSynonym){ + char *zIn = p->zTerm; + zQuoted[i++] = '"'; + while( *zIn ){ + if( *zIn=='"' ) zQuoted[i++] = '"'; + zQuoted[i++] = *zIn++; + } + zQuoted[i++] = '"'; + if( p->pSynonym ) zQuoted[i++] = '|'; + } + if( pTerm->bPrefix ){ + zQuoted[i++] = ' '; + zQuoted[i++] = '*'; + } + zQuoted[i++] = '\0'; + } + return zQuoted; +} + +static char *fts5PrintfAppend(char *zApp, const char *zFmt, ...){ + char *zNew; + va_list ap; + va_start(ap, zFmt); + zNew = sqlite3_vmprintf(zFmt, ap); + va_end(ap); + if( zApp && zNew ){ + char *zNew2 = sqlite3_mprintf("%s%s", zApp, zNew); + sqlite3_free(zNew); + zNew = zNew2; + } + sqlite3_free(zApp); + return zNew; +} + +/* +** Compose a tcl-readable representation of expression pExpr. Return a +** pointer to a buffer containing that representation. It is the +** responsibility of the caller to at some point free the buffer using +** sqlite3_free(). +*/ +static char *fts5ExprPrintTcl( + Fts5Config *pConfig, + const char *zNearsetCmd, + Fts5ExprNode *pExpr +){ + char *zRet = 0; + if( pExpr->eType==FTS5_STRING || pExpr->eType==FTS5_TERM ){ + Fts5ExprNearset *pNear = pExpr->pNear; + int i; + int iTerm; + + zRet = fts5PrintfAppend(zRet, "%s ", zNearsetCmd); + if( zRet==0 ) return 0; + if( pNear->pColset ){ + int *aiCol = pNear->pColset->aiCol; + int nCol = pNear->pColset->nCol; + if( nCol==1 ){ + zRet = fts5PrintfAppend(zRet, "-col %d ", aiCol[0]); + }else{ + zRet = fts5PrintfAppend(zRet, "-col {%d", aiCol[0]); + for(i=1; i<pNear->pColset->nCol; i++){ + zRet = fts5PrintfAppend(zRet, " %d", aiCol[i]); + } + zRet = fts5PrintfAppend(zRet, "} "); + } + if( zRet==0 ) return 0; + } + + if( pNear->nPhrase>1 ){ + zRet = fts5PrintfAppend(zRet, "-near %d ", pNear->nNear); + if( zRet==0 ) return 0; + } + + zRet = fts5PrintfAppend(zRet, "--"); + if( zRet==0 ) return 0; + + for(i=0; i<pNear->nPhrase; i++){ + Fts5ExprPhrase *pPhrase = pNear->apPhrase[i]; + + zRet = fts5PrintfAppend(zRet, " {"); + for(iTerm=0; zRet && iTerm<pPhrase->nTerm; iTerm++){ + char *zTerm = pPhrase->aTerm[iTerm].zTerm; + zRet = fts5PrintfAppend(zRet, "%s%s", iTerm==0?"":" ", zTerm); + if( pPhrase->aTerm[iTerm].bPrefix ){ + zRet = fts5PrintfAppend(zRet, "*"); + } + } + + if( zRet ) zRet = fts5PrintfAppend(zRet, "}"); + if( zRet==0 ) return 0; + } + + }else{ + char const *zOp = 0; + int i; + switch( pExpr->eType ){ + case FTS5_AND: zOp = "AND"; break; + case FTS5_NOT: zOp = "NOT"; break; + default: + assert( pExpr->eType==FTS5_OR ); + zOp = "OR"; + break; + } + + zRet = sqlite3_mprintf("%s", zOp); + for(i=0; zRet && i<pExpr->nChild; i++){ + char *z = fts5ExprPrintTcl(pConfig, zNearsetCmd, pExpr->apChild[i]); + if( !z ){ + sqlite3_free(zRet); + zRet = 0; + }else{ + zRet = fts5PrintfAppend(zRet, " [%z]", z); + } + } + } + + return zRet; +} + +static char *fts5ExprPrint(Fts5Config *pConfig, Fts5ExprNode *pExpr){ + char *zRet = 0; + if( pExpr->eType==FTS5_STRING || pExpr->eType==FTS5_TERM ){ + Fts5ExprNearset *pNear = pExpr->pNear; + int i; + int iTerm; + + if( pNear->pColset ){ + int iCol = pNear->pColset->aiCol[0]; + zRet = fts5PrintfAppend(zRet, "%s : ", pConfig->azCol[iCol]); + if( zRet==0 ) return 0; + } + + if( pNear->nPhrase>1 ){ + zRet = fts5PrintfAppend(zRet, "NEAR("); + if( zRet==0 ) return 0; + } + + for(i=0; i<pNear->nPhrase; i++){ + Fts5ExprPhrase *pPhrase = pNear->apPhrase[i]; + if( i!=0 ){ + zRet = fts5PrintfAppend(zRet, " "); + if( zRet==0 ) return 0; + } + for(iTerm=0; iTerm<pPhrase->nTerm; iTerm++){ + char *zTerm = fts5ExprTermPrint(&pPhrase->aTerm[iTerm]); + if( zTerm ){ + zRet = fts5PrintfAppend(zRet, "%s%s", iTerm==0?"":" + ", zTerm); + sqlite3_free(zTerm); + } + if( zTerm==0 || zRet==0 ){ + sqlite3_free(zRet); + return 0; + } + } + } + + if( pNear->nPhrase>1 ){ + zRet = fts5PrintfAppend(zRet, ", %d)", pNear->nNear); + if( zRet==0 ) return 0; + } + + }else{ + char const *zOp = 0; + int i; + + switch( pExpr->eType ){ + case FTS5_AND: zOp = " AND "; break; + case FTS5_NOT: zOp = " NOT "; break; + default: + assert( pExpr->eType==FTS5_OR ); + zOp = " OR "; + break; + } + + for(i=0; i<pExpr->nChild; i++){ + char *z = fts5ExprPrint(pConfig, pExpr->apChild[i]); + if( z==0 ){ + sqlite3_free(zRet); + zRet = 0; + }else{ + int e = pExpr->apChild[i]->eType; + int b = (e!=FTS5_STRING && e!=FTS5_TERM); + zRet = fts5PrintfAppend(zRet, "%s%s%z%s", + (i==0 ? "" : zOp), + (b?"(":""), z, (b?")":"") + ); + } + if( zRet==0 ) break; + } + } + + return zRet; +} + +/* +** The implementation of user-defined scalar functions fts5_expr() (bTcl==0) +** and fts5_expr_tcl() (bTcl!=0). +*/ +static void fts5ExprFunction( + sqlite3_context *pCtx, /* Function call context */ + int nArg, /* Number of args */ + sqlite3_value **apVal, /* Function arguments */ + int bTcl +){ + Fts5Global *pGlobal = (Fts5Global*)sqlite3_user_data(pCtx); + sqlite3 *db = sqlite3_context_db_handle(pCtx); + const char *zExpr = 0; + char *zErr = 0; + Fts5Expr *pExpr = 0; + int rc; + int i; + + const char **azConfig; /* Array of arguments for Fts5Config */ + const char *zNearsetCmd = "nearset"; + int nConfig; /* Size of azConfig[] */ + Fts5Config *pConfig = 0; + int iArg = 1; + + if( nArg<1 ){ + zErr = sqlite3_mprintf("wrong number of arguments to function %s", + bTcl ? "fts5_expr_tcl" : "fts5_expr" + ); + sqlite3_result_error(pCtx, zErr, -1); + sqlite3_free(zErr); + return; + } + + if( bTcl && nArg>1 ){ + zNearsetCmd = (const char*)sqlite3_value_text(apVal[1]); + iArg = 2; + } + + nConfig = 3 + (nArg-iArg); + azConfig = (const char**)sqlite3_malloc(sizeof(char*) * nConfig); + if( azConfig==0 ){ + sqlite3_result_error_nomem(pCtx); + return; + } + azConfig[0] = 0; + azConfig[1] = "main"; + azConfig[2] = "tbl"; + for(i=3; iArg<nArg; iArg++){ + azConfig[i++] = (const char*)sqlite3_value_text(apVal[iArg]); + } + + zExpr = (const char*)sqlite3_value_text(apVal[0]); + + rc = sqlite3Fts5ConfigParse(pGlobal, db, nConfig, azConfig, &pConfig, &zErr); + if( rc==SQLITE_OK ){ + rc = sqlite3Fts5ExprNew(pConfig, zExpr, &pExpr, &zErr); + } + if( rc==SQLITE_OK ){ + char *zText; + if( pExpr->pRoot->xNext==0 ){ + zText = sqlite3_mprintf(""); + }else if( bTcl ){ + zText = fts5ExprPrintTcl(pConfig, zNearsetCmd, pExpr->pRoot); + }else{ + zText = fts5ExprPrint(pConfig, pExpr->pRoot); + } + if( zText==0 ){ + rc = SQLITE_NOMEM; + }else{ + sqlite3_result_text(pCtx, zText, -1, SQLITE_TRANSIENT); + sqlite3_free(zText); + } + } + + if( rc!=SQLITE_OK ){ + if( zErr ){ + sqlite3_result_error(pCtx, zErr, -1); + sqlite3_free(zErr); + }else{ + sqlite3_result_error_code(pCtx, rc); + } + } + sqlite3_free((void *)azConfig); + sqlite3Fts5ConfigFree(pConfig); + sqlite3Fts5ExprFree(pExpr); +} + +static void fts5ExprFunctionHr( + sqlite3_context *pCtx, /* Function call context */ + int nArg, /* Number of args */ + sqlite3_value **apVal /* Function arguments */ +){ + fts5ExprFunction(pCtx, nArg, apVal, 0); +} +static void fts5ExprFunctionTcl( + sqlite3_context *pCtx, /* Function call context */ + int nArg, /* Number of args */ + sqlite3_value **apVal /* Function arguments */ +){ + fts5ExprFunction(pCtx, nArg, apVal, 1); +} + +/* +** The implementation of an SQLite user-defined-function that accepts a +** single integer as an argument. If the integer is an alpha-numeric +** unicode code point, 1 is returned. Otherwise 0. +*/ +static void fts5ExprIsAlnum( + sqlite3_context *pCtx, /* Function call context */ + int nArg, /* Number of args */ + sqlite3_value **apVal /* Function arguments */ +){ + int iCode; + if( nArg!=1 ){ + sqlite3_result_error(pCtx, + "wrong number of arguments to function fts5_isalnum", -1 + ); + return; + } + iCode = sqlite3_value_int(apVal[0]); + sqlite3_result_int(pCtx, sqlite3Fts5UnicodeIsalnum(iCode)); +} + +static void fts5ExprFold( + sqlite3_context *pCtx, /* Function call context */ + int nArg, /* Number of args */ + sqlite3_value **apVal /* Function arguments */ +){ + if( nArg!=1 && nArg!=2 ){ + sqlite3_result_error(pCtx, + "wrong number of arguments to function fts5_fold", -1 + ); + }else{ + int iCode; + int bRemoveDiacritics = 0; + iCode = sqlite3_value_int(apVal[0]); + if( nArg==2 ) bRemoveDiacritics = sqlite3_value_int(apVal[1]); + sqlite3_result_int(pCtx, sqlite3Fts5UnicodeFold(iCode, bRemoveDiacritics)); + } +} + +/* +** This is called during initialization to register the fts5_expr() scalar +** UDF with the SQLite handle passed as the only argument. +*/ +static int sqlite3Fts5ExprInit(Fts5Global *pGlobal, sqlite3 *db){ + struct Fts5ExprFunc { + const char *z; + void (*x)(sqlite3_context*,int,sqlite3_value**); + } aFunc[] = { + { "fts5_expr", fts5ExprFunctionHr }, + { "fts5_expr_tcl", fts5ExprFunctionTcl }, + { "fts5_isalnum", fts5ExprIsAlnum }, + { "fts5_fold", fts5ExprFold }, + }; + int i; + int rc = SQLITE_OK; + void *pCtx = (void*)pGlobal; + + for(i=0; rc==SQLITE_OK && i<ArraySize(aFunc); i++){ + struct Fts5ExprFunc *p = &aFunc[i]; + rc = sqlite3_create_function(db, p->z, -1, SQLITE_UTF8, pCtx, p->x, 0, 0); + } + + /* Avoid a warning indicating that sqlite3Fts5ParserTrace() is unused */ +#ifndef NDEBUG + (void)sqlite3Fts5ParserTrace; +#endif + + return rc; +} + +/* +** Return the number of phrases in expression pExpr. +*/ +static int sqlite3Fts5ExprPhraseCount(Fts5Expr *pExpr){ + return (pExpr ? pExpr->nPhrase : 0); +} + +/* +** Return the number of terms in the iPhrase'th phrase in pExpr. +*/ +static int sqlite3Fts5ExprPhraseSize(Fts5Expr *pExpr, int iPhrase){ + if( iPhrase<0 || iPhrase>=pExpr->nPhrase ) return 0; + return pExpr->apExprPhrase[iPhrase]->nTerm; +} + +/* +** This function is used to access the current position list for phrase +** iPhrase. +*/ +static int sqlite3Fts5ExprPoslist(Fts5Expr *pExpr, int iPhrase, const u8 **pa){ + int nRet; + Fts5ExprPhrase *pPhrase = pExpr->apExprPhrase[iPhrase]; + Fts5ExprNode *pNode = pPhrase->pNode; + if( pNode->bEof==0 && pNode->iRowid==pExpr->pRoot->iRowid ){ + *pa = pPhrase->poslist.p; + nRet = pPhrase->poslist.n; + }else{ + *pa = 0; + nRet = 0; + } + return nRet; +} + +struct Fts5PoslistPopulator { + Fts5PoslistWriter writer; + int bOk; /* True if ok to populate */ + int bMiss; +}; + +static Fts5PoslistPopulator *sqlite3Fts5ExprClearPoslists(Fts5Expr *pExpr, int bLive){ + Fts5PoslistPopulator *pRet; + pRet = sqlite3_malloc(sizeof(Fts5PoslistPopulator)*pExpr->nPhrase); + if( pRet ){ + int i; + memset(pRet, 0, sizeof(Fts5PoslistPopulator)*pExpr->nPhrase); + for(i=0; i<pExpr->nPhrase; i++){ + Fts5Buffer *pBuf = &pExpr->apExprPhrase[i]->poslist; + Fts5ExprNode *pNode = pExpr->apExprPhrase[i]->pNode; + assert( pExpr->apExprPhrase[i]->nTerm==1 ); + if( bLive && + (pBuf->n==0 || pNode->iRowid!=pExpr->pRoot->iRowid || pNode->bEof) + ){ + pRet[i].bMiss = 1; + }else{ + pBuf->n = 0; + } + } + } + return pRet; +} + +struct Fts5ExprCtx { + Fts5Expr *pExpr; + Fts5PoslistPopulator *aPopulator; + i64 iOff; +}; +typedef struct Fts5ExprCtx Fts5ExprCtx; + +/* +** TODO: Make this more efficient! +*/ +static int fts5ExprColsetTest(Fts5Colset *pColset, int iCol){ + int i; + for(i=0; i<pColset->nCol; i++){ + if( pColset->aiCol[i]==iCol ) return 1; + } + return 0; +} + +static int fts5ExprPopulatePoslistsCb( + void *pCtx, /* Copy of 2nd argument to xTokenize() */ + int tflags, /* Mask of FTS5_TOKEN_* flags */ + const char *pToken, /* Pointer to buffer containing token */ + int nToken, /* Size of token in bytes */ + int iUnused1, /* Byte offset of token within input text */ + int iUnused2 /* Byte offset of end of token within input text */ +){ + Fts5ExprCtx *p = (Fts5ExprCtx*)pCtx; + Fts5Expr *pExpr = p->pExpr; + int i; + + UNUSED_PARAM2(iUnused1, iUnused2); + + if( (tflags & FTS5_TOKEN_COLOCATED)==0 ) p->iOff++; + for(i=0; i<pExpr->nPhrase; i++){ + Fts5ExprTerm *pTerm; + if( p->aPopulator[i].bOk==0 ) continue; + for(pTerm=&pExpr->apExprPhrase[i]->aTerm[0]; pTerm; pTerm=pTerm->pSynonym){ + int nTerm = strlen(pTerm->zTerm); + if( (nTerm==nToken || (nTerm<nToken && pTerm->bPrefix)) + && memcmp(pTerm->zTerm, pToken, nTerm)==0 + ){ + int rc = sqlite3Fts5PoslistWriterAppend( + &pExpr->apExprPhrase[i]->poslist, &p->aPopulator[i].writer, p->iOff + ); + if( rc ) return rc; + break; + } + } + } + return SQLITE_OK; +} + +static int sqlite3Fts5ExprPopulatePoslists( + Fts5Config *pConfig, + Fts5Expr *pExpr, + Fts5PoslistPopulator *aPopulator, + int iCol, + const char *z, int n +){ + int i; + Fts5ExprCtx sCtx; + sCtx.pExpr = pExpr; + sCtx.aPopulator = aPopulator; + sCtx.iOff = (((i64)iCol) << 32) - 1; + + for(i=0; i<pExpr->nPhrase; i++){ + Fts5ExprNode *pNode = pExpr->apExprPhrase[i]->pNode; + Fts5Colset *pColset = pNode->pNear->pColset; + if( (pColset && 0==fts5ExprColsetTest(pColset, iCol)) + || aPopulator[i].bMiss + ){ + aPopulator[i].bOk = 0; + }else{ + aPopulator[i].bOk = 1; + } + } + + return sqlite3Fts5Tokenize(pConfig, + FTS5_TOKENIZE_DOCUMENT, z, n, (void*)&sCtx, fts5ExprPopulatePoslistsCb + ); +} + +static void fts5ExprClearPoslists(Fts5ExprNode *pNode){ + if( pNode->eType==FTS5_TERM || pNode->eType==FTS5_STRING ){ + pNode->pNear->apPhrase[0]->poslist.n = 0; + }else{ + int i; + for(i=0; i<pNode->nChild; i++){ + fts5ExprClearPoslists(pNode->apChild[i]); + } + } +} + +static int fts5ExprCheckPoslists(Fts5ExprNode *pNode, i64 iRowid){ + pNode->iRowid = iRowid; + pNode->bEof = 0; + switch( pNode->eType ){ + case FTS5_TERM: + case FTS5_STRING: + return (pNode->pNear->apPhrase[0]->poslist.n>0); + + case FTS5_AND: { + int i; + for(i=0; i<pNode->nChild; i++){ + if( fts5ExprCheckPoslists(pNode->apChild[i], iRowid)==0 ){ + fts5ExprClearPoslists(pNode); + return 0; + } + } + break; + } + + case FTS5_OR: { + int i; + int bRet = 0; + for(i=0; i<pNode->nChild; i++){ + if( fts5ExprCheckPoslists(pNode->apChild[i], iRowid) ){ + bRet = 1; + } + } + return bRet; + } + + default: { + assert( pNode->eType==FTS5_NOT ); + if( 0==fts5ExprCheckPoslists(pNode->apChild[0], iRowid) + || 0!=fts5ExprCheckPoslists(pNode->apChild[1], iRowid) + ){ + fts5ExprClearPoslists(pNode); + return 0; + } + break; + } + } + return 1; +} + +static void sqlite3Fts5ExprCheckPoslists(Fts5Expr *pExpr, i64 iRowid){ + fts5ExprCheckPoslists(pExpr->pRoot, iRowid); +} + +static void fts5ExprClearEof(Fts5ExprNode *pNode){ + int i; + for(i=0; i<pNode->nChild; i++){ + fts5ExprClearEof(pNode->apChild[i]); + } + pNode->bEof = 0; +} +static void sqlite3Fts5ExprClearEof(Fts5Expr *pExpr){ + fts5ExprClearEof(pExpr->pRoot); +} + +/* +** This function is only called for detail=columns tables. +*/ +static int sqlite3Fts5ExprPhraseCollist( + Fts5Expr *pExpr, + int iPhrase, + const u8 **ppCollist, + int *pnCollist +){ + Fts5ExprPhrase *pPhrase = pExpr->apExprPhrase[iPhrase]; + Fts5ExprNode *pNode = pPhrase->pNode; + int rc = SQLITE_OK; + + assert( iPhrase>=0 && iPhrase<pExpr->nPhrase ); + assert( pExpr->pConfig->eDetail==FTS5_DETAIL_COLUMNS ); + + if( pNode->bEof==0 + && pNode->iRowid==pExpr->pRoot->iRowid + && pPhrase->poslist.n>0 + ){ + Fts5ExprTerm *pTerm = &pPhrase->aTerm[0]; + if( pTerm->pSynonym ){ + Fts5Buffer *pBuf = (Fts5Buffer*)&pTerm->pSynonym[1]; + rc = fts5ExprSynonymList( + pTerm, pNode->iRowid, pBuf, (u8**)ppCollist, pnCollist + ); + }else{ + *ppCollist = pPhrase->aTerm[0].pIter->pData; + *pnCollist = pPhrase->aTerm[0].pIter->nData; + } + }else{ + *ppCollist = 0; + *pnCollist = 0; + } + + return rc; +} + + +#line 1 "fts5_hash.c" +/* +** 2014 August 11 +** +** The author disclaims copyright to this source code. In place of +** a legal notice, here is a blessing: +** +** May you do good and not evil. +** May you find forgiveness for yourself and forgive others. +** May you share freely, never taking more than you give. +** +****************************************************************************** +** +*/ + + + +/* #include "fts5Int.h" */ + +typedef struct Fts5HashEntry Fts5HashEntry; + +/* +** This file contains the implementation of an in-memory hash table used +** to accumuluate "term -> doclist" content before it is flused to a level-0 +** segment. +*/ + + +struct Fts5Hash { + int eDetail; /* Copy of Fts5Config.eDetail */ + int *pnByte; /* Pointer to bytes counter */ + int nEntry; /* Number of entries currently in hash */ + int nSlot; /* Size of aSlot[] array */ + Fts5HashEntry *pScan; /* Current ordered scan item */ + Fts5HashEntry **aSlot; /* Array of hash slots */ +}; + +/* +** Each entry in the hash table is represented by an object of the +** following type. Each object, its key (zKey[]) and its current data +** are stored in a single memory allocation. The position list data +** immediately follows the key data in memory. +** +** The data that follows the key is in a similar, but not identical format +** to the doclist data stored in the database. It is: +** +** * Rowid, as a varint +** * Position list, without 0x00 terminator. +** * Size of previous position list and rowid, as a 4 byte +** big-endian integer. +** +** iRowidOff: +** Offset of last rowid written to data area. Relative to first byte of +** structure. +** +** nData: +** Bytes of data written since iRowidOff. +*/ +struct Fts5HashEntry { + Fts5HashEntry *pHashNext; /* Next hash entry with same hash-key */ + Fts5HashEntry *pScanNext; /* Next entry in sorted order */ + + int nAlloc; /* Total size of allocation */ + int iSzPoslist; /* Offset of space for 4-byte poslist size */ + int nData; /* Total bytes of data (incl. structure) */ + int nKey; /* Length of zKey[] in bytes */ + u8 bDel; /* Set delete-flag @ iSzPoslist */ + u8 bContent; /* Set content-flag (detail=none mode) */ + i16 iCol; /* Column of last value written */ + int iPos; /* Position of last value written */ + i64 iRowid; /* Rowid of last value written */ + char zKey[8]; /* Nul-terminated entry key */ +}; + +/* +** Size of Fts5HashEntry without the zKey[] array. +*/ +#define FTS5_HASHENTRYSIZE (sizeof(Fts5HashEntry)-8) + + + +/* +** Allocate a new hash table. +*/ +static int sqlite3Fts5HashNew(Fts5Config *pConfig, Fts5Hash **ppNew, int *pnByte){ + int rc = SQLITE_OK; + Fts5Hash *pNew; + + *ppNew = pNew = (Fts5Hash*)sqlite3_malloc(sizeof(Fts5Hash)); + if( pNew==0 ){ + rc = SQLITE_NOMEM; + }else{ + int nByte; + memset(pNew, 0, sizeof(Fts5Hash)); + pNew->pnByte = pnByte; + pNew->eDetail = pConfig->eDetail; + + pNew->nSlot = 1024; + nByte = sizeof(Fts5HashEntry*) * pNew->nSlot; + pNew->aSlot = (Fts5HashEntry**)sqlite3_malloc(nByte); + if( pNew->aSlot==0 ){ + sqlite3_free(pNew); + *ppNew = 0; + rc = SQLITE_NOMEM; + }else{ + memset(pNew->aSlot, 0, nByte); + } + } + return rc; +} + +/* +** Free a hash table object. +*/ +static void sqlite3Fts5HashFree(Fts5Hash *pHash){ + if( pHash ){ + sqlite3Fts5HashClear(pHash); + sqlite3_free(pHash->aSlot); + sqlite3_free(pHash); + } +} + +/* +** Empty (but do not delete) a hash table. +*/ +static void sqlite3Fts5HashClear(Fts5Hash *pHash){ + int i; + for(i=0; i<pHash->nSlot; i++){ + Fts5HashEntry *pNext; + Fts5HashEntry *pSlot; + for(pSlot=pHash->aSlot[i]; pSlot; pSlot=pNext){ + pNext = pSlot->pHashNext; + sqlite3_free(pSlot); + } + } + memset(pHash->aSlot, 0, pHash->nSlot * sizeof(Fts5HashEntry*)); + pHash->nEntry = 0; +} + +static unsigned int fts5HashKey(int nSlot, const u8 *p, int n){ + int i; + unsigned int h = 13; + for(i=n-1; i>=0; i--){ + h = (h << 3) ^ h ^ p[i]; + } + return (h % nSlot); +} + +static unsigned int fts5HashKey2(int nSlot, u8 b, const u8 *p, int n){ + int i; + unsigned int h = 13; + for(i=n-1; i>=0; i--){ + h = (h << 3) ^ h ^ p[i]; + } + h = (h << 3) ^ h ^ b; + return (h % nSlot); +} + +/* +** Resize the hash table by doubling the number of slots. +*/ +static int fts5HashResize(Fts5Hash *pHash){ + int nNew = pHash->nSlot*2; + int i; + Fts5HashEntry **apNew; + Fts5HashEntry **apOld = pHash->aSlot; + + apNew = (Fts5HashEntry**)sqlite3_malloc(nNew*sizeof(Fts5HashEntry*)); + if( !apNew ) return SQLITE_NOMEM; + memset(apNew, 0, nNew*sizeof(Fts5HashEntry*)); + + for(i=0; i<pHash->nSlot; i++){ + while( apOld[i] ){ + int iHash; + Fts5HashEntry *p = apOld[i]; + apOld[i] = p->pHashNext; + iHash = fts5HashKey(nNew, (u8*)p->zKey, (int)strlen(p->zKey)); + p->pHashNext = apNew[iHash]; + apNew[iHash] = p; + } + } + + sqlite3_free(apOld); + pHash->nSlot = nNew; + pHash->aSlot = apNew; + return SQLITE_OK; +} + +static void fts5HashAddPoslistSize(Fts5Hash *pHash, Fts5HashEntry *p){ + if( p->iSzPoslist ){ + u8 *pPtr = (u8*)p; + if( pHash->eDetail==FTS5_DETAIL_NONE ){ + assert( p->nData==p->iSzPoslist ); + if( p->bDel ){ + pPtr[p->nData++] = 0x00; + if( p->bContent ){ + pPtr[p->nData++] = 0x00; + } + } + }else{ + int nSz = (p->nData - p->iSzPoslist - 1); /* Size in bytes */ + int nPos = nSz*2 + p->bDel; /* Value of nPos field */ + + assert( p->bDel==0 || p->bDel==1 ); + if( nPos<=127 ){ + pPtr[p->iSzPoslist] = (u8)nPos; + }else{ + int nByte = sqlite3Fts5GetVarintLen((u32)nPos); + memmove(&pPtr[p->iSzPoslist + nByte], &pPtr[p->iSzPoslist + 1], nSz); + sqlite3Fts5PutVarint(&pPtr[p->iSzPoslist], nPos); + p->nData += (nByte-1); + } + } + + p->iSzPoslist = 0; + p->bDel = 0; + p->bContent = 0; + } +} + +/* +** Add an entry to the in-memory hash table. The key is the concatenation +** of bByte and (pToken/nToken). The value is (iRowid/iCol/iPos). +** +** (bByte || pToken) -> (iRowid,iCol,iPos) +** +** Or, if iCol is negative, then the value is a delete marker. +*/ +static int sqlite3Fts5HashWrite( + Fts5Hash *pHash, + i64 iRowid, /* Rowid for this entry */ + int iCol, /* Column token appears in (-ve -> delete) */ + int iPos, /* Position of token within column */ + char bByte, /* First byte of token */ + const char *pToken, int nToken /* Token to add or remove to or from index */ +){ + unsigned int iHash; + Fts5HashEntry *p; + u8 *pPtr; + int nIncr = 0; /* Amount to increment (*pHash->pnByte) by */ + int bNew; /* If non-delete entry should be written */ + + bNew = (pHash->eDetail==FTS5_DETAIL_FULL); + + /* Attempt to locate an existing hash entry */ + iHash = fts5HashKey2(pHash->nSlot, (u8)bByte, (const u8*)pToken, nToken); + for(p=pHash->aSlot[iHash]; p; p=p->pHashNext){ + if( p->zKey[0]==bByte + && p->nKey==nToken + && memcmp(&p->zKey[1], pToken, nToken)==0 + ){ + break; + } + } + + /* If an existing hash entry cannot be found, create a new one. */ + if( p==0 ){ + /* Figure out how much space to allocate */ + int nByte = FTS5_HASHENTRYSIZE + (nToken+1) + 1 + 64; + if( nByte<128 ) nByte = 128; + + /* Grow the Fts5Hash.aSlot[] array if necessary. */ + if( (pHash->nEntry*2)>=pHash->nSlot ){ + int rc = fts5HashResize(pHash); + if( rc!=SQLITE_OK ) return rc; + iHash = fts5HashKey2(pHash->nSlot, (u8)bByte, (const u8*)pToken, nToken); + } + + /* Allocate new Fts5HashEntry and add it to the hash table. */ + p = (Fts5HashEntry*)sqlite3_malloc(nByte); + if( !p ) return SQLITE_NOMEM; + memset(p, 0, FTS5_HASHENTRYSIZE); + p->nAlloc = nByte; + p->zKey[0] = bByte; + memcpy(&p->zKey[1], pToken, nToken); + assert( iHash==fts5HashKey(pHash->nSlot, (u8*)p->zKey, nToken+1) ); + p->nKey = nToken; + p->zKey[nToken+1] = '\0'; + p->nData = nToken+1 + 1 + FTS5_HASHENTRYSIZE; + p->pHashNext = pHash->aSlot[iHash]; + pHash->aSlot[iHash] = p; + pHash->nEntry++; + + /* Add the first rowid field to the hash-entry */ + p->nData += sqlite3Fts5PutVarint(&((u8*)p)[p->nData], iRowid); + p->iRowid = iRowid; + + p->iSzPoslist = p->nData; + if( pHash->eDetail!=FTS5_DETAIL_NONE ){ + p->nData += 1; + p->iCol = (pHash->eDetail==FTS5_DETAIL_FULL ? 0 : -1); + } + + nIncr += p->nData; + }else{ + + /* Appending to an existing hash-entry. Check that there is enough + ** space to append the largest possible new entry. Worst case scenario + ** is: + ** + ** + 9 bytes for a new rowid, + ** + 4 byte reserved for the "poslist size" varint. + ** + 1 byte for a "new column" byte, + ** + 3 bytes for a new column number (16-bit max) as a varint, + ** + 5 bytes for the new position offset (32-bit max). + */ + if( (p->nAlloc - p->nData) < (9 + 4 + 1 + 3 + 5) ){ + int nNew = p->nAlloc * 2; + Fts5HashEntry *pNew; + Fts5HashEntry **pp; + pNew = (Fts5HashEntry*)sqlite3_realloc(p, nNew); + if( pNew==0 ) return SQLITE_NOMEM; + pNew->nAlloc = nNew; + for(pp=&pHash->aSlot[iHash]; *pp!=p; pp=&(*pp)->pHashNext); + *pp = pNew; + p = pNew; + } + nIncr -= p->nData; + } + assert( (p->nAlloc - p->nData) >= (9 + 4 + 1 + 3 + 5) ); + + pPtr = (u8*)p; + + /* If this is a new rowid, append the 4-byte size field for the previous + ** entry, and the new rowid for this entry. */ + if( iRowid!=p->iRowid ){ + fts5HashAddPoslistSize(pHash, p); + p->nData += sqlite3Fts5PutVarint(&pPtr[p->nData], iRowid - p->iRowid); + p->iRowid = iRowid; + bNew = 1; + p->iSzPoslist = p->nData; + if( pHash->eDetail!=FTS5_DETAIL_NONE ){ + p->nData += 1; + p->iCol = (pHash->eDetail==FTS5_DETAIL_FULL ? 0 : -1); + p->iPos = 0; + } + } + + if( iCol>=0 ){ + if( pHash->eDetail==FTS5_DETAIL_NONE ){ + p->bContent = 1; + }else{ + /* Append a new column value, if necessary */ + assert( iCol>=p->iCol ); + if( iCol!=p->iCol ){ + if( pHash->eDetail==FTS5_DETAIL_FULL ){ + pPtr[p->nData++] = 0x01; + p->nData += sqlite3Fts5PutVarint(&pPtr[p->nData], iCol); + p->iCol = iCol; + p->iPos = 0; + }else{ + bNew = 1; + p->iCol = iPos = iCol; + } + } + + /* Append the new position offset, if necessary */ + if( bNew ){ + p->nData += sqlite3Fts5PutVarint(&pPtr[p->nData], iPos - p->iPos + 2); + p->iPos = iPos; + } + } + }else{ + /* This is a delete. Set the delete flag. */ + p->bDel = 1; + } + + nIncr += p->nData; + *pHash->pnByte += nIncr; + return SQLITE_OK; +} + + +/* +** Arguments pLeft and pRight point to linked-lists of hash-entry objects, +** each sorted in key order. This function merges the two lists into a +** single list and returns a pointer to its first element. +*/ +static Fts5HashEntry *fts5HashEntryMerge( + Fts5HashEntry *pLeft, + Fts5HashEntry *pRight +){ + Fts5HashEntry *p1 = pLeft; + Fts5HashEntry *p2 = pRight; + Fts5HashEntry *pRet = 0; + Fts5HashEntry **ppOut = &pRet; + + while( p1 || p2 ){ + if( p1==0 ){ + *ppOut = p2; + p2 = 0; + }else if( p2==0 ){ + *ppOut = p1; + p1 = 0; + }else{ + int i = 0; + while( p1->zKey[i]==p2->zKey[i] ) i++; + + if( ((u8)p1->zKey[i])>((u8)p2->zKey[i]) ){ + /* p2 is smaller */ + *ppOut = p2; + ppOut = &p2->pScanNext; + p2 = p2->pScanNext; + }else{ + /* p1 is smaller */ + *ppOut = p1; + ppOut = &p1->pScanNext; + p1 = p1->pScanNext; + } + *ppOut = 0; + } + } + + return pRet; +} + +/* +** Extract all tokens from hash table iHash and link them into a list +** in sorted order. The hash table is cleared before returning. It is +** the responsibility of the caller to free the elements of the returned +** list. +*/ +static int fts5HashEntrySort( + Fts5Hash *pHash, + const char *pTerm, int nTerm, /* Query prefix, if any */ + Fts5HashEntry **ppSorted +){ + const int nMergeSlot = 32; + Fts5HashEntry **ap; + Fts5HashEntry *pList; + int iSlot; + int i; + + *ppSorted = 0; + ap = sqlite3_malloc(sizeof(Fts5HashEntry*) * nMergeSlot); + if( !ap ) return SQLITE_NOMEM; + memset(ap, 0, sizeof(Fts5HashEntry*) * nMergeSlot); + + for(iSlot=0; iSlot<pHash->nSlot; iSlot++){ + Fts5HashEntry *pIter; + for(pIter=pHash->aSlot[iSlot]; pIter; pIter=pIter->pHashNext){ + if( pTerm==0 || 0==memcmp(pIter->zKey, pTerm, nTerm) ){ + Fts5HashEntry *pEntry = pIter; + pEntry->pScanNext = 0; + for(i=0; ap[i]; i++){ + pEntry = fts5HashEntryMerge(pEntry, ap[i]); + ap[i] = 0; + } + ap[i] = pEntry; + } + } + } + + pList = 0; + for(i=0; i<nMergeSlot; i++){ + pList = fts5HashEntryMerge(pList, ap[i]); + } + + pHash->nEntry = 0; + sqlite3_free(ap); + *ppSorted = pList; + return SQLITE_OK; +} + +/* +** Query the hash table for a doclist associated with term pTerm/nTerm. +*/ +static int sqlite3Fts5HashQuery( + Fts5Hash *pHash, /* Hash table to query */ + const char *pTerm, int nTerm, /* Query term */ + const u8 **ppDoclist, /* OUT: Pointer to doclist for pTerm */ + int *pnDoclist /* OUT: Size of doclist in bytes */ +){ + unsigned int iHash = fts5HashKey(pHash->nSlot, (const u8*)pTerm, nTerm); + Fts5HashEntry *p; + + for(p=pHash->aSlot[iHash]; p; p=p->pHashNext){ + if( memcmp(p->zKey, pTerm, nTerm)==0 && p->zKey[nTerm]==0 ) break; + } + + if( p ){ + fts5HashAddPoslistSize(pHash, p); + *ppDoclist = (const u8*)&p->zKey[nTerm+1]; + *pnDoclist = p->nData - (FTS5_HASHENTRYSIZE + nTerm + 1); + }else{ + *ppDoclist = 0; + *pnDoclist = 0; + } + + return SQLITE_OK; +} + +static int sqlite3Fts5HashScanInit( + Fts5Hash *p, /* Hash table to query */ + const char *pTerm, int nTerm /* Query prefix */ +){ + return fts5HashEntrySort(p, pTerm, nTerm, &p->pScan); +} + +static void sqlite3Fts5HashScanNext(Fts5Hash *p){ + assert( !sqlite3Fts5HashScanEof(p) ); + p->pScan = p->pScan->pScanNext; +} + +static int sqlite3Fts5HashScanEof(Fts5Hash *p){ + return (p->pScan==0); +} + +static void sqlite3Fts5HashScanEntry( + Fts5Hash *pHash, + const char **pzTerm, /* OUT: term (nul-terminated) */ + const u8 **ppDoclist, /* OUT: pointer to doclist */ + int *pnDoclist /* OUT: size of doclist in bytes */ +){ + Fts5HashEntry *p; + if( (p = pHash->pScan) ){ + int nTerm = (int)strlen(p->zKey); + fts5HashAddPoslistSize(pHash, p); + *pzTerm = p->zKey; + *ppDoclist = (const u8*)&p->zKey[nTerm+1]; + *pnDoclist = p->nData - (FTS5_HASHENTRYSIZE + nTerm + 1); + }else{ + *pzTerm = 0; + *ppDoclist = 0; + *pnDoclist = 0; + } +} + + +#line 1 "fts5_index.c" +/* +** 2014 May 31 +** +** The author disclaims copyright to this source code. In place of +** a legal notice, here is a blessing: +** +** May you do good and not evil. +** May you find forgiveness for yourself and forgive others. +** May you share freely, never taking more than you give. +** +****************************************************************************** +** +** Low level access to the FTS index stored in the database file. The +** routines in this file file implement all read and write access to the +** %_data table. Other parts of the system access this functionality via +** the interface defined in fts5Int.h. +*/ + + +/* #include "fts5Int.h" */ + +/* +** Overview: +** +** The %_data table contains all the FTS indexes for an FTS5 virtual table. +** As well as the main term index, there may be up to 31 prefix indexes. +** The format is similar to FTS3/4, except that: +** +** * all segment b-tree leaf data is stored in fixed size page records +** (e.g. 1000 bytes). A single doclist may span multiple pages. Care is +** taken to ensure it is possible to iterate in either direction through +** the entries in a doclist, or to seek to a specific entry within a +** doclist, without loading it into memory. +** +** * large doclists that span many pages have associated "doclist index" +** records that contain a copy of the first rowid on each page spanned by +** the doclist. This is used to speed up seek operations, and merges of +** large doclists with very small doclists. +** +** * extra fields in the "structure record" record the state of ongoing +** incremental merge operations. +** +*/ + + +#define FTS5_OPT_WORK_UNIT 1000 /* Number of leaf pages per optimize step */ +#define FTS5_WORK_UNIT 64 /* Number of leaf pages in unit of work */ + +#define FTS5_MIN_DLIDX_SIZE 4 /* Add dlidx if this many empty pages */ + +#define FTS5_MAIN_PREFIX '0' + +#if FTS5_MAX_PREFIX_INDEXES > 31 +# error "FTS5_MAX_PREFIX_INDEXES is too large" +#endif + +/* +** Details: +** +** The %_data table managed by this module, +** +** CREATE TABLE %_data(id INTEGER PRIMARY KEY, block BLOB); +** +** , contains the following 5 types of records. See the comments surrounding +** the FTS5_*_ROWID macros below for a description of how %_data rowids are +** assigned to each fo them. +** +** 1. Structure Records: +** +** The set of segments that make up an index - the index structure - are +** recorded in a single record within the %_data table. The record consists +** of a single 32-bit configuration cookie value followed by a list of +** SQLite varints. If the FTS table features more than one index (because +** there are one or more prefix indexes), it is guaranteed that all share +** the same cookie value. +** +** Immediately following the configuration cookie, the record begins with +** three varints: +** +** + number of levels, +** + total number of segments on all levels, +** + value of write counter. +** +** Then, for each level from 0 to nMax: +** +** + number of input segments in ongoing merge. +** + total number of segments in level. +** + for each segment from oldest to newest: +** + segment id (always > 0) +** + first leaf page number (often 1, always greater than 0) +** + final leaf page number +** +** 2. The Averages Record: +** +** A single record within the %_data table. The data is a list of varints. +** The first value is the number of rows in the index. Then, for each column +** from left to right, the total number of tokens in the column for all +** rows of the table. +** +** 3. Segment leaves: +** +** TERM/DOCLIST FORMAT: +** +** Most of each segment leaf is taken up by term/doclist data. The +** general format of term/doclist, starting with the first term +** on the leaf page, is: +** +** varint : size of first term +** blob: first term data +** doclist: first doclist +** zero-or-more { +** varint: number of bytes in common with previous term +** varint: number of bytes of new term data (nNew) +** blob: nNew bytes of new term data +** doclist: next doclist +** } +** +** doclist format: +** +** varint: first rowid +** poslist: first poslist +** zero-or-more { +** varint: rowid delta (always > 0) +** poslist: next poslist +** } +** +** poslist format: +** +** varint: size of poslist in bytes multiplied by 2, not including +** this field. Plus 1 if this entry carries the "delete" flag. +** collist: collist for column 0 +** zero-or-more { +** 0x01 byte +** varint: column number (I) +** collist: collist for column I +** } +** +** collist format: +** +** varint: first offset + 2 +** zero-or-more { +** varint: offset delta + 2 +** } +** +** PAGE FORMAT +** +** Each leaf page begins with a 4-byte header containing 2 16-bit +** unsigned integer fields in big-endian format. They are: +** +** * The byte offset of the first rowid on the page, if it exists +** and occurs before the first term (otherwise 0). +** +** * The byte offset of the start of the page footer. If the page +** footer is 0 bytes in size, then this field is the same as the +** size of the leaf page in bytes. +** +** The page footer consists of a single varint for each term located +** on the page. Each varint is the byte offset of the current term +** within the page, delta-compressed against the previous value. In +** other words, the first varint in the footer is the byte offset of +** the first term, the second is the byte offset of the second less that +** of the first, and so on. +** +** The term/doclist format described above is accurate if the entire +** term/doclist data fits on a single leaf page. If this is not the case, +** the format is changed in two ways: +** +** + if the first rowid on a page occurs before the first term, it +** is stored as a literal value: +** +** varint: first rowid +** +** + the first term on each page is stored in the same way as the +** very first term of the segment: +** +** varint : size of first term +** blob: first term data +** +** 5. Segment doclist indexes: +** +** Doclist indexes are themselves b-trees, however they usually consist of +** a single leaf record only. The format of each doclist index leaf page +** is: +** +** * Flags byte. Bits are: +** 0x01: Clear if leaf is also the root page, otherwise set. +** +** * Page number of fts index leaf page. As a varint. +** +** * First rowid on page indicated by previous field. As a varint. +** +** * A list of varints, one for each subsequent termless page. A +** positive delta if the termless page contains at least one rowid, +** or an 0x00 byte otherwise. +** +** Internal doclist index nodes are: +** +** * Flags byte. Bits are: +** 0x01: Clear for root page, otherwise set. +** +** * Page number of first child page. As a varint. +** +** * Copy of first rowid on page indicated by previous field. As a varint. +** +** * A list of delta-encoded varints - the first rowid on each subsequent +** child page. +** +*/ + +/* +** Rowids for the averages and structure records in the %_data table. +*/ +#define FTS5_AVERAGES_ROWID 1 /* Rowid used for the averages record */ +#define FTS5_STRUCTURE_ROWID 10 /* The structure record */ + +/* +** Macros determining the rowids used by segment leaves and dlidx leaves +** and nodes. All nodes and leaves are stored in the %_data table with large +** positive rowids. +** +** Each segment has a unique non-zero 16-bit id. +** +** The rowid for each segment leaf is found by passing the segment id and +** the leaf page number to the FTS5_SEGMENT_ROWID macro. Leaves are numbered +** sequentially starting from 1. +*/ +#define FTS5_DATA_ID_B 16 /* Max seg id number 65535 */ +#define FTS5_DATA_DLI_B 1 /* Doclist-index flag (1 bit) */ +#define FTS5_DATA_HEIGHT_B 5 /* Max dlidx tree height of 32 */ +#define FTS5_DATA_PAGE_B 31 /* Max page number of 2147483648 */ + +#define fts5_dri(segid, dlidx, height, pgno) ( \ + ((i64)(segid) << (FTS5_DATA_PAGE_B+FTS5_DATA_HEIGHT_B+FTS5_DATA_DLI_B)) + \ + ((i64)(dlidx) << (FTS5_DATA_PAGE_B + FTS5_DATA_HEIGHT_B)) + \ + ((i64)(height) << (FTS5_DATA_PAGE_B)) + \ + ((i64)(pgno)) \ +) + +#define FTS5_SEGMENT_ROWID(segid, pgno) fts5_dri(segid, 0, 0, pgno) +#define FTS5_DLIDX_ROWID(segid, height, pgno) fts5_dri(segid, 1, height, pgno) + +/* +** Maximum segments permitted in a single index +*/ +#define FTS5_MAX_SEGMENT 2000 + +#ifdef SQLITE_DEBUG +static int sqlite3Fts5Corrupt() { return SQLITE_CORRUPT_VTAB; } +#endif + + +/* +** Each time a blob is read from the %_data table, it is padded with this +** many zero bytes. This makes it easier to decode the various record formats +** without overreading if the records are corrupt. +*/ +#define FTS5_DATA_ZERO_PADDING 8 +#define FTS5_DATA_PADDING 20 + +typedef struct Fts5Data Fts5Data; +typedef struct Fts5DlidxIter Fts5DlidxIter; +typedef struct Fts5DlidxLvl Fts5DlidxLvl; +typedef struct Fts5DlidxWriter Fts5DlidxWriter; +typedef struct Fts5Iter Fts5Iter; +typedef struct Fts5PageWriter Fts5PageWriter; +typedef struct Fts5SegIter Fts5SegIter; +typedef struct Fts5DoclistIter Fts5DoclistIter; +typedef struct Fts5SegWriter Fts5SegWriter; +typedef struct Fts5Structure Fts5Structure; +typedef struct Fts5StructureLevel Fts5StructureLevel; +typedef struct Fts5StructureSegment Fts5StructureSegment; + +struct Fts5Data { + u8 *p; /* Pointer to buffer containing record */ + int nn; /* Size of record in bytes */ + int szLeaf; /* Size of leaf without page-index */ +}; + +/* +** One object per %_data table. +*/ +struct Fts5Index { + Fts5Config *pConfig; /* Virtual table configuration */ + char *zDataTbl; /* Name of %_data table */ + int nWorkUnit; /* Leaf pages in a "unit" of work */ + + /* + ** Variables related to the accumulation of tokens and doclists within the + ** in-memory hash tables before they are flushed to disk. + */ + Fts5Hash *pHash; /* Hash table for in-memory data */ + int nPendingData; /* Current bytes of pending data */ + i64 iWriteRowid; /* Rowid for current doc being written */ + int bDelete; /* Current write is a delete */ + + /* Error state. */ + int rc; /* Current error code */ + + /* State used by the fts5DataXXX() functions. */ + sqlite3_blob *pReader; /* RO incr-blob open on %_data table */ + sqlite3_stmt *pWriter; /* "INSERT ... %_data VALUES(?,?)" */ + sqlite3_stmt *pDeleter; /* "DELETE FROM %_data ... id>=? AND id<=?" */ + sqlite3_stmt *pIdxWriter; /* "INSERT ... %_idx VALUES(?,?,?,?)" */ + sqlite3_stmt *pIdxDeleter; /* "DELETE FROM %_idx WHERE segid=? */ + sqlite3_stmt *pIdxSelect; + int nRead; /* Total number of blocks read */ +}; + +struct Fts5DoclistIter { + u8 *aEof; /* Pointer to 1 byte past end of doclist */ + + /* Output variables. aPoslist==0 at EOF */ + i64 iRowid; + u8 *aPoslist; + int nPoslist; + int nSize; +}; + +/* +** The contents of the "structure" record for each index are represented +** using an Fts5Structure record in memory. Which uses instances of the +** other Fts5StructureXXX types as components. +*/ +struct Fts5StructureSegment { + int iSegid; /* Segment id */ + int pgnoFirst; /* First leaf page number in segment */ + int pgnoLast; /* Last leaf page number in segment */ +}; +struct Fts5StructureLevel { + int nMerge; /* Number of segments in incr-merge */ + int nSeg; /* Total number of segments on level */ + Fts5StructureSegment *aSeg; /* Array of segments. aSeg[0] is oldest. */ +}; +struct Fts5Structure { + int nRef; /* Object reference count */ + u64 nWriteCounter; /* Total leaves written to level 0 */ + int nSegment; /* Total segments in this structure */ + int nLevel; /* Number of levels in this index */ + Fts5StructureLevel aLevel[1]; /* Array of nLevel level objects */ +}; + +/* +** An object of type Fts5SegWriter is used to write to segments. +*/ +struct Fts5PageWriter { + int pgno; /* Page number for this page */ + int iPrevPgidx; /* Previous value written into pgidx */ + Fts5Buffer buf; /* Buffer containing leaf data */ + Fts5Buffer pgidx; /* Buffer containing page-index */ + Fts5Buffer term; /* Buffer containing previous term on page */ +}; +struct Fts5DlidxWriter { + int pgno; /* Page number for this page */ + int bPrevValid; /* True if iPrev is valid */ + i64 iPrev; /* Previous rowid value written to page */ + Fts5Buffer buf; /* Buffer containing page data */ +}; +struct Fts5SegWriter { + int iSegid; /* Segid to write to */ + Fts5PageWriter writer; /* PageWriter object */ + i64 iPrevRowid; /* Previous rowid written to current leaf */ + u8 bFirstRowidInDoclist; /* True if next rowid is first in doclist */ + u8 bFirstRowidInPage; /* True if next rowid is first in page */ + /* TODO1: Can use (writer.pgidx.n==0) instead of bFirstTermInPage */ + u8 bFirstTermInPage; /* True if next term will be first in leaf */ + int nLeafWritten; /* Number of leaf pages written */ + int nEmpty; /* Number of contiguous term-less nodes */ + + int nDlidx; /* Allocated size of aDlidx[] array */ + Fts5DlidxWriter *aDlidx; /* Array of Fts5DlidxWriter objects */ + + /* Values to insert into the %_idx table */ + Fts5Buffer btterm; /* Next term to insert into %_idx table */ + int iBtPage; /* Page number corresponding to btterm */ +}; + +typedef struct Fts5CResult Fts5CResult; +struct Fts5CResult { + u16 iFirst; /* aSeg[] index of firstest iterator */ + u8 bTermEq; /* True if the terms are equal */ +}; + +/* +** Object for iterating through a single segment, visiting each term/rowid +** pair in the segment. +** +** pSeg: +** The segment to iterate through. +** +** iLeafPgno: +** Current leaf page number within segment. +** +** iLeafOffset: +** Byte offset within the current leaf that is the first byte of the +** position list data (one byte passed the position-list size field). +** rowid field of the current entry. Usually this is the size field of the +** position list data. The exception is if the rowid for the current entry +** is the last thing on the leaf page. +** +** pLeaf: +** Buffer containing current leaf page data. Set to NULL at EOF. +** +** iTermLeafPgno, iTermLeafOffset: +** Leaf page number containing the last term read from the segment. And +** the offset immediately following the term data. +** +** flags: +** Mask of FTS5_SEGITER_XXX values. Interpreted as follows: +** +** FTS5_SEGITER_ONETERM: +** If set, set the iterator to point to EOF after the current doclist +** has been exhausted. Do not proceed to the next term in the segment. +** +** FTS5_SEGITER_REVERSE: +** This flag is only ever set if FTS5_SEGITER_ONETERM is also set. If +** it is set, iterate through rowid in descending order instead of the +** default ascending order. +** +** iRowidOffset/nRowidOffset/aRowidOffset: +** These are used if the FTS5_SEGITER_REVERSE flag is set. +** +** For each rowid on the page corresponding to the current term, the +** corresponding aRowidOffset[] entry is set to the byte offset of the +** start of the "position-list-size" field within the page. +** +** iTermIdx: +** Index of current term on iTermLeafPgno. +*/ +struct Fts5SegIter { + Fts5StructureSegment *pSeg; /* Segment to iterate through */ + int flags; /* Mask of configuration flags */ + int iLeafPgno; /* Current leaf page number */ + Fts5Data *pLeaf; /* Current leaf data */ + Fts5Data *pNextLeaf; /* Leaf page (iLeafPgno+1) */ + int iLeafOffset; /* Byte offset within current leaf */ + + /* Next method */ + void (*xNext)(Fts5Index*, Fts5SegIter*, int*); + + /* The page and offset from which the current term was read. The offset + ** is the offset of the first rowid in the current doclist. */ + int iTermLeafPgno; + int iTermLeafOffset; + + int iPgidxOff; /* Next offset in pgidx */ + int iEndofDoclist; + + /* The following are only used if the FTS5_SEGITER_REVERSE flag is set. */ + int iRowidOffset; /* Current entry in aRowidOffset[] */ + int nRowidOffset; /* Allocated size of aRowidOffset[] array */ + int *aRowidOffset; /* Array of offset to rowid fields */ + + Fts5DlidxIter *pDlidx; /* If there is a doclist-index */ + + /* Variables populated based on current entry. */ + Fts5Buffer term; /* Current term */ + i64 iRowid; /* Current rowid */ + int nPos; /* Number of bytes in current position list */ + u8 bDel; /* True if the delete flag is set */ +}; + +/* +** Argument is a pointer to an Fts5Data structure that contains a +** leaf page. +*/ +#define ASSERT_SZLEAF_OK(x) assert( \ + (x)->szLeaf==(x)->nn || (x)->szLeaf==fts5GetU16(&(x)->p[2]) \ +) + +#define FTS5_SEGITER_ONETERM 0x01 +#define FTS5_SEGITER_REVERSE 0x02 + +/* +** Argument is a pointer to an Fts5Data structure that contains a leaf +** page. This macro evaluates to true if the leaf contains no terms, or +** false if it contains at least one term. +*/ +#define fts5LeafIsTermless(x) ((x)->szLeaf >= (x)->nn) + +#define fts5LeafTermOff(x, i) (fts5GetU16(&(x)->p[(x)->szLeaf + (i)*2])) + +#define fts5LeafFirstRowidOff(x) (fts5GetU16((x)->p)) + +/* +** Object for iterating through the merged results of one or more segments, +** visiting each term/rowid pair in the merged data. +** +** nSeg is always a power of two greater than or equal to the number of +** segments that this object is merging data from. Both the aSeg[] and +** aFirst[] arrays are sized at nSeg entries. The aSeg[] array is padded +** with zeroed objects - these are handled as if they were iterators opened +** on empty segments. +** +** The results of comparing segments aSeg[N] and aSeg[N+1], where N is an +** even number, is stored in aFirst[(nSeg+N)/2]. The "result" of the +** comparison in this context is the index of the iterator that currently +** points to the smaller term/rowid combination. Iterators at EOF are +** considered to be greater than all other iterators. +** +** aFirst[1] contains the index in aSeg[] of the iterator that points to +** the smallest key overall. aFirst[0] is unused. +** +** poslist: +** Used by sqlite3Fts5IterPoslist() when the poslist needs to be buffered. +** There is no way to tell if this is populated or not. +*/ +struct Fts5Iter { + Fts5IndexIter base; /* Base class containing output vars */ + + Fts5Index *pIndex; /* Index that owns this iterator */ + Fts5Structure *pStruct; /* Database structure for this iterator */ + Fts5Buffer poslist; /* Buffer containing current poslist */ + Fts5Colset *pColset; /* Restrict matches to these columns */ + + /* Invoked to set output variables. */ + void (*xSetOutputs)(Fts5Iter*, Fts5SegIter*); + + int nSeg; /* Size of aSeg[] array */ + int bRev; /* True to iterate in reverse order */ + u8 bSkipEmpty; /* True to skip deleted entries */ + + i64 iSwitchRowid; /* Firstest rowid of other than aFirst[1] */ + Fts5CResult *aFirst; /* Current merge state (see above) */ + Fts5SegIter aSeg[1]; /* Array of segment iterators */ +}; + + +/* +** An instance of the following type is used to iterate through the contents +** of a doclist-index record. +** +** pData: +** Record containing the doclist-index data. +** +** bEof: +** Set to true once iterator has reached EOF. +** +** iOff: +** Set to the current offset within record pData. +*/ +struct Fts5DlidxLvl { + Fts5Data *pData; /* Data for current page of this level */ + int iOff; /* Current offset into pData */ + int bEof; /* At EOF already */ + int iFirstOff; /* Used by reverse iterators */ + + /* Output variables */ + int iLeafPgno; /* Page number of current leaf page */ + i64 iRowid; /* First rowid on leaf iLeafPgno */ +}; +struct Fts5DlidxIter { + int nLvl; + int iSegid; + Fts5DlidxLvl aLvl[1]; +}; + +static void fts5PutU16(u8 *aOut, u16 iVal){ + aOut[0] = (iVal>>8); + aOut[1] = (iVal&0xFF); +} + +static u16 fts5GetU16(const u8 *aIn){ + return ((u16)aIn[0] << 8) + aIn[1]; +} + +/* +** Allocate and return a buffer at least nByte bytes in size. +** +** If an OOM error is encountered, return NULL and set the error code in +** the Fts5Index handle passed as the first argument. +*/ +static void *fts5IdxMalloc(Fts5Index *p, int nByte){ + return sqlite3Fts5MallocZero(&p->rc, nByte); +} + +/* +** Compare the contents of the pLeft buffer with the pRight/nRight blob. +** +** Return -ve if pLeft is smaller than pRight, 0 if they are equal or +** +ve if pRight is smaller than pLeft. In other words: +** +** res = *pLeft - *pRight +*/ +#ifdef SQLITE_DEBUG +static int fts5BufferCompareBlob( + Fts5Buffer *pLeft, /* Left hand side of comparison */ + const u8 *pRight, int nRight /* Right hand side of comparison */ +){ + int nCmp = MIN(pLeft->n, nRight); + int res = memcmp(pLeft->p, pRight, nCmp); + return (res==0 ? (pLeft->n - nRight) : res); +} +#endif + +/* +** Compare the contents of the two buffers using memcmp(). If one buffer +** is a prefix of the other, it is considered the lesser. +** +** Return -ve if pLeft is smaller than pRight, 0 if they are equal or +** +ve if pRight is smaller than pLeft. In other words: +** +** res = *pLeft - *pRight +*/ +static int fts5BufferCompare(Fts5Buffer *pLeft, Fts5Buffer *pRight){ + int nCmp = MIN(pLeft->n, pRight->n); + int res = memcmp(pLeft->p, pRight->p, nCmp); + return (res==0 ? (pLeft->n - pRight->n) : res); +} + +static int fts5LeafFirstTermOff(Fts5Data *pLeaf){ + int ret; + fts5GetVarint32(&pLeaf->p[pLeaf->szLeaf], ret); + return ret; +} + +/* +** Close the read-only blob handle, if it is open. +*/ +static void fts5CloseReader(Fts5Index *p){ + if( p->pReader ){ + sqlite3_blob *pReader = p->pReader; + p->pReader = 0; + sqlite3_blob_close(pReader); + } +} + + +/* +** Retrieve a record from the %_data table. +** +** If an error occurs, NULL is returned and an error left in the +** Fts5Index object. +*/ +static Fts5Data *fts5DataRead(Fts5Index *p, i64 iRowid){ + Fts5Data *pRet = 0; + if( p->rc==SQLITE_OK ){ + int rc = SQLITE_OK; + + if( p->pReader ){ + /* This call may return SQLITE_ABORT if there has been a savepoint + ** rollback since it was last used. In this case a new blob handle + ** is required. */ + sqlite3_blob *pBlob = p->pReader; + p->pReader = 0; + rc = sqlite3_blob_reopen(pBlob, iRowid); + assert( p->pReader==0 ); + p->pReader = pBlob; + if( rc!=SQLITE_OK ){ + fts5CloseReader(p); + } + if( rc==SQLITE_ABORT ) rc = SQLITE_OK; + } + + /* If the blob handle is not open at this point, open it and seek + ** to the requested entry. */ + if( p->pReader==0 && rc==SQLITE_OK ){ + Fts5Config *pConfig = p->pConfig; + rc = sqlite3_blob_open(pConfig->db, + pConfig->zDb, p->zDataTbl, "block", iRowid, 0, &p->pReader + ); + } + + /* If either of the sqlite3_blob_open() or sqlite3_blob_reopen() calls + ** above returned SQLITE_ERROR, return SQLITE_CORRUPT_VTAB instead. + ** All the reasons those functions might return SQLITE_ERROR - missing + ** table, missing row, non-blob/text in block column - indicate + ** backing store corruption. */ + if( rc==SQLITE_ERROR ) rc = FTS5_CORRUPT; + + if( rc==SQLITE_OK ){ + u8 *aOut = 0; /* Read blob data into this buffer */ + int nByte = sqlite3_blob_bytes(p->pReader); + int nAlloc = sizeof(Fts5Data) + nByte + FTS5_DATA_PADDING; + pRet = (Fts5Data*)sqlite3_malloc(nAlloc); + if( pRet ){ + pRet->nn = nByte; + aOut = pRet->p = (u8*)&pRet[1]; + }else{ + rc = SQLITE_NOMEM; + } + + if( rc==SQLITE_OK ){ + rc = sqlite3_blob_read(p->pReader, aOut, nByte, 0); + } + if( rc!=SQLITE_OK ){ + sqlite3_free(pRet); + pRet = 0; + }else{ + /* TODO1: Fix this */ + pRet->szLeaf = fts5GetU16(&pRet->p[2]); + } + } + p->rc = rc; + p->nRead++; + } + + assert( (pRet==0)==(p->rc!=SQLITE_OK) ); + return pRet; +} + +/* +** Release a reference to data record returned by an earlier call to +** fts5DataRead(). +*/ +static void fts5DataRelease(Fts5Data *pData){ + sqlite3_free(pData); +} + +static int fts5IndexPrepareStmt( + Fts5Index *p, + sqlite3_stmt **ppStmt, + char *zSql +){ + if( p->rc==SQLITE_OK ){ + if( zSql ){ + p->rc = sqlite3_prepare_v2(p->pConfig->db, zSql, -1, ppStmt, 0); + }else{ + p->rc = SQLITE_NOMEM; + } + } + sqlite3_free(zSql); + return p->rc; +} + + +/* +** INSERT OR REPLACE a record into the %_data table. +*/ +static void fts5DataWrite(Fts5Index *p, i64 iRowid, const u8 *pData, int nData){ + if( p->rc!=SQLITE_OK ) return; + + if( p->pWriter==0 ){ + Fts5Config *pConfig = p->pConfig; + fts5IndexPrepareStmt(p, &p->pWriter, sqlite3_mprintf( + "REPLACE INTO '%q'.'%q_data'(id, block) VALUES(?,?)", + pConfig->zDb, pConfig->zName + )); + if( p->rc ) return; + } + + sqlite3_bind_int64(p->pWriter, 1, iRowid); + sqlite3_bind_blob(p->pWriter, 2, pData, nData, SQLITE_STATIC); + sqlite3_step(p->pWriter); + p->rc = sqlite3_reset(p->pWriter); +} + +/* +** Execute the following SQL: +** +** DELETE FROM %_data WHERE id BETWEEN $iFirst AND $iLast +*/ +static void fts5DataDelete(Fts5Index *p, i64 iFirst, i64 iLast){ + if( p->rc!=SQLITE_OK ) return; + + if( p->pDeleter==0 ){ + int rc; + Fts5Config *pConfig = p->pConfig; + char *zSql = sqlite3_mprintf( + "DELETE FROM '%q'.'%q_data' WHERE id>=? AND id<=?", + pConfig->zDb, pConfig->zName + ); + if( zSql==0 ){ + rc = SQLITE_NOMEM; + }else{ + rc = sqlite3_prepare_v2(pConfig->db, zSql, -1, &p->pDeleter, 0); + sqlite3_free(zSql); + } + if( rc!=SQLITE_OK ){ + p->rc = rc; + return; + } + } + + sqlite3_bind_int64(p->pDeleter, 1, iFirst); + sqlite3_bind_int64(p->pDeleter, 2, iLast); + sqlite3_step(p->pDeleter); + p->rc = sqlite3_reset(p->pDeleter); +} + +/* +** Remove all records associated with segment iSegid. +*/ +static void fts5DataRemoveSegment(Fts5Index *p, int iSegid){ + i64 iFirst = FTS5_SEGMENT_ROWID(iSegid, 0); + i64 iLast = FTS5_SEGMENT_ROWID(iSegid+1, 0)-1; + fts5DataDelete(p, iFirst, iLast); + if( p->pIdxDeleter==0 ){ + Fts5Config *pConfig = p->pConfig; + fts5IndexPrepareStmt(p, &p->pIdxDeleter, sqlite3_mprintf( + "DELETE FROM '%q'.'%q_idx' WHERE segid=?", + pConfig->zDb, pConfig->zName + )); + } + if( p->rc==SQLITE_OK ){ + sqlite3_bind_int(p->pIdxDeleter, 1, iSegid); + sqlite3_step(p->pIdxDeleter); + p->rc = sqlite3_reset(p->pIdxDeleter); + } +} + +/* +** Release a reference to an Fts5Structure object returned by an earlier +** call to fts5StructureRead() or fts5StructureDecode(). +*/ +static void fts5StructureRelease(Fts5Structure *pStruct){ + if( pStruct && 0>=(--pStruct->nRef) ){ + int i; + assert( pStruct->nRef==0 ); + for(i=0; i<pStruct->nLevel; i++){ + sqlite3_free(pStruct->aLevel[i].aSeg); + } + sqlite3_free(pStruct); + } +} + +static void fts5StructureRef(Fts5Structure *pStruct){ + pStruct->nRef++; +} + +/* +** Deserialize and return the structure record currently stored in serialized +** form within buffer pData/nData. +** +** The Fts5Structure.aLevel[] and each Fts5StructureLevel.aSeg[] array +** are over-allocated by one slot. This allows the structure contents +** to be more easily edited. +** +** If an error occurs, *ppOut is set to NULL and an SQLite error code +** returned. Otherwise, *ppOut is set to point to the new object and +** SQLITE_OK returned. +*/ +static int fts5StructureDecode( + const u8 *pData, /* Buffer containing serialized structure */ + int nData, /* Size of buffer pData in bytes */ + int *piCookie, /* Configuration cookie value */ + Fts5Structure **ppOut /* OUT: Deserialized object */ +){ + int rc = SQLITE_OK; + int i = 0; + int iLvl; + int nLevel = 0; + int nSegment = 0; + int nByte; /* Bytes of space to allocate at pRet */ + Fts5Structure *pRet = 0; /* Structure object to return */ + + /* Grab the cookie value */ + if( piCookie ) *piCookie = sqlite3Fts5Get32(pData); + i = 4; + + /* Read the total number of levels and segments from the start of the + ** structure record. */ + i += fts5GetVarint32(&pData[i], nLevel); + i += fts5GetVarint32(&pData[i], nSegment); + nByte = ( + sizeof(Fts5Structure) + /* Main structure */ + sizeof(Fts5StructureLevel) * (nLevel-1) /* aLevel[] array */ + ); + pRet = (Fts5Structure*)sqlite3Fts5MallocZero(&rc, nByte); + + if( pRet ){ + pRet->nRef = 1; + pRet->nLevel = nLevel; + pRet->nSegment = nSegment; + i += sqlite3Fts5GetVarint(&pData[i], &pRet->nWriteCounter); + + for(iLvl=0; rc==SQLITE_OK && iLvl<nLevel; iLvl++){ + Fts5StructureLevel *pLvl = &pRet->aLevel[iLvl]; + int nTotal = 0; + int iSeg; + + if( i>=nData ){ + rc = FTS5_CORRUPT; + }else{ + i += fts5GetVarint32(&pData[i], pLvl->nMerge); + i += fts5GetVarint32(&pData[i], nTotal); + assert( nTotal>=pLvl->nMerge ); + pLvl->aSeg = (Fts5StructureSegment*)sqlite3Fts5MallocZero(&rc, + nTotal * sizeof(Fts5StructureSegment) + ); + } + + if( rc==SQLITE_OK ){ + pLvl->nSeg = nTotal; + for(iSeg=0; iSeg<nTotal; iSeg++){ + if( i>=nData ){ + rc = FTS5_CORRUPT; + break; + } + i += fts5GetVarint32(&pData[i], pLvl->aSeg[iSeg].iSegid); + i += fts5GetVarint32(&pData[i], pLvl->aSeg[iSeg].pgnoFirst); + i += fts5GetVarint32(&pData[i], pLvl->aSeg[iSeg].pgnoLast); + } + } + } + if( rc!=SQLITE_OK ){ + fts5StructureRelease(pRet); + pRet = 0; + } + } + + *ppOut = pRet; + return rc; +} + +/* +** +*/ +static void fts5StructureAddLevel(int *pRc, Fts5Structure **ppStruct){ + if( *pRc==SQLITE_OK ){ + Fts5Structure *pStruct = *ppStruct; + int nLevel = pStruct->nLevel; + int nByte = ( + sizeof(Fts5Structure) + /* Main structure */ + sizeof(Fts5StructureLevel) * (nLevel+1) /* aLevel[] array */ + ); + + pStruct = sqlite3_realloc(pStruct, nByte); + if( pStruct ){ + memset(&pStruct->aLevel[nLevel], 0, sizeof(Fts5StructureLevel)); + pStruct->nLevel++; + *ppStruct = pStruct; + }else{ + *pRc = SQLITE_NOMEM; + } + } +} + +/* +** Extend level iLvl so that there is room for at least nExtra more +** segments. +*/ +static void fts5StructureExtendLevel( + int *pRc, + Fts5Structure *pStruct, + int iLvl, + int nExtra, + int bInsert +){ + if( *pRc==SQLITE_OK ){ + Fts5StructureLevel *pLvl = &pStruct->aLevel[iLvl]; + Fts5StructureSegment *aNew; + int nByte; + + nByte = (pLvl->nSeg + nExtra) * sizeof(Fts5StructureSegment); + aNew = sqlite3_realloc(pLvl->aSeg, nByte); + if( aNew ){ + if( bInsert==0 ){ + memset(&aNew[pLvl->nSeg], 0, sizeof(Fts5StructureSegment) * nExtra); + }else{ + int nMove = pLvl->nSeg * sizeof(Fts5StructureSegment); + memmove(&aNew[nExtra], aNew, nMove); + memset(aNew, 0, sizeof(Fts5StructureSegment) * nExtra); + } + pLvl->aSeg = aNew; + }else{ + *pRc = SQLITE_NOMEM; + } + } +} + +/* +** Read, deserialize and return the structure record. +** +** The Fts5Structure.aLevel[] and each Fts5StructureLevel.aSeg[] array +** are over-allocated as described for function fts5StructureDecode() +** above. +** +** If an error occurs, NULL is returned and an error code left in the +** Fts5Index handle. If an error has already occurred when this function +** is called, it is a no-op. +*/ +static Fts5Structure *fts5StructureRead(Fts5Index *p){ + Fts5Config *pConfig = p->pConfig; + Fts5Structure *pRet = 0; /* Object to return */ + int iCookie; /* Configuration cookie */ + Fts5Data *pData; + + pData = fts5DataRead(p, FTS5_STRUCTURE_ROWID); + if( p->rc ) return 0; + /* TODO: Do we need this if the leaf-index is appended? Probably... */ + memset(&pData->p[pData->nn], 0, FTS5_DATA_PADDING); + p->rc = fts5StructureDecode(pData->p, pData->nn, &iCookie, &pRet); + if( p->rc==SQLITE_OK && pConfig->iCookie!=iCookie ){ + p->rc = sqlite3Fts5ConfigLoad(pConfig, iCookie); + } + + fts5DataRelease(pData); + if( p->rc!=SQLITE_OK ){ + fts5StructureRelease(pRet); + pRet = 0; + } + return pRet; +} + +/* +** Return the total number of segments in index structure pStruct. This +** function is only ever used as part of assert() conditions. +*/ +#ifdef SQLITE_DEBUG +static int fts5StructureCountSegments(Fts5Structure *pStruct){ + int nSegment = 0; /* Total number of segments */ + if( pStruct ){ + int iLvl; /* Used to iterate through levels */ + for(iLvl=0; iLvl<pStruct->nLevel; iLvl++){ + nSegment += pStruct->aLevel[iLvl].nSeg; + } + } + + return nSegment; +} +#endif + +#define fts5BufferSafeAppendBlob(pBuf, pBlob, nBlob) { \ + assert( (pBuf)->nSpace>=((pBuf)->n+nBlob) ); \ + memcpy(&(pBuf)->p[(pBuf)->n], pBlob, nBlob); \ + (pBuf)->n += nBlob; \ +} + +#define fts5BufferSafeAppendVarint(pBuf, iVal) { \ + (pBuf)->n += sqlite3Fts5PutVarint(&(pBuf)->p[(pBuf)->n], (iVal)); \ + assert( (pBuf)->nSpace>=(pBuf)->n ); \ +} + + +/* +** Serialize and store the "structure" record. +** +** If an error occurs, leave an error code in the Fts5Index object. If an +** error has already occurred, this function is a no-op. +*/ +static void fts5StructureWrite(Fts5Index *p, Fts5Structure *pStruct){ + if( p->rc==SQLITE_OK ){ + Fts5Buffer buf; /* Buffer to serialize record into */ + int iLvl; /* Used to iterate through levels */ + int iCookie; /* Cookie value to store */ + + assert( pStruct->nSegment==fts5StructureCountSegments(pStruct) ); + memset(&buf, 0, sizeof(Fts5Buffer)); + + /* Append the current configuration cookie */ + iCookie = p->pConfig->iCookie; + if( iCookie<0 ) iCookie = 0; + + if( 0==sqlite3Fts5BufferSize(&p->rc, &buf, 4+9+9+9) ){ + sqlite3Fts5Put32(buf.p, iCookie); + buf.n = 4; + fts5BufferSafeAppendVarint(&buf, pStruct->nLevel); + fts5BufferSafeAppendVarint(&buf, pStruct->nSegment); + fts5BufferSafeAppendVarint(&buf, (i64)pStruct->nWriteCounter); + } + + for(iLvl=0; iLvl<pStruct->nLevel; iLvl++){ + int iSeg; /* Used to iterate through segments */ + Fts5StructureLevel *pLvl = &pStruct->aLevel[iLvl]; + fts5BufferAppendVarint(&p->rc, &buf, pLvl->nMerge); + fts5BufferAppendVarint(&p->rc, &buf, pLvl->nSeg); + assert( pLvl->nMerge<=pLvl->nSeg ); + + for(iSeg=0; iSeg<pLvl->nSeg; iSeg++){ + fts5BufferAppendVarint(&p->rc, &buf, pLvl->aSeg[iSeg].iSegid); + fts5BufferAppendVarint(&p->rc, &buf, pLvl->aSeg[iSeg].pgnoFirst); + fts5BufferAppendVarint(&p->rc, &buf, pLvl->aSeg[iSeg].pgnoLast); + } + } + + fts5DataWrite(p, FTS5_STRUCTURE_ROWID, buf.p, buf.n); + fts5BufferFree(&buf); + } +} + +#if 0 +static void fts5DebugStructure(int*,Fts5Buffer*,Fts5Structure*); +static void fts5PrintStructure(const char *zCaption, Fts5Structure *pStruct){ + int rc = SQLITE_OK; + Fts5Buffer buf; + memset(&buf, 0, sizeof(buf)); + fts5DebugStructure(&rc, &buf, pStruct); + fprintf(stdout, "%s: %s\n", zCaption, buf.p); + fflush(stdout); + fts5BufferFree(&buf); +} +#else +# define fts5PrintStructure(x,y) +#endif + +static int fts5SegmentSize(Fts5StructureSegment *pSeg){ + return 1 + pSeg->pgnoLast - pSeg->pgnoFirst; +} + +/* +** Return a copy of index structure pStruct. Except, promote as many +** segments as possible to level iPromote. If an OOM occurs, NULL is +** returned. +*/ +static void fts5StructurePromoteTo( + Fts5Index *p, + int iPromote, + int szPromote, + Fts5Structure *pStruct +){ + int il, is; + Fts5StructureLevel *pOut = &pStruct->aLevel[iPromote]; + + if( pOut->nMerge==0 ){ + for(il=iPromote+1; il<pStruct->nLevel; il++){ + Fts5StructureLevel *pLvl = &pStruct->aLevel[il]; + if( pLvl->nMerge ) return; + for(is=pLvl->nSeg-1; is>=0; is--){ + int sz = fts5SegmentSize(&pLvl->aSeg[is]); + if( sz>szPromote ) return; + fts5StructureExtendLevel(&p->rc, pStruct, iPromote, 1, 1); + if( p->rc ) return; + memcpy(pOut->aSeg, &pLvl->aSeg[is], sizeof(Fts5StructureSegment)); + pOut->nSeg++; + pLvl->nSeg--; + } + } + } +} + +/* +** A new segment has just been written to level iLvl of index structure +** pStruct. This function determines if any segments should be promoted +** as a result. Segments are promoted in two scenarios: +** +** a) If the segment just written is smaller than one or more segments +** within the previous populated level, it is promoted to the previous +** populated level. +** +** b) If the segment just written is larger than the newest segment on +** the next populated level, then that segment, and any other adjacent +** segments that are also smaller than the one just written, are +** promoted. +** +** If one or more segments are promoted, the structure object is updated +** to reflect this. +*/ +static void fts5StructurePromote( + Fts5Index *p, /* FTS5 backend object */ + int iLvl, /* Index level just updated */ + Fts5Structure *pStruct /* Index structure */ +){ + if( p->rc==SQLITE_OK ){ + int iTst; + int iPromote = -1; + int szPromote = 0; /* Promote anything this size or smaller */ + Fts5StructureSegment *pSeg; /* Segment just written */ + int szSeg; /* Size of segment just written */ + int nSeg = pStruct->aLevel[iLvl].nSeg; + + if( nSeg==0 ) return; + pSeg = &pStruct->aLevel[iLvl].aSeg[pStruct->aLevel[iLvl].nSeg-1]; + szSeg = (1 + pSeg->pgnoLast - pSeg->pgnoFirst); + + /* Check for condition (a) */ + for(iTst=iLvl-1; iTst>=0 && pStruct->aLevel[iTst].nSeg==0; iTst--); + if( iTst>=0 ){ + int i; + int szMax = 0; + Fts5StructureLevel *pTst = &pStruct->aLevel[iTst]; + assert( pTst->nMerge==0 ); + for(i=0; i<pTst->nSeg; i++){ + int sz = pTst->aSeg[i].pgnoLast - pTst->aSeg[i].pgnoFirst + 1; + if( sz>szMax ) szMax = sz; + } + if( szMax>=szSeg ){ + /* Condition (a) is true. Promote the newest segment on level + ** iLvl to level iTst. */ + iPromote = iTst; + szPromote = szMax; + } + } + + /* If condition (a) is not met, assume (b) is true. StructurePromoteTo() + ** is a no-op if it is not. */ + if( iPromote<0 ){ + iPromote = iLvl; + szPromote = szSeg; + } + fts5StructurePromoteTo(p, iPromote, szPromote, pStruct); + } +} + + +/* +** Advance the iterator passed as the only argument. If the end of the +** doclist-index page is reached, return non-zero. +*/ +static int fts5DlidxLvlNext(Fts5DlidxLvl *pLvl){ + Fts5Data *pData = pLvl->pData; + + if( pLvl->iOff==0 ){ + assert( pLvl->bEof==0 ); + pLvl->iOff = 1; + pLvl->iOff += fts5GetVarint32(&pData->p[1], pLvl->iLeafPgno); + pLvl->iOff += fts5GetVarint(&pData->p[pLvl->iOff], (u64*)&pLvl->iRowid); + pLvl->iFirstOff = pLvl->iOff; + }else{ + int iOff; + for(iOff=pLvl->iOff; iOff<pData->nn; iOff++){ + if( pData->p[iOff] ) break; + } + + if( iOff<pData->nn ){ + i64 iVal; + pLvl->iLeafPgno += (iOff - pLvl->iOff) + 1; + iOff += fts5GetVarint(&pData->p[iOff], (u64*)&iVal); + pLvl->iRowid += iVal; + pLvl->iOff = iOff; + }else{ + pLvl->bEof = 1; + } + } + + return pLvl->bEof; +} + +/* +** Advance the iterator passed as the only argument. +*/ +static int fts5DlidxIterNextR(Fts5Index *p, Fts5DlidxIter *pIter, int iLvl){ + Fts5DlidxLvl *pLvl = &pIter->aLvl[iLvl]; + + assert( iLvl<pIter->nLvl ); + if( fts5DlidxLvlNext(pLvl) ){ + if( (iLvl+1) < pIter->nLvl ){ + fts5DlidxIterNextR(p, pIter, iLvl+1); + if( pLvl[1].bEof==0 ){ + fts5DataRelease(pLvl->pData); + memset(pLvl, 0, sizeof(Fts5DlidxLvl)); + pLvl->pData = fts5DataRead(p, + FTS5_DLIDX_ROWID(pIter->iSegid, iLvl, pLvl[1].iLeafPgno) + ); + if( pLvl->pData ) fts5DlidxLvlNext(pLvl); + } + } + } + + return pIter->aLvl[0].bEof; +} +static int fts5DlidxIterNext(Fts5Index *p, Fts5DlidxIter *pIter){ + return fts5DlidxIterNextR(p, pIter, 0); +} + +/* +** The iterator passed as the first argument has the following fields set +** as follows. This function sets up the rest of the iterator so that it +** points to the first rowid in the doclist-index. +** +** pData: +** pointer to doclist-index record, +** +** When this function is called pIter->iLeafPgno is the page number the +** doclist is associated with (the one featuring the term). +*/ +static int fts5DlidxIterFirst(Fts5DlidxIter *pIter){ + int i; + for(i=0; i<pIter->nLvl; i++){ + fts5DlidxLvlNext(&pIter->aLvl[i]); + } + return pIter->aLvl[0].bEof; +} + + +static int fts5DlidxIterEof(Fts5Index *p, Fts5DlidxIter *pIter){ + return p->rc!=SQLITE_OK || pIter->aLvl[0].bEof; +} + +static void fts5DlidxIterLast(Fts5Index *p, Fts5DlidxIter *pIter){ + int i; + + /* Advance each level to the last entry on the last page */ + for(i=pIter->nLvl-1; p->rc==SQLITE_OK && i>=0; i--){ + Fts5DlidxLvl *pLvl = &pIter->aLvl[i]; + while( fts5DlidxLvlNext(pLvl)==0 ); + pLvl->bEof = 0; + + if( i>0 ){ + Fts5DlidxLvl *pChild = &pLvl[-1]; + fts5DataRelease(pChild->pData); + memset(pChild, 0, sizeof(Fts5DlidxLvl)); + pChild->pData = fts5DataRead(p, + FTS5_DLIDX_ROWID(pIter->iSegid, i-1, pLvl->iLeafPgno) + ); + } + } +} + +/* +** Move the iterator passed as the only argument to the previous entry. +*/ +static int fts5DlidxLvlPrev(Fts5DlidxLvl *pLvl){ + int iOff = pLvl->iOff; + + assert( pLvl->bEof==0 ); + if( iOff<=pLvl->iFirstOff ){ + pLvl->bEof = 1; + }else{ + u8 *a = pLvl->pData->p; + i64 iVal; + int iLimit; + int ii; + int nZero = 0; + + /* Currently iOff points to the first byte of a varint. This block + ** decrements iOff until it points to the first byte of the previous + ** varint. Taking care not to read any memory locations that occur + ** before the buffer in memory. */ + iLimit = (iOff>9 ? iOff-9 : 0); + for(iOff--; iOff>iLimit; iOff--){ + if( (a[iOff-1] & 0x80)==0 ) break; + } + + fts5GetVarint(&a[iOff], (u64*)&iVal); + pLvl->iRowid -= iVal; + pLvl->iLeafPgno--; + + /* Skip backwards past any 0x00 varints. */ + for(ii=iOff-1; ii>=pLvl->iFirstOff && a[ii]==0x00; ii--){ + nZero++; + } + if( ii>=pLvl->iFirstOff && (a[ii] & 0x80) ){ + /* The byte immediately before the last 0x00 byte has the 0x80 bit + ** set. So the last 0x00 is only a varint 0 if there are 8 more 0x80 + ** bytes before a[ii]. */ + int bZero = 0; /* True if last 0x00 counts */ + if( (ii-8)>=pLvl->iFirstOff ){ + int j; + for(j=1; j<=8 && (a[ii-j] & 0x80); j++); + bZero = (j>8); + } + if( bZero==0 ) nZero--; + } + pLvl->iLeafPgno -= nZero; + pLvl->iOff = iOff - nZero; + } + + return pLvl->bEof; +} + +static int fts5DlidxIterPrevR(Fts5Index *p, Fts5DlidxIter *pIter, int iLvl){ + Fts5DlidxLvl *pLvl = &pIter->aLvl[iLvl]; + + assert( iLvl<pIter->nLvl ); + if( fts5DlidxLvlPrev(pLvl) ){ + if( (iLvl+1) < pIter->nLvl ){ + fts5DlidxIterPrevR(p, pIter, iLvl+1); + if( pLvl[1].bEof==0 ){ + fts5DataRelease(pLvl->pData); + memset(pLvl, 0, sizeof(Fts5DlidxLvl)); + pLvl->pData = fts5DataRead(p, + FTS5_DLIDX_ROWID(pIter->iSegid, iLvl, pLvl[1].iLeafPgno) + ); + if( pLvl->pData ){ + while( fts5DlidxLvlNext(pLvl)==0 ); + pLvl->bEof = 0; + } + } + } + } + + return pIter->aLvl[0].bEof; +} +static int fts5DlidxIterPrev(Fts5Index *p, Fts5DlidxIter *pIter){ + return fts5DlidxIterPrevR(p, pIter, 0); +} + +/* +** Free a doclist-index iterator object allocated by fts5DlidxIterInit(). +*/ +static void fts5DlidxIterFree(Fts5DlidxIter *pIter){ + if( pIter ){ + int i; + for(i=0; i<pIter->nLvl; i++){ + fts5DataRelease(pIter->aLvl[i].pData); + } + sqlite3_free(pIter); + } +} + +static Fts5DlidxIter *fts5DlidxIterInit( + Fts5Index *p, /* Fts5 Backend to iterate within */ + int bRev, /* True for ORDER BY ASC */ + int iSegid, /* Segment id */ + int iLeafPg /* Leaf page number to load dlidx for */ +){ + Fts5DlidxIter *pIter = 0; + int i; + int bDone = 0; + + for(i=0; p->rc==SQLITE_OK && bDone==0; i++){ + int nByte = sizeof(Fts5DlidxIter) + i * sizeof(Fts5DlidxLvl); + Fts5DlidxIter *pNew; + + pNew = (Fts5DlidxIter*)sqlite3_realloc(pIter, nByte); + if( pNew==0 ){ + p->rc = SQLITE_NOMEM; + }else{ + i64 iRowid = FTS5_DLIDX_ROWID(iSegid, i, iLeafPg); + Fts5DlidxLvl *pLvl = &pNew->aLvl[i]; + pIter = pNew; + memset(pLvl, 0, sizeof(Fts5DlidxLvl)); + pLvl->pData = fts5DataRead(p, iRowid); + if( pLvl->pData && (pLvl->pData->p[0] & 0x0001)==0 ){ + bDone = 1; + } + pIter->nLvl = i+1; + } + } + + if( p->rc==SQLITE_OK ){ + pIter->iSegid = iSegid; + if( bRev==0 ){ + fts5DlidxIterFirst(pIter); + }else{ + fts5DlidxIterLast(p, pIter); + } + } + + if( p->rc!=SQLITE_OK ){ + fts5DlidxIterFree(pIter); + pIter = 0; + } + + return pIter; +} + +static i64 fts5DlidxIterRowid(Fts5DlidxIter *pIter){ + return pIter->aLvl[0].iRowid; +} +static int fts5DlidxIterPgno(Fts5DlidxIter *pIter){ + return pIter->aLvl[0].iLeafPgno; +} + +/* +** Load the next leaf page into the segment iterator. +*/ +static void fts5SegIterNextPage( + Fts5Index *p, /* FTS5 backend object */ + Fts5SegIter *pIter /* Iterator to advance to next page */ +){ + Fts5Data *pLeaf; + Fts5StructureSegment *pSeg = pIter->pSeg; + fts5DataRelease(pIter->pLeaf); + pIter->iLeafPgno++; + if( pIter->pNextLeaf ){ + pIter->pLeaf = pIter->pNextLeaf; + pIter->pNextLeaf = 0; + }else if( pIter->iLeafPgno<=pSeg->pgnoLast ){ + pIter->pLeaf = fts5DataRead(p, + FTS5_SEGMENT_ROWID(pSeg->iSegid, pIter->iLeafPgno) + ); + }else{ + pIter->pLeaf = 0; + } + pLeaf = pIter->pLeaf; + + if( pLeaf ){ + pIter->iPgidxOff = pLeaf->szLeaf; + if( fts5LeafIsTermless(pLeaf) ){ + pIter->iEndofDoclist = pLeaf->nn+1; + }else{ + pIter->iPgidxOff += fts5GetVarint32(&pLeaf->p[pIter->iPgidxOff], + pIter->iEndofDoclist + ); + } + } +} + +/* +** Argument p points to a buffer containing a varint to be interpreted as a +** position list size field. Read the varint and return the number of bytes +** read. Before returning, set *pnSz to the number of bytes in the position +** list, and *pbDel to true if the delete flag is set, or false otherwise. +*/ +static int fts5GetPoslistSize(const u8 *p, int *pnSz, int *pbDel){ + int nSz; + int n = 0; + fts5FastGetVarint32(p, n, nSz); + assert_nc( nSz>=0 ); + *pnSz = nSz/2; + *pbDel = nSz & 0x0001; + return n; +} + +/* +** Fts5SegIter.iLeafOffset currently points to the first byte of a +** position-list size field. Read the value of the field and store it +** in the following variables: +** +** Fts5SegIter.nPos +** Fts5SegIter.bDel +** +** Leave Fts5SegIter.iLeafOffset pointing to the first byte of the +** position list content (if any). +*/ +static void fts5SegIterLoadNPos(Fts5Index *p, Fts5SegIter *pIter){ + if( p->rc==SQLITE_OK ){ + int iOff = pIter->iLeafOffset; /* Offset to read at */ + ASSERT_SZLEAF_OK(pIter->pLeaf); + if( p->pConfig->eDetail==FTS5_DETAIL_NONE ){ + int iEod = MIN(pIter->iEndofDoclist, pIter->pLeaf->szLeaf); + pIter->bDel = 0; + pIter->nPos = 1; + if( iOff<iEod && pIter->pLeaf->p[iOff]==0 ){ + pIter->bDel = 1; + iOff++; + if( iOff<iEod && pIter->pLeaf->p[iOff]==0 ){ + pIter->nPos = 1; + iOff++; + }else{ + pIter->nPos = 0; + } + } + }else{ + int nSz; + fts5FastGetVarint32(pIter->pLeaf->p, iOff, nSz); + pIter->bDel = (nSz & 0x0001); + pIter->nPos = nSz>>1; + assert_nc( pIter->nPos>=0 ); + } + pIter->iLeafOffset = iOff; + } +} + +static void fts5SegIterLoadRowid(Fts5Index *p, Fts5SegIter *pIter){ + u8 *a = pIter->pLeaf->p; /* Buffer to read data from */ + int iOff = pIter->iLeafOffset; + + ASSERT_SZLEAF_OK(pIter->pLeaf); + if( iOff>=pIter->pLeaf->szLeaf ){ + fts5SegIterNextPage(p, pIter); + if( pIter->pLeaf==0 ){ + if( p->rc==SQLITE_OK ) p->rc = FTS5_CORRUPT; + return; + } + iOff = 4; + a = pIter->pLeaf->p; + } + iOff += sqlite3Fts5GetVarint(&a[iOff], (u64*)&pIter->iRowid); + pIter->iLeafOffset = iOff; +} + +/* +** Fts5SegIter.iLeafOffset currently points to the first byte of the +** "nSuffix" field of a term. Function parameter nKeep contains the value +** of the "nPrefix" field (if there was one - it is passed 0 if this is +** the first term in the segment). +** +** This function populates: +** +** Fts5SegIter.term +** Fts5SegIter.rowid +** +** accordingly and leaves (Fts5SegIter.iLeafOffset) set to the content of +** the first position list. The position list belonging to document +** (Fts5SegIter.iRowid). +*/ +static void fts5SegIterLoadTerm(Fts5Index *p, Fts5SegIter *pIter, int nKeep){ + u8 *a = pIter->pLeaf->p; /* Buffer to read data from */ + int iOff = pIter->iLeafOffset; /* Offset to read at */ + int nNew; /* Bytes of new data */ + + iOff += fts5GetVarint32(&a[iOff], nNew); + if( iOff+nNew>pIter->pLeaf->nn ){ + p->rc = FTS5_CORRUPT; + return; + } + pIter->term.n = nKeep; + fts5BufferAppendBlob(&p->rc, &pIter->term, nNew, &a[iOff]); + iOff += nNew; + pIter->iTermLeafOffset = iOff; + pIter->iTermLeafPgno = pIter->iLeafPgno; + pIter->iLeafOffset = iOff; + + if( pIter->iPgidxOff>=pIter->pLeaf->nn ){ + pIter->iEndofDoclist = pIter->pLeaf->nn+1; + }else{ + int nExtra; + pIter->iPgidxOff += fts5GetVarint32(&a[pIter->iPgidxOff], nExtra); + pIter->iEndofDoclist += nExtra; + } + + fts5SegIterLoadRowid(p, pIter); +} + +static void fts5SegIterNext(Fts5Index*, Fts5SegIter*, int*); +static void fts5SegIterNext_Reverse(Fts5Index*, Fts5SegIter*, int*); +static void fts5SegIterNext_None(Fts5Index*, Fts5SegIter*, int*); + +static void fts5SegIterSetNext(Fts5Index *p, Fts5SegIter *pIter){ + if( pIter->flags & FTS5_SEGITER_REVERSE ){ + pIter->xNext = fts5SegIterNext_Reverse; + }else if( p->pConfig->eDetail==FTS5_DETAIL_NONE ){ + pIter->xNext = fts5SegIterNext_None; + }else{ + pIter->xNext = fts5SegIterNext; + } +} + +/* +** Initialize the iterator object pIter to iterate through the entries in +** segment pSeg. The iterator is left pointing to the first entry when +** this function returns. +** +** If an error occurs, Fts5Index.rc is set to an appropriate error code. If +** an error has already occurred when this function is called, it is a no-op. +*/ +static void fts5SegIterInit( + Fts5Index *p, /* FTS index object */ + Fts5StructureSegment *pSeg, /* Description of segment */ + Fts5SegIter *pIter /* Object to populate */ +){ + if( pSeg->pgnoFirst==0 ){ + /* This happens if the segment is being used as an input to an incremental + ** merge and all data has already been "trimmed". See function + ** fts5TrimSegments() for details. In this case leave the iterator empty. + ** The caller will see the (pIter->pLeaf==0) and assume the iterator is + ** at EOF already. */ + assert( pIter->pLeaf==0 ); + return; + } + + if( p->rc==SQLITE_OK ){ + memset(pIter, 0, sizeof(*pIter)); + fts5SegIterSetNext(p, pIter); + pIter->pSeg = pSeg; + pIter->iLeafPgno = pSeg->pgnoFirst-1; + fts5SegIterNextPage(p, pIter); + } + + if( p->rc==SQLITE_OK ){ + pIter->iLeafOffset = 4; + assert_nc( pIter->pLeaf->nn>4 ); + assert( fts5LeafFirstTermOff(pIter->pLeaf)==4 ); + pIter->iPgidxOff = pIter->pLeaf->szLeaf+1; + fts5SegIterLoadTerm(p, pIter, 0); + fts5SegIterLoadNPos(p, pIter); + } +} + +/* +** This function is only ever called on iterators created by calls to +** Fts5IndexQuery() with the FTS5INDEX_QUERY_DESC flag set. +** +** The iterator is in an unusual state when this function is called: the +** Fts5SegIter.iLeafOffset variable is set to the offset of the start of +** the position-list size field for the first relevant rowid on the page. +** Fts5SegIter.rowid is set, but nPos and bDel are not. +** +** This function advances the iterator so that it points to the last +** relevant rowid on the page and, if necessary, initializes the +** aRowidOffset[] and iRowidOffset variables. At this point the iterator +** is in its regular state - Fts5SegIter.iLeafOffset points to the first +** byte of the position list content associated with said rowid. +*/ +static void fts5SegIterReverseInitPage(Fts5Index *p, Fts5SegIter *pIter){ + int eDetail = p->pConfig->eDetail; + int n = pIter->pLeaf->szLeaf; + int i = pIter->iLeafOffset; + u8 *a = pIter->pLeaf->p; + int iRowidOffset = 0; + + if( n>pIter->iEndofDoclist ){ + n = pIter->iEndofDoclist; + } + + ASSERT_SZLEAF_OK(pIter->pLeaf); + while( 1 ){ + i64 iDelta = 0; + + if( eDetail==FTS5_DETAIL_NONE ){ + /* todo */ + if( i<n && a[i]==0 ){ + i++; + if( i<n && a[i]==0 ) i++; + } + }else{ + int nPos; + int bDummy; + i += fts5GetPoslistSize(&a[i], &nPos, &bDummy); + i += nPos; + } + if( i>=n ) break; + i += fts5GetVarint(&a[i], (u64*)&iDelta); + pIter->iRowid += iDelta; + + /* If necessary, grow the pIter->aRowidOffset[] array. */ + if( iRowidOffset>=pIter->nRowidOffset ){ + int nNew = pIter->nRowidOffset + 8; + int *aNew = (int*)sqlite3_realloc(pIter->aRowidOffset, nNew*sizeof(int)); + if( aNew==0 ){ + p->rc = SQLITE_NOMEM; + break; + } + pIter->aRowidOffset = aNew; + pIter->nRowidOffset = nNew; + } + + pIter->aRowidOffset[iRowidOffset++] = pIter->iLeafOffset; + pIter->iLeafOffset = i; + } + pIter->iRowidOffset = iRowidOffset; + fts5SegIterLoadNPos(p, pIter); +} + +/* +** +*/ +static void fts5SegIterReverseNewPage(Fts5Index *p, Fts5SegIter *pIter){ + assert( pIter->flags & FTS5_SEGITER_REVERSE ); + assert( pIter->flags & FTS5_SEGITER_ONETERM ); + + fts5DataRelease(pIter->pLeaf); + pIter->pLeaf = 0; + while( p->rc==SQLITE_OK && pIter->iLeafPgno>pIter->iTermLeafPgno ){ + Fts5Data *pNew; + pIter->iLeafPgno--; + pNew = fts5DataRead(p, FTS5_SEGMENT_ROWID( + pIter->pSeg->iSegid, pIter->iLeafPgno + )); + if( pNew ){ + /* iTermLeafOffset may be equal to szLeaf if the term is the last + ** thing on the page - i.e. the first rowid is on the following page. + ** In this case leave pIter->pLeaf==0, this iterator is at EOF. */ + if( pIter->iLeafPgno==pIter->iTermLeafPgno ){ + assert( pIter->pLeaf==0 ); + if( pIter->iTermLeafOffset<pNew->szLeaf ){ + pIter->pLeaf = pNew; + pIter->iLeafOffset = pIter->iTermLeafOffset; + } + }else{ + int iRowidOff; + iRowidOff = fts5LeafFirstRowidOff(pNew); + if( iRowidOff ){ + pIter->pLeaf = pNew; + pIter->iLeafOffset = iRowidOff; + } + } + + if( pIter->pLeaf ){ + u8 *a = &pIter->pLeaf->p[pIter->iLeafOffset]; + pIter->iLeafOffset += fts5GetVarint(a, (u64*)&pIter->iRowid); + break; + }else{ + fts5DataRelease(pNew); + } + } + } + + if( pIter->pLeaf ){ + pIter->iEndofDoclist = pIter->pLeaf->nn+1; + fts5SegIterReverseInitPage(p, pIter); + } +} + +/* +** Return true if the iterator passed as the second argument currently +** points to a delete marker. A delete marker is an entry with a 0 byte +** position-list. +*/ +static int fts5MultiIterIsEmpty(Fts5Index *p, Fts5Iter *pIter){ + Fts5SegIter *pSeg = &pIter->aSeg[pIter->aFirst[1].iFirst]; + return (p->rc==SQLITE_OK && pSeg->pLeaf && pSeg->nPos==0); +} + +/* +** Advance iterator pIter to the next entry. +** +** This version of fts5SegIterNext() is only used by reverse iterators. +*/ +static void fts5SegIterNext_Reverse( + Fts5Index *p, /* FTS5 backend object */ + Fts5SegIter *pIter, /* Iterator to advance */ + int *pbUnused /* Unused */ +){ + assert( pIter->flags & FTS5_SEGITER_REVERSE ); + assert( pIter->pNextLeaf==0 ); + UNUSED_PARAM(pbUnused); + + if( pIter->iRowidOffset>0 ){ + u8 *a = pIter->pLeaf->p; + int iOff; + i64 iDelta; + + pIter->iRowidOffset--; + pIter->iLeafOffset = pIter->aRowidOffset[pIter->iRowidOffset]; + fts5SegIterLoadNPos(p, pIter); + iOff = pIter->iLeafOffset; + if( p->pConfig->eDetail!=FTS5_DETAIL_NONE ){ + iOff += pIter->nPos; + } + fts5GetVarint(&a[iOff], (u64*)&iDelta); + pIter->iRowid -= iDelta; + }else{ + fts5SegIterReverseNewPage(p, pIter); + } +} + +/* +** Advance iterator pIter to the next entry. +** +** This version of fts5SegIterNext() is only used if detail=none and the +** iterator is not a reverse direction iterator. +*/ +static void fts5SegIterNext_None( + Fts5Index *p, /* FTS5 backend object */ + Fts5SegIter *pIter, /* Iterator to advance */ + int *pbNewTerm /* OUT: Set for new term */ +){ + int iOff; + + assert( p->rc==SQLITE_OK ); + assert( (pIter->flags & FTS5_SEGITER_REVERSE)==0 ); + assert( p->pConfig->eDetail==FTS5_DETAIL_NONE ); + + ASSERT_SZLEAF_OK(pIter->pLeaf); + iOff = pIter->iLeafOffset; + + /* Next entry is on the next page */ + if( pIter->pSeg && iOff>=pIter->pLeaf->szLeaf ){ + fts5SegIterNextPage(p, pIter); + if( p->rc || pIter->pLeaf==0 ) return; + pIter->iRowid = 0; + iOff = 4; + } + + if( iOff<pIter->iEndofDoclist ){ + /* Next entry is on the current page */ + i64 iDelta; + iOff += sqlite3Fts5GetVarint(&pIter->pLeaf->p[iOff], (u64*)&iDelta); + pIter->iLeafOffset = iOff; + pIter->iRowid += iDelta; + }else if( (pIter->flags & FTS5_SEGITER_ONETERM)==0 ){ + if( pIter->pSeg ){ + int nKeep = 0; + if( iOff!=fts5LeafFirstTermOff(pIter->pLeaf) ){ + iOff += fts5GetVarint32(&pIter->pLeaf->p[iOff], nKeep); + } + pIter->iLeafOffset = iOff; + fts5SegIterLoadTerm(p, pIter, nKeep); + }else{ + const u8 *pList = 0; + const char *zTerm = 0; + int nList; + sqlite3Fts5HashScanNext(p->pHash); + sqlite3Fts5HashScanEntry(p->pHash, &zTerm, &pList, &nList); + if( pList==0 ) goto next_none_eof; + pIter->pLeaf->p = (u8*)pList; + pIter->pLeaf->nn = nList; + pIter->pLeaf->szLeaf = nList; + pIter->iEndofDoclist = nList; + sqlite3Fts5BufferSet(&p->rc,&pIter->term, (int)strlen(zTerm), (u8*)zTerm); + pIter->iLeafOffset = fts5GetVarint(pList, (u64*)&pIter->iRowid); + } + + if( pbNewTerm ) *pbNewTerm = 1; + }else{ + goto next_none_eof; + } + + fts5SegIterLoadNPos(p, pIter); + + return; + next_none_eof: + fts5DataRelease(pIter->pLeaf); + pIter->pLeaf = 0; +} + + +/* +** Advance iterator pIter to the next entry. +** +** If an error occurs, Fts5Index.rc is set to an appropriate error code. It +** is not considered an error if the iterator reaches EOF. If an error has +** already occurred when this function is called, it is a no-op. +*/ +static void fts5SegIterNext( + Fts5Index *p, /* FTS5 backend object */ + Fts5SegIter *pIter, /* Iterator to advance */ + int *pbNewTerm /* OUT: Set for new term */ +){ + Fts5Data *pLeaf = pIter->pLeaf; + int iOff; + int bNewTerm = 0; + int nKeep = 0; + u8 *a; + int n; + + assert( pbNewTerm==0 || *pbNewTerm==0 ); + assert( p->pConfig->eDetail!=FTS5_DETAIL_NONE ); + + /* Search for the end of the position list within the current page. */ + a = pLeaf->p; + n = pLeaf->szLeaf; + + ASSERT_SZLEAF_OK(pLeaf); + iOff = pIter->iLeafOffset + pIter->nPos; + + if( iOff<n ){ + /* The next entry is on the current page. */ + assert_nc( iOff<=pIter->iEndofDoclist ); + if( iOff>=pIter->iEndofDoclist ){ + bNewTerm = 1; + if( iOff!=fts5LeafFirstTermOff(pLeaf) ){ + iOff += fts5GetVarint32(&a[iOff], nKeep); + } + }else{ + u64 iDelta; + iOff += sqlite3Fts5GetVarint(&a[iOff], &iDelta); + pIter->iRowid += iDelta; + assert_nc( iDelta>0 ); + } + pIter->iLeafOffset = iOff; + + }else if( pIter->pSeg==0 ){ + const u8 *pList = 0; + const char *zTerm = 0; + int nList = 0; + assert( (pIter->flags & FTS5_SEGITER_ONETERM) || pbNewTerm ); + if( 0==(pIter->flags & FTS5_SEGITER_ONETERM) ){ + sqlite3Fts5HashScanNext(p->pHash); + sqlite3Fts5HashScanEntry(p->pHash, &zTerm, &pList, &nList); + } + if( pList==0 ){ + fts5DataRelease(pIter->pLeaf); + pIter->pLeaf = 0; + }else{ + pIter->pLeaf->p = (u8*)pList; + pIter->pLeaf->nn = nList; + pIter->pLeaf->szLeaf = nList; + pIter->iEndofDoclist = nList+1; + sqlite3Fts5BufferSet(&p->rc, &pIter->term, (int)strlen(zTerm), + (u8*)zTerm); + pIter->iLeafOffset = fts5GetVarint(pList, (u64*)&pIter->iRowid); + *pbNewTerm = 1; + } + }else{ + iOff = 0; + /* Next entry is not on the current page */ + while( iOff==0 ){ + fts5SegIterNextPage(p, pIter); + pLeaf = pIter->pLeaf; + if( pLeaf==0 ) break; + ASSERT_SZLEAF_OK(pLeaf); + if( (iOff = fts5LeafFirstRowidOff(pLeaf)) && iOff<pLeaf->szLeaf ){ + iOff += sqlite3Fts5GetVarint(&pLeaf->p[iOff], (u64*)&pIter->iRowid); + pIter->iLeafOffset = iOff; + + if( pLeaf->nn>pLeaf->szLeaf ){ + pIter->iPgidxOff = pLeaf->szLeaf + fts5GetVarint32( + &pLeaf->p[pLeaf->szLeaf], pIter->iEndofDoclist + ); + } + + } + else if( pLeaf->nn>pLeaf->szLeaf ){ + pIter->iPgidxOff = pLeaf->szLeaf + fts5GetVarint32( + &pLeaf->p[pLeaf->szLeaf], iOff + ); + pIter->iLeafOffset = iOff; + pIter->iEndofDoclist = iOff; + bNewTerm = 1; + } + assert_nc( iOff<pLeaf->szLeaf ); + if( iOff>pLeaf->szLeaf ){ + p->rc = FTS5_CORRUPT; + return; + } + } + } + + /* Check if the iterator is now at EOF. If so, return early. */ + if( pIter->pLeaf ){ + if( bNewTerm ){ + if( pIter->flags & FTS5_SEGITER_ONETERM ){ + fts5DataRelease(pIter->pLeaf); + pIter->pLeaf = 0; + }else{ + fts5SegIterLoadTerm(p, pIter, nKeep); + fts5SegIterLoadNPos(p, pIter); + if( pbNewTerm ) *pbNewTerm = 1; + } + }else{ + /* The following could be done by calling fts5SegIterLoadNPos(). But + ** this block is particularly performance critical, so equivalent + ** code is inlined. + ** + ** Later: Switched back to fts5SegIterLoadNPos() because it supports + ** detail=none mode. Not ideal. + */ + int nSz; + assert( p->rc==SQLITE_OK ); + fts5FastGetVarint32(pIter->pLeaf->p, pIter->iLeafOffset, nSz); + pIter->bDel = (nSz & 0x0001); + pIter->nPos = nSz>>1; + assert_nc( pIter->nPos>=0 ); + } + } +} + +#define SWAPVAL(T, a, b) { T tmp; tmp=a; a=b; b=tmp; } + +#define fts5IndexSkipVarint(a, iOff) { \ + int iEnd = iOff+9; \ + while( (a[iOff++] & 0x80) && iOff<iEnd ); \ +} + +/* +** Iterator pIter currently points to the first rowid in a doclist. This +** function sets the iterator up so that iterates in reverse order through +** the doclist. +*/ +static void fts5SegIterReverse(Fts5Index *p, Fts5SegIter *pIter){ + Fts5DlidxIter *pDlidx = pIter->pDlidx; + Fts5Data *pLast = 0; + int pgnoLast = 0; + + if( pDlidx ){ + int iSegid = pIter->pSeg->iSegid; + pgnoLast = fts5DlidxIterPgno(pDlidx); + pLast = fts5DataRead(p, FTS5_SEGMENT_ROWID(iSegid, pgnoLast)); + }else{ + Fts5Data *pLeaf = pIter->pLeaf; /* Current leaf data */ + + /* Currently, Fts5SegIter.iLeafOffset points to the first byte of + ** position-list content for the current rowid. Back it up so that it + ** points to the start of the position-list size field. */ + int iPoslist; + if( pIter->iTermLeafPgno==pIter->iLeafPgno ){ + iPoslist = pIter->iTermLeafOffset; + }else{ + iPoslist = 4; + } + fts5IndexSkipVarint(pLeaf->p, iPoslist); + pIter->iLeafOffset = iPoslist; + + /* If this condition is true then the largest rowid for the current + ** term may not be stored on the current page. So search forward to + ** see where said rowid really is. */ + if( pIter->iEndofDoclist>=pLeaf->szLeaf ){ + int pgno; + Fts5StructureSegment *pSeg = pIter->pSeg; + + /* The last rowid in the doclist may not be on the current page. Search + ** forward to find the page containing the last rowid. */ + for(pgno=pIter->iLeafPgno+1; !p->rc && pgno<=pSeg->pgnoLast; pgno++){ + i64 iAbs = FTS5_SEGMENT_ROWID(pSeg->iSegid, pgno); + Fts5Data *pNew = fts5DataRead(p, iAbs); + if( pNew ){ + int iRowid, bTermless; + iRowid = fts5LeafFirstRowidOff(pNew); + bTermless = fts5LeafIsTermless(pNew); + if( iRowid ){ + SWAPVAL(Fts5Data*, pNew, pLast); + pgnoLast = pgno; + } + fts5DataRelease(pNew); + if( bTermless==0 ) break; + } + } + } + } + + /* If pLast is NULL at this point, then the last rowid for this doclist + ** lies on the page currently indicated by the iterator. In this case + ** pIter->iLeafOffset is already set to point to the position-list size + ** field associated with the first relevant rowid on the page. + ** + ** Or, if pLast is non-NULL, then it is the page that contains the last + ** rowid. In this case configure the iterator so that it points to the + ** first rowid on this page. + */ + if( pLast ){ + int iOff; + fts5DataRelease(pIter->pLeaf); + pIter->pLeaf = pLast; + pIter->iLeafPgno = pgnoLast; + iOff = fts5LeafFirstRowidOff(pLast); + iOff += fts5GetVarint(&pLast->p[iOff], (u64*)&pIter->iRowid); + pIter->iLeafOffset = iOff; + + if( fts5LeafIsTermless(pLast) ){ + pIter->iEndofDoclist = pLast->nn+1; + }else{ + pIter->iEndofDoclist = fts5LeafFirstTermOff(pLast); + } + + } + + fts5SegIterReverseInitPage(p, pIter); +} + +/* +** Iterator pIter currently points to the first rowid of a doclist. +** There is a doclist-index associated with the final term on the current +** page. If the current term is the last term on the page, load the +** doclist-index from disk and initialize an iterator at (pIter->pDlidx). +*/ +static void fts5SegIterLoadDlidx(Fts5Index *p, Fts5SegIter *pIter){ + int iSeg = pIter->pSeg->iSegid; + int bRev = (pIter->flags & FTS5_SEGITER_REVERSE); + Fts5Data *pLeaf = pIter->pLeaf; /* Current leaf data */ + + assert( pIter->flags & FTS5_SEGITER_ONETERM ); + assert( pIter->pDlidx==0 ); + + /* Check if the current doclist ends on this page. If it does, return + ** early without loading the doclist-index (as it belongs to a different + ** term. */ + if( pIter->iTermLeafPgno==pIter->iLeafPgno + && pIter->iEndofDoclist<pLeaf->szLeaf + ){ + return; + } + + pIter->pDlidx = fts5DlidxIterInit(p, bRev, iSeg, pIter->iTermLeafPgno); +} + +/* +** The iterator object passed as the second argument currently contains +** no valid values except for the Fts5SegIter.pLeaf member variable. This +** function searches the leaf page for a term matching (pTerm/nTerm). +** +** If the specified term is found on the page, then the iterator is left +** pointing to it. If argument bGe is zero and the term is not found, +** the iterator is left pointing at EOF. +** +** If bGe is non-zero and the specified term is not found, then the +** iterator is left pointing to the smallest term in the segment that +** is larger than the specified term, even if this term is not on the +** current page. +*/ +static void fts5LeafSeek( + Fts5Index *p, /* Leave any error code here */ + int bGe, /* True for a >= search */ + Fts5SegIter *pIter, /* Iterator to seek */ + const u8 *pTerm, int nTerm /* Term to search for */ +){ + int iOff; + const u8 *a = pIter->pLeaf->p; + int szLeaf = pIter->pLeaf->szLeaf; + int n = pIter->pLeaf->nn; + + int nMatch = 0; + int nKeep = 0; + int nNew = 0; + int iTermOff; + int iPgidx; /* Current offset in pgidx */ + int bEndOfPage = 0; + + assert( p->rc==SQLITE_OK ); + + iPgidx = szLeaf; + iPgidx += fts5GetVarint32(&a[iPgidx], iTermOff); + iOff = iTermOff; + + while( 1 ){ + + /* Figure out how many new bytes are in this term */ + fts5FastGetVarint32(a, iOff, nNew); + if( nKeep<nMatch ){ + goto search_failed; + } + + assert( nKeep>=nMatch ); + if( nKeep==nMatch ){ + int nCmp; + int i; + nCmp = MIN(nNew, nTerm-nMatch); + for(i=0; i<nCmp; i++){ + if( a[iOff+i]!=pTerm[nMatch+i] ) break; + } + nMatch += i; + + if( nTerm==nMatch ){ + if( i==nNew ){ + goto search_success; + }else{ + goto search_failed; + } + }else if( i<nNew && a[iOff+i]>pTerm[nMatch] ){ + goto search_failed; + } + } + + if( iPgidx>=n ){ + bEndOfPage = 1; + break; + } + + iPgidx += fts5GetVarint32(&a[iPgidx], nKeep); + iTermOff += nKeep; + iOff = iTermOff; + + /* Read the nKeep field of the next term. */ + fts5FastGetVarint32(a, iOff, nKeep); + } + + search_failed: + if( bGe==0 ){ + fts5DataRelease(pIter->pLeaf); + pIter->pLeaf = 0; + return; + }else if( bEndOfPage ){ + do { + fts5SegIterNextPage(p, pIter); + if( pIter->pLeaf==0 ) return; + a = pIter->pLeaf->p; + if( fts5LeafIsTermless(pIter->pLeaf)==0 ){ + iPgidx = pIter->pLeaf->szLeaf; + iPgidx += fts5GetVarint32(&pIter->pLeaf->p[iPgidx], iOff); + if( iOff<4 || iOff>=pIter->pLeaf->szLeaf ){ + p->rc = FTS5_CORRUPT; + }else{ + nKeep = 0; + iTermOff = iOff; + n = pIter->pLeaf->nn; + iOff += fts5GetVarint32(&a[iOff], nNew); + break; + } + } + }while( 1 ); + } + + search_success: + + pIter->iLeafOffset = iOff + nNew; + pIter->iTermLeafOffset = pIter->iLeafOffset; + pIter->iTermLeafPgno = pIter->iLeafPgno; + + fts5BufferSet(&p->rc, &pIter->term, nKeep, pTerm); + fts5BufferAppendBlob(&p->rc, &pIter->term, nNew, &a[iOff]); + + if( iPgidx>=n ){ + pIter->iEndofDoclist = pIter->pLeaf->nn+1; + }else{ + int nExtra; + iPgidx += fts5GetVarint32(&a[iPgidx], nExtra); + pIter->iEndofDoclist = iTermOff + nExtra; + } + pIter->iPgidxOff = iPgidx; + + fts5SegIterLoadRowid(p, pIter); + fts5SegIterLoadNPos(p, pIter); +} + +/* +** Initialize the object pIter to point to term pTerm/nTerm within segment +** pSeg. If there is no such term in the index, the iterator is set to EOF. +** +** If an error occurs, Fts5Index.rc is set to an appropriate error code. If +** an error has already occurred when this function is called, it is a no-op. +*/ +static void fts5SegIterSeekInit( + Fts5Index *p, /* FTS5 backend */ + const u8 *pTerm, int nTerm, /* Term to seek to */ + int flags, /* Mask of FTS5INDEX_XXX flags */ + Fts5StructureSegment *pSeg, /* Description of segment */ + Fts5SegIter *pIter /* Object to populate */ +){ + int iPg = 1; + int bGe = (flags & FTS5INDEX_QUERY_SCAN); + int bDlidx = 0; /* True if there is a doclist-index */ + + assert( bGe==0 || (flags & FTS5INDEX_QUERY_DESC)==0 ); + assert( pTerm && nTerm ); + memset(pIter, 0, sizeof(*pIter)); + pIter->pSeg = pSeg; + + /* This block sets stack variable iPg to the leaf page number that may + ** contain term (pTerm/nTerm), if it is present in the segment. */ + if( p->pIdxSelect==0 ){ + Fts5Config *pConfig = p->pConfig; + fts5IndexPrepareStmt(p, &p->pIdxSelect, sqlite3_mprintf( + "SELECT pgno FROM '%q'.'%q_idx' WHERE " + "segid=? AND term<=? ORDER BY term DESC LIMIT 1", + pConfig->zDb, pConfig->zName + )); + } + if( p->rc ) return; + sqlite3_bind_int(p->pIdxSelect, 1, pSeg->iSegid); + sqlite3_bind_blob(p->pIdxSelect, 2, pTerm, nTerm, SQLITE_STATIC); + if( SQLITE_ROW==sqlite3_step(p->pIdxSelect) ){ + i64 val = sqlite3_column_int(p->pIdxSelect, 0); + iPg = (int)(val>>1); + bDlidx = (val & 0x0001); + } + p->rc = sqlite3_reset(p->pIdxSelect); + + if( iPg<pSeg->pgnoFirst ){ + iPg = pSeg->pgnoFirst; + bDlidx = 0; + } + + pIter->iLeafPgno = iPg - 1; + fts5SegIterNextPage(p, pIter); + + if( pIter->pLeaf ){ + fts5LeafSeek(p, bGe, pIter, pTerm, nTerm); + } + + if( p->rc==SQLITE_OK && bGe==0 ){ + pIter->flags |= FTS5_SEGITER_ONETERM; + if( pIter->pLeaf ){ + if( flags & FTS5INDEX_QUERY_DESC ){ + pIter->flags |= FTS5_SEGITER_REVERSE; + } + if( bDlidx ){ + fts5SegIterLoadDlidx(p, pIter); + } + if( flags & FTS5INDEX_QUERY_DESC ){ + fts5SegIterReverse(p, pIter); + } + } + } + + fts5SegIterSetNext(p, pIter); + + /* Either: + ** + ** 1) an error has occurred, or + ** 2) the iterator points to EOF, or + ** 3) the iterator points to an entry with term (pTerm/nTerm), or + ** 4) the FTS5INDEX_QUERY_SCAN flag was set and the iterator points + ** to an entry with a term greater than or equal to (pTerm/nTerm). + */ + assert( p->rc!=SQLITE_OK /* 1 */ + || pIter->pLeaf==0 /* 2 */ + || fts5BufferCompareBlob(&pIter->term, pTerm, nTerm)==0 /* 3 */ + || (bGe && fts5BufferCompareBlob(&pIter->term, pTerm, nTerm)>0) /* 4 */ + ); +} + +/* +** Initialize the object pIter to point to term pTerm/nTerm within the +** in-memory hash table. If there is no such term in the hash-table, the +** iterator is set to EOF. +** +** If an error occurs, Fts5Index.rc is set to an appropriate error code. If +** an error has already occurred when this function is called, it is a no-op. +*/ +static void fts5SegIterHashInit( + Fts5Index *p, /* FTS5 backend */ + const u8 *pTerm, int nTerm, /* Term to seek to */ + int flags, /* Mask of FTS5INDEX_XXX flags */ + Fts5SegIter *pIter /* Object to populate */ +){ + const u8 *pList = 0; + int nList = 0; + const u8 *z = 0; + int n = 0; + + assert( p->pHash ); + assert( p->rc==SQLITE_OK ); + + if( pTerm==0 || (flags & FTS5INDEX_QUERY_SCAN) ){ + p->rc = sqlite3Fts5HashScanInit(p->pHash, (const char*)pTerm, nTerm); + sqlite3Fts5HashScanEntry(p->pHash, (const char**)&z, &pList, &nList); + n = (z ? (int)strlen((const char*)z) : 0); + }else{ + pIter->flags |= FTS5_SEGITER_ONETERM; + sqlite3Fts5HashQuery(p->pHash, (const char*)pTerm, nTerm, &pList, &nList); + z = pTerm; + n = nTerm; + } + + if( pList ){ + Fts5Data *pLeaf; + sqlite3Fts5BufferSet(&p->rc, &pIter->term, n, z); + pLeaf = fts5IdxMalloc(p, sizeof(Fts5Data)); + if( pLeaf==0 ) return; + pLeaf->p = (u8*)pList; + pLeaf->nn = pLeaf->szLeaf = nList; + pIter->pLeaf = pLeaf; + pIter->iLeafOffset = fts5GetVarint(pLeaf->p, (u64*)&pIter->iRowid); + pIter->iEndofDoclist = pLeaf->nn; + + if( flags & FTS5INDEX_QUERY_DESC ){ + pIter->flags |= FTS5_SEGITER_REVERSE; + fts5SegIterReverseInitPage(p, pIter); + }else{ + fts5SegIterLoadNPos(p, pIter); + } + } + + fts5SegIterSetNext(p, pIter); +} + +/* +** Zero the iterator passed as the only argument. +*/ +static void fts5SegIterClear(Fts5SegIter *pIter){ + fts5BufferFree(&pIter->term); + fts5DataRelease(pIter->pLeaf); + fts5DataRelease(pIter->pNextLeaf); + fts5DlidxIterFree(pIter->pDlidx); + sqlite3_free(pIter->aRowidOffset); + memset(pIter, 0, sizeof(Fts5SegIter)); +} + +#ifdef SQLITE_DEBUG + +/* +** This function is used as part of the big assert() procedure implemented by +** fts5AssertMultiIterSetup(). It ensures that the result currently stored +** in *pRes is the correct result of comparing the current positions of the +** two iterators. +*/ +static void fts5AssertComparisonResult( + Fts5Iter *pIter, + Fts5SegIter *p1, + Fts5SegIter *p2, + Fts5CResult *pRes +){ + int i1 = p1 - pIter->aSeg; + int i2 = p2 - pIter->aSeg; + + if( p1->pLeaf || p2->pLeaf ){ + if( p1->pLeaf==0 ){ + assert( pRes->iFirst==i2 ); + }else if( p2->pLeaf==0 ){ + assert( pRes->iFirst==i1 ); + }else{ + int nMin = MIN(p1->term.n, p2->term.n); + int res = memcmp(p1->term.p, p2->term.p, nMin); + if( res==0 ) res = p1->term.n - p2->term.n; + + if( res==0 ){ + assert( pRes->bTermEq==1 ); + assert( p1->iRowid!=p2->iRowid ); + res = ((p1->iRowid > p2->iRowid)==pIter->bRev) ? -1 : 1; + }else{ + assert( pRes->bTermEq==0 ); + } + + if( res<0 ){ + assert( pRes->iFirst==i1 ); + }else{ + assert( pRes->iFirst==i2 ); + } + } + } +} + +/* +** This function is a no-op unless SQLITE_DEBUG is defined when this module +** is compiled. In that case, this function is essentially an assert() +** statement used to verify that the contents of the pIter->aFirst[] array +** are correct. +*/ +static void fts5AssertMultiIterSetup(Fts5Index *p, Fts5Iter *pIter){ + if( p->rc==SQLITE_OK ){ + Fts5SegIter *pFirst = &pIter->aSeg[ pIter->aFirst[1].iFirst ]; + int i; + + assert( (pFirst->pLeaf==0)==pIter->base.bEof ); + + /* Check that pIter->iSwitchRowid is set correctly. */ + for(i=0; i<pIter->nSeg; i++){ + Fts5SegIter *p1 = &pIter->aSeg[i]; + assert( p1==pFirst + || p1->pLeaf==0 + || fts5BufferCompare(&pFirst->term, &p1->term) + || p1->iRowid==pIter->iSwitchRowid + || (p1->iRowid<pIter->iSwitchRowid)==pIter->bRev + ); + } + + for(i=0; i<pIter->nSeg; i+=2){ + Fts5SegIter *p1 = &pIter->aSeg[i]; + Fts5SegIter *p2 = &pIter->aSeg[i+1]; + Fts5CResult *pRes = &pIter->aFirst[(pIter->nSeg + i) / 2]; + fts5AssertComparisonResult(pIter, p1, p2, pRes); + } + + for(i=1; i<(pIter->nSeg / 2); i+=2){ + Fts5SegIter *p1 = &pIter->aSeg[ pIter->aFirst[i*2].iFirst ]; + Fts5SegIter *p2 = &pIter->aSeg[ pIter->aFirst[i*2+1].iFirst ]; + Fts5CResult *pRes = &pIter->aFirst[i]; + fts5AssertComparisonResult(pIter, p1, p2, pRes); + } + } +} +#else +# define fts5AssertMultiIterSetup(x,y) +#endif + +/* +** Do the comparison necessary to populate pIter->aFirst[iOut]. +** +** If the returned value is non-zero, then it is the index of an entry +** in the pIter->aSeg[] array that is (a) not at EOF, and (b) pointing +** to a key that is a duplicate of another, higher priority, +** segment-iterator in the pSeg->aSeg[] array. +*/ +static int fts5MultiIterDoCompare(Fts5Iter *pIter, int iOut){ + int i1; /* Index of left-hand Fts5SegIter */ + int i2; /* Index of right-hand Fts5SegIter */ + int iRes; + Fts5SegIter *p1; /* Left-hand Fts5SegIter */ + Fts5SegIter *p2; /* Right-hand Fts5SegIter */ + Fts5CResult *pRes = &pIter->aFirst[iOut]; + + assert( iOut<pIter->nSeg && iOut>0 ); + assert( pIter->bRev==0 || pIter->bRev==1 ); + + if( iOut>=(pIter->nSeg/2) ){ + i1 = (iOut - pIter->nSeg/2) * 2; + i2 = i1 + 1; + }else{ + i1 = pIter->aFirst[iOut*2].iFirst; + i2 = pIter->aFirst[iOut*2+1].iFirst; + } + p1 = &pIter->aSeg[i1]; + p2 = &pIter->aSeg[i2]; + + pRes->bTermEq = 0; + if( p1->pLeaf==0 ){ /* If p1 is at EOF */ + iRes = i2; + }else if( p2->pLeaf==0 ){ /* If p2 is at EOF */ + iRes = i1; + }else{ + int res = fts5BufferCompare(&p1->term, &p2->term); + if( res==0 ){ + assert( i2>i1 ); + assert( i2!=0 ); + pRes->bTermEq = 1; + if( p1->iRowid==p2->iRowid ){ + p1->bDel = p2->bDel; + return i2; + } + res = ((p1->iRowid > p2->iRowid)==pIter->bRev) ? -1 : +1; + } + assert( res!=0 ); + if( res<0 ){ + iRes = i1; + }else{ + iRes = i2; + } + } + + pRes->iFirst = (u16)iRes; + return 0; +} + +/* +** Move the seg-iter so that it points to the first rowid on page iLeafPgno. +** It is an error if leaf iLeafPgno does not exist or contains no rowids. +*/ +static void fts5SegIterGotoPage( + Fts5Index *p, /* FTS5 backend object */ + Fts5SegIter *pIter, /* Iterator to advance */ + int iLeafPgno +){ + assert( iLeafPgno>pIter->iLeafPgno ); + + if( iLeafPgno>pIter->pSeg->pgnoLast ){ + p->rc = FTS5_CORRUPT; + }else{ + fts5DataRelease(pIter->pNextLeaf); + pIter->pNextLeaf = 0; + pIter->iLeafPgno = iLeafPgno-1; + fts5SegIterNextPage(p, pIter); + assert( p->rc!=SQLITE_OK || pIter->iLeafPgno==iLeafPgno ); + + if( p->rc==SQLITE_OK ){ + int iOff; + u8 *a = pIter->pLeaf->p; + int n = pIter->pLeaf->szLeaf; + + iOff = fts5LeafFirstRowidOff(pIter->pLeaf); + if( iOff<4 || iOff>=n ){ + p->rc = FTS5_CORRUPT; + }else{ + iOff += fts5GetVarint(&a[iOff], (u64*)&pIter->iRowid); + pIter->iLeafOffset = iOff; + fts5SegIterLoadNPos(p, pIter); + } + } + } +} + +/* +** Advance the iterator passed as the second argument until it is at or +** past rowid iFrom. Regardless of the value of iFrom, the iterator is +** always advanced at least once. +*/ +static void fts5SegIterNextFrom( + Fts5Index *p, /* FTS5 backend object */ + Fts5SegIter *pIter, /* Iterator to advance */ + i64 iMatch /* Advance iterator at least this far */ +){ + int bRev = (pIter->flags & FTS5_SEGITER_REVERSE); + Fts5DlidxIter *pDlidx = pIter->pDlidx; + int iLeafPgno = pIter->iLeafPgno; + int bMove = 1; + + assert( pIter->flags & FTS5_SEGITER_ONETERM ); + assert( pIter->pDlidx ); + assert( pIter->pLeaf ); + + if( bRev==0 ){ + while( !fts5DlidxIterEof(p, pDlidx) && iMatch>fts5DlidxIterRowid(pDlidx) ){ + iLeafPgno = fts5DlidxIterPgno(pDlidx); + fts5DlidxIterNext(p, pDlidx); + } + assert_nc( iLeafPgno>=pIter->iLeafPgno || p->rc ); + if( iLeafPgno>pIter->iLeafPgno ){ + fts5SegIterGotoPage(p, pIter, iLeafPgno); + bMove = 0; + } + }else{ + assert( pIter->pNextLeaf==0 ); + assert( iMatch<pIter->iRowid ); + while( !fts5DlidxIterEof(p, pDlidx) && iMatch<fts5DlidxIterRowid(pDlidx) ){ + fts5DlidxIterPrev(p, pDlidx); + } + iLeafPgno = fts5DlidxIterPgno(pDlidx); + + assert( fts5DlidxIterEof(p, pDlidx) || iLeafPgno<=pIter->iLeafPgno ); + + if( iLeafPgno<pIter->iLeafPgno ){ + pIter->iLeafPgno = iLeafPgno+1; + fts5SegIterReverseNewPage(p, pIter); + bMove = 0; + } + } + + do{ + if( bMove && p->rc==SQLITE_OK ) pIter->xNext(p, pIter, 0); + if( pIter->pLeaf==0 ) break; + if( bRev==0 && pIter->iRowid>=iMatch ) break; + if( bRev!=0 && pIter->iRowid<=iMatch ) break; + bMove = 1; + }while( p->rc==SQLITE_OK ); +} + + +/* +** Free the iterator object passed as the second argument. +*/ +static void fts5MultiIterFree(Fts5Iter *pIter){ + if( pIter ){ + int i; + for(i=0; i<pIter->nSeg; i++){ + fts5SegIterClear(&pIter->aSeg[i]); + } + fts5StructureRelease(pIter->pStruct); + fts5BufferFree(&pIter->poslist); + sqlite3_free(pIter); + } +} + +static void fts5MultiIterAdvanced( + Fts5Index *p, /* FTS5 backend to iterate within */ + Fts5Iter *pIter, /* Iterator to update aFirst[] array for */ + int iChanged, /* Index of sub-iterator just advanced */ + int iMinset /* Minimum entry in aFirst[] to set */ +){ + int i; + for(i=(pIter->nSeg+iChanged)/2; i>=iMinset && p->rc==SQLITE_OK; i=i/2){ + int iEq; + if( (iEq = fts5MultiIterDoCompare(pIter, i)) ){ + Fts5SegIter *pSeg = &pIter->aSeg[iEq]; + assert( p->rc==SQLITE_OK ); + pSeg->xNext(p, pSeg, 0); + i = pIter->nSeg + iEq; + } + } +} + +/* +** Sub-iterator iChanged of iterator pIter has just been advanced. It still +** points to the same term though - just a different rowid. This function +** attempts to update the contents of the pIter->aFirst[] accordingly. +** If it does so successfully, 0 is returned. Otherwise 1. +** +** If non-zero is returned, the caller should call fts5MultiIterAdvanced() +** on the iterator instead. That function does the same as this one, except +** that it deals with more complicated cases as well. +*/ +static int fts5MultiIterAdvanceRowid( + Fts5Iter *pIter, /* Iterator to update aFirst[] array for */ + int iChanged, /* Index of sub-iterator just advanced */ + Fts5SegIter **ppFirst +){ + Fts5SegIter *pNew = &pIter->aSeg[iChanged]; + + if( pNew->iRowid==pIter->iSwitchRowid + || (pNew->iRowid<pIter->iSwitchRowid)==pIter->bRev + ){ + int i; + Fts5SegIter *pOther = &pIter->aSeg[iChanged ^ 0x0001]; + pIter->iSwitchRowid = pIter->bRev ? SMALLEST_INT64 : LARGEST_INT64; + for(i=(pIter->nSeg+iChanged)/2; 1; i=i/2){ + Fts5CResult *pRes = &pIter->aFirst[i]; + + assert( pNew->pLeaf ); + assert( pRes->bTermEq==0 || pOther->pLeaf ); + + if( pRes->bTermEq ){ + if( pNew->iRowid==pOther->iRowid ){ + return 1; + }else if( (pOther->iRowid>pNew->iRowid)==pIter->bRev ){ + pIter->iSwitchRowid = pOther->iRowid; + pNew = pOther; + }else if( (pOther->iRowid>pIter->iSwitchRowid)==pIter->bRev ){ + pIter->iSwitchRowid = pOther->iRowid; + } + } + pRes->iFirst = (u16)(pNew - pIter->aSeg); + if( i==1 ) break; + + pOther = &pIter->aSeg[ pIter->aFirst[i ^ 0x0001].iFirst ]; + } + } + + *ppFirst = pNew; + return 0; +} + +/* +** Set the pIter->bEof variable based on the state of the sub-iterators. +*/ +static void fts5MultiIterSetEof(Fts5Iter *pIter){ + Fts5SegIter *pSeg = &pIter->aSeg[ pIter->aFirst[1].iFirst ]; + pIter->base.bEof = pSeg->pLeaf==0; + pIter->iSwitchRowid = pSeg->iRowid; +} + +/* +** Move the iterator to the next entry. +** +** If an error occurs, an error code is left in Fts5Index.rc. It is not +** considered an error if the iterator reaches EOF, or if it is already at +** EOF when this function is called. +*/ +static void fts5MultiIterNext( + Fts5Index *p, + Fts5Iter *pIter, + int bFrom, /* True if argument iFrom is valid */ + i64 iFrom /* Advance at least as far as this */ +){ + int bUseFrom = bFrom; + while( p->rc==SQLITE_OK ){ + int iFirst = pIter->aFirst[1].iFirst; + int bNewTerm = 0; + Fts5SegIter *pSeg = &pIter->aSeg[iFirst]; + assert( p->rc==SQLITE_OK ); + if( bUseFrom && pSeg->pDlidx ){ + fts5SegIterNextFrom(p, pSeg, iFrom); + }else{ + pSeg->xNext(p, pSeg, &bNewTerm); + } + + if( pSeg->pLeaf==0 || bNewTerm + || fts5MultiIterAdvanceRowid(pIter, iFirst, &pSeg) + ){ + fts5MultiIterAdvanced(p, pIter, iFirst, 1); + fts5MultiIterSetEof(pIter); + pSeg = &pIter->aSeg[pIter->aFirst[1].iFirst]; + if( pSeg->pLeaf==0 ) return; + } + + fts5AssertMultiIterSetup(p, pIter); + assert( pSeg==&pIter->aSeg[pIter->aFirst[1].iFirst] && pSeg->pLeaf ); + if( pIter->bSkipEmpty==0 || pSeg->nPos ){ + pIter->xSetOutputs(pIter, pSeg); + return; + } + bUseFrom = 0; + } +} + +static void fts5MultiIterNext2( + Fts5Index *p, + Fts5Iter *pIter, + int *pbNewTerm /* OUT: True if *might* be new term */ +){ + assert( pIter->bSkipEmpty ); + if( p->rc==SQLITE_OK ){ + do { + int iFirst = pIter->aFirst[1].iFirst; + Fts5SegIter *pSeg = &pIter->aSeg[iFirst]; + int bNewTerm = 0; + + assert( p->rc==SQLITE_OK ); + pSeg->xNext(p, pSeg, &bNewTerm); + if( pSeg->pLeaf==0 || bNewTerm + || fts5MultiIterAdvanceRowid(pIter, iFirst, &pSeg) + ){ + fts5MultiIterAdvanced(p, pIter, iFirst, 1); + fts5MultiIterSetEof(pIter); + *pbNewTerm = 1; + }else{ + *pbNewTerm = 0; + } + fts5AssertMultiIterSetup(p, pIter); + + }while( fts5MultiIterIsEmpty(p, pIter) ); + } +} + +static void fts5IterSetOutputs_Noop(Fts5Iter *pUnused1, Fts5SegIter *pUnused2){ + UNUSED_PARAM2(pUnused1, pUnused2); +} + +static Fts5Iter *fts5MultiIterAlloc( + Fts5Index *p, /* FTS5 backend to iterate within */ + int nSeg +){ + Fts5Iter *pNew; + int nSlot; /* Power of two >= nSeg */ + + for(nSlot=2; nSlot<nSeg; nSlot=nSlot*2); + pNew = fts5IdxMalloc(p, + sizeof(Fts5Iter) + /* pNew */ + sizeof(Fts5SegIter) * (nSlot-1) + /* pNew->aSeg[] */ + sizeof(Fts5CResult) * nSlot /* pNew->aFirst[] */ + ); + if( pNew ){ + pNew->nSeg = nSlot; + pNew->aFirst = (Fts5CResult*)&pNew->aSeg[nSlot]; + pNew->pIndex = p; + pNew->xSetOutputs = fts5IterSetOutputs_Noop; + } + return pNew; +} + +static void fts5PoslistCallback( + Fts5Index *pUnused, + void *pContext, + const u8 *pChunk, int nChunk +){ + UNUSED_PARAM(pUnused); + assert_nc( nChunk>=0 ); + if( nChunk>0 ){ + fts5BufferSafeAppendBlob((Fts5Buffer*)pContext, pChunk, nChunk); + } +} + +typedef struct PoslistCallbackCtx PoslistCallbackCtx; +struct PoslistCallbackCtx { + Fts5Buffer *pBuf; /* Append to this buffer */ + Fts5Colset *pColset; /* Restrict matches to this column */ + int eState; /* See above */ +}; + +typedef struct PoslistOffsetsCtx PoslistOffsetsCtx; +struct PoslistOffsetsCtx { + Fts5Buffer *pBuf; /* Append to this buffer */ + Fts5Colset *pColset; /* Restrict matches to this column */ + int iRead; + int iWrite; +}; + +/* +** TODO: Make this more efficient! +*/ +static int fts5IndexColsetTest(Fts5Colset *pColset, int iCol){ + int i; + for(i=0; i<pColset->nCol; i++){ + if( pColset->aiCol[i]==iCol ) return 1; + } + return 0; +} + +static void fts5PoslistOffsetsCallback( + Fts5Index *pUnused, + void *pContext, + const u8 *pChunk, int nChunk +){ + PoslistOffsetsCtx *pCtx = (PoslistOffsetsCtx*)pContext; + UNUSED_PARAM(pUnused); + assert_nc( nChunk>=0 ); + if( nChunk>0 ){ + int i = 0; + while( i<nChunk ){ + int iVal; + i += fts5GetVarint32(&pChunk[i], iVal); + iVal += pCtx->iRead - 2; + pCtx->iRead = iVal; + if( fts5IndexColsetTest(pCtx->pColset, iVal) ){ + fts5BufferSafeAppendVarint(pCtx->pBuf, iVal + 2 - pCtx->iWrite); + pCtx->iWrite = iVal; + } + } + } +} + +static void fts5PoslistFilterCallback( + Fts5Index *pUnused, + void *pContext, + const u8 *pChunk, int nChunk +){ + PoslistCallbackCtx *pCtx = (PoslistCallbackCtx*)pContext; + UNUSED_PARAM(pUnused); + assert_nc( nChunk>=0 ); + if( nChunk>0 ){ + /* Search through to find the first varint with value 1. This is the + ** start of the next columns hits. */ + int i = 0; + int iStart = 0; + + if( pCtx->eState==2 ){ + int iCol; + fts5FastGetVarint32(pChunk, i, iCol); + if( fts5IndexColsetTest(pCtx->pColset, iCol) ){ + pCtx->eState = 1; + fts5BufferSafeAppendVarint(pCtx->pBuf, 1); + }else{ + pCtx->eState = 0; + } + } + + do { + while( i<nChunk && pChunk[i]!=0x01 ){ + while( pChunk[i] & 0x80 ) i++; + i++; + } + if( pCtx->eState ){ + fts5BufferSafeAppendBlob(pCtx->pBuf, &pChunk[iStart], i-iStart); + } + if( i<nChunk ){ + int iCol; + iStart = i; + i++; + if( i>=nChunk ){ + pCtx->eState = 2; + }else{ + fts5FastGetVarint32(pChunk, i, iCol); + pCtx->eState = fts5IndexColsetTest(pCtx->pColset, iCol); + if( pCtx->eState ){ + fts5BufferSafeAppendBlob(pCtx->pBuf, &pChunk[iStart], i-iStart); + iStart = i; + } + } + } + }while( i<nChunk ); + } +} + +static void fts5ChunkIterate( + Fts5Index *p, /* Index object */ + Fts5SegIter *pSeg, /* Poslist of this iterator */ + void *pCtx, /* Context pointer for xChunk callback */ + void (*xChunk)(Fts5Index*, void*, const u8*, int) +){ + int nRem = pSeg->nPos; /* Number of bytes still to come */ + Fts5Data *pData = 0; + u8 *pChunk = &pSeg->pLeaf->p[pSeg->iLeafOffset]; + int nChunk = MIN(nRem, pSeg->pLeaf->szLeaf - pSeg->iLeafOffset); + int pgno = pSeg->iLeafPgno; + int pgnoSave = 0; + + /* This function does notmwork with detail=none databases. */ + assert( p->pConfig->eDetail!=FTS5_DETAIL_NONE ); + + if( (pSeg->flags & FTS5_SEGITER_REVERSE)==0 ){ + pgnoSave = pgno+1; + } + + while( 1 ){ + xChunk(p, pCtx, pChunk, nChunk); + nRem -= nChunk; + fts5DataRelease(pData); + if( nRem<=0 ){ + break; + }else{ + pgno++; + pData = fts5DataRead(p, FTS5_SEGMENT_ROWID(pSeg->pSeg->iSegid, pgno)); + if( pData==0 ) break; + pChunk = &pData->p[4]; + nChunk = MIN(nRem, pData->szLeaf - 4); + if( pgno==pgnoSave ){ + assert( pSeg->pNextLeaf==0 ); + pSeg->pNextLeaf = pData; + pData = 0; + } + } + } +} + +/* +** Iterator pIter currently points to a valid entry (not EOF). This +** function appends the position list data for the current entry to +** buffer pBuf. It does not make a copy of the position-list size +** field. +*/ +static void fts5SegiterPoslist( + Fts5Index *p, + Fts5SegIter *pSeg, + Fts5Colset *pColset, + Fts5Buffer *pBuf +){ + if( 0==fts5BufferGrow(&p->rc, pBuf, pSeg->nPos) ){ + if( pColset==0 ){ + fts5ChunkIterate(p, pSeg, (void*)pBuf, fts5PoslistCallback); + }else{ + if( p->pConfig->eDetail==FTS5_DETAIL_FULL ){ + PoslistCallbackCtx sCtx; + sCtx.pBuf = pBuf; + sCtx.pColset = pColset; + sCtx.eState = fts5IndexColsetTest(pColset, 0); + assert( sCtx.eState==0 || sCtx.eState==1 ); + fts5ChunkIterate(p, pSeg, (void*)&sCtx, fts5PoslistFilterCallback); + }else{ + PoslistOffsetsCtx sCtx; + memset(&sCtx, 0, sizeof(sCtx)); + sCtx.pBuf = pBuf; + sCtx.pColset = pColset; + fts5ChunkIterate(p, pSeg, (void*)&sCtx, fts5PoslistOffsetsCallback); + } + } + } +} + +/* +** IN/OUT parameter (*pa) points to a position list n bytes in size. If +** the position list contains entries for column iCol, then (*pa) is set +** to point to the sub-position-list for that column and the number of +** bytes in it returned. Or, if the argument position list does not +** contain any entries for column iCol, return 0. +*/ +static int fts5IndexExtractCol( + const u8 **pa, /* IN/OUT: Pointer to poslist */ + int n, /* IN: Size of poslist in bytes */ + int iCol /* Column to extract from poslist */ +){ + int iCurrent = 0; /* Anything before the first 0x01 is col 0 */ + const u8 *p = *pa; + const u8 *pEnd = &p[n]; /* One byte past end of position list */ + + while( iCol>iCurrent ){ + /* Advance pointer p until it points to pEnd or an 0x01 byte that is + ** not part of a varint. Note that it is not possible for a negative + ** or extremely large varint to occur within an uncorrupted position + ** list. So the last byte of each varint may be assumed to have a clear + ** 0x80 bit. */ + while( *p!=0x01 ){ + while( *p++ & 0x80 ); + if( p>=pEnd ) return 0; + } + *pa = p++; + iCurrent = *p++; + if( iCurrent & 0x80 ){ + p--; + p += fts5GetVarint32(p, iCurrent); + } + } + if( iCol!=iCurrent ) return 0; + + /* Advance pointer p until it points to pEnd or an 0x01 byte that is + ** not part of a varint */ + while( p<pEnd && *p!=0x01 ){ + while( *p++ & 0x80 ); + } + + return p - (*pa); +} + +static int fts5IndexExtractColset ( + Fts5Colset *pColset, /* Colset to filter on */ + const u8 *pPos, int nPos, /* Position list */ + Fts5Buffer *pBuf /* Output buffer */ +){ + int rc = SQLITE_OK; + int i; + + fts5BufferZero(pBuf); + for(i=0; i<pColset->nCol; i++){ + const u8 *pSub = pPos; + int nSub = fts5IndexExtractCol(&pSub, nPos, pColset->aiCol[i]); + if( nSub ){ + fts5BufferAppendBlob(&rc, pBuf, nSub, pSub); + } + } + return rc; +} + +/* +** xSetOutputs callback used by detail=none tables. +*/ +static void fts5IterSetOutputs_None(Fts5Iter *pIter, Fts5SegIter *pSeg){ + assert( pIter->pIndex->pConfig->eDetail==FTS5_DETAIL_NONE ); + pIter->base.iRowid = pSeg->iRowid; + pIter->base.nData = pSeg->nPos; +} + +/* +** xSetOutputs callback used by detail=full and detail=col tables when no +** column filters are specified. +*/ +static void fts5IterSetOutputs_Nocolset(Fts5Iter *pIter, Fts5SegIter *pSeg){ + pIter->base.iRowid = pSeg->iRowid; + pIter->base.nData = pSeg->nPos; + + assert( pIter->pIndex->pConfig->eDetail!=FTS5_DETAIL_NONE ); + assert( pIter->pColset==0 ); + + if( pSeg->iLeafOffset+pSeg->nPos<=pSeg->pLeaf->szLeaf ){ + /* All data is stored on the current page. Populate the output + ** variables to point into the body of the page object. */ + pIter->base.pData = &pSeg->pLeaf->p[pSeg->iLeafOffset]; + }else{ + /* The data is distributed over two or more pages. Copy it into the + ** Fts5Iter.poslist buffer and then set the output pointer to point + ** to this buffer. */ + fts5BufferZero(&pIter->poslist); + fts5SegiterPoslist(pIter->pIndex, pSeg, 0, &pIter->poslist); + pIter->base.pData = pIter->poslist.p; + } +} + +/* +** xSetOutputs callback used by detail=col when there is a column filter +** and there are 100 or more columns. Also called as a fallback from +** fts5IterSetOutputs_Col100 if the column-list spans more than one page. +*/ +static void fts5IterSetOutputs_Col(Fts5Iter *pIter, Fts5SegIter *pSeg){ + fts5BufferZero(&pIter->poslist); + fts5SegiterPoslist(pIter->pIndex, pSeg, pIter->pColset, &pIter->poslist); + pIter->base.iRowid = pSeg->iRowid; + pIter->base.pData = pIter->poslist.p; + pIter->base.nData = pIter->poslist.n; +} + +/* +** xSetOutputs callback used when: +** +** * detail=col, +** * there is a column filter, and +** * the table contains 100 or fewer columns. +** +** The last point is to ensure all column numbers are stored as +** single-byte varints. +*/ +static void fts5IterSetOutputs_Col100(Fts5Iter *pIter, Fts5SegIter *pSeg){ + + assert( pIter->pIndex->pConfig->eDetail==FTS5_DETAIL_COLUMNS ); + assert( pIter->pColset ); + + if( pSeg->iLeafOffset+pSeg->nPos>pSeg->pLeaf->szLeaf ){ + fts5IterSetOutputs_Col(pIter, pSeg); + }else{ + u8 *a = (u8*)&pSeg->pLeaf->p[pSeg->iLeafOffset]; + u8 *pEnd = (u8*)&a[pSeg->nPos]; + int iPrev = 0; + int *aiCol = pIter->pColset->aiCol; + int *aiColEnd = &aiCol[pIter->pColset->nCol]; + + u8 *aOut = pIter->poslist.p; + int iPrevOut = 0; + + pIter->base.iRowid = pSeg->iRowid; + + while( a<pEnd ){ + iPrev += (int)a++[0] - 2; + while( *aiCol<iPrev ){ + aiCol++; + if( aiCol==aiColEnd ) goto setoutputs_col_out; + } + if( *aiCol==iPrev ){ + *aOut++ = (iPrev - iPrevOut) + 2; + iPrevOut = iPrev; + } + } + +setoutputs_col_out: + pIter->base.pData = pIter->poslist.p; + pIter->base.nData = aOut - pIter->poslist.p; + } +} + +/* +** xSetOutputs callback used by detail=full when there is a column filter. +*/ +static void fts5IterSetOutputs_Full(Fts5Iter *pIter, Fts5SegIter *pSeg){ + Fts5Colset *pColset = pIter->pColset; + pIter->base.iRowid = pSeg->iRowid; + + assert( pIter->pIndex->pConfig->eDetail==FTS5_DETAIL_FULL ); + assert( pColset ); + + if( pSeg->iLeafOffset+pSeg->nPos<=pSeg->pLeaf->szLeaf ){ + /* All data is stored on the current page. Populate the output + ** variables to point into the body of the page object. */ + const u8 *a = &pSeg->pLeaf->p[pSeg->iLeafOffset]; + if( pColset->nCol==1 ){ + pIter->base.nData = fts5IndexExtractCol(&a, pSeg->nPos,pColset->aiCol[0]); + pIter->base.pData = a; + }else{ + fts5BufferZero(&pIter->poslist); + fts5IndexExtractColset(pColset, a, pSeg->nPos, &pIter->poslist); + pIter->base.pData = pIter->poslist.p; + pIter->base.nData = pIter->poslist.n; + } + }else{ + /* The data is distributed over two or more pages. Copy it into the + ** Fts5Iter.poslist buffer and then set the output pointer to point + ** to this buffer. */ + fts5BufferZero(&pIter->poslist); + fts5SegiterPoslist(pIter->pIndex, pSeg, pColset, &pIter->poslist); + pIter->base.pData = pIter->poslist.p; + pIter->base.nData = pIter->poslist.n; + } +} + +static void fts5IterSetOutputCb(int *pRc, Fts5Iter *pIter){ + if( *pRc==SQLITE_OK ){ + Fts5Config *pConfig = pIter->pIndex->pConfig; + if( pConfig->eDetail==FTS5_DETAIL_NONE ){ + pIter->xSetOutputs = fts5IterSetOutputs_None; + } + + else if( pIter->pColset==0 ){ + pIter->xSetOutputs = fts5IterSetOutputs_Nocolset; + } + + else if( pConfig->eDetail==FTS5_DETAIL_FULL ){ + pIter->xSetOutputs = fts5IterSetOutputs_Full; + } + + else{ + assert( pConfig->eDetail==FTS5_DETAIL_COLUMNS ); + if( pConfig->nCol<=100 ){ + pIter->xSetOutputs = fts5IterSetOutputs_Col100; + sqlite3Fts5BufferSize(pRc, &pIter->poslist, pConfig->nCol); + }else{ + pIter->xSetOutputs = fts5IterSetOutputs_Col; + } + } + } +} + + +/* +** Allocate a new Fts5Iter object. +** +** The new object will be used to iterate through data in structure pStruct. +** If iLevel is -ve, then all data in all segments is merged. Or, if iLevel +** is zero or greater, data from the first nSegment segments on level iLevel +** is merged. +** +** The iterator initially points to the first term/rowid entry in the +** iterated data. +*/ +static void fts5MultiIterNew( + Fts5Index *p, /* FTS5 backend to iterate within */ + Fts5Structure *pStruct, /* Structure of specific index */ + int flags, /* FTS5INDEX_QUERY_XXX flags */ + Fts5Colset *pColset, /* Colset to filter on (or NULL) */ + const u8 *pTerm, int nTerm, /* Term to seek to (or NULL/0) */ + int iLevel, /* Level to iterate (-1 for all) */ + int nSegment, /* Number of segments to merge (iLevel>=0) */ + Fts5Iter **ppOut /* New object */ +){ + int nSeg = 0; /* Number of segment-iters in use */ + int iIter = 0; /* */ + int iSeg; /* Used to iterate through segments */ + Fts5StructureLevel *pLvl; + Fts5Iter *pNew; + + assert( (pTerm==0 && nTerm==0) || iLevel<0 ); + + /* Allocate space for the new multi-seg-iterator. */ + if( p->rc==SQLITE_OK ){ + if( iLevel<0 ){ + assert( pStruct->nSegment==fts5StructureCountSegments(pStruct) ); + nSeg = pStruct->nSegment; + nSeg += (p->pHash ? 1 : 0); + }else{ + nSeg = MIN(pStruct->aLevel[iLevel].nSeg, nSegment); + } + } + *ppOut = pNew = fts5MultiIterAlloc(p, nSeg); + if( pNew==0 ) return; + pNew->bRev = (0!=(flags & FTS5INDEX_QUERY_DESC)); + pNew->bSkipEmpty = (0!=(flags & FTS5INDEX_QUERY_SKIPEMPTY)); + pNew->pStruct = pStruct; + pNew->pColset = pColset; + fts5StructureRef(pStruct); + if( (flags & FTS5INDEX_QUERY_NOOUTPUT)==0 ){ + fts5IterSetOutputCb(&p->rc, pNew); + } + + /* Initialize each of the component segment iterators. */ + if( p->rc==SQLITE_OK ){ + if( iLevel<0 ){ + Fts5StructureLevel *pEnd = &pStruct->aLevel[pStruct->nLevel]; + if( p->pHash ){ + /* Add a segment iterator for the current contents of the hash table. */ + Fts5SegIter *pIter = &pNew->aSeg[iIter++]; + fts5SegIterHashInit(p, pTerm, nTerm, flags, pIter); + } + for(pLvl=&pStruct->aLevel[0]; pLvl<pEnd; pLvl++){ + for(iSeg=pLvl->nSeg-1; iSeg>=0; iSeg--){ + Fts5StructureSegment *pSeg = &pLvl->aSeg[iSeg]; + Fts5SegIter *pIter = &pNew->aSeg[iIter++]; + if( pTerm==0 ){ + fts5SegIterInit(p, pSeg, pIter); + }else{ + fts5SegIterSeekInit(p, pTerm, nTerm, flags, pSeg, pIter); + } + } + } + }else{ + pLvl = &pStruct->aLevel[iLevel]; + for(iSeg=nSeg-1; iSeg>=0; iSeg--){ + fts5SegIterInit(p, &pLvl->aSeg[iSeg], &pNew->aSeg[iIter++]); + } + } + assert( iIter==nSeg ); + } + + /* If the above was successful, each component iterators now points + ** to the first entry in its segment. In this case initialize the + ** aFirst[] array. Or, if an error has occurred, free the iterator + ** object and set the output variable to NULL. */ + if( p->rc==SQLITE_OK ){ + for(iIter=pNew->nSeg-1; iIter>0; iIter--){ + int iEq; + if( (iEq = fts5MultiIterDoCompare(pNew, iIter)) ){ + Fts5SegIter *pSeg = &pNew->aSeg[iEq]; + if( p->rc==SQLITE_OK ) pSeg->xNext(p, pSeg, 0); + fts5MultiIterAdvanced(p, pNew, iEq, iIter); + } + } + fts5MultiIterSetEof(pNew); + fts5AssertMultiIterSetup(p, pNew); + + if( pNew->bSkipEmpty && fts5MultiIterIsEmpty(p, pNew) ){ + fts5MultiIterNext(p, pNew, 0, 0); + }else if( pNew->base.bEof==0 ){ + Fts5SegIter *pSeg = &pNew->aSeg[pNew->aFirst[1].iFirst]; + pNew->xSetOutputs(pNew, pSeg); + } + + }else{ + fts5MultiIterFree(pNew); + *ppOut = 0; + } +} + +/* +** Create an Fts5Iter that iterates through the doclist provided +** as the second argument. +*/ +static void fts5MultiIterNew2( + Fts5Index *p, /* FTS5 backend to iterate within */ + Fts5Data *pData, /* Doclist to iterate through */ + int bDesc, /* True for descending rowid order */ + Fts5Iter **ppOut /* New object */ +){ + Fts5Iter *pNew; + pNew = fts5MultiIterAlloc(p, 2); + if( pNew ){ + Fts5SegIter *pIter = &pNew->aSeg[1]; + + pIter->flags = FTS5_SEGITER_ONETERM; + if( pData->szLeaf>0 ){ + pIter->pLeaf = pData; + pIter->iLeafOffset = fts5GetVarint(pData->p, (u64*)&pIter->iRowid); + pIter->iEndofDoclist = pData->nn; + pNew->aFirst[1].iFirst = 1; + if( bDesc ){ + pNew->bRev = 1; + pIter->flags |= FTS5_SEGITER_REVERSE; + fts5SegIterReverseInitPage(p, pIter); + }else{ + fts5SegIterLoadNPos(p, pIter); + } + pData = 0; + }else{ + pNew->base.bEof = 1; + } + fts5SegIterSetNext(p, pIter); + + *ppOut = pNew; + } + + fts5DataRelease(pData); +} + +/* +** Return true if the iterator is at EOF or if an error has occurred. +** False otherwise. +*/ +static int fts5MultiIterEof(Fts5Index *p, Fts5Iter *pIter){ + assert( p->rc + || (pIter->aSeg[ pIter->aFirst[1].iFirst ].pLeaf==0)==pIter->base.bEof + ); + return (p->rc || pIter->base.bEof); +} + +/* +** Return the rowid of the entry that the iterator currently points +** to. If the iterator points to EOF when this function is called the +** results are undefined. +*/ +static i64 fts5MultiIterRowid(Fts5Iter *pIter){ + assert( pIter->aSeg[ pIter->aFirst[1].iFirst ].pLeaf ); + return pIter->aSeg[ pIter->aFirst[1].iFirst ].iRowid; +} + +/* +** Move the iterator to the next entry at or following iMatch. +*/ +static void fts5MultiIterNextFrom( + Fts5Index *p, + Fts5Iter *pIter, + i64 iMatch +){ + while( 1 ){ + i64 iRowid; + fts5MultiIterNext(p, pIter, 1, iMatch); + if( fts5MultiIterEof(p, pIter) ) break; + iRowid = fts5MultiIterRowid(pIter); + if( pIter->bRev==0 && iRowid>=iMatch ) break; + if( pIter->bRev!=0 && iRowid<=iMatch ) break; + } +} + +/* +** Return a pointer to a buffer containing the term associated with the +** entry that the iterator currently points to. +*/ +static const u8 *fts5MultiIterTerm(Fts5Iter *pIter, int *pn){ + Fts5SegIter *p = &pIter->aSeg[ pIter->aFirst[1].iFirst ]; + *pn = p->term.n; + return p->term.p; +} + +/* +** Allocate a new segment-id for the structure pStruct. The new segment +** id must be between 1 and 65335 inclusive, and must not be used by +** any currently existing segment. If a free segment id cannot be found, +** SQLITE_FULL is returned. +** +** If an error has already occurred, this function is a no-op. 0 is +** returned in this case. +*/ +static int fts5AllocateSegid(Fts5Index *p, Fts5Structure *pStruct){ + int iSegid = 0; + + if( p->rc==SQLITE_OK ){ + if( pStruct->nSegment>=FTS5_MAX_SEGMENT ){ + p->rc = SQLITE_FULL; + }else{ + while( iSegid==0 ){ + int iLvl, iSeg; + sqlite3_randomness(sizeof(u32), (void*)&iSegid); + iSegid = iSegid & ((1 << FTS5_DATA_ID_B)-1); + for(iLvl=0; iLvl<pStruct->nLevel; iLvl++){ + for(iSeg=0; iSeg<pStruct->aLevel[iLvl].nSeg; iSeg++){ + if( iSegid==pStruct->aLevel[iLvl].aSeg[iSeg].iSegid ){ + iSegid = 0; + } + } + } + } + } + } + + return iSegid; +} + +/* +** Discard all data currently cached in the hash-tables. +*/ +static void fts5IndexDiscardData(Fts5Index *p){ + assert( p->pHash || p->nPendingData==0 ); + if( p->pHash ){ + sqlite3Fts5HashClear(p->pHash); + p->nPendingData = 0; + } +} + +/* +** Return the size of the prefix, in bytes, that buffer +** (pNew/<length-unknown>) shares with buffer (pOld/nOld). +** +** Buffer (pNew/<length-unknown>) is guaranteed to be greater +** than buffer (pOld/nOld). +*/ +static int fts5PrefixCompress(int nOld, const u8 *pOld, const u8 *pNew){ + int i; + for(i=0; i<nOld; i++){ + if( pOld[i]!=pNew[i] ) break; + } + return i; +} + +static void fts5WriteDlidxClear( + Fts5Index *p, + Fts5SegWriter *pWriter, + int bFlush /* If true, write dlidx to disk */ +){ + int i; + assert( bFlush==0 || (pWriter->nDlidx>0 && pWriter->aDlidx[0].buf.n>0) ); + for(i=0; i<pWriter->nDlidx; i++){ + Fts5DlidxWriter *pDlidx = &pWriter->aDlidx[i]; + if( pDlidx->buf.n==0 ) break; + if( bFlush ){ + assert( pDlidx->pgno!=0 ); + fts5DataWrite(p, + FTS5_DLIDX_ROWID(pWriter->iSegid, i, pDlidx->pgno), + pDlidx->buf.p, pDlidx->buf.n + ); + } + sqlite3Fts5BufferZero(&pDlidx->buf); + pDlidx->bPrevValid = 0; + } +} + +/* +** Grow the pWriter->aDlidx[] array to at least nLvl elements in size. +** Any new array elements are zeroed before returning. +*/ +static int fts5WriteDlidxGrow( + Fts5Index *p, + Fts5SegWriter *pWriter, + int nLvl +){ + if( p->rc==SQLITE_OK && nLvl>=pWriter->nDlidx ){ + Fts5DlidxWriter *aDlidx = (Fts5DlidxWriter*)sqlite3_realloc( + pWriter->aDlidx, sizeof(Fts5DlidxWriter) * nLvl + ); + if( aDlidx==0 ){ + p->rc = SQLITE_NOMEM; + }else{ + int nByte = sizeof(Fts5DlidxWriter) * (nLvl - pWriter->nDlidx); + memset(&aDlidx[pWriter->nDlidx], 0, nByte); + pWriter->aDlidx = aDlidx; + pWriter->nDlidx = nLvl; + } + } + return p->rc; +} + +/* +** If the current doclist-index accumulating in pWriter->aDlidx[] is large +** enough, flush it to disk and return 1. Otherwise discard it and return +** zero. +*/ +static int fts5WriteFlushDlidx(Fts5Index *p, Fts5SegWriter *pWriter){ + int bFlag = 0; + + /* If there were FTS5_MIN_DLIDX_SIZE or more empty leaf pages written + ** to the database, also write the doclist-index to disk. */ + if( pWriter->aDlidx[0].buf.n>0 && pWriter->nEmpty>=FTS5_MIN_DLIDX_SIZE ){ + bFlag = 1; + } + fts5WriteDlidxClear(p, pWriter, bFlag); + pWriter->nEmpty = 0; + return bFlag; +} + +/* +** This function is called whenever processing of the doclist for the +** last term on leaf page (pWriter->iBtPage) is completed. +** +** The doclist-index for that term is currently stored in-memory within the +** Fts5SegWriter.aDlidx[] array. If it is large enough, this function +** writes it out to disk. Or, if it is too small to bother with, discards +** it. +** +** Fts5SegWriter.btterm currently contains the first term on page iBtPage. +*/ +static void fts5WriteFlushBtree(Fts5Index *p, Fts5SegWriter *pWriter){ + int bFlag; + + assert( pWriter->iBtPage || pWriter->nEmpty==0 ); + if( pWriter->iBtPage==0 ) return; + bFlag = fts5WriteFlushDlidx(p, pWriter); + + if( p->rc==SQLITE_OK ){ + const char *z = (pWriter->btterm.n>0?(const char*)pWriter->btterm.p:""); + /* The following was already done in fts5WriteInit(): */ + /* sqlite3_bind_int(p->pIdxWriter, 1, pWriter->iSegid); */ + sqlite3_bind_blob(p->pIdxWriter, 2, z, pWriter->btterm.n, SQLITE_STATIC); + sqlite3_bind_int64(p->pIdxWriter, 3, bFlag + ((i64)pWriter->iBtPage<<1)); + sqlite3_step(p->pIdxWriter); + p->rc = sqlite3_reset(p->pIdxWriter); + } + pWriter->iBtPage = 0; +} + +/* +** This is called once for each leaf page except the first that contains +** at least one term. Argument (nTerm/pTerm) is the split-key - a term that +** is larger than all terms written to earlier leaves, and equal to or +** smaller than the first term on the new leaf. +** +** If an error occurs, an error code is left in Fts5Index.rc. If an error +** has already occurred when this function is called, it is a no-op. +*/ +static void fts5WriteBtreeTerm( + Fts5Index *p, /* FTS5 backend object */ + Fts5SegWriter *pWriter, /* Writer object */ + int nTerm, const u8 *pTerm /* First term on new page */ +){ + fts5WriteFlushBtree(p, pWriter); + fts5BufferSet(&p->rc, &pWriter->btterm, nTerm, pTerm); + pWriter->iBtPage = pWriter->writer.pgno; +} + +/* +** This function is called when flushing a leaf page that contains no +** terms at all to disk. +*/ +static void fts5WriteBtreeNoTerm( + Fts5Index *p, /* FTS5 backend object */ + Fts5SegWriter *pWriter /* Writer object */ +){ + /* If there were no rowids on the leaf page either and the doclist-index + ** has already been started, append an 0x00 byte to it. */ + if( pWriter->bFirstRowidInPage && pWriter->aDlidx[0].buf.n>0 ){ + Fts5DlidxWriter *pDlidx = &pWriter->aDlidx[0]; + assert( pDlidx->bPrevValid ); + sqlite3Fts5BufferAppendVarint(&p->rc, &pDlidx->buf, 0); + } + + /* Increment the "number of sequential leaves without a term" counter. */ + pWriter->nEmpty++; +} + +static i64 fts5DlidxExtractFirstRowid(Fts5Buffer *pBuf){ + i64 iRowid; + int iOff; + + iOff = 1 + fts5GetVarint(&pBuf->p[1], (u64*)&iRowid); + fts5GetVarint(&pBuf->p[iOff], (u64*)&iRowid); + return iRowid; +} + +/* +** Rowid iRowid has just been appended to the current leaf page. It is the +** first on the page. This function appends an appropriate entry to the current +** doclist-index. +*/ +static void fts5WriteDlidxAppend( + Fts5Index *p, + Fts5SegWriter *pWriter, + i64 iRowid +){ + int i; + int bDone = 0; + + for(i=0; p->rc==SQLITE_OK && bDone==0; i++){ + i64 iVal; + Fts5DlidxWriter *pDlidx = &pWriter->aDlidx[i]; + + if( pDlidx->buf.n>=p->pConfig->pgsz ){ + /* The current doclist-index page is full. Write it to disk and push + ** a copy of iRowid (which will become the first rowid on the next + ** doclist-index leaf page) up into the next level of the b-tree + ** hierarchy. If the node being flushed is currently the root node, + ** also push its first rowid upwards. */ + pDlidx->buf.p[0] = 0x01; /* Not the root node */ + fts5DataWrite(p, + FTS5_DLIDX_ROWID(pWriter->iSegid, i, pDlidx->pgno), + pDlidx->buf.p, pDlidx->buf.n + ); + fts5WriteDlidxGrow(p, pWriter, i+2); + pDlidx = &pWriter->aDlidx[i]; + if( p->rc==SQLITE_OK && pDlidx[1].buf.n==0 ){ + i64 iFirst = fts5DlidxExtractFirstRowid(&pDlidx->buf); + + /* This was the root node. Push its first rowid up to the new root. */ + pDlidx[1].pgno = pDlidx->pgno; + sqlite3Fts5BufferAppendVarint(&p->rc, &pDlidx[1].buf, 0); + sqlite3Fts5BufferAppendVarint(&p->rc, &pDlidx[1].buf, pDlidx->pgno); + sqlite3Fts5BufferAppendVarint(&p->rc, &pDlidx[1].buf, iFirst); + pDlidx[1].bPrevValid = 1; + pDlidx[1].iPrev = iFirst; + } + + sqlite3Fts5BufferZero(&pDlidx->buf); + pDlidx->bPrevValid = 0; + pDlidx->pgno++; + }else{ + bDone = 1; + } + + if( pDlidx->bPrevValid ){ + iVal = iRowid - pDlidx->iPrev; + }else{ + i64 iPgno = (i==0 ? pWriter->writer.pgno : pDlidx[-1].pgno); + assert( pDlidx->buf.n==0 ); + sqlite3Fts5BufferAppendVarint(&p->rc, &pDlidx->buf, !bDone); + sqlite3Fts5BufferAppendVarint(&p->rc, &pDlidx->buf, iPgno); + iVal = iRowid; + } + + sqlite3Fts5BufferAppendVarint(&p->rc, &pDlidx->buf, iVal); + pDlidx->bPrevValid = 1; + pDlidx->iPrev = iRowid; + } +} + +static void fts5WriteFlushLeaf(Fts5Index *p, Fts5SegWriter *pWriter){ + static const u8 zero[] = { 0x00, 0x00, 0x00, 0x00 }; + Fts5PageWriter *pPage = &pWriter->writer; + i64 iRowid; + + assert( (pPage->pgidx.n==0)==(pWriter->bFirstTermInPage) ); + + /* Set the szLeaf header field. */ + assert( 0==fts5GetU16(&pPage->buf.p[2]) ); + fts5PutU16(&pPage->buf.p[2], (u16)pPage->buf.n); + + if( pWriter->bFirstTermInPage ){ + /* No term was written to this page. */ + assert( pPage->pgidx.n==0 ); + fts5WriteBtreeNoTerm(p, pWriter); + }else{ + /* Append the pgidx to the page buffer. Set the szLeaf header field. */ + fts5BufferAppendBlob(&p->rc, &pPage->buf, pPage->pgidx.n, pPage->pgidx.p); + } + + /* Write the page out to disk */ + iRowid = FTS5_SEGMENT_ROWID(pWriter->iSegid, pPage->pgno); + fts5DataWrite(p, iRowid, pPage->buf.p, pPage->buf.n); + + /* Initialize the next page. */ + fts5BufferZero(&pPage->buf); + fts5BufferZero(&pPage->pgidx); + fts5BufferAppendBlob(&p->rc, &pPage->buf, 4, zero); + pPage->iPrevPgidx = 0; + pPage->pgno++; + + /* Increase the leaves written counter */ + pWriter->nLeafWritten++; + + /* The new leaf holds no terms or rowids */ + pWriter->bFirstTermInPage = 1; + pWriter->bFirstRowidInPage = 1; +} + +/* +** Append term pTerm/nTerm to the segment being written by the writer passed +** as the second argument. +** +** If an error occurs, set the Fts5Index.rc error code. If an error has +** already occurred, this function is a no-op. +*/ +static void fts5WriteAppendTerm( + Fts5Index *p, + Fts5SegWriter *pWriter, + int nTerm, const u8 *pTerm +){ + int nPrefix; /* Bytes of prefix compression for term */ + Fts5PageWriter *pPage = &pWriter->writer; + Fts5Buffer *pPgidx = &pWriter->writer.pgidx; + + assert( p->rc==SQLITE_OK ); + assert( pPage->buf.n>=4 ); + assert( pPage->buf.n>4 || pWriter->bFirstTermInPage ); + + /* If the current leaf page is full, flush it to disk. */ + if( (pPage->buf.n + pPgidx->n + nTerm + 2)>=p->pConfig->pgsz ){ + if( pPage->buf.n>4 ){ + fts5WriteFlushLeaf(p, pWriter); + } + fts5BufferGrow(&p->rc, &pPage->buf, nTerm+FTS5_DATA_PADDING); + } + + /* TODO1: Updating pgidx here. */ + pPgidx->n += sqlite3Fts5PutVarint( + &pPgidx->p[pPgidx->n], pPage->buf.n - pPage->iPrevPgidx + ); + pPage->iPrevPgidx = pPage->buf.n; +#if 0 + fts5PutU16(&pPgidx->p[pPgidx->n], pPage->buf.n); + pPgidx->n += 2; +#endif + + if( pWriter->bFirstTermInPage ){ + nPrefix = 0; + if( pPage->pgno!=1 ){ + /* This is the first term on a leaf that is not the leftmost leaf in + ** the segment b-tree. In this case it is necessary to add a term to + ** the b-tree hierarchy that is (a) larger than the largest term + ** already written to the segment and (b) smaller than or equal to + ** this term. In other words, a prefix of (pTerm/nTerm) that is one + ** byte longer than the longest prefix (pTerm/nTerm) shares with the + ** previous term. + ** + ** Usually, the previous term is available in pPage->term. The exception + ** is if this is the first term written in an incremental-merge step. + ** In this case the previous term is not available, so just write a + ** copy of (pTerm/nTerm) into the parent node. This is slightly + ** inefficient, but still correct. */ + int n = nTerm; + if( pPage->term.n ){ + n = 1 + fts5PrefixCompress(pPage->term.n, pPage->term.p, pTerm); + } + fts5WriteBtreeTerm(p, pWriter, n, pTerm); + pPage = &pWriter->writer; + } + }else{ + nPrefix = fts5PrefixCompress(pPage->term.n, pPage->term.p, pTerm); + fts5BufferAppendVarint(&p->rc, &pPage->buf, nPrefix); + } + + /* Append the number of bytes of new data, then the term data itself + ** to the page. */ + fts5BufferAppendVarint(&p->rc, &pPage->buf, nTerm - nPrefix); + fts5BufferAppendBlob(&p->rc, &pPage->buf, nTerm - nPrefix, &pTerm[nPrefix]); + + /* Update the Fts5PageWriter.term field. */ + fts5BufferSet(&p->rc, &pPage->term, nTerm, pTerm); + pWriter->bFirstTermInPage = 0; + + pWriter->bFirstRowidInPage = 0; + pWriter->bFirstRowidInDoclist = 1; + + assert( p->rc || (pWriter->nDlidx>0 && pWriter->aDlidx[0].buf.n==0) ); + pWriter->aDlidx[0].pgno = pPage->pgno; +} + +/* +** Append a rowid and position-list size field to the writers output. +*/ +static void fts5WriteAppendRowid( + Fts5Index *p, + Fts5SegWriter *pWriter, + i64 iRowid +){ + if( p->rc==SQLITE_OK ){ + Fts5PageWriter *pPage = &pWriter->writer; + + if( (pPage->buf.n + pPage->pgidx.n)>=p->pConfig->pgsz ){ + fts5WriteFlushLeaf(p, pWriter); + } + + /* If this is to be the first rowid written to the page, set the + ** rowid-pointer in the page-header. Also append a value to the dlidx + ** buffer, in case a doclist-index is required. */ + if( pWriter->bFirstRowidInPage ){ + fts5PutU16(pPage->buf.p, (u16)pPage->buf.n); + fts5WriteDlidxAppend(p, pWriter, iRowid); + } + + /* Write the rowid. */ + if( pWriter->bFirstRowidInDoclist || pWriter->bFirstRowidInPage ){ + fts5BufferAppendVarint(&p->rc, &pPage->buf, iRowid); + }else{ + assert( p->rc || iRowid>pWriter->iPrevRowid ); + fts5BufferAppendVarint(&p->rc, &pPage->buf, iRowid - pWriter->iPrevRowid); + } + pWriter->iPrevRowid = iRowid; + pWriter->bFirstRowidInDoclist = 0; + pWriter->bFirstRowidInPage = 0; + } +} + +static void fts5WriteAppendPoslistData( + Fts5Index *p, + Fts5SegWriter *pWriter, + const u8 *aData, + int nData +){ + Fts5PageWriter *pPage = &pWriter->writer; + const u8 *a = aData; + int n = nData; + + assert( p->pConfig->pgsz>0 ); + while( p->rc==SQLITE_OK + && (pPage->buf.n + pPage->pgidx.n + n)>=p->pConfig->pgsz + ){ + int nReq = p->pConfig->pgsz - pPage->buf.n - pPage->pgidx.n; + int nCopy = 0; + while( nCopy<nReq ){ + i64 dummy; + nCopy += fts5GetVarint(&a[nCopy], (u64*)&dummy); + } + fts5BufferAppendBlob(&p->rc, &pPage->buf, nCopy, a); + a += nCopy; + n -= nCopy; + fts5WriteFlushLeaf(p, pWriter); + } + if( n>0 ){ + fts5BufferAppendBlob(&p->rc, &pPage->buf, n, a); + } +} + +/* +** Flush any data cached by the writer object to the database. Free any +** allocations associated with the writer. +*/ +static void fts5WriteFinish( + Fts5Index *p, + Fts5SegWriter *pWriter, /* Writer object */ + int *pnLeaf /* OUT: Number of leaf pages in b-tree */ +){ + int i; + Fts5PageWriter *pLeaf = &pWriter->writer; + if( p->rc==SQLITE_OK ){ + assert( pLeaf->pgno>=1 ); + if( pLeaf->buf.n>4 ){ + fts5WriteFlushLeaf(p, pWriter); + } + *pnLeaf = pLeaf->pgno-1; + fts5WriteFlushBtree(p, pWriter); + } + fts5BufferFree(&pLeaf->term); + fts5BufferFree(&pLeaf->buf); + fts5BufferFree(&pLeaf->pgidx); + fts5BufferFree(&pWriter->btterm); + + for(i=0; i<pWriter->nDlidx; i++){ + sqlite3Fts5BufferFree(&pWriter->aDlidx[i].buf); + } + sqlite3_free(pWriter->aDlidx); +} + +static void fts5WriteInit( + Fts5Index *p, + Fts5SegWriter *pWriter, + int iSegid +){ + const int nBuffer = p->pConfig->pgsz + FTS5_DATA_PADDING; + + memset(pWriter, 0, sizeof(Fts5SegWriter)); + pWriter->iSegid = iSegid; + + fts5WriteDlidxGrow(p, pWriter, 1); + pWriter->writer.pgno = 1; + pWriter->bFirstTermInPage = 1; + pWriter->iBtPage = 1; + + assert( pWriter->writer.buf.n==0 ); + assert( pWriter->writer.pgidx.n==0 ); + + /* Grow the two buffers to pgsz + padding bytes in size. */ + sqlite3Fts5BufferSize(&p->rc, &pWriter->writer.pgidx, nBuffer); + sqlite3Fts5BufferSize(&p->rc, &pWriter->writer.buf, nBuffer); + + if( p->pIdxWriter==0 ){ + Fts5Config *pConfig = p->pConfig; + fts5IndexPrepareStmt(p, &p->pIdxWriter, sqlite3_mprintf( + "INSERT INTO '%q'.'%q_idx'(segid,term,pgno) VALUES(?,?,?)", + pConfig->zDb, pConfig->zName + )); + } + + if( p->rc==SQLITE_OK ){ + /* Initialize the 4-byte leaf-page header to 0x00. */ + memset(pWriter->writer.buf.p, 0, 4); + pWriter->writer.buf.n = 4; + + /* Bind the current output segment id to the index-writer. This is an + ** optimization over binding the same value over and over as rows are + ** inserted into %_idx by the current writer. */ + sqlite3_bind_int(p->pIdxWriter, 1, pWriter->iSegid); + } +} + +/* +** Iterator pIter was used to iterate through the input segments of on an +** incremental merge operation. This function is called if the incremental +** merge step has finished but the input has not been completely exhausted. +*/ +static void fts5TrimSegments(Fts5Index *p, Fts5Iter *pIter){ + int i; + Fts5Buffer buf; + memset(&buf, 0, sizeof(Fts5Buffer)); + for(i=0; i<pIter->nSeg; i++){ + Fts5SegIter *pSeg = &pIter->aSeg[i]; + if( pSeg->pSeg==0 ){ + /* no-op */ + }else if( pSeg->pLeaf==0 ){ + /* All keys from this input segment have been transfered to the output. + ** Set both the first and last page-numbers to 0 to indicate that the + ** segment is now empty. */ + pSeg->pSeg->pgnoLast = 0; + pSeg->pSeg->pgnoFirst = 0; + }else{ + int iOff = pSeg->iTermLeafOffset; /* Offset on new first leaf page */ + i64 iLeafRowid; + Fts5Data *pData; + int iId = pSeg->pSeg->iSegid; + u8 aHdr[4] = {0x00, 0x00, 0x00, 0x00}; + + iLeafRowid = FTS5_SEGMENT_ROWID(iId, pSeg->iTermLeafPgno); + pData = fts5DataRead(p, iLeafRowid); + if( pData ){ + fts5BufferZero(&buf); + fts5BufferGrow(&p->rc, &buf, pData->nn); + fts5BufferAppendBlob(&p->rc, &buf, sizeof(aHdr), aHdr); + fts5BufferAppendVarint(&p->rc, &buf, pSeg->term.n); + fts5BufferAppendBlob(&p->rc, &buf, pSeg->term.n, pSeg->term.p); + fts5BufferAppendBlob(&p->rc, &buf, pData->szLeaf-iOff, &pData->p[iOff]); + if( p->rc==SQLITE_OK ){ + /* Set the szLeaf field */ + fts5PutU16(&buf.p[2], (u16)buf.n); + } + + /* Set up the new page-index array */ + fts5BufferAppendVarint(&p->rc, &buf, 4); + if( pSeg->iLeafPgno==pSeg->iTermLeafPgno + && pSeg->iEndofDoclist<pData->szLeaf + ){ + int nDiff = pData->szLeaf - pSeg->iEndofDoclist; + fts5BufferAppendVarint(&p->rc, &buf, buf.n - 1 - nDiff - 4); + fts5BufferAppendBlob(&p->rc, &buf, + pData->nn - pSeg->iPgidxOff, &pData->p[pSeg->iPgidxOff] + ); + } + + fts5DataRelease(pData); + pSeg->pSeg->pgnoFirst = pSeg->iTermLeafPgno; + fts5DataDelete(p, FTS5_SEGMENT_ROWID(iId, 1), iLeafRowid); + fts5DataWrite(p, iLeafRowid, buf.p, buf.n); + } + } + } + fts5BufferFree(&buf); +} + +static void fts5MergeChunkCallback( + Fts5Index *p, + void *pCtx, + const u8 *pChunk, int nChunk +){ + Fts5SegWriter *pWriter = (Fts5SegWriter*)pCtx; + fts5WriteAppendPoslistData(p, pWriter, pChunk, nChunk); +} + +/* +** +*/ +static void fts5IndexMergeLevel( + Fts5Index *p, /* FTS5 backend object */ + Fts5Structure **ppStruct, /* IN/OUT: Stucture of index */ + int iLvl, /* Level to read input from */ + int *pnRem /* Write up to this many output leaves */ +){ + Fts5Structure *pStruct = *ppStruct; + Fts5StructureLevel *pLvl = &pStruct->aLevel[iLvl]; + Fts5StructureLevel *pLvlOut; + Fts5Iter *pIter = 0; /* Iterator to read input data */ + int nRem = pnRem ? *pnRem : 0; /* Output leaf pages left to write */ + int nInput; /* Number of input segments */ + Fts5SegWriter writer; /* Writer object */ + Fts5StructureSegment *pSeg; /* Output segment */ + Fts5Buffer term; + int bOldest; /* True if the output segment is the oldest */ + int eDetail = p->pConfig->eDetail; + const int flags = FTS5INDEX_QUERY_NOOUTPUT; + + assert( iLvl<pStruct->nLevel ); + assert( pLvl->nMerge<=pLvl->nSeg ); + + memset(&writer, 0, sizeof(Fts5SegWriter)); + memset(&term, 0, sizeof(Fts5Buffer)); + if( pLvl->nMerge ){ + pLvlOut = &pStruct->aLevel[iLvl+1]; + assert( pLvlOut->nSeg>0 ); + nInput = pLvl->nMerge; + pSeg = &pLvlOut->aSeg[pLvlOut->nSeg-1]; + + fts5WriteInit(p, &writer, pSeg->iSegid); + writer.writer.pgno = pSeg->pgnoLast+1; + writer.iBtPage = 0; + }else{ + int iSegid = fts5AllocateSegid(p, pStruct); + + /* Extend the Fts5Structure object as required to ensure the output + ** segment exists. */ + if( iLvl==pStruct->nLevel-1 ){ + fts5StructureAddLevel(&p->rc, ppStruct); + pStruct = *ppStruct; + } + fts5StructureExtendLevel(&p->rc, pStruct, iLvl+1, 1, 0); + if( p->rc ) return; + pLvl = &pStruct->aLevel[iLvl]; + pLvlOut = &pStruct->aLevel[iLvl+1]; + + fts5WriteInit(p, &writer, iSegid); + + /* Add the new segment to the output level */ + pSeg = &pLvlOut->aSeg[pLvlOut->nSeg]; + pLvlOut->nSeg++; + pSeg->pgnoFirst = 1; + pSeg->iSegid = iSegid; + pStruct->nSegment++; + + /* Read input from all segments in the input level */ + nInput = pLvl->nSeg; + } + bOldest = (pLvlOut->nSeg==1 && pStruct->nLevel==iLvl+2); + + assert( iLvl>=0 ); + for(fts5MultiIterNew(p, pStruct, flags, 0, 0, 0, iLvl, nInput, &pIter); + fts5MultiIterEof(p, pIter)==0; + fts5MultiIterNext(p, pIter, 0, 0) + ){ + Fts5SegIter *pSegIter = &pIter->aSeg[ pIter->aFirst[1].iFirst ]; + int nPos; /* position-list size field value */ + int nTerm; + const u8 *pTerm; + + /* Check for key annihilation. */ + if( pSegIter->nPos==0 && (bOldest || pSegIter->bDel==0) ) continue; + + pTerm = fts5MultiIterTerm(pIter, &nTerm); + if( nTerm!=term.n || memcmp(pTerm, term.p, nTerm) ){ + if( pnRem && writer.nLeafWritten>nRem ){ + break; + } + + /* This is a new term. Append a term to the output segment. */ + fts5WriteAppendTerm(p, &writer, nTerm, pTerm); + fts5BufferSet(&p->rc, &term, nTerm, pTerm); + } + + /* Append the rowid to the output */ + /* WRITEPOSLISTSIZE */ + fts5WriteAppendRowid(p, &writer, fts5MultiIterRowid(pIter)); + + if( eDetail==FTS5_DETAIL_NONE ){ + if( pSegIter->bDel ){ + fts5BufferAppendVarint(&p->rc, &writer.writer.buf, 0); + if( pSegIter->nPos>0 ){ + fts5BufferAppendVarint(&p->rc, &writer.writer.buf, 0); + } + } + }else{ + /* Append the position-list data to the output */ + nPos = pSegIter->nPos*2 + pSegIter->bDel; + fts5BufferAppendVarint(&p->rc, &writer.writer.buf, nPos); + fts5ChunkIterate(p, pSegIter, (void*)&writer, fts5MergeChunkCallback); + } + } + + /* Flush the last leaf page to disk. Set the output segment b-tree height + ** and last leaf page number at the same time. */ + fts5WriteFinish(p, &writer, &pSeg->pgnoLast); + + if( fts5MultiIterEof(p, pIter) ){ + int i; + + /* Remove the redundant segments from the %_data table */ + for(i=0; i<nInput; i++){ + fts5DataRemoveSegment(p, pLvl->aSeg[i].iSegid); + } + + /* Remove the redundant segments from the input level */ + if( pLvl->nSeg!=nInput ){ + int nMove = (pLvl->nSeg - nInput) * sizeof(Fts5StructureSegment); + memmove(pLvl->aSeg, &pLvl->aSeg[nInput], nMove); + } + pStruct->nSegment -= nInput; + pLvl->nSeg -= nInput; + pLvl->nMerge = 0; + if( pSeg->pgnoLast==0 ){ + pLvlOut->nSeg--; + pStruct->nSegment--; + } + }else{ + assert( pSeg->pgnoLast>0 ); + fts5TrimSegments(p, pIter); + pLvl->nMerge = nInput; + } + + fts5MultiIterFree(pIter); + fts5BufferFree(&term); + if( pnRem ) *pnRem -= writer.nLeafWritten; +} + +/* +** Do up to nPg pages of automerge work on the index. +*/ +static void fts5IndexMerge( + Fts5Index *p, /* FTS5 backend object */ + Fts5Structure **ppStruct, /* IN/OUT: Current structure of index */ + int nPg /* Pages of work to do */ +){ + int nRem = nPg; + Fts5Structure *pStruct = *ppStruct; + while( nRem>0 && p->rc==SQLITE_OK ){ + int iLvl; /* To iterate through levels */ + int iBestLvl = 0; /* Level offering the most input segments */ + int nBest = 0; /* Number of input segments on best level */ + + /* Set iBestLvl to the level to read input segments from. */ + assert( pStruct->nLevel>0 ); + for(iLvl=0; iLvl<pStruct->nLevel; iLvl++){ + Fts5StructureLevel *pLvl = &pStruct->aLevel[iLvl]; + if( pLvl->nMerge ){ + if( pLvl->nMerge>nBest ){ + iBestLvl = iLvl; + nBest = pLvl->nMerge; + } + break; + } + if( pLvl->nSeg>nBest ){ + nBest = pLvl->nSeg; + iBestLvl = iLvl; + } + } + + /* If nBest is still 0, then the index must be empty. */ +#ifdef SQLITE_DEBUG + for(iLvl=0; nBest==0 && iLvl<pStruct->nLevel; iLvl++){ + assert( pStruct->aLevel[iLvl].nSeg==0 ); + } +#endif + + if( nBest<p->pConfig->nAutomerge + && pStruct->aLevel[iBestLvl].nMerge==0 + ){ + break; + } + fts5IndexMergeLevel(p, &pStruct, iBestLvl, &nRem); + if( p->rc==SQLITE_OK && pStruct->aLevel[iBestLvl].nMerge==0 ){ + fts5StructurePromote(p, iBestLvl+1, pStruct); + } + } + *ppStruct = pStruct; +} + +/* +** A total of nLeaf leaf pages of data has just been flushed to a level-0 +** segment. This function updates the write-counter accordingly and, if +** necessary, performs incremental merge work. +** +** If an error occurs, set the Fts5Index.rc error code. If an error has +** already occurred, this function is a no-op. +*/ +static void fts5IndexAutomerge( + Fts5Index *p, /* FTS5 backend object */ + Fts5Structure **ppStruct, /* IN/OUT: Current structure of index */ + int nLeaf /* Number of output leaves just written */ +){ + if( p->rc==SQLITE_OK && p->pConfig->nAutomerge>0 ){ + Fts5Structure *pStruct = *ppStruct; + u64 nWrite; /* Initial value of write-counter */ + int nWork; /* Number of work-quanta to perform */ + int nRem; /* Number of leaf pages left to write */ + + /* Update the write-counter. While doing so, set nWork. */ + nWrite = pStruct->nWriteCounter; + nWork = (int)(((nWrite + nLeaf) / p->nWorkUnit) - (nWrite / p->nWorkUnit)); + pStruct->nWriteCounter += nLeaf; + nRem = (int)(p->nWorkUnit * nWork * pStruct->nLevel); + + fts5IndexMerge(p, ppStruct, nRem); + } +} + +static void fts5IndexCrisismerge( + Fts5Index *p, /* FTS5 backend object */ + Fts5Structure **ppStruct /* IN/OUT: Current structure of index */ +){ + const int nCrisis = p->pConfig->nCrisisMerge; + Fts5Structure *pStruct = *ppStruct; + int iLvl = 0; + + assert( p->rc!=SQLITE_OK || pStruct->nLevel>0 ); + while( p->rc==SQLITE_OK && pStruct->aLevel[iLvl].nSeg>=nCrisis ){ + fts5IndexMergeLevel(p, &pStruct, iLvl, 0); + assert( p->rc!=SQLITE_OK || pStruct->nLevel>(iLvl+1) ); + fts5StructurePromote(p, iLvl+1, pStruct); + iLvl++; + } + *ppStruct = pStruct; +} + +static int fts5IndexReturn(Fts5Index *p){ + int rc = p->rc; + p->rc = SQLITE_OK; + return rc; +} + +typedef struct Fts5FlushCtx Fts5FlushCtx; +struct Fts5FlushCtx { + Fts5Index *pIdx; + Fts5SegWriter writer; +}; + +/* +** Buffer aBuf[] contains a list of varints, all small enough to fit +** in a 32-bit integer. Return the size of the largest prefix of this +** list nMax bytes or less in size. +*/ +static int fts5PoslistPrefix(const u8 *aBuf, int nMax){ + int ret; + u32 dummy; + ret = fts5GetVarint32(aBuf, dummy); + if( ret<nMax ){ + while( 1 ){ + int i = fts5GetVarint32(&aBuf[ret], dummy); + if( (ret + i) > nMax ) break; + ret += i; + } + } + return ret; +} + +/* +** Flush the contents of in-memory hash table iHash to a new level-0 +** segment on disk. Also update the corresponding structure record. +** +** If an error occurs, set the Fts5Index.rc error code. If an error has +** already occurred, this function is a no-op. +*/ +static void fts5FlushOneHash(Fts5Index *p){ + Fts5Hash *pHash = p->pHash; + Fts5Structure *pStruct; + int iSegid; + int pgnoLast = 0; /* Last leaf page number in segment */ + + /* Obtain a reference to the index structure and allocate a new segment-id + ** for the new level-0 segment. */ + pStruct = fts5StructureRead(p); + iSegid = fts5AllocateSegid(p, pStruct); + + if( iSegid ){ + const int pgsz = p->pConfig->pgsz; + int eDetail = p->pConfig->eDetail; + Fts5StructureSegment *pSeg; /* New segment within pStruct */ + Fts5Buffer *pBuf; /* Buffer in which to assemble leaf page */ + Fts5Buffer *pPgidx; /* Buffer in which to assemble pgidx */ + + Fts5SegWriter writer; + fts5WriteInit(p, &writer, iSegid); + + pBuf = &writer.writer.buf; + pPgidx = &writer.writer.pgidx; + + /* fts5WriteInit() should have initialized the buffers to (most likely) + ** the maximum space required. */ + assert( p->rc || pBuf->nSpace>=(pgsz + FTS5_DATA_PADDING) ); + assert( p->rc || pPgidx->nSpace>=(pgsz + FTS5_DATA_PADDING) ); + + /* Begin scanning through hash table entries. This loop runs once for each + ** term/doclist currently stored within the hash table. */ + if( p->rc==SQLITE_OK ){ + p->rc = sqlite3Fts5HashScanInit(pHash, 0, 0); + } + while( p->rc==SQLITE_OK && 0==sqlite3Fts5HashScanEof(pHash) ){ + const char *zTerm; /* Buffer containing term */ + const u8 *pDoclist; /* Pointer to doclist for this term */ + int nDoclist; /* Size of doclist in bytes */ + + /* Write the term for this entry to disk. */ + sqlite3Fts5HashScanEntry(pHash, &zTerm, &pDoclist, &nDoclist); + fts5WriteAppendTerm(p, &writer, (int)strlen(zTerm), (const u8*)zTerm); + + assert( writer.bFirstRowidInPage==0 ); + if( pgsz>=(pBuf->n + pPgidx->n + nDoclist + 1) ){ + /* The entire doclist will fit on the current leaf. */ + fts5BufferSafeAppendBlob(pBuf, pDoclist, nDoclist); + }else{ + i64 iRowid = 0; + i64 iDelta = 0; + int iOff = 0; + + /* The entire doclist will not fit on this leaf. The following + ** loop iterates through the poslists that make up the current + ** doclist. */ + while( p->rc==SQLITE_OK && iOff<nDoclist ){ + iOff += fts5GetVarint(&pDoclist[iOff], (u64*)&iDelta); + iRowid += iDelta; + + if( writer.bFirstRowidInPage ){ + fts5PutU16(&pBuf->p[0], (u16)pBuf->n); /* first rowid on page */ + pBuf->n += sqlite3Fts5PutVarint(&pBuf->p[pBuf->n], iRowid); + writer.bFirstRowidInPage = 0; + fts5WriteDlidxAppend(p, &writer, iRowid); + }else{ + pBuf->n += sqlite3Fts5PutVarint(&pBuf->p[pBuf->n], iDelta); + } + assert( pBuf->n<=pBuf->nSpace ); + + if( eDetail==FTS5_DETAIL_NONE ){ + if( iOff<nDoclist && pDoclist[iOff]==0 ){ + pBuf->p[pBuf->n++] = 0; + iOff++; + if( iOff<nDoclist && pDoclist[iOff]==0 ){ + pBuf->p[pBuf->n++] = 0; + iOff++; + } + } + if( (pBuf->n + pPgidx->n)>=pgsz ){ + fts5WriteFlushLeaf(p, &writer); + } + }else{ + int bDummy; + int nPos; + int nCopy = fts5GetPoslistSize(&pDoclist[iOff], &nPos, &bDummy); + nCopy += nPos; + if( (pBuf->n + pPgidx->n + nCopy) <= pgsz ){ + /* The entire poslist will fit on the current leaf. So copy + ** it in one go. */ + fts5BufferSafeAppendBlob(pBuf, &pDoclist[iOff], nCopy); + }else{ + /* The entire poslist will not fit on this leaf. So it needs + ** to be broken into sections. The only qualification being + ** that each varint must be stored contiguously. */ + const u8 *pPoslist = &pDoclist[iOff]; + int iPos = 0; + while( p->rc==SQLITE_OK ){ + int nSpace = pgsz - pBuf->n - pPgidx->n; + int n = 0; + if( (nCopy - iPos)<=nSpace ){ + n = nCopy - iPos; + }else{ + n = fts5PoslistPrefix(&pPoslist[iPos], nSpace); + } + assert( n>0 ); + fts5BufferSafeAppendBlob(pBuf, &pPoslist[iPos], n); + iPos += n; + if( (pBuf->n + pPgidx->n)>=pgsz ){ + fts5WriteFlushLeaf(p, &writer); + } + if( iPos>=nCopy ) break; + } + } + iOff += nCopy; + } + } + } + + /* TODO2: Doclist terminator written here. */ + /* pBuf->p[pBuf->n++] = '\0'; */ + assert( pBuf->n<=pBuf->nSpace ); + sqlite3Fts5HashScanNext(pHash); + } + sqlite3Fts5HashClear(pHash); + fts5WriteFinish(p, &writer, &pgnoLast); + + /* Update the Fts5Structure. It is written back to the database by the + ** fts5StructureRelease() call below. */ + if( pStruct->nLevel==0 ){ + fts5StructureAddLevel(&p->rc, &pStruct); + } + fts5StructureExtendLevel(&p->rc, pStruct, 0, 1, 0); + if( p->rc==SQLITE_OK ){ + pSeg = &pStruct->aLevel[0].aSeg[ pStruct->aLevel[0].nSeg++ ]; + pSeg->iSegid = iSegid; + pSeg->pgnoFirst = 1; + pSeg->pgnoLast = pgnoLast; + pStruct->nSegment++; + } + fts5StructurePromote(p, 0, pStruct); + } + + fts5IndexAutomerge(p, &pStruct, pgnoLast); + fts5IndexCrisismerge(p, &pStruct); + fts5StructureWrite(p, pStruct); + fts5StructureRelease(pStruct); +} + +/* +** Flush any data stored in the in-memory hash tables to the database. +*/ +static void fts5IndexFlush(Fts5Index *p){ + /* Unless it is empty, flush the hash table to disk */ + if( p->nPendingData ){ + assert( p->pHash ); + p->nPendingData = 0; + fts5FlushOneHash(p); + } +} + + +static int sqlite3Fts5IndexOptimize(Fts5Index *p){ + Fts5Structure *pStruct; + Fts5Structure *pNew = 0; + int nSeg = 0; + + assert( p->rc==SQLITE_OK ); + fts5IndexFlush(p); + pStruct = fts5StructureRead(p); + + if( pStruct ){ + assert( pStruct->nSegment==fts5StructureCountSegments(pStruct) ); + nSeg = pStruct->nSegment; + if( nSeg>1 ){ + int nByte = sizeof(Fts5Structure); + nByte += (pStruct->nLevel+1) * sizeof(Fts5StructureLevel); + pNew = (Fts5Structure*)sqlite3Fts5MallocZero(&p->rc, nByte); + } + } + if( pNew ){ + Fts5StructureLevel *pLvl; + int nByte = nSeg * sizeof(Fts5StructureSegment); + pNew->nLevel = pStruct->nLevel+1; + pNew->nRef = 1; + pNew->nWriteCounter = pStruct->nWriteCounter; + pLvl = &pNew->aLevel[pStruct->nLevel]; + pLvl->aSeg = (Fts5StructureSegment*)sqlite3Fts5MallocZero(&p->rc, nByte); + if( pLvl->aSeg ){ + int iLvl, iSeg; + int iSegOut = 0; + for(iLvl=0; iLvl<pStruct->nLevel; iLvl++){ + for(iSeg=0; iSeg<pStruct->aLevel[iLvl].nSeg; iSeg++){ + pLvl->aSeg[iSegOut] = pStruct->aLevel[iLvl].aSeg[iSeg]; + iSegOut++; + } + } + pNew->nSegment = pLvl->nSeg = nSeg; + }else{ + sqlite3_free(pNew); + pNew = 0; + } + } + + if( pNew ){ + int iLvl = pNew->nLevel-1; + while( p->rc==SQLITE_OK && pNew->aLevel[iLvl].nSeg>0 ){ + int nRem = FTS5_OPT_WORK_UNIT; + fts5IndexMergeLevel(p, &pNew, iLvl, &nRem); + } + + fts5StructureWrite(p, pNew); + fts5StructureRelease(pNew); + } + + fts5StructureRelease(pStruct); + return fts5IndexReturn(p); +} + +static int sqlite3Fts5IndexMerge(Fts5Index *p, int nMerge){ + Fts5Structure *pStruct; + + pStruct = fts5StructureRead(p); + if( pStruct && pStruct->nLevel ){ + fts5IndexMerge(p, &pStruct, nMerge); + fts5StructureWrite(p, pStruct); + } + fts5StructureRelease(pStruct); + + return fts5IndexReturn(p); +} + +static void fts5AppendRowid( + Fts5Index *p, + i64 iDelta, + Fts5Iter *pUnused, + Fts5Buffer *pBuf +){ + UNUSED_PARAM(pUnused); + fts5BufferAppendVarint(&p->rc, pBuf, iDelta); +} + +static void fts5AppendPoslist( + Fts5Index *p, + i64 iDelta, + Fts5Iter *pMulti, + Fts5Buffer *pBuf +){ + int nData = pMulti->base.nData; + assert( nData>0 ); + if( p->rc==SQLITE_OK && 0==fts5BufferGrow(&p->rc, pBuf, nData+9+9) ){ + fts5BufferSafeAppendVarint(pBuf, iDelta); + fts5BufferSafeAppendVarint(pBuf, nData*2); + fts5BufferSafeAppendBlob(pBuf, pMulti->base.pData, nData); + } +} + + +static void fts5DoclistIterNext(Fts5DoclistIter *pIter){ + u8 *p = pIter->aPoslist + pIter->nSize + pIter->nPoslist; + + assert( pIter->aPoslist ); + if( p>=pIter->aEof ){ + pIter->aPoslist = 0; + }else{ + i64 iDelta; + + p += fts5GetVarint(p, (u64*)&iDelta); + pIter->iRowid += iDelta; + + /* Read position list size */ + if( p[0] & 0x80 ){ + int nPos; + pIter->nSize = fts5GetVarint32(p, nPos); + pIter->nPoslist = (nPos>>1); + }else{ + pIter->nPoslist = ((int)(p[0])) >> 1; + pIter->nSize = 1; + } + + pIter->aPoslist = p; + } +} + +static void fts5DoclistIterInit( + Fts5Buffer *pBuf, + Fts5DoclistIter *pIter +){ + memset(pIter, 0, sizeof(*pIter)); + pIter->aPoslist = pBuf->p; + pIter->aEof = &pBuf->p[pBuf->n]; + fts5DoclistIterNext(pIter); +} + +#if 0 +/* +** Append a doclist to buffer pBuf. +** +** This function assumes that space within the buffer has already been +** allocated. +*/ +static void fts5MergeAppendDocid( + Fts5Buffer *pBuf, /* Buffer to write to */ + i64 *piLastRowid, /* IN/OUT: Previous rowid written (if any) */ + i64 iRowid /* Rowid to append */ +){ + assert( pBuf->n!=0 || (*piLastRowid)==0 ); + fts5BufferSafeAppendVarint(pBuf, iRowid - *piLastRowid); + *piLastRowid = iRowid; +} +#endif + +#define fts5MergeAppendDocid(pBuf, iLastRowid, iRowid) { \ + assert( (pBuf)->n!=0 || (iLastRowid)==0 ); \ + fts5BufferSafeAppendVarint((pBuf), (iRowid) - (iLastRowid)); \ + (iLastRowid) = (iRowid); \ +} + +/* +** Swap the contents of buffer *p1 with that of *p2. +*/ +static void fts5BufferSwap(Fts5Buffer *p1, Fts5Buffer *p2){ + Fts5Buffer tmp = *p1; + *p1 = *p2; + *p2 = tmp; +} + +static void fts5NextRowid(Fts5Buffer *pBuf, int *piOff, i64 *piRowid){ + int i = *piOff; + if( i>=pBuf->n ){ + *piOff = -1; + }else{ + u64 iVal; + *piOff = i + sqlite3Fts5GetVarint(&pBuf->p[i], &iVal); + *piRowid += iVal; + } +} + +/* +** This is the equivalent of fts5MergePrefixLists() for detail=none mode. +** In this case the buffers consist of a delta-encoded list of rowids only. +*/ +static void fts5MergeRowidLists( + Fts5Index *p, /* FTS5 backend object */ + Fts5Buffer *p1, /* First list to merge */ + Fts5Buffer *p2 /* Second list to merge */ +){ + int i1 = 0; + int i2 = 0; + i64 iRowid1 = 0; + i64 iRowid2 = 0; + i64 iOut = 0; + + Fts5Buffer out; + memset(&out, 0, sizeof(out)); + sqlite3Fts5BufferSize(&p->rc, &out, p1->n + p2->n); + if( p->rc ) return; + + fts5NextRowid(p1, &i1, &iRowid1); + fts5NextRowid(p2, &i2, &iRowid2); + while( i1>=0 || i2>=0 ){ + if( i1>=0 && (i2<0 || iRowid1<iRowid2) ){ + assert( iOut==0 || iRowid1>iOut ); + fts5BufferSafeAppendVarint(&out, iRowid1 - iOut); + iOut = iRowid1; + fts5NextRowid(p1, &i1, &iRowid1); + }else{ + assert( iOut==0 || iRowid2>iOut ); + fts5BufferSafeAppendVarint(&out, iRowid2 - iOut); + iOut = iRowid2; + if( i1>=0 && iRowid1==iRowid2 ){ + fts5NextRowid(p1, &i1, &iRowid1); + } + fts5NextRowid(p2, &i2, &iRowid2); + } + } + + fts5BufferSwap(&out, p1); + fts5BufferFree(&out); +} + +/* +** Buffers p1 and p2 contain doclists. This function merges the content +** of the two doclists together and sets buffer p1 to the result before +** returning. +** +** If an error occurs, an error code is left in p->rc. If an error has +** already occurred, this function is a no-op. +*/ +static void fts5MergePrefixLists( + Fts5Index *p, /* FTS5 backend object */ + Fts5Buffer *p1, /* First list to merge */ + Fts5Buffer *p2 /* Second list to merge */ +){ + if( p2->n ){ + i64 iLastRowid = 0; + Fts5DoclistIter i1; + Fts5DoclistIter i2; + Fts5Buffer out = {0, 0, 0}; + Fts5Buffer tmp = {0, 0, 0}; + + if( sqlite3Fts5BufferSize(&p->rc, &out, p1->n + p2->n) ) return; + fts5DoclistIterInit(p1, &i1); + fts5DoclistIterInit(p2, &i2); + + while( 1 ){ + if( i1.iRowid<i2.iRowid ){ + /* Copy entry from i1 */ + fts5MergeAppendDocid(&out, iLastRowid, i1.iRowid); + fts5BufferSafeAppendBlob(&out, i1.aPoslist, i1.nPoslist+i1.nSize); + fts5DoclistIterNext(&i1); + if( i1.aPoslist==0 ) break; + } + else if( i2.iRowid!=i1.iRowid ){ + /* Copy entry from i2 */ + fts5MergeAppendDocid(&out, iLastRowid, i2.iRowid); + fts5BufferSafeAppendBlob(&out, i2.aPoslist, i2.nPoslist+i2.nSize); + fts5DoclistIterNext(&i2); + if( i2.aPoslist==0 ) break; + } + else{ + /* Merge the two position lists. */ + i64 iPos1 = 0; + i64 iPos2 = 0; + int iOff1 = 0; + int iOff2 = 0; + u8 *a1 = &i1.aPoslist[i1.nSize]; + u8 *a2 = &i2.aPoslist[i2.nSize]; + + i64 iPrev = 0; + Fts5PoslistWriter writer; + memset(&writer, 0, sizeof(writer)); + + fts5MergeAppendDocid(&out, iLastRowid, i2.iRowid); + fts5BufferZero(&tmp); + sqlite3Fts5BufferSize(&p->rc, &tmp, i1.nPoslist + i2.nPoslist); + if( p->rc ) break; + + sqlite3Fts5PoslistNext64(a1, i1.nPoslist, &iOff1, &iPos1); + sqlite3Fts5PoslistNext64(a2, i2.nPoslist, &iOff2, &iPos2); + assert( iPos1>=0 && iPos2>=0 ); + + if( iPos1<iPos2 ){ + sqlite3Fts5PoslistSafeAppend(&tmp, &iPrev, iPos1); + sqlite3Fts5PoslistNext64(a1, i1.nPoslist, &iOff1, &iPos1); + }else{ + sqlite3Fts5PoslistSafeAppend(&tmp, &iPrev, iPos2); + sqlite3Fts5PoslistNext64(a2, i2.nPoslist, &iOff2, &iPos2); + } + + if( iPos1>=0 && iPos2>=0 ){ + while( 1 ){ + if( iPos1<iPos2 ){ + if( iPos1!=iPrev ){ + sqlite3Fts5PoslistSafeAppend(&tmp, &iPrev, iPos1); + } + sqlite3Fts5PoslistNext64(a1, i1.nPoslist, &iOff1, &iPos1); + if( iPos1<0 ) break; + }else{ + assert( iPos2!=iPrev ); + sqlite3Fts5PoslistSafeAppend(&tmp, &iPrev, iPos2); + sqlite3Fts5PoslistNext64(a2, i2.nPoslist, &iOff2, &iPos2); + if( iPos2<0 ) break; + } + } + } + + if( iPos1>=0 ){ + if( iPos1!=iPrev ){ + sqlite3Fts5PoslistSafeAppend(&tmp, &iPrev, iPos1); + } + fts5BufferSafeAppendBlob(&tmp, &a1[iOff1], i1.nPoslist-iOff1); + }else{ + assert( iPos2>=0 && iPos2!=iPrev ); + sqlite3Fts5PoslistSafeAppend(&tmp, &iPrev, iPos2); + fts5BufferSafeAppendBlob(&tmp, &a2[iOff2], i2.nPoslist-iOff2); + } + + /* WRITEPOSLISTSIZE */ + fts5BufferSafeAppendVarint(&out, tmp.n * 2); + fts5BufferSafeAppendBlob(&out, tmp.p, tmp.n); + fts5DoclistIterNext(&i1); + fts5DoclistIterNext(&i2); + if( i1.aPoslist==0 || i2.aPoslist==0 ) break; + } + } + + if( i1.aPoslist ){ + fts5MergeAppendDocid(&out, iLastRowid, i1.iRowid); + fts5BufferSafeAppendBlob(&out, i1.aPoslist, i1.aEof - i1.aPoslist); + } + else if( i2.aPoslist ){ + fts5MergeAppendDocid(&out, iLastRowid, i2.iRowid); + fts5BufferSafeAppendBlob(&out, i2.aPoslist, i2.aEof - i2.aPoslist); + } + + fts5BufferSet(&p->rc, p1, out.n, out.p); + fts5BufferFree(&tmp); + fts5BufferFree(&out); + } +} + +static void fts5SetupPrefixIter( + Fts5Index *p, /* Index to read from */ + int bDesc, /* True for "ORDER BY rowid DESC" */ + const u8 *pToken, /* Buffer containing prefix to match */ + int nToken, /* Size of buffer pToken in bytes */ + Fts5Colset *pColset, /* Restrict matches to these columns */ + Fts5Iter **ppIter /* OUT: New iterator */ +){ + Fts5Structure *pStruct; + Fts5Buffer *aBuf; + const int nBuf = 32; + + void (*xMerge)(Fts5Index*, Fts5Buffer*, Fts5Buffer*); + void (*xAppend)(Fts5Index*, i64, Fts5Iter*, Fts5Buffer*); + if( p->pConfig->eDetail==FTS5_DETAIL_NONE ){ + xMerge = fts5MergeRowidLists; + xAppend = fts5AppendRowid; + }else{ + xMerge = fts5MergePrefixLists; + xAppend = fts5AppendPoslist; + } + + aBuf = (Fts5Buffer*)fts5IdxMalloc(p, sizeof(Fts5Buffer)*nBuf); + pStruct = fts5StructureRead(p); + + if( aBuf && pStruct ){ + const int flags = FTS5INDEX_QUERY_SCAN + | FTS5INDEX_QUERY_SKIPEMPTY + | FTS5INDEX_QUERY_NOOUTPUT; + int i; + i64 iLastRowid = 0; + Fts5Iter *p1 = 0; /* Iterator used to gather data from index */ + Fts5Data *pData; + Fts5Buffer doclist; + int bNewTerm = 1; + + memset(&doclist, 0, sizeof(doclist)); + fts5MultiIterNew(p, pStruct, flags, pColset, pToken, nToken, -1, 0, &p1); + fts5IterSetOutputCb(&p->rc, p1); + for( /* no-op */ ; + fts5MultiIterEof(p, p1)==0; + fts5MultiIterNext2(p, p1, &bNewTerm) + ){ + Fts5SegIter *pSeg = &p1->aSeg[ p1->aFirst[1].iFirst ]; + int nTerm = pSeg->term.n; + const u8 *pTerm = pSeg->term.p; + p1->xSetOutputs(p1, pSeg); + + assert_nc( memcmp(pToken, pTerm, MIN(nToken, nTerm))<=0 ); + if( bNewTerm ){ + if( nTerm<nToken || memcmp(pToken, pTerm, nToken) ) break; + } + + if( p1->base.nData==0 ) continue; + + if( p1->base.iRowid<=iLastRowid && doclist.n>0 ){ + for(i=0; p->rc==SQLITE_OK && doclist.n; i++){ + assert( i<nBuf ); + if( aBuf[i].n==0 ){ + fts5BufferSwap(&doclist, &aBuf[i]); + fts5BufferZero(&doclist); + }else{ + xMerge(p, &doclist, &aBuf[i]); + fts5BufferZero(&aBuf[i]); + } + } + iLastRowid = 0; + } + + xAppend(p, p1->base.iRowid-iLastRowid, p1, &doclist); + iLastRowid = p1->base.iRowid; + } + + for(i=0; i<nBuf; i++){ + if( p->rc==SQLITE_OK ){ + xMerge(p, &doclist, &aBuf[i]); + } + fts5BufferFree(&aBuf[i]); + } + fts5MultiIterFree(p1); + + pData = fts5IdxMalloc(p, sizeof(Fts5Data) + doclist.n); + if( pData ){ + pData->p = (u8*)&pData[1]; + pData->nn = pData->szLeaf = doclist.n; + memcpy(pData->p, doclist.p, doclist.n); + fts5MultiIterNew2(p, pData, bDesc, ppIter); + } + fts5BufferFree(&doclist); + } + + fts5StructureRelease(pStruct); + sqlite3_free(aBuf); +} + + +/* +** Indicate that all subsequent calls to sqlite3Fts5IndexWrite() pertain +** to the document with rowid iRowid. +*/ +static int sqlite3Fts5IndexBeginWrite(Fts5Index *p, int bDelete, i64 iRowid){ + assert( p->rc==SQLITE_OK ); + + /* Allocate the hash table if it has not already been allocated */ + if( p->pHash==0 ){ + p->rc = sqlite3Fts5HashNew(p->pConfig, &p->pHash, &p->nPendingData); + } + + /* Flush the hash table to disk if required */ + if( iRowid<p->iWriteRowid + || (iRowid==p->iWriteRowid && p->bDelete==0) + || (p->nPendingData > p->pConfig->nHashSize) + ){ + fts5IndexFlush(p); + } + + p->iWriteRowid = iRowid; + p->bDelete = bDelete; + return fts5IndexReturn(p); +} + +/* +** Commit data to disk. +*/ +static int sqlite3Fts5IndexSync(Fts5Index *p, int bCommit){ + assert( p->rc==SQLITE_OK ); + fts5IndexFlush(p); + if( bCommit ) fts5CloseReader(p); + return fts5IndexReturn(p); +} + +/* +** Discard any data stored in the in-memory hash tables. Do not write it +** to the database. Additionally, assume that the contents of the %_data +** table may have changed on disk. So any in-memory caches of %_data +** records must be invalidated. +*/ +static int sqlite3Fts5IndexRollback(Fts5Index *p){ + fts5CloseReader(p); + fts5IndexDiscardData(p); + /* assert( p->rc==SQLITE_OK ); */ + return SQLITE_OK; +} + +/* +** The %_data table is completely empty when this function is called. This +** function populates it with the initial structure objects for each index, +** and the initial version of the "averages" record (a zero-byte blob). +*/ +static int sqlite3Fts5IndexReinit(Fts5Index *p){ + Fts5Structure s; + memset(&s, 0, sizeof(Fts5Structure)); + fts5DataWrite(p, FTS5_AVERAGES_ROWID, (const u8*)"", 0); + fts5StructureWrite(p, &s); + return fts5IndexReturn(p); +} + +/* +** Open a new Fts5Index handle. If the bCreate argument is true, create +** and initialize the underlying %_data table. +** +** If successful, set *pp to point to the new object and return SQLITE_OK. +** Otherwise, set *pp to NULL and return an SQLite error code. +*/ +static int sqlite3Fts5IndexOpen( + Fts5Config *pConfig, + int bCreate, + Fts5Index **pp, + char **pzErr +){ + int rc = SQLITE_OK; + Fts5Index *p; /* New object */ + + *pp = p = (Fts5Index*)sqlite3Fts5MallocZero(&rc, sizeof(Fts5Index)); + if( rc==SQLITE_OK ){ + p->pConfig = pConfig; + p->nWorkUnit = FTS5_WORK_UNIT; + p->zDataTbl = sqlite3Fts5Mprintf(&rc, "%s_data", pConfig->zName); + if( p->zDataTbl && bCreate ){ + rc = sqlite3Fts5CreateTable( + pConfig, "data", "id INTEGER PRIMARY KEY, block BLOB", 0, pzErr + ); + if( rc==SQLITE_OK ){ + rc = sqlite3Fts5CreateTable(pConfig, "idx", + "segid, term, pgno, PRIMARY KEY(segid, term)", + 1, pzErr + ); + } + if( rc==SQLITE_OK ){ + rc = sqlite3Fts5IndexReinit(p); + } + } + } + + assert( rc!=SQLITE_OK || p->rc==SQLITE_OK ); + if( rc ){ + sqlite3Fts5IndexClose(p); + *pp = 0; + } + return rc; +} + +/* +** Close a handle opened by an earlier call to sqlite3Fts5IndexOpen(). +*/ +static int sqlite3Fts5IndexClose(Fts5Index *p){ + int rc = SQLITE_OK; + if( p ){ + assert( p->pReader==0 ); + sqlite3_finalize(p->pWriter); + sqlite3_finalize(p->pDeleter); + sqlite3_finalize(p->pIdxWriter); + sqlite3_finalize(p->pIdxDeleter); + sqlite3_finalize(p->pIdxSelect); + sqlite3Fts5HashFree(p->pHash); + sqlite3_free(p->zDataTbl); + sqlite3_free(p); + } + return rc; +} + +/* +** Argument p points to a buffer containing utf-8 text that is n bytes in +** size. Return the number of bytes in the nChar character prefix of the +** buffer, or 0 if there are less than nChar characters in total. +*/ +static int sqlite3Fts5IndexCharlenToBytelen( + const char *p, + int nByte, + int nChar +){ + int n = 0; + int i; + for(i=0; i<nChar; i++){ + if( n>=nByte ) return 0; /* Input contains fewer than nChar chars */ + if( (unsigned char)p[n++]>=0xc0 ){ + while( (p[n] & 0xc0)==0x80 ) n++; + } + } + return n; +} + +/* +** pIn is a UTF-8 encoded string, nIn bytes in size. Return the number of +** unicode characters in the string. +*/ +static int fts5IndexCharlen(const char *pIn, int nIn){ + int nChar = 0; + int i = 0; + while( i<nIn ){ + if( (unsigned char)pIn[i++]>=0xc0 ){ + while( i<nIn && (pIn[i] & 0xc0)==0x80 ) i++; + } + nChar++; + } + return nChar; +} + +/* +** Insert or remove data to or from the index. Each time a document is +** added to or removed from the index, this function is called one or more +** times. +** +** For an insert, it must be called once for each token in the new document. +** If the operation is a delete, it must be called (at least) once for each +** unique token in the document with an iCol value less than zero. The iPos +** argument is ignored for a delete. +*/ +static int sqlite3Fts5IndexWrite( + Fts5Index *p, /* Index to write to */ + int iCol, /* Column token appears in (-ve -> delete) */ + int iPos, /* Position of token within column */ + const char *pToken, int nToken /* Token to add or remove to or from index */ +){ + int i; /* Used to iterate through indexes */ + int rc = SQLITE_OK; /* Return code */ + Fts5Config *pConfig = p->pConfig; + + assert( p->rc==SQLITE_OK ); + assert( (iCol<0)==p->bDelete ); + + /* Add the entry to the main terms index. */ + rc = sqlite3Fts5HashWrite( + p->pHash, p->iWriteRowid, iCol, iPos, FTS5_MAIN_PREFIX, pToken, nToken + ); + + for(i=0; i<pConfig->nPrefix && rc==SQLITE_OK; i++){ + const int nChar = pConfig->aPrefix[i]; + int nByte = sqlite3Fts5IndexCharlenToBytelen(pToken, nToken, nChar); + if( nByte ){ + rc = sqlite3Fts5HashWrite(p->pHash, + p->iWriteRowid, iCol, iPos, (char)(FTS5_MAIN_PREFIX+i+1), pToken, + nByte + ); + } + } + + return rc; +} + +/* +** Open a new iterator to iterate though all rowid that match the +** specified token or token prefix. +*/ +static int sqlite3Fts5IndexQuery( + Fts5Index *p, /* FTS index to query */ + const char *pToken, int nToken, /* Token (or prefix) to query for */ + int flags, /* Mask of FTS5INDEX_QUERY_X flags */ + Fts5Colset *pColset, /* Match these columns only */ + Fts5IndexIter **ppIter /* OUT: New iterator object */ +){ + Fts5Config *pConfig = p->pConfig; + Fts5Iter *pRet = 0; + Fts5Buffer buf = {0, 0, 0}; + + /* If the QUERY_SCAN flag is set, all other flags must be clear. */ + assert( (flags & FTS5INDEX_QUERY_SCAN)==0 || flags==FTS5INDEX_QUERY_SCAN ); + + if( sqlite3Fts5BufferSize(&p->rc, &buf, nToken+1)==0 ){ + int iIdx = 0; /* Index to search */ + memcpy(&buf.p[1], pToken, nToken); + + /* Figure out which index to search and set iIdx accordingly. If this + ** is a prefix query for which there is no prefix index, set iIdx to + ** greater than pConfig->nPrefix to indicate that the query will be + ** satisfied by scanning multiple terms in the main index. + ** + ** If the QUERY_TEST_NOIDX flag was specified, then this must be a + ** prefix-query. Instead of using a prefix-index (if one exists), + ** evaluate the prefix query using the main FTS index. This is used + ** for internal sanity checking by the integrity-check in debug + ** mode only. */ +#ifdef SQLITE_DEBUG + if( pConfig->bPrefixIndex==0 || (flags & FTS5INDEX_QUERY_TEST_NOIDX) ){ + assert( flags & FTS5INDEX_QUERY_PREFIX ); + iIdx = 1+pConfig->nPrefix; + }else +#endif + if( flags & FTS5INDEX_QUERY_PREFIX ){ + int nChar = fts5IndexCharlen(pToken, nToken); + for(iIdx=1; iIdx<=pConfig->nPrefix; iIdx++){ + if( pConfig->aPrefix[iIdx-1]==nChar ) break; + } + } + + if( iIdx<=pConfig->nPrefix ){ + /* Straight index lookup */ + Fts5Structure *pStruct = fts5StructureRead(p); + buf.p[0] = (u8)(FTS5_MAIN_PREFIX + iIdx); + if( pStruct ){ + fts5MultiIterNew(p, pStruct, flags | FTS5INDEX_QUERY_SKIPEMPTY, + pColset, buf.p, nToken+1, -1, 0, &pRet + ); + fts5StructureRelease(pStruct); + } + }else{ + /* Scan multiple terms in the main index */ + int bDesc = (flags & FTS5INDEX_QUERY_DESC)!=0; + buf.p[0] = FTS5_MAIN_PREFIX; + fts5SetupPrefixIter(p, bDesc, buf.p, nToken+1, pColset, &pRet); + assert( p->rc!=SQLITE_OK || pRet->pColset==0 ); + fts5IterSetOutputCb(&p->rc, pRet); + if( p->rc==SQLITE_OK ){ + Fts5SegIter *pSeg = &pRet->aSeg[pRet->aFirst[1].iFirst]; + if( pSeg->pLeaf ) pRet->xSetOutputs(pRet, pSeg); + } + } + + if( p->rc ){ + sqlite3Fts5IterClose(&pRet->base); + pRet = 0; + fts5CloseReader(p); + } + + *ppIter = &pRet->base; + sqlite3Fts5BufferFree(&buf); + } + return fts5IndexReturn(p); +} + +/* +** Return true if the iterator passed as the only argument is at EOF. +*/ +/* +** Move to the next matching rowid. +*/ +static int sqlite3Fts5IterNext(Fts5IndexIter *pIndexIter){ + Fts5Iter *pIter = (Fts5Iter*)pIndexIter; + assert( pIter->pIndex->rc==SQLITE_OK ); + fts5MultiIterNext(pIter->pIndex, pIter, 0, 0); + return fts5IndexReturn(pIter->pIndex); +} + +/* +** Move to the next matching term/rowid. Used by the fts5vocab module. +*/ +static int sqlite3Fts5IterNextScan(Fts5IndexIter *pIndexIter){ + Fts5Iter *pIter = (Fts5Iter*)pIndexIter; + Fts5Index *p = pIter->pIndex; + + assert( pIter->pIndex->rc==SQLITE_OK ); + + fts5MultiIterNext(p, pIter, 0, 0); + if( p->rc==SQLITE_OK ){ + Fts5SegIter *pSeg = &pIter->aSeg[ pIter->aFirst[1].iFirst ]; + if( pSeg->pLeaf && pSeg->term.p[0]!=FTS5_MAIN_PREFIX ){ + fts5DataRelease(pSeg->pLeaf); + pSeg->pLeaf = 0; + pIter->base.bEof = 1; + } + } + + return fts5IndexReturn(pIter->pIndex); +} + +/* +** Move to the next matching rowid that occurs at or after iMatch. The +** definition of "at or after" depends on whether this iterator iterates +** in ascending or descending rowid order. +*/ +static int sqlite3Fts5IterNextFrom(Fts5IndexIter *pIndexIter, i64 iMatch){ + Fts5Iter *pIter = (Fts5Iter*)pIndexIter; + fts5MultiIterNextFrom(pIter->pIndex, pIter, iMatch); + return fts5IndexReturn(pIter->pIndex); +} + +/* +** Return the current term. +*/ +static const char *sqlite3Fts5IterTerm(Fts5IndexIter *pIndexIter, int *pn){ + int n; + const char *z = (const char*)fts5MultiIterTerm((Fts5Iter*)pIndexIter, &n); + *pn = n-1; + return &z[1]; +} + +/* +** Close an iterator opened by an earlier call to sqlite3Fts5IndexQuery(). +*/ +static void sqlite3Fts5IterClose(Fts5IndexIter *pIndexIter){ + if( pIndexIter ){ + Fts5Iter *pIter = (Fts5Iter*)pIndexIter; + Fts5Index *pIndex = pIter->pIndex; + fts5MultiIterFree(pIter); + fts5CloseReader(pIndex); + } +} + +/* +** Read and decode the "averages" record from the database. +** +** Parameter anSize must point to an array of size nCol, where nCol is +** the number of user defined columns in the FTS table. +*/ +static int sqlite3Fts5IndexGetAverages(Fts5Index *p, i64 *pnRow, i64 *anSize){ + int nCol = p->pConfig->nCol; + Fts5Data *pData; + + *pnRow = 0; + memset(anSize, 0, sizeof(i64) * nCol); + pData = fts5DataRead(p, FTS5_AVERAGES_ROWID); + if( p->rc==SQLITE_OK && pData->nn ){ + int i = 0; + int iCol; + i += fts5GetVarint(&pData->p[i], (u64*)pnRow); + for(iCol=0; i<pData->nn && iCol<nCol; iCol++){ + i += fts5GetVarint(&pData->p[i], (u64*)&anSize[iCol]); + } + } + + fts5DataRelease(pData); + return fts5IndexReturn(p); +} + +/* +** Replace the current "averages" record with the contents of the buffer +** supplied as the second argument. +*/ +static int sqlite3Fts5IndexSetAverages(Fts5Index *p, const u8 *pData, int nData){ + assert( p->rc==SQLITE_OK ); + fts5DataWrite(p, FTS5_AVERAGES_ROWID, pData, nData); + return fts5IndexReturn(p); +} + +/* +** Return the total number of blocks this module has read from the %_data +** table since it was created. +*/ +static int sqlite3Fts5IndexReads(Fts5Index *p){ + return p->nRead; +} + +/* +** Set the 32-bit cookie value stored at the start of all structure +** records to the value passed as the second argument. +** +** Return SQLITE_OK if successful, or an SQLite error code if an error +** occurs. +*/ +static int sqlite3Fts5IndexSetCookie(Fts5Index *p, int iNew){ + int rc; /* Return code */ + Fts5Config *pConfig = p->pConfig; /* Configuration object */ + u8 aCookie[4]; /* Binary representation of iNew */ + sqlite3_blob *pBlob = 0; + + assert( p->rc==SQLITE_OK ); + sqlite3Fts5Put32(aCookie, iNew); + + rc = sqlite3_blob_open(pConfig->db, pConfig->zDb, p->zDataTbl, + "block", FTS5_STRUCTURE_ROWID, 1, &pBlob + ); + if( rc==SQLITE_OK ){ + sqlite3_blob_write(pBlob, aCookie, 4, 0); + rc = sqlite3_blob_close(pBlob); + } + + return rc; +} + +static int sqlite3Fts5IndexLoadConfig(Fts5Index *p){ + Fts5Structure *pStruct; + pStruct = fts5StructureRead(p); + fts5StructureRelease(pStruct); + return fts5IndexReturn(p); +} + + +/************************************************************************* +************************************************************************** +** Below this point is the implementation of the integrity-check +** functionality. +*/ + +/* +** Return a simple checksum value based on the arguments. +*/ +static u64 sqlite3Fts5IndexEntryCksum( + i64 iRowid, + int iCol, + int iPos, + int iIdx, + const char *pTerm, + int nTerm +){ + int i; + u64 ret = iRowid; + ret += (ret<<3) + iCol; + ret += (ret<<3) + iPos; + if( iIdx>=0 ) ret += (ret<<3) + (FTS5_MAIN_PREFIX + iIdx); + for(i=0; i<nTerm; i++) ret += (ret<<3) + pTerm[i]; + return ret; +} + +#ifdef SQLITE_DEBUG +/* +** This function is purely an internal test. It does not contribute to +** FTS functionality, or even the integrity-check, in any way. +** +** Instead, it tests that the same set of pgno/rowid combinations are +** visited regardless of whether the doclist-index identified by parameters +** iSegid/iLeaf is iterated in forwards or reverse order. +*/ +static void fts5TestDlidxReverse( + Fts5Index *p, + int iSegid, /* Segment id to load from */ + int iLeaf /* Load doclist-index for this leaf */ +){ + Fts5DlidxIter *pDlidx = 0; + u64 cksum1 = 13; + u64 cksum2 = 13; + + for(pDlidx=fts5DlidxIterInit(p, 0, iSegid, iLeaf); + fts5DlidxIterEof(p, pDlidx)==0; + fts5DlidxIterNext(p, pDlidx) + ){ + i64 iRowid = fts5DlidxIterRowid(pDlidx); + int pgno = fts5DlidxIterPgno(pDlidx); + assert( pgno>iLeaf ); + cksum1 += iRowid + ((i64)pgno<<32); + } + fts5DlidxIterFree(pDlidx); + pDlidx = 0; + + for(pDlidx=fts5DlidxIterInit(p, 1, iSegid, iLeaf); + fts5DlidxIterEof(p, pDlidx)==0; + fts5DlidxIterPrev(p, pDlidx) + ){ + i64 iRowid = fts5DlidxIterRowid(pDlidx); + int pgno = fts5DlidxIterPgno(pDlidx); + assert( fts5DlidxIterPgno(pDlidx)>iLeaf ); + cksum2 += iRowid + ((i64)pgno<<32); + } + fts5DlidxIterFree(pDlidx); + pDlidx = 0; + + if( p->rc==SQLITE_OK && cksum1!=cksum2 ) p->rc = FTS5_CORRUPT; +} + +static int fts5QueryCksum( + Fts5Index *p, /* Fts5 index object */ + int iIdx, + const char *z, /* Index key to query for */ + int n, /* Size of index key in bytes */ + int flags, /* Flags for Fts5IndexQuery */ + u64 *pCksum /* IN/OUT: Checksum value */ +){ + int eDetail = p->pConfig->eDetail; + u64 cksum = *pCksum; + Fts5IndexIter *pIter = 0; + int rc = sqlite3Fts5IndexQuery(p, z, n, flags, 0, &pIter); + + while( rc==SQLITE_OK && 0==sqlite3Fts5IterEof(pIter) ){ + i64 rowid = pIter->iRowid; + + if( eDetail==FTS5_DETAIL_NONE ){ + cksum ^= sqlite3Fts5IndexEntryCksum(rowid, 0, 0, iIdx, z, n); + }else{ + Fts5PoslistReader sReader; + for(sqlite3Fts5PoslistReaderInit(pIter->pData, pIter->nData, &sReader); + sReader.bEof==0; + sqlite3Fts5PoslistReaderNext(&sReader) + ){ + int iCol = FTS5_POS2COLUMN(sReader.iPos); + int iOff = FTS5_POS2OFFSET(sReader.iPos); + cksum ^= sqlite3Fts5IndexEntryCksum(rowid, iCol, iOff, iIdx, z, n); + } + } + if( rc==SQLITE_OK ){ + rc = sqlite3Fts5IterNext(pIter); + } + } + sqlite3Fts5IterClose(pIter); + + *pCksum = cksum; + return rc; +} + + +/* +** This function is also purely an internal test. It does not contribute to +** FTS functionality, or even the integrity-check, in any way. +*/ +static void fts5TestTerm( + Fts5Index *p, + Fts5Buffer *pPrev, /* Previous term */ + const char *z, int n, /* Possibly new term to test */ + u64 expected, + u64 *pCksum +){ + int rc = p->rc; + if( pPrev->n==0 ){ + fts5BufferSet(&rc, pPrev, n, (const u8*)z); + }else + if( rc==SQLITE_OK && (pPrev->n!=n || memcmp(pPrev->p, z, n)) ){ + u64 cksum3 = *pCksum; + const char *zTerm = (const char*)&pPrev->p[1]; /* term sans prefix-byte */ + int nTerm = pPrev->n-1; /* Size of zTerm in bytes */ + int iIdx = (pPrev->p[0] - FTS5_MAIN_PREFIX); + int flags = (iIdx==0 ? 0 : FTS5INDEX_QUERY_PREFIX); + u64 ck1 = 0; + u64 ck2 = 0; + + /* Check that the results returned for ASC and DESC queries are + ** the same. If not, call this corruption. */ + rc = fts5QueryCksum(p, iIdx, zTerm, nTerm, flags, &ck1); + if( rc==SQLITE_OK ){ + int f = flags|FTS5INDEX_QUERY_DESC; + rc = fts5QueryCksum(p, iIdx, zTerm, nTerm, f, &ck2); + } + if( rc==SQLITE_OK && ck1!=ck2 ) rc = FTS5_CORRUPT; + + /* If this is a prefix query, check that the results returned if the + ** the index is disabled are the same. In both ASC and DESC order. + ** + ** This check may only be performed if the hash table is empty. This + ** is because the hash table only supports a single scan query at + ** a time, and the multi-iter loop from which this function is called + ** is already performing such a scan. */ + if( p->nPendingData==0 ){ + if( iIdx>0 && rc==SQLITE_OK ){ + int f = flags|FTS5INDEX_QUERY_TEST_NOIDX; + ck2 = 0; + rc = fts5QueryCksum(p, iIdx, zTerm, nTerm, f, &ck2); + if( rc==SQLITE_OK && ck1!=ck2 ) rc = FTS5_CORRUPT; + } + if( iIdx>0 && rc==SQLITE_OK ){ + int f = flags|FTS5INDEX_QUERY_TEST_NOIDX|FTS5INDEX_QUERY_DESC; + ck2 = 0; + rc = fts5QueryCksum(p, iIdx, zTerm, nTerm, f, &ck2); + if( rc==SQLITE_OK && ck1!=ck2 ) rc = FTS5_CORRUPT; + } + } + + cksum3 ^= ck1; + fts5BufferSet(&rc, pPrev, n, (const u8*)z); + + if( rc==SQLITE_OK && cksum3!=expected ){ + rc = FTS5_CORRUPT; + } + *pCksum = cksum3; + } + p->rc = rc; +} + +#else +# define fts5TestDlidxReverse(x,y,z) +# define fts5TestTerm(u,v,w,x,y,z) +#endif + +/* +** Check that: +** +** 1) All leaves of pSeg between iFirst and iLast (inclusive) exist and +** contain zero terms. +** 2) All leaves of pSeg between iNoRowid and iLast (inclusive) exist and +** contain zero rowids. +*/ +static void fts5IndexIntegrityCheckEmpty( + Fts5Index *p, + Fts5StructureSegment *pSeg, /* Segment to check internal consistency */ + int iFirst, + int iNoRowid, + int iLast +){ + int i; + + /* Now check that the iter.nEmpty leaves following the current leaf + ** (a) exist and (b) contain no terms. */ + for(i=iFirst; p->rc==SQLITE_OK && i<=iLast; i++){ + Fts5Data *pLeaf = fts5DataRead(p, FTS5_SEGMENT_ROWID(pSeg->iSegid, i)); + if( pLeaf ){ + if( !fts5LeafIsTermless(pLeaf) ) p->rc = FTS5_CORRUPT; + if( i>=iNoRowid && 0!=fts5LeafFirstRowidOff(pLeaf) ) p->rc = FTS5_CORRUPT; + } + fts5DataRelease(pLeaf); + } +} + +static void fts5IntegrityCheckPgidx(Fts5Index *p, Fts5Data *pLeaf){ + int iTermOff = 0; + int ii; + + Fts5Buffer buf1 = {0,0,0}; + Fts5Buffer buf2 = {0,0,0}; + + ii = pLeaf->szLeaf; + while( ii<pLeaf->nn && p->rc==SQLITE_OK ){ + int res; + int iOff; + int nIncr; + + ii += fts5GetVarint32(&pLeaf->p[ii], nIncr); + iTermOff += nIncr; + iOff = iTermOff; + + if( iOff>=pLeaf->szLeaf ){ + p->rc = FTS5_CORRUPT; + }else if( iTermOff==nIncr ){ + int nByte; + iOff += fts5GetVarint32(&pLeaf->p[iOff], nByte); + if( (iOff+nByte)>pLeaf->szLeaf ){ + p->rc = FTS5_CORRUPT; + }else{ + fts5BufferSet(&p->rc, &buf1, nByte, &pLeaf->p[iOff]); + } + }else{ + int nKeep, nByte; + iOff += fts5GetVarint32(&pLeaf->p[iOff], nKeep); + iOff += fts5GetVarint32(&pLeaf->p[iOff], nByte); + if( nKeep>buf1.n || (iOff+nByte)>pLeaf->szLeaf ){ + p->rc = FTS5_CORRUPT; + }else{ + buf1.n = nKeep; + fts5BufferAppendBlob(&p->rc, &buf1, nByte, &pLeaf->p[iOff]); + } + + if( p->rc==SQLITE_OK ){ + res = fts5BufferCompare(&buf1, &buf2); + if( res<=0 ) p->rc = FTS5_CORRUPT; + } + } + fts5BufferSet(&p->rc, &buf2, buf1.n, buf1.p); + } + + fts5BufferFree(&buf1); + fts5BufferFree(&buf2); +} + +static void fts5IndexIntegrityCheckSegment( + Fts5Index *p, /* FTS5 backend object */ + Fts5StructureSegment *pSeg /* Segment to check internal consistency */ +){ + Fts5Config *pConfig = p->pConfig; + sqlite3_stmt *pStmt = 0; + int rc2; + int iIdxPrevLeaf = pSeg->pgnoFirst-1; + int iDlidxPrevLeaf = pSeg->pgnoLast; + + if( pSeg->pgnoFirst==0 ) return; + + fts5IndexPrepareStmt(p, &pStmt, sqlite3_mprintf( + "SELECT segid, term, (pgno>>1), (pgno&1) FROM %Q.'%q_idx' WHERE segid=%d", + pConfig->zDb, pConfig->zName, pSeg->iSegid + )); + + /* Iterate through the b-tree hierarchy. */ + while( p->rc==SQLITE_OK && SQLITE_ROW==sqlite3_step(pStmt) ){ + i64 iRow; /* Rowid for this leaf */ + Fts5Data *pLeaf; /* Data for this leaf */ + + int nIdxTerm = sqlite3_column_bytes(pStmt, 1); + const char *zIdxTerm = (const char*)sqlite3_column_text(pStmt, 1); + int iIdxLeaf = sqlite3_column_int(pStmt, 2); + int bIdxDlidx = sqlite3_column_int(pStmt, 3); + + /* If the leaf in question has already been trimmed from the segment, + ** ignore this b-tree entry. Otherwise, load it into memory. */ + if( iIdxLeaf<pSeg->pgnoFirst ) continue; + iRow = FTS5_SEGMENT_ROWID(pSeg->iSegid, iIdxLeaf); + pLeaf = fts5DataRead(p, iRow); + if( pLeaf==0 ) break; + + /* Check that the leaf contains at least one term, and that it is equal + ** to or larger than the split-key in zIdxTerm. Also check that if there + ** is also a rowid pointer within the leaf page header, it points to a + ** location before the term. */ + if( pLeaf->nn<=pLeaf->szLeaf ){ + p->rc = FTS5_CORRUPT; + }else{ + int iOff; /* Offset of first term on leaf */ + int iRowidOff; /* Offset of first rowid on leaf */ + int nTerm; /* Size of term on leaf in bytes */ + int res; /* Comparison of term and split-key */ + + iOff = fts5LeafFirstTermOff(pLeaf); + iRowidOff = fts5LeafFirstRowidOff(pLeaf); + if( iRowidOff>=iOff ){ + p->rc = FTS5_CORRUPT; + }else{ + iOff += fts5GetVarint32(&pLeaf->p[iOff], nTerm); + res = memcmp(&pLeaf->p[iOff], zIdxTerm, MIN(nTerm, nIdxTerm)); + if( res==0 ) res = nTerm - nIdxTerm; + if( res<0 ) p->rc = FTS5_CORRUPT; + } + + fts5IntegrityCheckPgidx(p, pLeaf); + } + fts5DataRelease(pLeaf); + if( p->rc ) break; + + /* Now check that the iter.nEmpty leaves following the current leaf + ** (a) exist and (b) contain no terms. */ + fts5IndexIntegrityCheckEmpty( + p, pSeg, iIdxPrevLeaf+1, iDlidxPrevLeaf+1, iIdxLeaf-1 + ); + if( p->rc ) break; + + /* If there is a doclist-index, check that it looks right. */ + if( bIdxDlidx ){ + Fts5DlidxIter *pDlidx = 0; /* For iterating through doclist index */ + int iPrevLeaf = iIdxLeaf; + int iSegid = pSeg->iSegid; + int iPg = 0; + i64 iKey; + + for(pDlidx=fts5DlidxIterInit(p, 0, iSegid, iIdxLeaf); + fts5DlidxIterEof(p, pDlidx)==0; + fts5DlidxIterNext(p, pDlidx) + ){ + + /* Check any rowid-less pages that occur before the current leaf. */ + for(iPg=iPrevLeaf+1; iPg<fts5DlidxIterPgno(pDlidx); iPg++){ + iKey = FTS5_SEGMENT_ROWID(iSegid, iPg); + pLeaf = fts5DataRead(p, iKey); + if( pLeaf ){ + if( fts5LeafFirstRowidOff(pLeaf)!=0 ) p->rc = FTS5_CORRUPT; + fts5DataRelease(pLeaf); + } + } + iPrevLeaf = fts5DlidxIterPgno(pDlidx); + + /* Check that the leaf page indicated by the iterator really does + ** contain the rowid suggested by the same. */ + iKey = FTS5_SEGMENT_ROWID(iSegid, iPrevLeaf); + pLeaf = fts5DataRead(p, iKey); + if( pLeaf ){ + i64 iRowid; + int iRowidOff = fts5LeafFirstRowidOff(pLeaf); + ASSERT_SZLEAF_OK(pLeaf); + if( iRowidOff>=pLeaf->szLeaf ){ + p->rc = FTS5_CORRUPT; + }else{ + fts5GetVarint(&pLeaf->p[iRowidOff], (u64*)&iRowid); + if( iRowid!=fts5DlidxIterRowid(pDlidx) ) p->rc = FTS5_CORRUPT; + } + fts5DataRelease(pLeaf); + } + } + + iDlidxPrevLeaf = iPg; + fts5DlidxIterFree(pDlidx); + fts5TestDlidxReverse(p, iSegid, iIdxLeaf); + }else{ + iDlidxPrevLeaf = pSeg->pgnoLast; + /* TODO: Check there is no doclist index */ + } + + iIdxPrevLeaf = iIdxLeaf; + } + + rc2 = sqlite3_finalize(pStmt); + if( p->rc==SQLITE_OK ) p->rc = rc2; + + /* Page iter.iLeaf must now be the rightmost leaf-page in the segment */ +#if 0 + if( p->rc==SQLITE_OK && iter.iLeaf!=pSeg->pgnoLast ){ + p->rc = FTS5_CORRUPT; + } +#endif +} + + +/* +** Run internal checks to ensure that the FTS index (a) is internally +** consistent and (b) contains entries for which the XOR of the checksums +** as calculated by sqlite3Fts5IndexEntryCksum() is cksum. +** +** Return SQLITE_CORRUPT if any of the internal checks fail, or if the +** checksum does not match. Return SQLITE_OK if all checks pass without +** error, or some other SQLite error code if another error (e.g. OOM) +** occurs. +*/ +static int sqlite3Fts5IndexIntegrityCheck(Fts5Index *p, u64 cksum){ + int eDetail = p->pConfig->eDetail; + u64 cksum2 = 0; /* Checksum based on contents of indexes */ + Fts5Buffer poslist = {0,0,0}; /* Buffer used to hold a poslist */ + Fts5Iter *pIter; /* Used to iterate through entire index */ + Fts5Structure *pStruct; /* Index structure */ + +#ifdef SQLITE_DEBUG + /* Used by extra internal tests only run if NDEBUG is not defined */ + u64 cksum3 = 0; /* Checksum based on contents of indexes */ + Fts5Buffer term = {0,0,0}; /* Buffer used to hold most recent term */ +#endif + const int flags = FTS5INDEX_QUERY_NOOUTPUT; + + /* Load the FTS index structure */ + pStruct = fts5StructureRead(p); + + /* Check that the internal nodes of each segment match the leaves */ + if( pStruct ){ + int iLvl, iSeg; + for(iLvl=0; iLvl<pStruct->nLevel; iLvl++){ + for(iSeg=0; iSeg<pStruct->aLevel[iLvl].nSeg; iSeg++){ + Fts5StructureSegment *pSeg = &pStruct->aLevel[iLvl].aSeg[iSeg]; + fts5IndexIntegrityCheckSegment(p, pSeg); + } + } + } + + /* The cksum argument passed to this function is a checksum calculated + ** based on all expected entries in the FTS index (including prefix index + ** entries). This block checks that a checksum calculated based on the + ** actual contents of FTS index is identical. + ** + ** Two versions of the same checksum are calculated. The first (stack + ** variable cksum2) based on entries extracted from the full-text index + ** while doing a linear scan of each individual index in turn. + ** + ** As each term visited by the linear scans, a separate query for the + ** same term is performed. cksum3 is calculated based on the entries + ** extracted by these queries. + */ + for(fts5MultiIterNew(p, pStruct, flags, 0, 0, 0, -1, 0, &pIter); + fts5MultiIterEof(p, pIter)==0; + fts5MultiIterNext(p, pIter, 0, 0) + ){ + int n; /* Size of term in bytes */ + i64 iPos = 0; /* Position read from poslist */ + int iOff = 0; /* Offset within poslist */ + i64 iRowid = fts5MultiIterRowid(pIter); + char *z = (char*)fts5MultiIterTerm(pIter, &n); + + /* If this is a new term, query for it. Update cksum3 with the results. */ + fts5TestTerm(p, &term, z, n, cksum2, &cksum3); + + if( eDetail==FTS5_DETAIL_NONE ){ + if( 0==fts5MultiIterIsEmpty(p, pIter) ){ + cksum2 ^= sqlite3Fts5IndexEntryCksum(iRowid, 0, 0, -1, z, n); + } + }else{ + poslist.n = 0; + fts5SegiterPoslist(p, &pIter->aSeg[pIter->aFirst[1].iFirst], 0, &poslist); + while( 0==sqlite3Fts5PoslistNext64(poslist.p, poslist.n, &iOff, &iPos) ){ + int iCol = FTS5_POS2COLUMN(iPos); + int iTokOff = FTS5_POS2OFFSET(iPos); + cksum2 ^= sqlite3Fts5IndexEntryCksum(iRowid, iCol, iTokOff, -1, z, n); + } + } + } + fts5TestTerm(p, &term, 0, 0, cksum2, &cksum3); + + fts5MultiIterFree(pIter); + if( p->rc==SQLITE_OK && cksum!=cksum2 ) p->rc = FTS5_CORRUPT; + + fts5StructureRelease(pStruct); +#ifdef SQLITE_DEBUG + fts5BufferFree(&term); +#endif + fts5BufferFree(&poslist); + return fts5IndexReturn(p); +} + +/************************************************************************* +************************************************************************** +** Below this point is the implementation of the fts5_decode() scalar +** function only. +*/ + +/* +** Decode a segment-data rowid from the %_data table. This function is +** the opposite of macro FTS5_SEGMENT_ROWID(). +*/ +static void fts5DecodeRowid( + i64 iRowid, /* Rowid from %_data table */ + int *piSegid, /* OUT: Segment id */ + int *pbDlidx, /* OUT: Dlidx flag */ + int *piHeight, /* OUT: Height */ + int *piPgno /* OUT: Page number */ +){ + *piPgno = (int)(iRowid & (((i64)1 << FTS5_DATA_PAGE_B) - 1)); + iRowid >>= FTS5_DATA_PAGE_B; + + *piHeight = (int)(iRowid & (((i64)1 << FTS5_DATA_HEIGHT_B) - 1)); + iRowid >>= FTS5_DATA_HEIGHT_B; + + *pbDlidx = (int)(iRowid & 0x0001); + iRowid >>= FTS5_DATA_DLI_B; + + *piSegid = (int)(iRowid & (((i64)1 << FTS5_DATA_ID_B) - 1)); +} + +static void fts5DebugRowid(int *pRc, Fts5Buffer *pBuf, i64 iKey){ + int iSegid, iHeight, iPgno, bDlidx; /* Rowid compenents */ + fts5DecodeRowid(iKey, &iSegid, &bDlidx, &iHeight, &iPgno); + + if( iSegid==0 ){ + if( iKey==FTS5_AVERAGES_ROWID ){ + sqlite3Fts5BufferAppendPrintf(pRc, pBuf, "{averages} "); + }else{ + sqlite3Fts5BufferAppendPrintf(pRc, pBuf, "{structure}"); + } + } + else{ + sqlite3Fts5BufferAppendPrintf(pRc, pBuf, "{%ssegid=%d h=%d pgno=%d}", + bDlidx ? "dlidx " : "", iSegid, iHeight, iPgno + ); + } +} + +static void fts5DebugStructure( + int *pRc, /* IN/OUT: error code */ + Fts5Buffer *pBuf, + Fts5Structure *p +){ + int iLvl, iSeg; /* Iterate through levels, segments */ + + for(iLvl=0; iLvl<p->nLevel; iLvl++){ + Fts5StructureLevel *pLvl = &p->aLevel[iLvl]; + sqlite3Fts5BufferAppendPrintf(pRc, pBuf, + " {lvl=%d nMerge=%d nSeg=%d", iLvl, pLvl->nMerge, pLvl->nSeg + ); + for(iSeg=0; iSeg<pLvl->nSeg; iSeg++){ + Fts5StructureSegment *pSeg = &pLvl->aSeg[iSeg]; + sqlite3Fts5BufferAppendPrintf(pRc, pBuf, " {id=%d leaves=%d..%d}", + pSeg->iSegid, pSeg->pgnoFirst, pSeg->pgnoLast + ); + } + sqlite3Fts5BufferAppendPrintf(pRc, pBuf, "}"); + } +} + +/* +** This is part of the fts5_decode() debugging aid. +** +** Arguments pBlob/nBlob contain a serialized Fts5Structure object. This +** function appends a human-readable representation of the same object +** to the buffer passed as the second argument. +*/ +static void fts5DecodeStructure( + int *pRc, /* IN/OUT: error code */ + Fts5Buffer *pBuf, + const u8 *pBlob, int nBlob +){ + int rc; /* Return code */ + Fts5Structure *p = 0; /* Decoded structure object */ + + rc = fts5StructureDecode(pBlob, nBlob, 0, &p); + if( rc!=SQLITE_OK ){ + *pRc = rc; + return; + } + + fts5DebugStructure(pRc, pBuf, p); + fts5StructureRelease(p); +} + +/* +** This is part of the fts5_decode() debugging aid. +** +** Arguments pBlob/nBlob contain an "averages" record. This function +** appends a human-readable representation of record to the buffer passed +** as the second argument. +*/ +static void fts5DecodeAverages( + int *pRc, /* IN/OUT: error code */ + Fts5Buffer *pBuf, + const u8 *pBlob, int nBlob +){ + int i = 0; + const char *zSpace = ""; + + while( i<nBlob ){ + u64 iVal; + i += sqlite3Fts5GetVarint(&pBlob[i], &iVal); + sqlite3Fts5BufferAppendPrintf(pRc, pBuf, "%s%d", zSpace, (int)iVal); + zSpace = " "; + } +} + +/* +** Buffer (a/n) is assumed to contain a list of serialized varints. Read +** each varint and append its string representation to buffer pBuf. Return +** after either the input buffer is exhausted or a 0 value is read. +** +** The return value is the number of bytes read from the input buffer. +*/ +static int fts5DecodePoslist(int *pRc, Fts5Buffer *pBuf, const u8 *a, int n){ + int iOff = 0; + while( iOff<n ){ + int iVal; + iOff += fts5GetVarint32(&a[iOff], iVal); + sqlite3Fts5BufferAppendPrintf(pRc, pBuf, " %d", iVal); + } + return iOff; +} + +/* +** The start of buffer (a/n) contains the start of a doclist. The doclist +** may or may not finish within the buffer. This function appends a text +** representation of the part of the doclist that is present to buffer +** pBuf. +** +** The return value is the number of bytes read from the input buffer. +*/ +static int fts5DecodeDoclist(int *pRc, Fts5Buffer *pBuf, const u8 *a, int n){ + i64 iDocid = 0; + int iOff = 0; + + if( n>0 ){ + iOff = sqlite3Fts5GetVarint(a, (u64*)&iDocid); + sqlite3Fts5BufferAppendPrintf(pRc, pBuf, " id=%lld", iDocid); + } + while( iOff<n ){ + int nPos; + int bDel; + iOff += fts5GetPoslistSize(&a[iOff], &nPos, &bDel); + sqlite3Fts5BufferAppendPrintf(pRc, pBuf, " nPos=%d%s", nPos, bDel?"*":""); + iOff += fts5DecodePoslist(pRc, pBuf, &a[iOff], MIN(n-iOff, nPos)); + if( iOff<n ){ + i64 iDelta; + iOff += sqlite3Fts5GetVarint(&a[iOff], (u64*)&iDelta); + iDocid += iDelta; + sqlite3Fts5BufferAppendPrintf(pRc, pBuf, " id=%lld", iDocid); + } + } + + return iOff; +} + +/* +** This function is part of the fts5_decode() debugging function. It is +** only ever used with detail=none tables. +** +** Buffer (pData/nData) contains a doclist in the format used by detail=none +** tables. This function appends a human-readable version of that list to +** buffer pBuf. +** +** If *pRc is other than SQLITE_OK when this function is called, it is a +** no-op. If an OOM or other error occurs within this function, *pRc is +** set to an SQLite error code before returning. The final state of buffer +** pBuf is undefined in this case. +*/ +static void fts5DecodeRowidList( + int *pRc, /* IN/OUT: Error code */ + Fts5Buffer *pBuf, /* Buffer to append text to */ + const u8 *pData, int nData /* Data to decode list-of-rowids from */ +){ + int i = 0; + i64 iRowid = 0; + + while( i<nData ){ + const char *zApp = ""; + u64 iVal; + i += sqlite3Fts5GetVarint(&pData[i], &iVal); + iRowid += iVal; + + if( i<nData && pData[i]==0x00 ){ + i++; + if( i<nData && pData[i]==0x00 ){ + i++; + zApp = "+"; + }else{ + zApp = "*"; + } + } + + sqlite3Fts5BufferAppendPrintf(pRc, pBuf, " %lld%s", iRowid, zApp); + } +} + +/* +** The implementation of user-defined scalar function fts5_decode(). +*/ +static void fts5DecodeFunction( + sqlite3_context *pCtx, /* Function call context */ + int nArg, /* Number of args (always 2) */ + sqlite3_value **apVal /* Function arguments */ +){ + i64 iRowid; /* Rowid for record being decoded */ + int iSegid,iHeight,iPgno,bDlidx;/* Rowid components */ + const u8 *aBlob; int n; /* Record to decode */ + u8 *a = 0; + Fts5Buffer s; /* Build up text to return here */ + int rc = SQLITE_OK; /* Return code */ + int nSpace = 0; + int eDetailNone = (sqlite3_user_data(pCtx)!=0); + + assert( nArg==2 ); + UNUSED_PARAM(nArg); + memset(&s, 0, sizeof(Fts5Buffer)); + iRowid = sqlite3_value_int64(apVal[0]); + + /* Make a copy of the second argument (a blob) in aBlob[]. The aBlob[] + ** copy is followed by FTS5_DATA_ZERO_PADDING 0x00 bytes, which prevents + ** buffer overreads even if the record is corrupt. */ + n = sqlite3_value_bytes(apVal[1]); + aBlob = sqlite3_value_blob(apVal[1]); + nSpace = n + FTS5_DATA_ZERO_PADDING; + a = (u8*)sqlite3Fts5MallocZero(&rc, nSpace); + if( a==0 ) goto decode_out; + memcpy(a, aBlob, n); + + + fts5DecodeRowid(iRowid, &iSegid, &bDlidx, &iHeight, &iPgno); + + fts5DebugRowid(&rc, &s, iRowid); + if( bDlidx ){ + Fts5Data dlidx; + Fts5DlidxLvl lvl; + + dlidx.p = a; + dlidx.nn = n; + + memset(&lvl, 0, sizeof(Fts5DlidxLvl)); + lvl.pData = &dlidx; + lvl.iLeafPgno = iPgno; + + for(fts5DlidxLvlNext(&lvl); lvl.bEof==0; fts5DlidxLvlNext(&lvl)){ + sqlite3Fts5BufferAppendPrintf(&rc, &s, + " %d(%lld)", lvl.iLeafPgno, lvl.iRowid + ); + } + }else if( iSegid==0 ){ + if( iRowid==FTS5_AVERAGES_ROWID ){ + fts5DecodeAverages(&rc, &s, a, n); + }else{ + fts5DecodeStructure(&rc, &s, a, n); + } + }else if( eDetailNone ){ + Fts5Buffer term; /* Current term read from page */ + int szLeaf; + int iPgidxOff = szLeaf = fts5GetU16(&a[2]); + int iTermOff; + int nKeep = 0; + int iOff; + + memset(&term, 0, sizeof(Fts5Buffer)); + + /* Decode any entries that occur before the first term. */ + if( szLeaf<n ){ + iPgidxOff += fts5GetVarint32(&a[iPgidxOff], iTermOff); + }else{ + iTermOff = szLeaf; + } + fts5DecodeRowidList(&rc, &s, &a[4], iTermOff-4); + + iOff = iTermOff; + while( iOff<szLeaf ){ + int nAppend; + + /* Read the term data for the next term*/ + iOff += fts5GetVarint32(&a[iOff], nAppend); + term.n = nKeep; + fts5BufferAppendBlob(&rc, &term, nAppend, &a[iOff]); + sqlite3Fts5BufferAppendPrintf( + &rc, &s, " term=%.*s", term.n, (const char*)term.p + ); + iOff += nAppend; + + /* Figure out where the doclist for this term ends */ + if( iPgidxOff<n ){ + int nIncr; + iPgidxOff += fts5GetVarint32(&a[iPgidxOff], nIncr); + iTermOff += nIncr; + }else{ + iTermOff = szLeaf; + } + + fts5DecodeRowidList(&rc, &s, &a[iOff], iTermOff-iOff); + iOff = iTermOff; + if( iOff<szLeaf ){ + iOff += fts5GetVarint32(&a[iOff], nKeep); + } + } + + fts5BufferFree(&term); + }else{ + Fts5Buffer term; /* Current term read from page */ + int szLeaf; /* Offset of pgidx in a[] */ + int iPgidxOff; + int iPgidxPrev = 0; /* Previous value read from pgidx */ + int iTermOff = 0; + int iRowidOff = 0; + int iOff; + int nDoclist; + + memset(&term, 0, sizeof(Fts5Buffer)); + + if( n<4 ){ + sqlite3Fts5BufferSet(&rc, &s, 7, (const u8*)"corrupt"); + goto decode_out; + }else{ + iRowidOff = fts5GetU16(&a[0]); + iPgidxOff = szLeaf = fts5GetU16(&a[2]); + if( iPgidxOff<n ){ + fts5GetVarint32(&a[iPgidxOff], iTermOff); + } + } + + /* Decode the position list tail at the start of the page */ + if( iRowidOff!=0 ){ + iOff = iRowidOff; + }else if( iTermOff!=0 ){ + iOff = iTermOff; + }else{ + iOff = szLeaf; + } + fts5DecodePoslist(&rc, &s, &a[4], iOff-4); + + /* Decode any more doclist data that appears on the page before the + ** first term. */ + nDoclist = (iTermOff ? iTermOff : szLeaf) - iOff; + fts5DecodeDoclist(&rc, &s, &a[iOff], nDoclist); + + while( iPgidxOff<n ){ + int bFirst = (iPgidxOff==szLeaf); /* True for first term on page */ + int nByte; /* Bytes of data */ + int iEnd; + + iPgidxOff += fts5GetVarint32(&a[iPgidxOff], nByte); + iPgidxPrev += nByte; + iOff = iPgidxPrev; + + if( iPgidxOff<n ){ + fts5GetVarint32(&a[iPgidxOff], nByte); + iEnd = iPgidxPrev + nByte; + }else{ + iEnd = szLeaf; + } + + if( bFirst==0 ){ + iOff += fts5GetVarint32(&a[iOff], nByte); + term.n = nByte; + } + iOff += fts5GetVarint32(&a[iOff], nByte); + fts5BufferAppendBlob(&rc, &term, nByte, &a[iOff]); + iOff += nByte; + + sqlite3Fts5BufferAppendPrintf( + &rc, &s, " term=%.*s", term.n, (const char*)term.p + ); + iOff += fts5DecodeDoclist(&rc, &s, &a[iOff], iEnd-iOff); + } + + fts5BufferFree(&term); + } + + decode_out: + sqlite3_free(a); + if( rc==SQLITE_OK ){ + sqlite3_result_text(pCtx, (const char*)s.p, s.n, SQLITE_TRANSIENT); + }else{ + sqlite3_result_error_code(pCtx, rc); + } + fts5BufferFree(&s); +} + +/* +** The implementation of user-defined scalar function fts5_rowid(). +*/ +static void fts5RowidFunction( + sqlite3_context *pCtx, /* Function call context */ + int nArg, /* Number of args (always 2) */ + sqlite3_value **apVal /* Function arguments */ +){ + const char *zArg; + if( nArg==0 ){ + sqlite3_result_error(pCtx, "should be: fts5_rowid(subject, ....)", -1); + }else{ + zArg = (const char*)sqlite3_value_text(apVal[0]); + if( 0==sqlite3_stricmp(zArg, "segment") ){ + i64 iRowid; + int segid, pgno; + if( nArg!=3 ){ + sqlite3_result_error(pCtx, + "should be: fts5_rowid('segment', segid, pgno))", -1 + ); + }else{ + segid = sqlite3_value_int(apVal[1]); + pgno = sqlite3_value_int(apVal[2]); + iRowid = FTS5_SEGMENT_ROWID(segid, pgno); + sqlite3_result_int64(pCtx, iRowid); + } + }else{ + sqlite3_result_error(pCtx, + "first arg to fts5_rowid() must be 'segment'" , -1 + ); + } + } +} + +/* +** This is called as part of registering the FTS5 module with database +** connection db. It registers several user-defined scalar functions useful +** with FTS5. +** +** If successful, SQLITE_OK is returned. If an error occurs, some other +** SQLite error code is returned instead. +*/ +static int sqlite3Fts5IndexInit(sqlite3 *db){ + int rc = sqlite3_create_function( + db, "fts5_decode", 2, SQLITE_UTF8, 0, fts5DecodeFunction, 0, 0 + ); + + if( rc==SQLITE_OK ){ + rc = sqlite3_create_function( + db, "fts5_decode_none", 2, + SQLITE_UTF8, (void*)db, fts5DecodeFunction, 0, 0 + ); + } + + if( rc==SQLITE_OK ){ + rc = sqlite3_create_function( + db, "fts5_rowid", -1, SQLITE_UTF8, 0, fts5RowidFunction, 0, 0 + ); + } + return rc; +} + +#line 1 "fts5_main.c" +/* +** 2014 Jun 09 +** +** The author disclaims copyright to this source code. In place of +** a legal notice, here is a blessing: +** +** May you do good and not evil. +** May you find forgiveness for yourself and forgive others. +** May you share freely, never taking more than you give. +** +****************************************************************************** +** +** This is an SQLite module implementing full-text search. +*/ + + +/* #include "fts5Int.h" */ + +/* +** This variable is set to false when running tests for which the on disk +** structures should not be corrupt. Otherwise, true. If it is false, extra +** assert() conditions in the fts5 code are activated - conditions that are +** only true if it is guaranteed that the fts5 database is not corrupt. +*/ +int sqlite3_fts5_may_be_corrupt = 1; + + +typedef struct Fts5Auxdata Fts5Auxdata; +typedef struct Fts5Auxiliary Fts5Auxiliary; +typedef struct Fts5Cursor Fts5Cursor; +typedef struct Fts5Sorter Fts5Sorter; +typedef struct Fts5Table Fts5Table; +typedef struct Fts5TokenizerModule Fts5TokenizerModule; + +/* +** NOTES ON TRANSACTIONS: +** +** SQLite invokes the following virtual table methods as transactions are +** opened and closed by the user: +** +** xBegin(): Start of a new transaction. +** xSync(): Initial part of two-phase commit. +** xCommit(): Final part of two-phase commit. +** xRollback(): Rollback the transaction. +** +** Anything that is required as part of a commit that may fail is performed +** in the xSync() callback. Current versions of SQLite ignore any errors +** returned by xCommit(). +** +** And as sub-transactions are opened/closed: +** +** xSavepoint(int S): Open savepoint S. +** xRelease(int S): Commit and close savepoint S. +** xRollbackTo(int S): Rollback to start of savepoint S. +** +** During a write-transaction the fts5_index.c module may cache some data +** in-memory. It is flushed to disk whenever xSync(), xRelease() or +** xSavepoint() is called. And discarded whenever xRollback() or xRollbackTo() +** is called. +** +** Additionally, if SQLITE_DEBUG is defined, an instance of the following +** structure is used to record the current transaction state. This information +** is not required, but it is used in the assert() statements executed by +** function fts5CheckTransactionState() (see below). +*/ +struct Fts5TransactionState { + int eState; /* 0==closed, 1==open, 2==synced */ + int iSavepoint; /* Number of open savepoints (0 -> none) */ +}; + +/* +** A single object of this type is allocated when the FTS5 module is +** registered with a database handle. It is used to store pointers to +** all registered FTS5 extensions - tokenizers and auxiliary functions. +*/ +struct Fts5Global { + fts5_api api; /* User visible part of object (see fts5.h) */ + sqlite3 *db; /* Associated database connection */ + i64 iNextId; /* Used to allocate unique cursor ids */ + Fts5Auxiliary *pAux; /* First in list of all aux. functions */ + Fts5TokenizerModule *pTok; /* First in list of all tokenizer modules */ + Fts5TokenizerModule *pDfltTok; /* Default tokenizer module */ + Fts5Cursor *pCsr; /* First in list of all open cursors */ +}; + +/* +** Each auxiliary function registered with the FTS5 module is represented +** by an object of the following type. All such objects are stored as part +** of the Fts5Global.pAux list. +*/ +struct Fts5Auxiliary { + Fts5Global *pGlobal; /* Global context for this function */ + char *zFunc; /* Function name (nul-terminated) */ + void *pUserData; /* User-data pointer */ + fts5_extension_function xFunc; /* Callback function */ + void (*xDestroy)(void*); /* Destructor function */ + Fts5Auxiliary *pNext; /* Next registered auxiliary function */ +}; + +/* +** Each tokenizer module registered with the FTS5 module is represented +** by an object of the following type. All such objects are stored as part +** of the Fts5Global.pTok list. +*/ +struct Fts5TokenizerModule { + char *zName; /* Name of tokenizer */ + void *pUserData; /* User pointer passed to xCreate() */ + fts5_tokenizer x; /* Tokenizer functions */ + void (*xDestroy)(void*); /* Destructor function */ + Fts5TokenizerModule *pNext; /* Next registered tokenizer module */ +}; + +/* +** Virtual-table object. +*/ +struct Fts5Table { + sqlite3_vtab base; /* Base class used by SQLite core */ + Fts5Config *pConfig; /* Virtual table configuration */ + Fts5Index *pIndex; /* Full-text index */ + Fts5Storage *pStorage; /* Document store */ + Fts5Global *pGlobal; /* Global (connection wide) data */ + Fts5Cursor *pSortCsr; /* Sort data from this cursor */ +#ifdef SQLITE_DEBUG + struct Fts5TransactionState ts; +#endif +}; + +struct Fts5MatchPhrase { + Fts5Buffer *pPoslist; /* Pointer to current poslist */ + int nTerm; /* Size of phrase in terms */ +}; + +/* +** pStmt: +** SELECT rowid, <fts> FROM <fts> ORDER BY +rank; +** +** aIdx[]: +** There is one entry in the aIdx[] array for each phrase in the query, +** the value of which is the offset within aPoslist[] following the last +** byte of the position list for the corresponding phrase. +*/ +struct Fts5Sorter { + sqlite3_stmt *pStmt; + i64 iRowid; /* Current rowid */ + const u8 *aPoslist; /* Position lists for current row */ + int nIdx; /* Number of entries in aIdx[] */ + int aIdx[1]; /* Offsets into aPoslist for current row */ +}; + + +/* +** Virtual-table cursor object. +** +** iSpecial: +** If this is a 'special' query (refer to function fts5SpecialMatch()), +** then this variable contains the result of the query. +** +** iFirstRowid, iLastRowid: +** These variables are only used for FTS5_PLAN_MATCH cursors. Assuming the +** cursor iterates in ascending order of rowids, iFirstRowid is the lower +** limit of rowids to return, and iLastRowid the upper. In other words, the +** WHERE clause in the user's query might have been: +** +** <tbl> MATCH <expr> AND rowid BETWEEN $iFirstRowid AND $iLastRowid +** +** If the cursor iterates in descending order of rowid, iFirstRowid +** is the upper limit (i.e. the "first" rowid visited) and iLastRowid +** the lower. +*/ +struct Fts5Cursor { + sqlite3_vtab_cursor base; /* Base class used by SQLite core */ + Fts5Cursor *pNext; /* Next cursor in Fts5Cursor.pCsr list */ + int *aColumnSize; /* Values for xColumnSize() */ + i64 iCsrId; /* Cursor id */ + + /* Zero from this point onwards on cursor reset */ + int ePlan; /* FTS5_PLAN_XXX value */ + int bDesc; /* True for "ORDER BY rowid DESC" queries */ + i64 iFirstRowid; /* Return no rowids earlier than this */ + i64 iLastRowid; /* Return no rowids later than this */ + sqlite3_stmt *pStmt; /* Statement used to read %_content */ + Fts5Expr *pExpr; /* Expression for MATCH queries */ + Fts5Sorter *pSorter; /* Sorter for "ORDER BY rank" queries */ + int csrflags; /* Mask of cursor flags (see below) */ + i64 iSpecial; /* Result of special query */ + + /* "rank" function. Populated on demand from vtab.xColumn(). */ + char *zRank; /* Custom rank function */ + char *zRankArgs; /* Custom rank function args */ + Fts5Auxiliary *pRank; /* Rank callback (or NULL) */ + int nRankArg; /* Number of trailing arguments for rank() */ + sqlite3_value **apRankArg; /* Array of trailing arguments */ + sqlite3_stmt *pRankArgStmt; /* Origin of objects in apRankArg[] */ + + /* Auxiliary data storage */ + Fts5Auxiliary *pAux; /* Currently executing extension function */ + Fts5Auxdata *pAuxdata; /* First in linked list of saved aux-data */ + + /* Cache used by auxiliary functions xInst() and xInstCount() */ + Fts5PoslistReader *aInstIter; /* One for each phrase */ + int nInstAlloc; /* Size of aInst[] array (entries / 3) */ + int nInstCount; /* Number of phrase instances */ + int *aInst; /* 3 integers per phrase instance */ +}; + +/* +** Bits that make up the "idxNum" parameter passed indirectly by +** xBestIndex() to xFilter(). +*/ +#define FTS5_BI_MATCH 0x0001 /* <tbl> MATCH ? */ +#define FTS5_BI_RANK 0x0002 /* rank MATCH ? */ +#define FTS5_BI_ROWID_EQ 0x0004 /* rowid == ? */ +#define FTS5_BI_ROWID_LE 0x0008 /* rowid <= ? */ +#define FTS5_BI_ROWID_GE 0x0010 /* rowid >= ? */ + +#define FTS5_BI_ORDER_RANK 0x0020 +#define FTS5_BI_ORDER_ROWID 0x0040 +#define FTS5_BI_ORDER_DESC 0x0080 + +/* +** Values for Fts5Cursor.csrflags +*/ +#define FTS5CSR_EOF 0x01 +#define FTS5CSR_REQUIRE_CONTENT 0x02 +#define FTS5CSR_REQUIRE_DOCSIZE 0x04 +#define FTS5CSR_REQUIRE_INST 0x08 +#define FTS5CSR_FREE_ZRANK 0x10 +#define FTS5CSR_REQUIRE_RESEEK 0x20 +#define FTS5CSR_REQUIRE_POSLIST 0x40 + +#define BitFlagAllTest(x,y) (((x) & (y))==(y)) +#define BitFlagTest(x,y) (((x) & (y))!=0) + + +/* +** Macros to Set(), Clear() and Test() cursor flags. +*/ +#define CsrFlagSet(pCsr, flag) ((pCsr)->csrflags |= (flag)) +#define CsrFlagClear(pCsr, flag) ((pCsr)->csrflags &= ~(flag)) +#define CsrFlagTest(pCsr, flag) ((pCsr)->csrflags & (flag)) + +struct Fts5Auxdata { + Fts5Auxiliary *pAux; /* Extension to which this belongs */ + void *pPtr; /* Pointer value */ + void(*xDelete)(void*); /* Destructor */ + Fts5Auxdata *pNext; /* Next object in linked list */ +}; + +#ifdef SQLITE_DEBUG +#define FTS5_BEGIN 1 +#define FTS5_SYNC 2 +#define FTS5_COMMIT 3 +#define FTS5_ROLLBACK 4 +#define FTS5_SAVEPOINT 5 +#define FTS5_RELEASE 6 +#define FTS5_ROLLBACKTO 7 +static void fts5CheckTransactionState(Fts5Table *p, int op, int iSavepoint){ + switch( op ){ + case FTS5_BEGIN: + assert( p->ts.eState==0 ); + p->ts.eState = 1; + p->ts.iSavepoint = -1; + break; + + case FTS5_SYNC: + assert( p->ts.eState==1 ); + p->ts.eState = 2; + break; + + case FTS5_COMMIT: + assert( p->ts.eState==2 ); + p->ts.eState = 0; + break; + + case FTS5_ROLLBACK: + assert( p->ts.eState==1 || p->ts.eState==2 || p->ts.eState==0 ); + p->ts.eState = 0; + break; + + case FTS5_SAVEPOINT: + assert( p->ts.eState==1 ); + assert( iSavepoint>=0 ); + assert( iSavepoint>p->ts.iSavepoint ); + p->ts.iSavepoint = iSavepoint; + break; + + case FTS5_RELEASE: + assert( p->ts.eState==1 ); + assert( iSavepoint>=0 ); + assert( iSavepoint<=p->ts.iSavepoint ); + p->ts.iSavepoint = iSavepoint-1; + break; + + case FTS5_ROLLBACKTO: + assert( p->ts.eState==1 ); + assert( iSavepoint>=0 ); + assert( iSavepoint<=p->ts.iSavepoint ); + p->ts.iSavepoint = iSavepoint; + break; + } +} +#else +# define fts5CheckTransactionState(x,y,z) +#endif + +/* +** Return true if pTab is a contentless table. +*/ +static int fts5IsContentless(Fts5Table *pTab){ + return pTab->pConfig->eContent==FTS5_CONTENT_NONE; +} + +/* +** Delete a virtual table handle allocated by fts5InitVtab(). +*/ +static void fts5FreeVtab(Fts5Table *pTab){ + if( pTab ){ + sqlite3Fts5IndexClose(pTab->pIndex); + sqlite3Fts5StorageClose(pTab->pStorage); + sqlite3Fts5ConfigFree(pTab->pConfig); + sqlite3_free(pTab); + } +} + +/* +** The xDisconnect() virtual table method. +*/ +static int fts5DisconnectMethod(sqlite3_vtab *pVtab){ + fts5FreeVtab((Fts5Table*)pVtab); + return SQLITE_OK; +} + +/* +** The xDestroy() virtual table method. +*/ +static int fts5DestroyMethod(sqlite3_vtab *pVtab){ + Fts5Table *pTab = (Fts5Table*)pVtab; + int rc = sqlite3Fts5DropAll(pTab->pConfig); + if( rc==SQLITE_OK ){ + fts5FreeVtab((Fts5Table*)pVtab); + } + return rc; +} + +/* +** This function is the implementation of both the xConnect and xCreate +** methods of the FTS3 virtual table. +** +** The argv[] array contains the following: +** +** argv[0] -> module name ("fts5") +** argv[1] -> database name +** argv[2] -> table name +** argv[...] -> "column name" and other module argument fields. +*/ +static int fts5InitVtab( + int bCreate, /* True for xCreate, false for xConnect */ + sqlite3 *db, /* The SQLite database connection */ + void *pAux, /* Hash table containing tokenizers */ + int argc, /* Number of elements in argv array */ + const char * const *argv, /* xCreate/xConnect argument array */ + sqlite3_vtab **ppVTab, /* Write the resulting vtab structure here */ + char **pzErr /* Write any error message here */ +){ + Fts5Global *pGlobal = (Fts5Global*)pAux; + const char **azConfig = (const char**)argv; + int rc = SQLITE_OK; /* Return code */ + Fts5Config *pConfig = 0; /* Results of parsing argc/argv */ + Fts5Table *pTab = 0; /* New virtual table object */ + + /* Allocate the new vtab object and parse the configuration */ + pTab = (Fts5Table*)sqlite3Fts5MallocZero(&rc, sizeof(Fts5Table)); + if( rc==SQLITE_OK ){ + rc = sqlite3Fts5ConfigParse(pGlobal, db, argc, azConfig, &pConfig, pzErr); + assert( (rc==SQLITE_OK && *pzErr==0) || pConfig==0 ); + } + if( rc==SQLITE_OK ){ + pTab->pConfig = pConfig; + pTab->pGlobal = pGlobal; + } + + /* Open the index sub-system */ + if( rc==SQLITE_OK ){ + rc = sqlite3Fts5IndexOpen(pConfig, bCreate, &pTab->pIndex, pzErr); + } + + /* Open the storage sub-system */ + if( rc==SQLITE_OK ){ + rc = sqlite3Fts5StorageOpen( + pConfig, pTab->pIndex, bCreate, &pTab->pStorage, pzErr + ); + } + + /* Call sqlite3_declare_vtab() */ + if( rc==SQLITE_OK ){ + rc = sqlite3Fts5ConfigDeclareVtab(pConfig); + } + + /* Load the initial configuration */ + if( rc==SQLITE_OK ){ + assert( pConfig->pzErrmsg==0 ); + pConfig->pzErrmsg = pzErr; + rc = sqlite3Fts5IndexLoadConfig(pTab->pIndex); + sqlite3Fts5IndexRollback(pTab->pIndex); + pConfig->pzErrmsg = 0; + } + + if( rc!=SQLITE_OK ){ + fts5FreeVtab(pTab); + pTab = 0; + }else if( bCreate ){ + fts5CheckTransactionState(pTab, FTS5_BEGIN, 0); + } + *ppVTab = (sqlite3_vtab*)pTab; + return rc; +} + +/* +** The xConnect() and xCreate() methods for the virtual table. All the +** work is done in function fts5InitVtab(). +*/ +static int fts5ConnectMethod( + sqlite3 *db, /* Database connection */ + void *pAux, /* Pointer to tokenizer hash table */ + int argc, /* Number of elements in argv array */ + const char * const *argv, /* xCreate/xConnect argument array */ + sqlite3_vtab **ppVtab, /* OUT: New sqlite3_vtab object */ + char **pzErr /* OUT: sqlite3_malloc'd error message */ +){ + return fts5InitVtab(0, db, pAux, argc, argv, ppVtab, pzErr); +} +static int fts5CreateMethod( + sqlite3 *db, /* Database connection */ + void *pAux, /* Pointer to tokenizer hash table */ + int argc, /* Number of elements in argv array */ + const char * const *argv, /* xCreate/xConnect argument array */ + sqlite3_vtab **ppVtab, /* OUT: New sqlite3_vtab object */ + char **pzErr /* OUT: sqlite3_malloc'd error message */ +){ + return fts5InitVtab(1, db, pAux, argc, argv, ppVtab, pzErr); +} + +/* +** The different query plans. +*/ +#define FTS5_PLAN_MATCH 1 /* (<tbl> MATCH ?) */ +#define FTS5_PLAN_SOURCE 2 /* A source cursor for SORTED_MATCH */ +#define FTS5_PLAN_SPECIAL 3 /* An internal query */ +#define FTS5_PLAN_SORTED_MATCH 4 /* (<tbl> MATCH ? ORDER BY rank) */ +#define FTS5_PLAN_SCAN 5 /* No usable constraint */ +#define FTS5_PLAN_ROWID 6 /* (rowid = ?) */ + +/* +** Set the SQLITE_INDEX_SCAN_UNIQUE flag in pIdxInfo->flags. Unless this +** extension is currently being used by a version of SQLite too old to +** support index-info flags. In that case this function is a no-op. +*/ +static void fts5SetUniqueFlag(sqlite3_index_info *pIdxInfo){ +#if SQLITE_VERSION_NUMBER>=3008012 +#ifndef SQLITE_CORE + if( sqlite3_libversion_number()>=3008012 ) +#endif + { + pIdxInfo->idxFlags |= SQLITE_INDEX_SCAN_UNIQUE; + } +#endif +} + +/* +** Implementation of the xBestIndex method for FTS5 tables. Within the +** WHERE constraint, it searches for the following: +** +** 1. A MATCH constraint against the special column. +** 2. A MATCH constraint against the "rank" column. +** 3. An == constraint against the rowid column. +** 4. A < or <= constraint against the rowid column. +** 5. A > or >= constraint against the rowid column. +** +** Within the ORDER BY, either: +** +** 5. ORDER BY rank [ASC|DESC] +** 6. ORDER BY rowid [ASC|DESC] +** +** Costs are assigned as follows: +** +** a) If an unusable MATCH operator is present in the WHERE clause, the +** cost is unconditionally set to 1e50 (a really big number). +** +** a) If a MATCH operator is present, the cost depends on the other +** constraints also present. As follows: +** +** * No other constraints: cost=1000.0 +** * One rowid range constraint: cost=750.0 +** * Both rowid range constraints: cost=500.0 +** * An == rowid constraint: cost=100.0 +** +** b) Otherwise, if there is no MATCH: +** +** * No other constraints: cost=1000000.0 +** * One rowid range constraint: cost=750000.0 +** * Both rowid range constraints: cost=250000.0 +** * An == rowid constraint: cost=10.0 +** +** Costs are not modified by the ORDER BY clause. +*/ +static int fts5BestIndexMethod(sqlite3_vtab *pVTab, sqlite3_index_info *pInfo){ + Fts5Table *pTab = (Fts5Table*)pVTab; + Fts5Config *pConfig = pTab->pConfig; + int idxFlags = 0; /* Parameter passed through to xFilter() */ + int bHasMatch; + int iNext; + int i; + + struct Constraint { + int op; /* Mask against sqlite3_index_constraint.op */ + int fts5op; /* FTS5 mask for idxFlags */ + int iCol; /* 0==rowid, 1==tbl, 2==rank */ + int omit; /* True to omit this if found */ + int iConsIndex; /* Index in pInfo->aConstraint[] */ + } aConstraint[] = { + {SQLITE_INDEX_CONSTRAINT_MATCH|SQLITE_INDEX_CONSTRAINT_EQ, + FTS5_BI_MATCH, 1, 1, -1}, + {SQLITE_INDEX_CONSTRAINT_MATCH|SQLITE_INDEX_CONSTRAINT_EQ, + FTS5_BI_RANK, 2, 1, -1}, + {SQLITE_INDEX_CONSTRAINT_EQ, FTS5_BI_ROWID_EQ, 0, 0, -1}, + {SQLITE_INDEX_CONSTRAINT_LT|SQLITE_INDEX_CONSTRAINT_LE, + FTS5_BI_ROWID_LE, 0, 0, -1}, + {SQLITE_INDEX_CONSTRAINT_GT|SQLITE_INDEX_CONSTRAINT_GE, + FTS5_BI_ROWID_GE, 0, 0, -1}, + }; + + int aColMap[3]; + aColMap[0] = -1; + aColMap[1] = pConfig->nCol; + aColMap[2] = pConfig->nCol+1; + + /* Set idxFlags flags for all WHERE clause terms that will be used. */ + for(i=0; i<pInfo->nConstraint; i++){ + struct sqlite3_index_constraint *p = &pInfo->aConstraint[i]; + int j; + for(j=0; j<ArraySize(aConstraint); j++){ + struct Constraint *pC = &aConstraint[j]; + if( p->iColumn==aColMap[pC->iCol] && p->op & pC->op ){ + if( p->usable ){ + pC->iConsIndex = i; + idxFlags |= pC->fts5op; + }else if( j==0 ){ + /* As there exists an unusable MATCH constraint this is an + ** unusable plan. Set a prohibitively high cost. */ + pInfo->estimatedCost = 1e50; + return SQLITE_OK; + } + } + } + } + + /* Set idxFlags flags for the ORDER BY clause */ + if( pInfo->nOrderBy==1 ){ + int iSort = pInfo->aOrderBy[0].iColumn; + if( iSort==(pConfig->nCol+1) && BitFlagTest(idxFlags, FTS5_BI_MATCH) ){ + idxFlags |= FTS5_BI_ORDER_RANK; + }else if( iSort==-1 ){ + idxFlags |= FTS5_BI_ORDER_ROWID; + } + if( BitFlagTest(idxFlags, FTS5_BI_ORDER_RANK|FTS5_BI_ORDER_ROWID) ){ + pInfo->orderByConsumed = 1; + if( pInfo->aOrderBy[0].desc ){ + idxFlags |= FTS5_BI_ORDER_DESC; + } + } + } + + /* Calculate the estimated cost based on the flags set in idxFlags. */ + bHasMatch = BitFlagTest(idxFlags, FTS5_BI_MATCH); + if( BitFlagTest(idxFlags, FTS5_BI_ROWID_EQ) ){ + pInfo->estimatedCost = bHasMatch ? 100.0 : 10.0; + if( bHasMatch==0 ) fts5SetUniqueFlag(pInfo); + }else if( BitFlagAllTest(idxFlags, FTS5_BI_ROWID_LE|FTS5_BI_ROWID_GE) ){ + pInfo->estimatedCost = bHasMatch ? 500.0 : 250000.0; + }else if( BitFlagTest(idxFlags, FTS5_BI_ROWID_LE|FTS5_BI_ROWID_GE) ){ + pInfo->estimatedCost = bHasMatch ? 750.0 : 750000.0; + }else{ + pInfo->estimatedCost = bHasMatch ? 1000.0 : 1000000.0; + } + + /* Assign argvIndex values to each constraint in use. */ + iNext = 1; + for(i=0; i<ArraySize(aConstraint); i++){ + struct Constraint *pC = &aConstraint[i]; + if( pC->iConsIndex>=0 ){ + pInfo->aConstraintUsage[pC->iConsIndex].argvIndex = iNext++; + pInfo->aConstraintUsage[pC->iConsIndex].omit = (unsigned char)pC->omit; + } + } + + pInfo->idxNum = idxFlags; + return SQLITE_OK; +} + +/* +** Implementation of xOpen method. +*/ +static int fts5OpenMethod(sqlite3_vtab *pVTab, sqlite3_vtab_cursor **ppCsr){ + Fts5Table *pTab = (Fts5Table*)pVTab; + Fts5Config *pConfig = pTab->pConfig; + Fts5Cursor *pCsr; /* New cursor object */ + int nByte; /* Bytes of space to allocate */ + int rc = SQLITE_OK; /* Return code */ + + nByte = sizeof(Fts5Cursor) + pConfig->nCol * sizeof(int); + pCsr = (Fts5Cursor*)sqlite3_malloc(nByte); + if( pCsr ){ + Fts5Global *pGlobal = pTab->pGlobal; + memset(pCsr, 0, nByte); + pCsr->aColumnSize = (int*)&pCsr[1]; + pCsr->pNext = pGlobal->pCsr; + pGlobal->pCsr = pCsr; + pCsr->iCsrId = ++pGlobal->iNextId; + }else{ + rc = SQLITE_NOMEM; + } + *ppCsr = (sqlite3_vtab_cursor*)pCsr; + return rc; +} + +static int fts5StmtType(Fts5Cursor *pCsr){ + if( pCsr->ePlan==FTS5_PLAN_SCAN ){ + return (pCsr->bDesc) ? FTS5_STMT_SCAN_DESC : FTS5_STMT_SCAN_ASC; + } + return FTS5_STMT_LOOKUP; +} + +/* +** This function is called after the cursor passed as the only argument +** is moved to point at a different row. It clears all cached data +** specific to the previous row stored by the cursor object. +*/ +static void fts5CsrNewrow(Fts5Cursor *pCsr){ + CsrFlagSet(pCsr, + FTS5CSR_REQUIRE_CONTENT + | FTS5CSR_REQUIRE_DOCSIZE + | FTS5CSR_REQUIRE_INST + | FTS5CSR_REQUIRE_POSLIST + ); +} + +static void fts5FreeCursorComponents(Fts5Cursor *pCsr){ + Fts5Table *pTab = (Fts5Table*)(pCsr->base.pVtab); + Fts5Auxdata *pData; + Fts5Auxdata *pNext; + + sqlite3_free(pCsr->aInstIter); + sqlite3_free(pCsr->aInst); + if( pCsr->pStmt ){ + int eStmt = fts5StmtType(pCsr); + sqlite3Fts5StorageStmtRelease(pTab->pStorage, eStmt, pCsr->pStmt); + } + if( pCsr->pSorter ){ + Fts5Sorter *pSorter = pCsr->pSorter; + sqlite3_finalize(pSorter->pStmt); + sqlite3_free(pSorter); + } + + if( pCsr->ePlan!=FTS5_PLAN_SOURCE ){ + sqlite3Fts5ExprFree(pCsr->pExpr); + } + + for(pData=pCsr->pAuxdata; pData; pData=pNext){ + pNext = pData->pNext; + if( pData->xDelete ) pData->xDelete(pData->pPtr); + sqlite3_free(pData); + } + + sqlite3_finalize(pCsr->pRankArgStmt); + sqlite3_free(pCsr->apRankArg); + + if( CsrFlagTest(pCsr, FTS5CSR_FREE_ZRANK) ){ + sqlite3_free(pCsr->zRank); + sqlite3_free(pCsr->zRankArgs); + } + + memset(&pCsr->ePlan, 0, sizeof(Fts5Cursor) - ((u8*)&pCsr->ePlan - (u8*)pCsr)); +} + + +/* +** Close the cursor. For additional information see the documentation +** on the xClose method of the virtual table interface. +*/ +static int fts5CloseMethod(sqlite3_vtab_cursor *pCursor){ + if( pCursor ){ + Fts5Table *pTab = (Fts5Table*)(pCursor->pVtab); + Fts5Cursor *pCsr = (Fts5Cursor*)pCursor; + Fts5Cursor **pp; + + fts5FreeCursorComponents(pCsr); + /* Remove the cursor from the Fts5Global.pCsr list */ + for(pp=&pTab->pGlobal->pCsr; (*pp)!=pCsr; pp=&(*pp)->pNext); + *pp = pCsr->pNext; + + sqlite3_free(pCsr); + } + return SQLITE_OK; +} + +static int fts5SorterNext(Fts5Cursor *pCsr){ + Fts5Sorter *pSorter = pCsr->pSorter; + int rc; + + rc = sqlite3_step(pSorter->pStmt); + if( rc==SQLITE_DONE ){ + rc = SQLITE_OK; + CsrFlagSet(pCsr, FTS5CSR_EOF); + }else if( rc==SQLITE_ROW ){ + const u8 *a; + const u8 *aBlob; + int nBlob; + int i; + int iOff = 0; + rc = SQLITE_OK; + + pSorter->iRowid = sqlite3_column_int64(pSorter->pStmt, 0); + nBlob = sqlite3_column_bytes(pSorter->pStmt, 1); + aBlob = a = sqlite3_column_blob(pSorter->pStmt, 1); + + /* nBlob==0 in detail=none mode. */ + if( nBlob>0 ){ + for(i=0; i<(pSorter->nIdx-1); i++){ + int iVal; + a += fts5GetVarint32(a, iVal); + iOff += iVal; + pSorter->aIdx[i] = iOff; + } + pSorter->aIdx[i] = &aBlob[nBlob] - a; + pSorter->aPoslist = a; + } + + fts5CsrNewrow(pCsr); + } + + return rc; +} + + +/* +** Set the FTS5CSR_REQUIRE_RESEEK flag on all FTS5_PLAN_MATCH cursors +** open on table pTab. +*/ +static void fts5TripCursors(Fts5Table *pTab){ + Fts5Cursor *pCsr; + for(pCsr=pTab->pGlobal->pCsr; pCsr; pCsr=pCsr->pNext){ + if( pCsr->ePlan==FTS5_PLAN_MATCH + && pCsr->base.pVtab==(sqlite3_vtab*)pTab + ){ + CsrFlagSet(pCsr, FTS5CSR_REQUIRE_RESEEK); + } + } +} + +/* +** If the REQUIRE_RESEEK flag is set on the cursor passed as the first +** argument, close and reopen all Fts5IndexIter iterators that the cursor +** is using. Then attempt to move the cursor to a rowid equal to or laster +** (in the cursors sort order - ASC or DESC) than the current rowid. +** +** If the new rowid is not equal to the old, set output parameter *pbSkip +** to 1 before returning. Otherwise, leave it unchanged. +** +** Return SQLITE_OK if successful or if no reseek was required, or an +** error code if an error occurred. +*/ +static int fts5CursorReseek(Fts5Cursor *pCsr, int *pbSkip){ + int rc = SQLITE_OK; + assert( *pbSkip==0 ); + if( CsrFlagTest(pCsr, FTS5CSR_REQUIRE_RESEEK) ){ + Fts5Table *pTab = (Fts5Table*)(pCsr->base.pVtab); + int bDesc = pCsr->bDesc; + i64 iRowid = sqlite3Fts5ExprRowid(pCsr->pExpr); + + rc = sqlite3Fts5ExprFirst(pCsr->pExpr, pTab->pIndex, iRowid, bDesc); + if( rc==SQLITE_OK && iRowid!=sqlite3Fts5ExprRowid(pCsr->pExpr) ){ + *pbSkip = 1; + } + + CsrFlagClear(pCsr, FTS5CSR_REQUIRE_RESEEK); + fts5CsrNewrow(pCsr); + if( sqlite3Fts5ExprEof(pCsr->pExpr) ){ + CsrFlagSet(pCsr, FTS5CSR_EOF); + *pbSkip = 1; + } + } + return rc; +} + + +/* +** Advance the cursor to the next row in the table that matches the +** search criteria. +** +** Return SQLITE_OK if nothing goes wrong. SQLITE_OK is returned +** even if we reach end-of-file. The fts5EofMethod() will be called +** subsequently to determine whether or not an EOF was hit. +*/ +static int fts5NextMethod(sqlite3_vtab_cursor *pCursor){ + Fts5Cursor *pCsr = (Fts5Cursor*)pCursor; + int rc; + + assert( (pCsr->ePlan<3)== + (pCsr->ePlan==FTS5_PLAN_MATCH || pCsr->ePlan==FTS5_PLAN_SOURCE) + ); + assert( !CsrFlagTest(pCsr, FTS5CSR_EOF) ); + + if( pCsr->ePlan<3 ){ + int bSkip = 0; + if( (rc = fts5CursorReseek(pCsr, &bSkip)) || bSkip ) return rc; + rc = sqlite3Fts5ExprNext(pCsr->pExpr, pCsr->iLastRowid); + CsrFlagSet(pCsr, sqlite3Fts5ExprEof(pCsr->pExpr)); + fts5CsrNewrow(pCsr); + }else{ + switch( pCsr->ePlan ){ + case FTS5_PLAN_SPECIAL: { + CsrFlagSet(pCsr, FTS5CSR_EOF); + rc = SQLITE_OK; + break; + } + + case FTS5_PLAN_SORTED_MATCH: { + rc = fts5SorterNext(pCsr); + break; + } + + default: + rc = sqlite3_step(pCsr->pStmt); + if( rc!=SQLITE_ROW ){ + CsrFlagSet(pCsr, FTS5CSR_EOF); + rc = sqlite3_reset(pCsr->pStmt); + }else{ + rc = SQLITE_OK; + } + break; + } + } + + return rc; +} + + +static int fts5PrepareStatement( + sqlite3_stmt **ppStmt, + Fts5Config *pConfig, + const char *zFmt, + ... +){ + sqlite3_stmt *pRet = 0; + int rc; + char *zSql; + va_list ap; + + va_start(ap, zFmt); + zSql = sqlite3_vmprintf(zFmt, ap); + if( zSql==0 ){ + rc = SQLITE_NOMEM; + }else{ + rc = sqlite3_prepare_v2(pConfig->db, zSql, -1, &pRet, 0); + if( rc!=SQLITE_OK ){ + *pConfig->pzErrmsg = sqlite3_mprintf("%s", sqlite3_errmsg(pConfig->db)); + } + sqlite3_free(zSql); + } + + va_end(ap); + *ppStmt = pRet; + return rc; +} + +static int fts5CursorFirstSorted(Fts5Table *pTab, Fts5Cursor *pCsr, int bDesc){ + Fts5Config *pConfig = pTab->pConfig; + Fts5Sorter *pSorter; + int nPhrase; + int nByte; + int rc; + const char *zRank = pCsr->zRank; + const char *zRankArgs = pCsr->zRankArgs; + + nPhrase = sqlite3Fts5ExprPhraseCount(pCsr->pExpr); + nByte = sizeof(Fts5Sorter) + sizeof(int) * (nPhrase-1); + pSorter = (Fts5Sorter*)sqlite3_malloc(nByte); + if( pSorter==0 ) return SQLITE_NOMEM; + memset(pSorter, 0, nByte); + pSorter->nIdx = nPhrase; + + /* TODO: It would be better to have some system for reusing statement + ** handles here, rather than preparing a new one for each query. But that + ** is not possible as SQLite reference counts the virtual table objects. + ** And since the statement required here reads from this very virtual + ** table, saving it creates a circular reference. + ** + ** If SQLite a built-in statement cache, this wouldn't be a problem. */ + rc = fts5PrepareStatement(&pSorter->pStmt, pConfig, + "SELECT rowid, rank FROM %Q.%Q ORDER BY %s(%s%s%s) %s", + pConfig->zDb, pConfig->zName, zRank, pConfig->zName, + (zRankArgs ? ", " : ""), + (zRankArgs ? zRankArgs : ""), + bDesc ? "DESC" : "ASC" + ); + + pCsr->pSorter = pSorter; + if( rc==SQLITE_OK ){ + assert( pTab->pSortCsr==0 ); + pTab->pSortCsr = pCsr; + rc = fts5SorterNext(pCsr); + pTab->pSortCsr = 0; + } + + if( rc!=SQLITE_OK ){ + sqlite3_finalize(pSorter->pStmt); + sqlite3_free(pSorter); + pCsr->pSorter = 0; + } + + return rc; +} + +static int fts5CursorFirst(Fts5Table *pTab, Fts5Cursor *pCsr, int bDesc){ + int rc; + Fts5Expr *pExpr = pCsr->pExpr; + rc = sqlite3Fts5ExprFirst(pExpr, pTab->pIndex, pCsr->iFirstRowid, bDesc); + if( sqlite3Fts5ExprEof(pExpr) ){ + CsrFlagSet(pCsr, FTS5CSR_EOF); + } + fts5CsrNewrow(pCsr); + return rc; +} + +/* +** Process a "special" query. A special query is identified as one with a +** MATCH expression that begins with a '*' character. The remainder of +** the text passed to the MATCH operator are used as the special query +** parameters. +*/ +static int fts5SpecialMatch( + Fts5Table *pTab, + Fts5Cursor *pCsr, + const char *zQuery +){ + int rc = SQLITE_OK; /* Return code */ + const char *z = zQuery; /* Special query text */ + int n; /* Number of bytes in text at z */ + + while( z[0]==' ' ) z++; + for(n=0; z[n] && z[n]!=' '; n++); + + assert( pTab->base.zErrMsg==0 ); + pCsr->ePlan = FTS5_PLAN_SPECIAL; + + if( 0==sqlite3_strnicmp("reads", z, n) ){ + pCsr->iSpecial = sqlite3Fts5IndexReads(pTab->pIndex); + } + else if( 0==sqlite3_strnicmp("id", z, n) ){ + pCsr->iSpecial = pCsr->iCsrId; + } + else{ + /* An unrecognized directive. Return an error message. */ + pTab->base.zErrMsg = sqlite3_mprintf("unknown special query: %.*s", n, z); + rc = SQLITE_ERROR; + } + + return rc; +} + +/* +** Search for an auxiliary function named zName that can be used with table +** pTab. If one is found, return a pointer to the corresponding Fts5Auxiliary +** structure. Otherwise, if no such function exists, return NULL. +*/ +static Fts5Auxiliary *fts5FindAuxiliary(Fts5Table *pTab, const char *zName){ + Fts5Auxiliary *pAux; + + for(pAux=pTab->pGlobal->pAux; pAux; pAux=pAux->pNext){ + if( sqlite3_stricmp(zName, pAux->zFunc)==0 ) return pAux; + } + + /* No function of the specified name was found. Return 0. */ + return 0; +} + + +static int fts5FindRankFunction(Fts5Cursor *pCsr){ + Fts5Table *pTab = (Fts5Table*)(pCsr->base.pVtab); + Fts5Config *pConfig = pTab->pConfig; + int rc = SQLITE_OK; + Fts5Auxiliary *pAux = 0; + const char *zRank = pCsr->zRank; + const char *zRankArgs = pCsr->zRankArgs; + + if( zRankArgs ){ + char *zSql = sqlite3Fts5Mprintf(&rc, "SELECT %s", zRankArgs); + if( zSql ){ + sqlite3_stmt *pStmt = 0; + rc = sqlite3_prepare_v2(pConfig->db, zSql, -1, &pStmt, 0); + sqlite3_free(zSql); + assert( rc==SQLITE_OK || pCsr->pRankArgStmt==0 ); + if( rc==SQLITE_OK ){ + if( SQLITE_ROW==sqlite3_step(pStmt) ){ + int nByte; + pCsr->nRankArg = sqlite3_column_count(pStmt); + nByte = sizeof(sqlite3_value*)*pCsr->nRankArg; + pCsr->apRankArg = (sqlite3_value**)sqlite3Fts5MallocZero(&rc, nByte); + if( rc==SQLITE_OK ){ + int i; + for(i=0; i<pCsr->nRankArg; i++){ + pCsr->apRankArg[i] = sqlite3_column_value(pStmt, i); + } + } + pCsr->pRankArgStmt = pStmt; + }else{ + rc = sqlite3_finalize(pStmt); + assert( rc!=SQLITE_OK ); + } + } + } + } + + if( rc==SQLITE_OK ){ + pAux = fts5FindAuxiliary(pTab, zRank); + if( pAux==0 ){ + assert( pTab->base.zErrMsg==0 ); + pTab->base.zErrMsg = sqlite3_mprintf("no such function: %s", zRank); + rc = SQLITE_ERROR; + } + } + + pCsr->pRank = pAux; + return rc; +} + + +static int fts5CursorParseRank( + Fts5Config *pConfig, + Fts5Cursor *pCsr, + sqlite3_value *pRank +){ + int rc = SQLITE_OK; + if( pRank ){ + const char *z = (const char*)sqlite3_value_text(pRank); + char *zRank = 0; + char *zRankArgs = 0; + + if( z==0 ){ + if( sqlite3_value_type(pRank)==SQLITE_NULL ) rc = SQLITE_ERROR; + }else{ + rc = sqlite3Fts5ConfigParseRank(z, &zRank, &zRankArgs); + } + if( rc==SQLITE_OK ){ + pCsr->zRank = zRank; + pCsr->zRankArgs = zRankArgs; + CsrFlagSet(pCsr, FTS5CSR_FREE_ZRANK); + }else if( rc==SQLITE_ERROR ){ + pCsr->base.pVtab->zErrMsg = sqlite3_mprintf( + "parse error in rank function: %s", z + ); + } + }else{ + if( pConfig->zRank ){ + pCsr->zRank = (char*)pConfig->zRank; + pCsr->zRankArgs = (char*)pConfig->zRankArgs; + }else{ + pCsr->zRank = (char*)FTS5_DEFAULT_RANK; + pCsr->zRankArgs = 0; + } + } + return rc; +} + +static i64 fts5GetRowidLimit(sqlite3_value *pVal, i64 iDefault){ + if( pVal ){ + int eType = sqlite3_value_numeric_type(pVal); + if( eType==SQLITE_INTEGER ){ + return sqlite3_value_int64(pVal); + } + } + return iDefault; +} + +/* +** This is the xFilter interface for the virtual table. See +** the virtual table xFilter method documentation for additional +** information. +** +** There are three possible query strategies: +** +** 1. Full-text search using a MATCH operator. +** 2. A by-rowid lookup. +** 3. A full-table scan. +*/ +static int fts5FilterMethod( + sqlite3_vtab_cursor *pCursor, /* The cursor used for this query */ + int idxNum, /* Strategy index */ + const char *zUnused, /* Unused */ + int nVal, /* Number of elements in apVal */ + sqlite3_value **apVal /* Arguments for the indexing scheme */ +){ + Fts5Table *pTab = (Fts5Table*)(pCursor->pVtab); + Fts5Config *pConfig = pTab->pConfig; + Fts5Cursor *pCsr = (Fts5Cursor*)pCursor; + int rc = SQLITE_OK; /* Error code */ + int iVal = 0; /* Counter for apVal[] */ + int bDesc; /* True if ORDER BY [rank|rowid] DESC */ + int bOrderByRank; /* True if ORDER BY rank */ + sqlite3_value *pMatch = 0; /* <tbl> MATCH ? expression (or NULL) */ + sqlite3_value *pRank = 0; /* rank MATCH ? expression (or NULL) */ + sqlite3_value *pRowidEq = 0; /* rowid = ? expression (or NULL) */ + sqlite3_value *pRowidLe = 0; /* rowid <= ? expression (or NULL) */ + sqlite3_value *pRowidGe = 0; /* rowid >= ? expression (or NULL) */ + char **pzErrmsg = pConfig->pzErrmsg; + + UNUSED_PARAM(zUnused); + UNUSED_PARAM(nVal); + + if( pCsr->ePlan ){ + fts5FreeCursorComponents(pCsr); + memset(&pCsr->ePlan, 0, sizeof(Fts5Cursor) - ((u8*)&pCsr->ePlan-(u8*)pCsr)); + } + + assert( pCsr->pStmt==0 ); + assert( pCsr->pExpr==0 ); + assert( pCsr->csrflags==0 ); + assert( pCsr->pRank==0 ); + assert( pCsr->zRank==0 ); + assert( pCsr->zRankArgs==0 ); + + assert( pzErrmsg==0 || pzErrmsg==&pTab->base.zErrMsg ); + pConfig->pzErrmsg = &pTab->base.zErrMsg; + + /* Decode the arguments passed through to this function. + ** + ** Note: The following set of if(...) statements must be in the same + ** order as the corresponding entries in the struct at the top of + ** fts5BestIndexMethod(). */ + if( BitFlagTest(idxNum, FTS5_BI_MATCH) ) pMatch = apVal[iVal++]; + if( BitFlagTest(idxNum, FTS5_BI_RANK) ) pRank = apVal[iVal++]; + if( BitFlagTest(idxNum, FTS5_BI_ROWID_EQ) ) pRowidEq = apVal[iVal++]; + if( BitFlagTest(idxNum, FTS5_BI_ROWID_LE) ) pRowidLe = apVal[iVal++]; + if( BitFlagTest(idxNum, FTS5_BI_ROWID_GE) ) pRowidGe = apVal[iVal++]; + assert( iVal==nVal ); + bOrderByRank = ((idxNum & FTS5_BI_ORDER_RANK) ? 1 : 0); + pCsr->bDesc = bDesc = ((idxNum & FTS5_BI_ORDER_DESC) ? 1 : 0); + + /* Set the cursor upper and lower rowid limits. Only some strategies + ** actually use them. This is ok, as the xBestIndex() method leaves the + ** sqlite3_index_constraint.omit flag clear for range constraints + ** on the rowid field. */ + if( pRowidEq ){ + pRowidLe = pRowidGe = pRowidEq; + } + if( bDesc ){ + pCsr->iFirstRowid = fts5GetRowidLimit(pRowidLe, LARGEST_INT64); + pCsr->iLastRowid = fts5GetRowidLimit(pRowidGe, SMALLEST_INT64); + }else{ + pCsr->iLastRowid = fts5GetRowidLimit(pRowidLe, LARGEST_INT64); + pCsr->iFirstRowid = fts5GetRowidLimit(pRowidGe, SMALLEST_INT64); + } + + if( pTab->pSortCsr ){ + /* If pSortCsr is non-NULL, then this call is being made as part of + ** processing for a "... MATCH <expr> ORDER BY rank" query (ePlan is + ** set to FTS5_PLAN_SORTED_MATCH). pSortCsr is the cursor that will + ** return results to the user for this query. The current cursor + ** (pCursor) is used to execute the query issued by function + ** fts5CursorFirstSorted() above. */ + assert( pRowidEq==0 && pRowidLe==0 && pRowidGe==0 && pRank==0 ); + assert( nVal==0 && pMatch==0 && bOrderByRank==0 && bDesc==0 ); + assert( pCsr->iLastRowid==LARGEST_INT64 ); + assert( pCsr->iFirstRowid==SMALLEST_INT64 ); + pCsr->ePlan = FTS5_PLAN_SOURCE; + pCsr->pExpr = pTab->pSortCsr->pExpr; + rc = fts5CursorFirst(pTab, pCsr, bDesc); + sqlite3Fts5ExprClearEof(pCsr->pExpr); + }else if( pMatch ){ + const char *zExpr = (const char*)sqlite3_value_text(apVal[0]); + if( zExpr==0 ) zExpr = ""; + + rc = fts5CursorParseRank(pConfig, pCsr, pRank); + if( rc==SQLITE_OK ){ + if( zExpr[0]=='*' ){ + /* The user has issued a query of the form "MATCH '*...'". This + ** indicates that the MATCH expression is not a full text query, + ** but a request for an internal parameter. */ + rc = fts5SpecialMatch(pTab, pCsr, &zExpr[1]); + }else{ + char **pzErr = &pTab->base.zErrMsg; + rc = sqlite3Fts5ExprNew(pConfig, zExpr, &pCsr->pExpr, pzErr); + if( rc==SQLITE_OK ){ + if( bOrderByRank ){ + pCsr->ePlan = FTS5_PLAN_SORTED_MATCH; + rc = fts5CursorFirstSorted(pTab, pCsr, bDesc); + }else{ + pCsr->ePlan = FTS5_PLAN_MATCH; + rc = fts5CursorFirst(pTab, pCsr, bDesc); + } + } + } + } + }else if( pConfig->zContent==0 ){ + *pConfig->pzErrmsg = sqlite3_mprintf( + "%s: table does not support scanning", pConfig->zName + ); + rc = SQLITE_ERROR; + }else{ + /* This is either a full-table scan (ePlan==FTS5_PLAN_SCAN) or a lookup + ** by rowid (ePlan==FTS5_PLAN_ROWID). */ + pCsr->ePlan = (pRowidEq ? FTS5_PLAN_ROWID : FTS5_PLAN_SCAN); + rc = sqlite3Fts5StorageStmt( + pTab->pStorage, fts5StmtType(pCsr), &pCsr->pStmt, &pTab->base.zErrMsg + ); + if( rc==SQLITE_OK ){ + if( pCsr->ePlan==FTS5_PLAN_ROWID ){ + sqlite3_bind_value(pCsr->pStmt, 1, apVal[0]); + }else{ + sqlite3_bind_int64(pCsr->pStmt, 1, pCsr->iFirstRowid); + sqlite3_bind_int64(pCsr->pStmt, 2, pCsr->iLastRowid); + } + rc = fts5NextMethod(pCursor); + } + } + + pConfig->pzErrmsg = pzErrmsg; + return rc; +} + +/* +** This is the xEof method of the virtual table. SQLite calls this +** routine to find out if it has reached the end of a result set. +*/ +static int fts5EofMethod(sqlite3_vtab_cursor *pCursor){ + Fts5Cursor *pCsr = (Fts5Cursor*)pCursor; + return (CsrFlagTest(pCsr, FTS5CSR_EOF) ? 1 : 0); +} + +/* +** Return the rowid that the cursor currently points to. +*/ +static i64 fts5CursorRowid(Fts5Cursor *pCsr){ + assert( pCsr->ePlan==FTS5_PLAN_MATCH + || pCsr->ePlan==FTS5_PLAN_SORTED_MATCH + || pCsr->ePlan==FTS5_PLAN_SOURCE + ); + if( pCsr->pSorter ){ + return pCsr->pSorter->iRowid; + }else{ + return sqlite3Fts5ExprRowid(pCsr->pExpr); + } +} + +/* +** This is the xRowid method. The SQLite core calls this routine to +** retrieve the rowid for the current row of the result set. fts5 +** exposes %_content.rowid as the rowid for the virtual table. The +** rowid should be written to *pRowid. +*/ +static int fts5RowidMethod(sqlite3_vtab_cursor *pCursor, sqlite_int64 *pRowid){ + Fts5Cursor *pCsr = (Fts5Cursor*)pCursor; + int ePlan = pCsr->ePlan; + + assert( CsrFlagTest(pCsr, FTS5CSR_EOF)==0 ); + switch( ePlan ){ + case FTS5_PLAN_SPECIAL: + *pRowid = 0; + break; + + case FTS5_PLAN_SOURCE: + case FTS5_PLAN_MATCH: + case FTS5_PLAN_SORTED_MATCH: + *pRowid = fts5CursorRowid(pCsr); + break; + + default: + *pRowid = sqlite3_column_int64(pCsr->pStmt, 0); + break; + } + + return SQLITE_OK; +} + +/* +** If the cursor requires seeking (bSeekRequired flag is set), seek it. +** Return SQLITE_OK if no error occurs, or an SQLite error code otherwise. +** +** If argument bErrormsg is true and an error occurs, an error message may +** be left in sqlite3_vtab.zErrMsg. +*/ +static int fts5SeekCursor(Fts5Cursor *pCsr, int bErrormsg){ + int rc = SQLITE_OK; + + /* If the cursor does not yet have a statement handle, obtain one now. */ + if( pCsr->pStmt==0 ){ + Fts5Table *pTab = (Fts5Table*)(pCsr->base.pVtab); + int eStmt = fts5StmtType(pCsr); + rc = sqlite3Fts5StorageStmt( + pTab->pStorage, eStmt, &pCsr->pStmt, (bErrormsg?&pTab->base.zErrMsg:0) + ); + assert( rc!=SQLITE_OK || pTab->base.zErrMsg==0 ); + assert( CsrFlagTest(pCsr, FTS5CSR_REQUIRE_CONTENT) ); + } + + if( rc==SQLITE_OK && CsrFlagTest(pCsr, FTS5CSR_REQUIRE_CONTENT) ){ + assert( pCsr->pExpr ); + sqlite3_reset(pCsr->pStmt); + sqlite3_bind_int64(pCsr->pStmt, 1, fts5CursorRowid(pCsr)); + rc = sqlite3_step(pCsr->pStmt); + if( rc==SQLITE_ROW ){ + rc = SQLITE_OK; + CsrFlagClear(pCsr, FTS5CSR_REQUIRE_CONTENT); + }else{ + rc = sqlite3_reset(pCsr->pStmt); + if( rc==SQLITE_OK ){ + rc = FTS5_CORRUPT; + } + } + } + return rc; +} + +static void fts5SetVtabError(Fts5Table *p, const char *zFormat, ...){ + va_list ap; /* ... printf arguments */ + va_start(ap, zFormat); + assert( p->base.zErrMsg==0 ); + p->base.zErrMsg = sqlite3_vmprintf(zFormat, ap); + va_end(ap); +} + +/* +** This function is called to handle an FTS INSERT command. In other words, +** an INSERT statement of the form: +** +** INSERT INTO fts(fts) VALUES($pCmd) +** INSERT INTO fts(fts, rank) VALUES($pCmd, $pVal) +** +** Argument pVal is the value assigned to column "fts" by the INSERT +** statement. This function returns SQLITE_OK if successful, or an SQLite +** error code if an error occurs. +** +** The commands implemented by this function are documented in the "Special +** INSERT Directives" section of the documentation. It should be updated if +** more commands are added to this function. +*/ +static int fts5SpecialInsert( + Fts5Table *pTab, /* Fts5 table object */ + const char *zCmd, /* Text inserted into table-name column */ + sqlite3_value *pVal /* Value inserted into rank column */ +){ + Fts5Config *pConfig = pTab->pConfig; + int rc = SQLITE_OK; + int bError = 0; + + if( 0==sqlite3_stricmp("delete-all", zCmd) ){ + if( pConfig->eContent==FTS5_CONTENT_NORMAL ){ + fts5SetVtabError(pTab, + "'delete-all' may only be used with a " + "contentless or external content fts5 table" + ); + rc = SQLITE_ERROR; + }else{ + rc = sqlite3Fts5StorageDeleteAll(pTab->pStorage); + } + }else if( 0==sqlite3_stricmp("rebuild", zCmd) ){ + if( pConfig->eContent==FTS5_CONTENT_NONE ){ + fts5SetVtabError(pTab, + "'rebuild' may not be used with a contentless fts5 table" + ); + rc = SQLITE_ERROR; + }else{ + rc = sqlite3Fts5StorageRebuild(pTab->pStorage); + } + }else if( 0==sqlite3_stricmp("optimize", zCmd) ){ + rc = sqlite3Fts5StorageOptimize(pTab->pStorage); + }else if( 0==sqlite3_stricmp("merge", zCmd) ){ + int nMerge = sqlite3_value_int(pVal); + rc = sqlite3Fts5StorageMerge(pTab->pStorage, nMerge); + }else if( 0==sqlite3_stricmp("integrity-check", zCmd) ){ + rc = sqlite3Fts5StorageIntegrity(pTab->pStorage); +#ifdef SQLITE_DEBUG + }else if( 0==sqlite3_stricmp("prefix-index", zCmd) ){ + pConfig->bPrefixIndex = sqlite3_value_int(pVal); +#endif + }else{ + rc = sqlite3Fts5IndexLoadConfig(pTab->pIndex); + if( rc==SQLITE_OK ){ + rc = sqlite3Fts5ConfigSetValue(pTab->pConfig, zCmd, pVal, &bError); + } + if( rc==SQLITE_OK ){ + if( bError ){ + rc = SQLITE_ERROR; + }else{ + rc = sqlite3Fts5StorageConfigValue(pTab->pStorage, zCmd, pVal, 0); + } + } + } + return rc; +} + +static int fts5SpecialDelete( + Fts5Table *pTab, + sqlite3_value **apVal +){ + int rc = SQLITE_OK; + int eType1 = sqlite3_value_type(apVal[1]); + if( eType1==SQLITE_INTEGER ){ + sqlite3_int64 iDel = sqlite3_value_int64(apVal[1]); + rc = sqlite3Fts5StorageDelete(pTab->pStorage, iDel, &apVal[2]); + } + return rc; +} + +static void fts5StorageInsert( + int *pRc, + Fts5Table *pTab, + sqlite3_value **apVal, + i64 *piRowid +){ + int rc = *pRc; + if( rc==SQLITE_OK ){ + rc = sqlite3Fts5StorageContentInsert(pTab->pStorage, apVal, piRowid); + } + if( rc==SQLITE_OK ){ + rc = sqlite3Fts5StorageIndexInsert(pTab->pStorage, apVal, *piRowid); + } + *pRc = rc; +} + +/* +** This function is the implementation of the xUpdate callback used by +** FTS3 virtual tables. It is invoked by SQLite each time a row is to be +** inserted, updated or deleted. +** +** A delete specifies a single argument - the rowid of the row to remove. +** +** Update and insert operations pass: +** +** 1. The "old" rowid, or NULL. +** 2. The "new" rowid. +** 3. Values for each of the nCol matchable columns. +** 4. Values for the two hidden columns (<tablename> and "rank"). +*/ +static int fts5UpdateMethod( + sqlite3_vtab *pVtab, /* Virtual table handle */ + int nArg, /* Size of argument array */ + sqlite3_value **apVal, /* Array of arguments */ + sqlite_int64 *pRowid /* OUT: The affected (or effected) rowid */ +){ + Fts5Table *pTab = (Fts5Table*)pVtab; + Fts5Config *pConfig = pTab->pConfig; + int eType0; /* value_type() of apVal[0] */ + int rc = SQLITE_OK; /* Return code */ + + /* A transaction must be open when this is called. */ + assert( pTab->ts.eState==1 ); + + assert( pVtab->zErrMsg==0 ); + assert( nArg==1 || nArg==(2+pConfig->nCol+2) ); + assert( nArg==1 + || sqlite3_value_type(apVal[1])==SQLITE_INTEGER + || sqlite3_value_type(apVal[1])==SQLITE_NULL + ); + assert( pTab->pConfig->pzErrmsg==0 ); + pTab->pConfig->pzErrmsg = &pTab->base.zErrMsg; + + /* Put any active cursors into REQUIRE_SEEK state. */ + fts5TripCursors(pTab); + + eType0 = sqlite3_value_type(apVal[0]); + if( eType0==SQLITE_NULL + && sqlite3_value_type(apVal[2+pConfig->nCol])!=SQLITE_NULL + ){ + /* A "special" INSERT op. These are handled separately. */ + const char *z = (const char*)sqlite3_value_text(apVal[2+pConfig->nCol]); + if( pConfig->eContent!=FTS5_CONTENT_NORMAL + && 0==sqlite3_stricmp("delete", z) + ){ + rc = fts5SpecialDelete(pTab, apVal); + }else{ + rc = fts5SpecialInsert(pTab, z, apVal[2 + pConfig->nCol + 1]); + } + }else{ + /* A regular INSERT, UPDATE or DELETE statement. The trick here is that + ** any conflict on the rowid value must be detected before any + ** modifications are made to the database file. There are 4 cases: + ** + ** 1) DELETE + ** 2) UPDATE (rowid not modified) + ** 3) UPDATE (rowid modified) + ** 4) INSERT + ** + ** Cases 3 and 4 may violate the rowid constraint. + */ + int eConflict = SQLITE_ABORT; + if( pConfig->eContent==FTS5_CONTENT_NORMAL ){ + eConflict = sqlite3_vtab_on_conflict(pConfig->db); + } + + assert( eType0==SQLITE_INTEGER || eType0==SQLITE_NULL ); + assert( nArg!=1 || eType0==SQLITE_INTEGER ); + + /* Filter out attempts to run UPDATE or DELETE on contentless tables. + ** This is not suported. */ + if( eType0==SQLITE_INTEGER && fts5IsContentless(pTab) ){ + pTab->base.zErrMsg = sqlite3_mprintf( + "cannot %s contentless fts5 table: %s", + (nArg>1 ? "UPDATE" : "DELETE from"), pConfig->zName + ); + rc = SQLITE_ERROR; + } + + /* Case 1: DELETE */ + else if( nArg==1 ){ + i64 iDel = sqlite3_value_int64(apVal[0]); /* Rowid to delete */ + rc = sqlite3Fts5StorageDelete(pTab->pStorage, iDel, 0); + } + + /* Case 2: INSERT */ + else if( eType0!=SQLITE_INTEGER ){ + /* If this is a REPLACE, first remove the current entry (if any) */ + if( eConflict==SQLITE_REPLACE + && sqlite3_value_type(apVal[1])==SQLITE_INTEGER + ){ + i64 iNew = sqlite3_value_int64(apVal[1]); /* Rowid to delete */ + rc = sqlite3Fts5StorageDelete(pTab->pStorage, iNew, 0); + } + fts5StorageInsert(&rc, pTab, apVal, pRowid); + } + + /* Case 2: UPDATE */ + else{ + i64 iOld = sqlite3_value_int64(apVal[0]); /* Old rowid */ + i64 iNew = sqlite3_value_int64(apVal[1]); /* New rowid */ + if( iOld!=iNew ){ + if( eConflict==SQLITE_REPLACE ){ + rc = sqlite3Fts5StorageDelete(pTab->pStorage, iOld, 0); + if( rc==SQLITE_OK ){ + rc = sqlite3Fts5StorageDelete(pTab->pStorage, iNew, 0); + } + fts5StorageInsert(&rc, pTab, apVal, pRowid); + }else{ + rc = sqlite3Fts5StorageContentInsert(pTab->pStorage, apVal, pRowid); + if( rc==SQLITE_OK ){ + rc = sqlite3Fts5StorageDelete(pTab->pStorage, iOld, 0); + } + if( rc==SQLITE_OK ){ + rc = sqlite3Fts5StorageIndexInsert(pTab->pStorage, apVal, *pRowid); + } + } + }else{ + rc = sqlite3Fts5StorageDelete(pTab->pStorage, iOld, 0); + fts5StorageInsert(&rc, pTab, apVal, pRowid); + } + } + } + + pTab->pConfig->pzErrmsg = 0; + return rc; +} + +/* +** Implementation of xSync() method. +*/ +static int fts5SyncMethod(sqlite3_vtab *pVtab){ + int rc; + Fts5Table *pTab = (Fts5Table*)pVtab; + fts5CheckTransactionState(pTab, FTS5_SYNC, 0); + pTab->pConfig->pzErrmsg = &pTab->base.zErrMsg; + fts5TripCursors(pTab); + rc = sqlite3Fts5StorageSync(pTab->pStorage, 1); + pTab->pConfig->pzErrmsg = 0; + return rc; +} + +/* +** Implementation of xBegin() method. +*/ +static int fts5BeginMethod(sqlite3_vtab *pVtab){ + UNUSED_PARAM(pVtab); /* Call below is a no-op for NDEBUG builds */ + fts5CheckTransactionState((Fts5Table*)pVtab, FTS5_BEGIN, 0); + return SQLITE_OK; +} + +/* +** Implementation of xCommit() method. This is a no-op. The contents of +** the pending-terms hash-table have already been flushed into the database +** by fts5SyncMethod(). +*/ +static int fts5CommitMethod(sqlite3_vtab *pVtab){ + UNUSED_PARAM(pVtab); /* Call below is a no-op for NDEBUG builds */ + fts5CheckTransactionState((Fts5Table*)pVtab, FTS5_COMMIT, 0); + return SQLITE_OK; +} + +/* +** Implementation of xRollback(). Discard the contents of the pending-terms +** hash-table. Any changes made to the database are reverted by SQLite. +*/ +static int fts5RollbackMethod(sqlite3_vtab *pVtab){ + int rc; + Fts5Table *pTab = (Fts5Table*)pVtab; + fts5CheckTransactionState(pTab, FTS5_ROLLBACK, 0); + rc = sqlite3Fts5StorageRollback(pTab->pStorage); + return rc; +} + +static int fts5CsrPoslist(Fts5Cursor*, int, const u8**, int*); + +static void *fts5ApiUserData(Fts5Context *pCtx){ + Fts5Cursor *pCsr = (Fts5Cursor*)pCtx; + return pCsr->pAux->pUserData; +} + +static int fts5ApiColumnCount(Fts5Context *pCtx){ + Fts5Cursor *pCsr = (Fts5Cursor*)pCtx; + return ((Fts5Table*)(pCsr->base.pVtab))->pConfig->nCol; +} + +static int fts5ApiColumnTotalSize( + Fts5Context *pCtx, + int iCol, + sqlite3_int64 *pnToken +){ + Fts5Cursor *pCsr = (Fts5Cursor*)pCtx; + Fts5Table *pTab = (Fts5Table*)(pCsr->base.pVtab); + return sqlite3Fts5StorageSize(pTab->pStorage, iCol, pnToken); +} + +static int fts5ApiRowCount(Fts5Context *pCtx, i64 *pnRow){ + Fts5Cursor *pCsr = (Fts5Cursor*)pCtx; + Fts5Table *pTab = (Fts5Table*)(pCsr->base.pVtab); + return sqlite3Fts5StorageRowCount(pTab->pStorage, pnRow); +} + +static int fts5ApiTokenize( + Fts5Context *pCtx, + const char *pText, int nText, + void *pUserData, + int (*xToken)(void*, int, const char*, int, int, int) +){ + Fts5Cursor *pCsr = (Fts5Cursor*)pCtx; + Fts5Table *pTab = (Fts5Table*)(pCsr->base.pVtab); + return sqlite3Fts5Tokenize( + pTab->pConfig, FTS5_TOKENIZE_AUX, pText, nText, pUserData, xToken + ); +} + +static int fts5ApiPhraseCount(Fts5Context *pCtx){ + Fts5Cursor *pCsr = (Fts5Cursor*)pCtx; + return sqlite3Fts5ExprPhraseCount(pCsr->pExpr); +} + +static int fts5ApiPhraseSize(Fts5Context *pCtx, int iPhrase){ + Fts5Cursor *pCsr = (Fts5Cursor*)pCtx; + return sqlite3Fts5ExprPhraseSize(pCsr->pExpr, iPhrase); +} + +static int fts5ApiColumnText( + Fts5Context *pCtx, + int iCol, + const char **pz, + int *pn +){ + int rc = SQLITE_OK; + Fts5Cursor *pCsr = (Fts5Cursor*)pCtx; + if( fts5IsContentless((Fts5Table*)(pCsr->base.pVtab)) ){ + *pz = 0; + *pn = 0; + }else{ + rc = fts5SeekCursor(pCsr, 0); + if( rc==SQLITE_OK ){ + *pz = (const char*)sqlite3_column_text(pCsr->pStmt, iCol+1); + *pn = sqlite3_column_bytes(pCsr->pStmt, iCol+1); + } + } + return rc; +} + +static int fts5CsrPoslist( + Fts5Cursor *pCsr, + int iPhrase, + const u8 **pa, + int *pn +){ + Fts5Config *pConfig = ((Fts5Table*)(pCsr->base.pVtab))->pConfig; + int rc = SQLITE_OK; + int bLive = (pCsr->pSorter==0); + + if( CsrFlagTest(pCsr, FTS5CSR_REQUIRE_POSLIST) ){ + + if( pConfig->eDetail!=FTS5_DETAIL_FULL ){ + Fts5PoslistPopulator *aPopulator; + int i; + aPopulator = sqlite3Fts5ExprClearPoslists(pCsr->pExpr, bLive); + if( aPopulator==0 ) rc = SQLITE_NOMEM; + for(i=0; i<pConfig->nCol && rc==SQLITE_OK; i++){ + int n; const char *z; + rc = fts5ApiColumnText((Fts5Context*)pCsr, i, &z, &n); + if( rc==SQLITE_OK ){ + rc = sqlite3Fts5ExprPopulatePoslists( + pConfig, pCsr->pExpr, aPopulator, i, z, n + ); + } + } + sqlite3_free(aPopulator); + + if( pCsr->pSorter ){ + sqlite3Fts5ExprCheckPoslists(pCsr->pExpr, pCsr->pSorter->iRowid); + } + } + CsrFlagClear(pCsr, FTS5CSR_REQUIRE_POSLIST); + } + + if( pCsr->pSorter && pConfig->eDetail==FTS5_DETAIL_FULL ){ + Fts5Sorter *pSorter = pCsr->pSorter; + int i1 = (iPhrase==0 ? 0 : pSorter->aIdx[iPhrase-1]); + *pn = pSorter->aIdx[iPhrase] - i1; + *pa = &pSorter->aPoslist[i1]; + }else{ + *pn = sqlite3Fts5ExprPoslist(pCsr->pExpr, iPhrase, pa); + } + + return rc; +} + +/* +** Ensure that the Fts5Cursor.nInstCount and aInst[] variables are populated +** correctly for the current view. Return SQLITE_OK if successful, or an +** SQLite error code otherwise. +*/ +static int fts5CacheInstArray(Fts5Cursor *pCsr){ + int rc = SQLITE_OK; + Fts5PoslistReader *aIter; /* One iterator for each phrase */ + int nIter; /* Number of iterators/phrases */ + + nIter = sqlite3Fts5ExprPhraseCount(pCsr->pExpr); + if( pCsr->aInstIter==0 ){ + int nByte = sizeof(Fts5PoslistReader) * nIter; + pCsr->aInstIter = (Fts5PoslistReader*)sqlite3Fts5MallocZero(&rc, nByte); + } + aIter = pCsr->aInstIter; + + if( aIter ){ + int nInst = 0; /* Number instances seen so far */ + int i; + + /* Initialize all iterators */ + for(i=0; i<nIter && rc==SQLITE_OK; i++){ + const u8 *a; + int n; + rc = fts5CsrPoslist(pCsr, i, &a, &n); + if( rc==SQLITE_OK ){ + sqlite3Fts5PoslistReaderInit(a, n, &aIter[i]); + } + } + + if( rc==SQLITE_OK ){ + while( 1 ){ + int *aInst; + int iBest = -1; + for(i=0; i<nIter; i++){ + if( (aIter[i].bEof==0) + && (iBest<0 || aIter[i].iPos<aIter[iBest].iPos) + ){ + iBest = i; + } + } + if( iBest<0 ) break; + + nInst++; + if( nInst>=pCsr->nInstAlloc ){ + pCsr->nInstAlloc = pCsr->nInstAlloc ? pCsr->nInstAlloc*2 : 32; + aInst = (int*)sqlite3_realloc( + pCsr->aInst, pCsr->nInstAlloc*sizeof(int)*3 + ); + if( aInst ){ + pCsr->aInst = aInst; + }else{ + rc = SQLITE_NOMEM; + break; + } + } + + aInst = &pCsr->aInst[3 * (nInst-1)]; + aInst[0] = iBest; + aInst[1] = FTS5_POS2COLUMN(aIter[iBest].iPos); + aInst[2] = FTS5_POS2OFFSET(aIter[iBest].iPos); + sqlite3Fts5PoslistReaderNext(&aIter[iBest]); + } + } + + pCsr->nInstCount = nInst; + CsrFlagClear(pCsr, FTS5CSR_REQUIRE_INST); + } + return rc; +} + +static int fts5ApiInstCount(Fts5Context *pCtx, int *pnInst){ + Fts5Cursor *pCsr = (Fts5Cursor*)pCtx; + int rc = SQLITE_OK; + if( CsrFlagTest(pCsr, FTS5CSR_REQUIRE_INST)==0 + || SQLITE_OK==(rc = fts5CacheInstArray(pCsr)) ){ + *pnInst = pCsr->nInstCount; + } + return rc; +} + +static int fts5ApiInst( + Fts5Context *pCtx, + int iIdx, + int *piPhrase, + int *piCol, + int *piOff +){ + Fts5Cursor *pCsr = (Fts5Cursor*)pCtx; + int rc = SQLITE_OK; + if( CsrFlagTest(pCsr, FTS5CSR_REQUIRE_INST)==0 + || SQLITE_OK==(rc = fts5CacheInstArray(pCsr)) + ){ + if( iIdx<0 || iIdx>=pCsr->nInstCount ){ + rc = SQLITE_RANGE; +#if 0 + }else if( fts5IsOffsetless((Fts5Table*)pCsr->base.pVtab) ){ + *piPhrase = pCsr->aInst[iIdx*3]; + *piCol = pCsr->aInst[iIdx*3 + 2]; + *piOff = -1; +#endif + }else{ + *piPhrase = pCsr->aInst[iIdx*3]; + *piCol = pCsr->aInst[iIdx*3 + 1]; + *piOff = pCsr->aInst[iIdx*3 + 2]; + } + } + return rc; +} + +static sqlite3_int64 fts5ApiRowid(Fts5Context *pCtx){ + return fts5CursorRowid((Fts5Cursor*)pCtx); +} + +static int fts5ColumnSizeCb( + void *pContext, /* Pointer to int */ + int tflags, + const char *pUnused, /* Buffer containing token */ + int nUnused, /* Size of token in bytes */ + int iUnused1, /* Start offset of token */ + int iUnused2 /* End offset of token */ +){ + int *pCnt = (int*)pContext; + UNUSED_PARAM2(pUnused, nUnused); + UNUSED_PARAM2(iUnused1, iUnused2); + if( (tflags & FTS5_TOKEN_COLOCATED)==0 ){ + (*pCnt)++; + } + return SQLITE_OK; +} + +static int fts5ApiColumnSize(Fts5Context *pCtx, int iCol, int *pnToken){ + Fts5Cursor *pCsr = (Fts5Cursor*)pCtx; + Fts5Table *pTab = (Fts5Table*)(pCsr->base.pVtab); + Fts5Config *pConfig = pTab->pConfig; + int rc = SQLITE_OK; + + if( CsrFlagTest(pCsr, FTS5CSR_REQUIRE_DOCSIZE) ){ + if( pConfig->bColumnsize ){ + i64 iRowid = fts5CursorRowid(pCsr); + rc = sqlite3Fts5StorageDocsize(pTab->pStorage, iRowid, pCsr->aColumnSize); + }else if( pConfig->zContent==0 ){ + int i; + for(i=0; i<pConfig->nCol; i++){ + if( pConfig->abUnindexed[i]==0 ){ + pCsr->aColumnSize[i] = -1; + } + } + }else{ + int i; + for(i=0; rc==SQLITE_OK && i<pConfig->nCol; i++){ + if( pConfig->abUnindexed[i]==0 ){ + const char *z; int n; + void *p = (void*)(&pCsr->aColumnSize[i]); + pCsr->aColumnSize[i] = 0; + rc = fts5ApiColumnText(pCtx, i, &z, &n); + if( rc==SQLITE_OK ){ + rc = sqlite3Fts5Tokenize( + pConfig, FTS5_TOKENIZE_AUX, z, n, p, fts5ColumnSizeCb + ); + } + } + } + } + CsrFlagClear(pCsr, FTS5CSR_REQUIRE_DOCSIZE); + } + if( iCol<0 ){ + int i; + *pnToken = 0; + for(i=0; i<pConfig->nCol; i++){ + *pnToken += pCsr->aColumnSize[i]; + } + }else if( iCol<pConfig->nCol ){ + *pnToken = pCsr->aColumnSize[iCol]; + }else{ + *pnToken = 0; + rc = SQLITE_RANGE; + } + return rc; +} + +/* +** Implementation of the xSetAuxdata() method. +*/ +static int fts5ApiSetAuxdata( + Fts5Context *pCtx, /* Fts5 context */ + void *pPtr, /* Pointer to save as auxdata */ + void(*xDelete)(void*) /* Destructor for pPtr (or NULL) */ +){ + Fts5Cursor *pCsr = (Fts5Cursor*)pCtx; + Fts5Auxdata *pData; + + /* Search through the cursors list of Fts5Auxdata objects for one that + ** corresponds to the currently executing auxiliary function. */ + for(pData=pCsr->pAuxdata; pData; pData=pData->pNext){ + if( pData->pAux==pCsr->pAux ) break; + } + + if( pData ){ + if( pData->xDelete ){ + pData->xDelete(pData->pPtr); + } + }else{ + int rc = SQLITE_OK; + pData = (Fts5Auxdata*)sqlite3Fts5MallocZero(&rc, sizeof(Fts5Auxdata)); + if( pData==0 ){ + if( xDelete ) xDelete(pPtr); + return rc; + } + pData->pAux = pCsr->pAux; + pData->pNext = pCsr->pAuxdata; + pCsr->pAuxdata = pData; + } + + pData->xDelete = xDelete; + pData->pPtr = pPtr; + return SQLITE_OK; +} + +static void *fts5ApiGetAuxdata(Fts5Context *pCtx, int bClear){ + Fts5Cursor *pCsr = (Fts5Cursor*)pCtx; + Fts5Auxdata *pData; + void *pRet = 0; + + for(pData=pCsr->pAuxdata; pData; pData=pData->pNext){ + if( pData->pAux==pCsr->pAux ) break; + } + + if( pData ){ + pRet = pData->pPtr; + if( bClear ){ + pData->pPtr = 0; + pData->xDelete = 0; + } + } + + return pRet; +} + +static void fts5ApiPhraseNext( + Fts5Context *pUnused, + Fts5PhraseIter *pIter, + int *piCol, int *piOff +){ + UNUSED_PARAM(pUnused); + if( pIter->a>=pIter->b ){ + *piCol = -1; + *piOff = -1; + }else{ + int iVal; + pIter->a += fts5GetVarint32(pIter->a, iVal); + if( iVal==1 ){ + pIter->a += fts5GetVarint32(pIter->a, iVal); + *piCol = iVal; + *piOff = 0; + pIter->a += fts5GetVarint32(pIter->a, iVal); + } + *piOff += (iVal-2); + } +} + +static int fts5ApiPhraseFirst( + Fts5Context *pCtx, + int iPhrase, + Fts5PhraseIter *pIter, + int *piCol, int *piOff +){ + Fts5Cursor *pCsr = (Fts5Cursor*)pCtx; + int n; + int rc = fts5CsrPoslist(pCsr, iPhrase, &pIter->a, &n); + if( rc==SQLITE_OK ){ + pIter->b = &pIter->a[n]; + *piCol = 0; + *piOff = 0; + fts5ApiPhraseNext(pCtx, pIter, piCol, piOff); + } + return rc; +} + +static void fts5ApiPhraseNextColumn( + Fts5Context *pCtx, + Fts5PhraseIter *pIter, + int *piCol +){ + Fts5Cursor *pCsr = (Fts5Cursor*)pCtx; + Fts5Config *pConfig = ((Fts5Table*)(pCsr->base.pVtab))->pConfig; + + if( pConfig->eDetail==FTS5_DETAIL_COLUMNS ){ + if( pIter->a>=pIter->b ){ + *piCol = -1; + }else{ + int iIncr; + pIter->a += fts5GetVarint32(&pIter->a[0], iIncr); + *piCol += (iIncr-2); + } + }else{ + while( 1 ){ + int dummy; + if( pIter->a>=pIter->b ){ + *piCol = -1; + return; + } + if( pIter->a[0]==0x01 ) break; + pIter->a += fts5GetVarint32(pIter->a, dummy); + } + pIter->a += 1 + fts5GetVarint32(&pIter->a[1], *piCol); + } +} + +static int fts5ApiPhraseFirstColumn( + Fts5Context *pCtx, + int iPhrase, + Fts5PhraseIter *pIter, + int *piCol +){ + int rc = SQLITE_OK; + Fts5Cursor *pCsr = (Fts5Cursor*)pCtx; + Fts5Config *pConfig = ((Fts5Table*)(pCsr->base.pVtab))->pConfig; + + if( pConfig->eDetail==FTS5_DETAIL_COLUMNS ){ + Fts5Sorter *pSorter = pCsr->pSorter; + int n; + if( pSorter ){ + int i1 = (iPhrase==0 ? 0 : pSorter->aIdx[iPhrase-1]); + n = pSorter->aIdx[iPhrase] - i1; + pIter->a = &pSorter->aPoslist[i1]; + }else{ + rc = sqlite3Fts5ExprPhraseCollist(pCsr->pExpr, iPhrase, &pIter->a, &n); + } + if( rc==SQLITE_OK ){ + pIter->b = &pIter->a[n]; + *piCol = 0; + fts5ApiPhraseNextColumn(pCtx, pIter, piCol); + } + }else{ + int n; + rc = fts5CsrPoslist(pCsr, iPhrase, &pIter->a, &n); + if( rc==SQLITE_OK ){ + pIter->b = &pIter->a[n]; + if( n<=0 ){ + *piCol = -1; + }else if( pIter->a[0]==0x01 ){ + pIter->a += 1 + fts5GetVarint32(&pIter->a[1], *piCol); + }else{ + *piCol = 0; + } + } + } + + return rc; +} + + +static int fts5ApiQueryPhrase(Fts5Context*, int, void*, + int(*)(const Fts5ExtensionApi*, Fts5Context*, void*) +); + +static const Fts5ExtensionApi sFts5Api = { + 2, /* iVersion */ + fts5ApiUserData, + fts5ApiColumnCount, + fts5ApiRowCount, + fts5ApiColumnTotalSize, + fts5ApiTokenize, + fts5ApiPhraseCount, + fts5ApiPhraseSize, + fts5ApiInstCount, + fts5ApiInst, + fts5ApiRowid, + fts5ApiColumnText, + fts5ApiColumnSize, + fts5ApiQueryPhrase, + fts5ApiSetAuxdata, + fts5ApiGetAuxdata, + fts5ApiPhraseFirst, + fts5ApiPhraseNext, + fts5ApiPhraseFirstColumn, + fts5ApiPhraseNextColumn, +}; + +/* +** Implementation of API function xQueryPhrase(). +*/ +static int fts5ApiQueryPhrase( + Fts5Context *pCtx, + int iPhrase, + void *pUserData, + int(*xCallback)(const Fts5ExtensionApi*, Fts5Context*, void*) +){ + Fts5Cursor *pCsr = (Fts5Cursor*)pCtx; + Fts5Table *pTab = (Fts5Table*)(pCsr->base.pVtab); + int rc; + Fts5Cursor *pNew = 0; + + rc = fts5OpenMethod(pCsr->base.pVtab, (sqlite3_vtab_cursor**)&pNew); + if( rc==SQLITE_OK ){ + pNew->ePlan = FTS5_PLAN_MATCH; + pNew->iFirstRowid = SMALLEST_INT64; + pNew->iLastRowid = LARGEST_INT64; + pNew->base.pVtab = (sqlite3_vtab*)pTab; + rc = sqlite3Fts5ExprClonePhrase(pCsr->pExpr, iPhrase, &pNew->pExpr); + } + + if( rc==SQLITE_OK ){ + for(rc = fts5CursorFirst(pTab, pNew, 0); + rc==SQLITE_OK && CsrFlagTest(pNew, FTS5CSR_EOF)==0; + rc = fts5NextMethod((sqlite3_vtab_cursor*)pNew) + ){ + rc = xCallback(&sFts5Api, (Fts5Context*)pNew, pUserData); + if( rc!=SQLITE_OK ){ + if( rc==SQLITE_DONE ) rc = SQLITE_OK; + break; + } + } + } + + fts5CloseMethod((sqlite3_vtab_cursor*)pNew); + return rc; +} + +static void fts5ApiInvoke( + Fts5Auxiliary *pAux, + Fts5Cursor *pCsr, + sqlite3_context *context, + int argc, + sqlite3_value **argv +){ + assert( pCsr->pAux==0 ); + pCsr->pAux = pAux; + pAux->xFunc(&sFts5Api, (Fts5Context*)pCsr, context, argc, argv); + pCsr->pAux = 0; +} + +static Fts5Cursor *fts5CursorFromCsrid(Fts5Global *pGlobal, i64 iCsrId){ + Fts5Cursor *pCsr; + for(pCsr=pGlobal->pCsr; pCsr; pCsr=pCsr->pNext){ + if( pCsr->iCsrId==iCsrId ) break; + } + return pCsr; +} + +static void fts5ApiCallback( + sqlite3_context *context, + int argc, + sqlite3_value **argv +){ + + Fts5Auxiliary *pAux; + Fts5Cursor *pCsr; + i64 iCsrId; + + assert( argc>=1 ); + pAux = (Fts5Auxiliary*)sqlite3_user_data(context); + iCsrId = sqlite3_value_int64(argv[0]); + + pCsr = fts5CursorFromCsrid(pAux->pGlobal, iCsrId); + if( pCsr==0 ){ + char *zErr = sqlite3_mprintf("no such cursor: %lld", iCsrId); + sqlite3_result_error(context, zErr, -1); + sqlite3_free(zErr); + }else{ + fts5ApiInvoke(pAux, pCsr, context, argc-1, &argv[1]); + } +} + + +/* +** Given cursor id iId, return a pointer to the corresponding Fts5Index +** object. Or NULL If the cursor id does not exist. +** +** If successful, set *ppConfig to point to the associated config object +** before returning. +*/ +static Fts5Index *sqlite3Fts5IndexFromCsrid( + Fts5Global *pGlobal, /* FTS5 global context for db handle */ + i64 iCsrId, /* Id of cursor to find */ + Fts5Config **ppConfig /* OUT: Configuration object */ +){ + Fts5Cursor *pCsr; + Fts5Table *pTab; + + pCsr = fts5CursorFromCsrid(pGlobal, iCsrId); + pTab = (Fts5Table*)pCsr->base.pVtab; + *ppConfig = pTab->pConfig; + + return pTab->pIndex; +} + +/* +** Return a "position-list blob" corresponding to the current position of +** cursor pCsr via sqlite3_result_blob(). A position-list blob contains +** the current position-list for each phrase in the query associated with +** cursor pCsr. +** +** A position-list blob begins with (nPhrase-1) varints, where nPhrase is +** the number of phrases in the query. Following the varints are the +** concatenated position lists for each phrase, in order. +** +** The first varint (if it exists) contains the size of the position list +** for phrase 0. The second (same disclaimer) contains the size of position +** list 1. And so on. There is no size field for the final position list, +** as it can be derived from the total size of the blob. +*/ +static int fts5PoslistBlob(sqlite3_context *pCtx, Fts5Cursor *pCsr){ + int i; + int rc = SQLITE_OK; + int nPhrase = sqlite3Fts5ExprPhraseCount(pCsr->pExpr); + Fts5Buffer val; + + memset(&val, 0, sizeof(Fts5Buffer)); + switch( ((Fts5Table*)(pCsr->base.pVtab))->pConfig->eDetail ){ + case FTS5_DETAIL_FULL: + + /* Append the varints */ + for(i=0; i<(nPhrase-1); i++){ + const u8 *dummy; + int nByte = sqlite3Fts5ExprPoslist(pCsr->pExpr, i, &dummy); + sqlite3Fts5BufferAppendVarint(&rc, &val, nByte); + } + + /* Append the position lists */ + for(i=0; i<nPhrase; i++){ + const u8 *pPoslist; + int nPoslist; + nPoslist = sqlite3Fts5ExprPoslist(pCsr->pExpr, i, &pPoslist); + sqlite3Fts5BufferAppendBlob(&rc, &val, nPoslist, pPoslist); + } + break; + + case FTS5_DETAIL_COLUMNS: + + /* Append the varints */ + for(i=0; rc==SQLITE_OK && i<(nPhrase-1); i++){ + const u8 *dummy; + int nByte; + rc = sqlite3Fts5ExprPhraseCollist(pCsr->pExpr, i, &dummy, &nByte); + sqlite3Fts5BufferAppendVarint(&rc, &val, nByte); + } + + /* Append the position lists */ + for(i=0; rc==SQLITE_OK && i<nPhrase; i++){ + const u8 *pPoslist; + int nPoslist; + rc = sqlite3Fts5ExprPhraseCollist(pCsr->pExpr, i, &pPoslist, &nPoslist); + sqlite3Fts5BufferAppendBlob(&rc, &val, nPoslist, pPoslist); + } + break; + + default: + break; + } + + sqlite3_result_blob(pCtx, val.p, val.n, sqlite3_free); + return rc; +} + +/* +** This is the xColumn method, called by SQLite to request a value from +** the row that the supplied cursor currently points to. +*/ +static int fts5ColumnMethod( + sqlite3_vtab_cursor *pCursor, /* Cursor to retrieve value from */ + sqlite3_context *pCtx, /* Context for sqlite3_result_xxx() calls */ + int iCol /* Index of column to read value from */ +){ + Fts5Table *pTab = (Fts5Table*)(pCursor->pVtab); + Fts5Config *pConfig = pTab->pConfig; + Fts5Cursor *pCsr = (Fts5Cursor*)pCursor; + int rc = SQLITE_OK; + + assert( CsrFlagTest(pCsr, FTS5CSR_EOF)==0 ); + + if( pCsr->ePlan==FTS5_PLAN_SPECIAL ){ + if( iCol==pConfig->nCol ){ + sqlite3_result_int64(pCtx, pCsr->iSpecial); + } + }else + + if( iCol==pConfig->nCol ){ + /* User is requesting the value of the special column with the same name + ** as the table. Return the cursor integer id number. This value is only + ** useful in that it may be passed as the first argument to an FTS5 + ** auxiliary function. */ + sqlite3_result_int64(pCtx, pCsr->iCsrId); + }else if( iCol==pConfig->nCol+1 ){ + + /* The value of the "rank" column. */ + if( pCsr->ePlan==FTS5_PLAN_SOURCE ){ + fts5PoslistBlob(pCtx, pCsr); + }else if( + pCsr->ePlan==FTS5_PLAN_MATCH + || pCsr->ePlan==FTS5_PLAN_SORTED_MATCH + ){ + if( pCsr->pRank || SQLITE_OK==(rc = fts5FindRankFunction(pCsr)) ){ + fts5ApiInvoke(pCsr->pRank, pCsr, pCtx, pCsr->nRankArg, pCsr->apRankArg); + } + } + }else if( !fts5IsContentless(pTab) ){ + rc = fts5SeekCursor(pCsr, 1); + if( rc==SQLITE_OK ){ + sqlite3_result_value(pCtx, sqlite3_column_value(pCsr->pStmt, iCol+1)); + } + } + return rc; +} + + +/* +** This routine implements the xFindFunction method for the FTS3 +** virtual table. +*/ +static int fts5FindFunctionMethod( + sqlite3_vtab *pVtab, /* Virtual table handle */ + int nUnused, /* Number of SQL function arguments */ + const char *zName, /* Name of SQL function */ + void (**pxFunc)(sqlite3_context*,int,sqlite3_value**), /* OUT: Result */ + void **ppArg /* OUT: User data for *pxFunc */ +){ + Fts5Table *pTab = (Fts5Table*)pVtab; + Fts5Auxiliary *pAux; + + UNUSED_PARAM(nUnused); + pAux = fts5FindAuxiliary(pTab, zName); + if( pAux ){ + *pxFunc = fts5ApiCallback; + *ppArg = (void*)pAux; + return 1; + } + + /* No function of the specified name was found. Return 0. */ + return 0; +} + +/* +** Implementation of FTS5 xRename method. Rename an fts5 table. +*/ +static int fts5RenameMethod( + sqlite3_vtab *pVtab, /* Virtual table handle */ + const char *zName /* New name of table */ +){ + Fts5Table *pTab = (Fts5Table*)pVtab; + return sqlite3Fts5StorageRename(pTab->pStorage, zName); +} + +/* +** The xSavepoint() method. +** +** Flush the contents of the pending-terms table to disk. +*/ +static int fts5SavepointMethod(sqlite3_vtab *pVtab, int iSavepoint){ + Fts5Table *pTab = (Fts5Table*)pVtab; + UNUSED_PARAM(iSavepoint); /* Call below is a no-op for NDEBUG builds */ + fts5CheckTransactionState(pTab, FTS5_SAVEPOINT, iSavepoint); + fts5TripCursors(pTab); + return sqlite3Fts5StorageSync(pTab->pStorage, 0); +} + +/* +** The xRelease() method. +** +** This is a no-op. +*/ +static int fts5ReleaseMethod(sqlite3_vtab *pVtab, int iSavepoint){ + Fts5Table *pTab = (Fts5Table*)pVtab; + UNUSED_PARAM(iSavepoint); /* Call below is a no-op for NDEBUG builds */ + fts5CheckTransactionState(pTab, FTS5_RELEASE, iSavepoint); + fts5TripCursors(pTab); + return sqlite3Fts5StorageSync(pTab->pStorage, 0); +} + +/* +** The xRollbackTo() method. +** +** Discard the contents of the pending terms table. +*/ +static int fts5RollbackToMethod(sqlite3_vtab *pVtab, int iSavepoint){ + Fts5Table *pTab = (Fts5Table*)pVtab; + UNUSED_PARAM(iSavepoint); /* Call below is a no-op for NDEBUG builds */ + fts5CheckTransactionState(pTab, FTS5_ROLLBACKTO, iSavepoint); + fts5TripCursors(pTab); + return sqlite3Fts5StorageRollback(pTab->pStorage); +} + +/* +** Register a new auxiliary function with global context pGlobal. +*/ +static int fts5CreateAux( + fts5_api *pApi, /* Global context (one per db handle) */ + const char *zName, /* Name of new function */ + void *pUserData, /* User data for aux. function */ + fts5_extension_function xFunc, /* Aux. function implementation */ + void(*xDestroy)(void*) /* Destructor for pUserData */ +){ + Fts5Global *pGlobal = (Fts5Global*)pApi; + int rc = sqlite3_overload_function(pGlobal->db, zName, -1); + if( rc==SQLITE_OK ){ + Fts5Auxiliary *pAux; + int nName; /* Size of zName in bytes, including \0 */ + int nByte; /* Bytes of space to allocate */ + + nName = (int)strlen(zName) + 1; + nByte = sizeof(Fts5Auxiliary) + nName; + pAux = (Fts5Auxiliary*)sqlite3_malloc(nByte); + if( pAux ){ + memset(pAux, 0, nByte); + pAux->zFunc = (char*)&pAux[1]; + memcpy(pAux->zFunc, zName, nName); + pAux->pGlobal = pGlobal; + pAux->pUserData = pUserData; + pAux->xFunc = xFunc; + pAux->xDestroy = xDestroy; + pAux->pNext = pGlobal->pAux; + pGlobal->pAux = pAux; + }else{ + rc = SQLITE_NOMEM; + } + } + + return rc; +} + +/* +** Register a new tokenizer. This is the implementation of the +** fts5_api.xCreateTokenizer() method. +*/ +static int fts5CreateTokenizer( + fts5_api *pApi, /* Global context (one per db handle) */ + const char *zName, /* Name of new function */ + void *pUserData, /* User data for aux. function */ + fts5_tokenizer *pTokenizer, /* Tokenizer implementation */ + void(*xDestroy)(void*) /* Destructor for pUserData */ +){ + Fts5Global *pGlobal = (Fts5Global*)pApi; + Fts5TokenizerModule *pNew; + int nName; /* Size of zName and its \0 terminator */ + int nByte; /* Bytes of space to allocate */ + int rc = SQLITE_OK; + + nName = (int)strlen(zName) + 1; + nByte = sizeof(Fts5TokenizerModule) + nName; + pNew = (Fts5TokenizerModule*)sqlite3_malloc(nByte); + if( pNew ){ + memset(pNew, 0, nByte); + pNew->zName = (char*)&pNew[1]; + memcpy(pNew->zName, zName, nName); + pNew->pUserData = pUserData; + pNew->x = *pTokenizer; + pNew->xDestroy = xDestroy; + pNew->pNext = pGlobal->pTok; + pGlobal->pTok = pNew; + if( pNew->pNext==0 ){ + pGlobal->pDfltTok = pNew; + } + }else{ + rc = SQLITE_NOMEM; + } + + return rc; +} + +static Fts5TokenizerModule *fts5LocateTokenizer( + Fts5Global *pGlobal, + const char *zName +){ + Fts5TokenizerModule *pMod = 0; + + if( zName==0 ){ + pMod = pGlobal->pDfltTok; + }else{ + for(pMod=pGlobal->pTok; pMod; pMod=pMod->pNext){ + if( sqlite3_stricmp(zName, pMod->zName)==0 ) break; + } + } + + return pMod; +} + +/* +** Find a tokenizer. This is the implementation of the +** fts5_api.xFindTokenizer() method. +*/ +static int fts5FindTokenizer( + fts5_api *pApi, /* Global context (one per db handle) */ + const char *zName, /* Name of new function */ + void **ppUserData, + fts5_tokenizer *pTokenizer /* Populate this object */ +){ + int rc = SQLITE_OK; + Fts5TokenizerModule *pMod; + + pMod = fts5LocateTokenizer((Fts5Global*)pApi, zName); + if( pMod ){ + *pTokenizer = pMod->x; + *ppUserData = pMod->pUserData; + }else{ + memset(pTokenizer, 0, sizeof(fts5_tokenizer)); + rc = SQLITE_ERROR; + } + + return rc; +} + +static int sqlite3Fts5GetTokenizer( + Fts5Global *pGlobal, + const char **azArg, + int nArg, + Fts5Tokenizer **ppTok, + fts5_tokenizer **ppTokApi, + char **pzErr +){ + Fts5TokenizerModule *pMod; + int rc = SQLITE_OK; + + pMod = fts5LocateTokenizer(pGlobal, nArg==0 ? 0 : azArg[0]); + if( pMod==0 ){ + assert( nArg>0 ); + rc = SQLITE_ERROR; + *pzErr = sqlite3_mprintf("no such tokenizer: %s", azArg[0]); + }else{ + rc = pMod->x.xCreate(pMod->pUserData, &azArg[1], (nArg?nArg-1:0), ppTok); + *ppTokApi = &pMod->x; + if( rc!=SQLITE_OK && pzErr ){ + *pzErr = sqlite3_mprintf("error in tokenizer constructor"); + } + } + + if( rc!=SQLITE_OK ){ + *ppTokApi = 0; + *ppTok = 0; + } + + return rc; +} + +static void fts5ModuleDestroy(void *pCtx){ + Fts5TokenizerModule *pTok, *pNextTok; + Fts5Auxiliary *pAux, *pNextAux; + Fts5Global *pGlobal = (Fts5Global*)pCtx; + + for(pAux=pGlobal->pAux; pAux; pAux=pNextAux){ + pNextAux = pAux->pNext; + if( pAux->xDestroy ) pAux->xDestroy(pAux->pUserData); + sqlite3_free(pAux); + } + + for(pTok=pGlobal->pTok; pTok; pTok=pNextTok){ + pNextTok = pTok->pNext; + if( pTok->xDestroy ) pTok->xDestroy(pTok->pUserData); + sqlite3_free(pTok); + } + + sqlite3_free(pGlobal); +} + +static void fts5Fts5Func( + sqlite3_context *pCtx, /* Function call context */ + int nArg, /* Number of args */ + sqlite3_value **apUnused /* Function arguments */ +){ + Fts5Global *pGlobal = (Fts5Global*)sqlite3_user_data(pCtx); + char buf[8]; + UNUSED_PARAM2(nArg, apUnused); + assert( nArg==0 ); + assert( sizeof(buf)>=sizeof(pGlobal) ); + memcpy(buf, (void*)&pGlobal, sizeof(pGlobal)); + sqlite3_result_blob(pCtx, buf, sizeof(pGlobal), SQLITE_TRANSIENT); +} + +/* +** Implementation of fts5_source_id() function. +*/ +static void fts5SourceIdFunc( + sqlite3_context *pCtx, /* Function call context */ + int nArg, /* Number of args */ + sqlite3_value **apUnused /* Function arguments */ +){ + assert( nArg==0 ); + UNUSED_PARAM2(nArg, apUnused); + sqlite3_result_text(pCtx, "fts5: 2016-02-19 16:19:23 0cb728c15c66f1bf09cc1e0731a95ba937c6c71c", -1, SQLITE_TRANSIENT); +} + +static int fts5Init(sqlite3 *db){ + static const sqlite3_module fts5Mod = { + /* iVersion */ 2, + /* xCreate */ fts5CreateMethod, + /* xConnect */ fts5ConnectMethod, + /* xBestIndex */ fts5BestIndexMethod, + /* xDisconnect */ fts5DisconnectMethod, + /* xDestroy */ fts5DestroyMethod, + /* xOpen */ fts5OpenMethod, + /* xClose */ fts5CloseMethod, + /* xFilter */ fts5FilterMethod, + /* xNext */ fts5NextMethod, + /* xEof */ fts5EofMethod, + /* xColumn */ fts5ColumnMethod, + /* xRowid */ fts5RowidMethod, + /* xUpdate */ fts5UpdateMethod, + /* xBegin */ fts5BeginMethod, + /* xSync */ fts5SyncMethod, + /* xCommit */ fts5CommitMethod, + /* xRollback */ fts5RollbackMethod, + /* xFindFunction */ fts5FindFunctionMethod, + /* xRename */ fts5RenameMethod, + /* xSavepoint */ fts5SavepointMethod, + /* xRelease */ fts5ReleaseMethod, + /* xRollbackTo */ fts5RollbackToMethod, + }; + + int rc; + Fts5Global *pGlobal = 0; + + pGlobal = (Fts5Global*)sqlite3_malloc(sizeof(Fts5Global)); + if( pGlobal==0 ){ + rc = SQLITE_NOMEM; + }else{ + void *p = (void*)pGlobal; + memset(pGlobal, 0, sizeof(Fts5Global)); + pGlobal->db = db; + pGlobal->api.iVersion = 2; + pGlobal->api.xCreateFunction = fts5CreateAux; + pGlobal->api.xCreateTokenizer = fts5CreateTokenizer; + pGlobal->api.xFindTokenizer = fts5FindTokenizer; + rc = sqlite3_create_module_v2(db, "fts5", &fts5Mod, p, fts5ModuleDestroy); + if( rc==SQLITE_OK ) rc = sqlite3Fts5IndexInit(db); + if( rc==SQLITE_OK ) rc = sqlite3Fts5ExprInit(pGlobal, db); + if( rc==SQLITE_OK ) rc = sqlite3Fts5AuxInit(&pGlobal->api); + if( rc==SQLITE_OK ) rc = sqlite3Fts5TokenizerInit(&pGlobal->api); + if( rc==SQLITE_OK ) rc = sqlite3Fts5VocabInit(pGlobal, db); + if( rc==SQLITE_OK ){ + rc = sqlite3_create_function( + db, "fts5", 0, SQLITE_UTF8, p, fts5Fts5Func, 0, 0 + ); + } + if( rc==SQLITE_OK ){ + rc = sqlite3_create_function( + db, "fts5_source_id", 0, SQLITE_UTF8, p, fts5SourceIdFunc, 0, 0 + ); + } + } + + /* If SQLITE_FTS5_ENABLE_TEST_MI is defined, assume that the file + ** fts5_test_mi.c is compiled and linked into the executable. And call + ** its entry point to enable the matchinfo() demo. */ +#ifdef SQLITE_FTS5_ENABLE_TEST_MI + if( rc==SQLITE_OK ){ + extern int sqlite3Fts5TestRegisterMatchinfo(sqlite3*); + rc = sqlite3Fts5TestRegisterMatchinfo(db); + } +#endif + + return rc; +} + +/* +** The following functions are used to register the module with SQLite. If +** this module is being built as part of the SQLite core (SQLITE_CORE is +** defined), then sqlite3_open() will call sqlite3Fts5Init() directly. +** +** Or, if this module is being built as a loadable extension, +** sqlite3Fts5Init() is omitted and the two standard entry points +** sqlite3_fts_init() and sqlite3_fts5_init() defined instead. +*/ +#ifndef SQLITE_CORE +#ifdef _WIN32 +__declspec(dllexport) +#endif +int sqlite3_fts_init( + sqlite3 *db, + char **pzErrMsg, + const sqlite3_api_routines *pApi +){ + SQLITE_EXTENSION_INIT2(pApi); + (void)pzErrMsg; /* Unused parameter */ + return fts5Init(db); +} + +#ifdef _WIN32 +__declspec(dllexport) +#endif +int sqlite3_fts5_init( + sqlite3 *db, + char **pzErrMsg, + const sqlite3_api_routines *pApi +){ + SQLITE_EXTENSION_INIT2(pApi); + (void)pzErrMsg; /* Unused parameter */ + return fts5Init(db); +} +#else +int sqlite3Fts5Init(sqlite3 *db){ + return fts5Init(db); +} +#endif + +#line 1 "fts5_storage.c" +/* +** 2014 May 31 +** +** The author disclaims copyright to this source code. In place of +** a legal notice, here is a blessing: +** +** May you do good and not evil. +** May you find forgiveness for yourself and forgive others. +** May you share freely, never taking more than you give. +** +****************************************************************************** +** +*/ + + + +/* #include "fts5Int.h" */ + +struct Fts5Storage { + Fts5Config *pConfig; + Fts5Index *pIndex; + int bTotalsValid; /* True if nTotalRow/aTotalSize[] are valid */ + i64 nTotalRow; /* Total number of rows in FTS table */ + i64 *aTotalSize; /* Total sizes of each column */ + sqlite3_stmt *aStmt[11]; +}; + + +#if FTS5_STMT_SCAN_ASC!=0 +# error "FTS5_STMT_SCAN_ASC mismatch" +#endif +#if FTS5_STMT_SCAN_DESC!=1 +# error "FTS5_STMT_SCAN_DESC mismatch" +#endif +#if FTS5_STMT_LOOKUP!=2 +# error "FTS5_STMT_LOOKUP mismatch" +#endif + +#define FTS5_STMT_INSERT_CONTENT 3 +#define FTS5_STMT_REPLACE_CONTENT 4 +#define FTS5_STMT_DELETE_CONTENT 5 +#define FTS5_STMT_REPLACE_DOCSIZE 6 +#define FTS5_STMT_DELETE_DOCSIZE 7 +#define FTS5_STMT_LOOKUP_DOCSIZE 8 +#define FTS5_STMT_REPLACE_CONFIG 9 +#define FTS5_STMT_SCAN 10 + +/* +** Prepare the two insert statements - Fts5Storage.pInsertContent and +** Fts5Storage.pInsertDocsize - if they have not already been prepared. +** Return SQLITE_OK if successful, or an SQLite error code if an error +** occurs. +*/ +static int fts5StorageGetStmt( + Fts5Storage *p, /* Storage handle */ + int eStmt, /* FTS5_STMT_XXX constant */ + sqlite3_stmt **ppStmt, /* OUT: Prepared statement handle */ + char **pzErrMsg /* OUT: Error message (if any) */ +){ + int rc = SQLITE_OK; + + /* If there is no %_docsize table, there should be no requests for + ** statements to operate on it. */ + assert( p->pConfig->bColumnsize || ( + eStmt!=FTS5_STMT_REPLACE_DOCSIZE + && eStmt!=FTS5_STMT_DELETE_DOCSIZE + && eStmt!=FTS5_STMT_LOOKUP_DOCSIZE + )); + + assert( eStmt>=0 && eStmt<ArraySize(p->aStmt) ); + if( p->aStmt[eStmt]==0 ){ + const char *azStmt[] = { + "SELECT %s FROM %s T WHERE T.%Q >= ? AND T.%Q <= ? ORDER BY T.%Q ASC", + "SELECT %s FROM %s T WHERE T.%Q <= ? AND T.%Q >= ? ORDER BY T.%Q DESC", + "SELECT %s FROM %s T WHERE T.%Q=?", /* LOOKUP */ + + "INSERT INTO %Q.'%q_content' VALUES(%s)", /* INSERT_CONTENT */ + "REPLACE INTO %Q.'%q_content' VALUES(%s)", /* REPLACE_CONTENT */ + "DELETE FROM %Q.'%q_content' WHERE id=?", /* DELETE_CONTENT */ + "REPLACE INTO %Q.'%q_docsize' VALUES(?,?)", /* REPLACE_DOCSIZE */ + "DELETE FROM %Q.'%q_docsize' WHERE id=?", /* DELETE_DOCSIZE */ + + "SELECT sz FROM %Q.'%q_docsize' WHERE id=?", /* LOOKUP_DOCSIZE */ + + "REPLACE INTO %Q.'%q_config' VALUES(?,?)", /* REPLACE_CONFIG */ + "SELECT %s FROM %s AS T", /* SCAN */ + }; + Fts5Config *pC = p->pConfig; + char *zSql = 0; + + switch( eStmt ){ + case FTS5_STMT_SCAN: + zSql = sqlite3_mprintf(azStmt[eStmt], + pC->zContentExprlist, pC->zContent + ); + break; + + case FTS5_STMT_SCAN_ASC: + case FTS5_STMT_SCAN_DESC: + zSql = sqlite3_mprintf(azStmt[eStmt], pC->zContentExprlist, + pC->zContent, pC->zContentRowid, pC->zContentRowid, + pC->zContentRowid + ); + break; + + case FTS5_STMT_LOOKUP: + zSql = sqlite3_mprintf(azStmt[eStmt], + pC->zContentExprlist, pC->zContent, pC->zContentRowid + ); + break; + + case FTS5_STMT_INSERT_CONTENT: + case FTS5_STMT_REPLACE_CONTENT: { + int nCol = pC->nCol + 1; + char *zBind; + int i; + + zBind = sqlite3_malloc(1 + nCol*2); + if( zBind ){ + for(i=0; i<nCol; i++){ + zBind[i*2] = '?'; + zBind[i*2 + 1] = ','; + } + zBind[i*2-1] = '\0'; + zSql = sqlite3_mprintf(azStmt[eStmt], pC->zDb, pC->zName, zBind); + sqlite3_free(zBind); + } + break; + } + + default: + zSql = sqlite3_mprintf(azStmt[eStmt], pC->zDb, pC->zName); + break; + } + + if( zSql==0 ){ + rc = SQLITE_NOMEM; + }else{ + rc = sqlite3_prepare_v2(pC->db, zSql, -1, &p->aStmt[eStmt], 0); + sqlite3_free(zSql); + if( rc!=SQLITE_OK && pzErrMsg ){ + *pzErrMsg = sqlite3_mprintf("%s", sqlite3_errmsg(pC->db)); + } + } + } + + *ppStmt = p->aStmt[eStmt]; + return rc; +} + + +static int fts5ExecPrintf( + sqlite3 *db, + char **pzErr, + const char *zFormat, + ... +){ + int rc; + va_list ap; /* ... printf arguments */ + char *zSql; + + va_start(ap, zFormat); + zSql = sqlite3_vmprintf(zFormat, ap); + + if( zSql==0 ){ + rc = SQLITE_NOMEM; + }else{ + rc = sqlite3_exec(db, zSql, 0, 0, pzErr); + sqlite3_free(zSql); + } + + va_end(ap); + return rc; +} + +/* +** Drop all shadow tables. Return SQLITE_OK if successful or an SQLite error +** code otherwise. +*/ +static int sqlite3Fts5DropAll(Fts5Config *pConfig){ + int rc = fts5ExecPrintf(pConfig->db, 0, + "DROP TABLE IF EXISTS %Q.'%q_data';" + "DROP TABLE IF EXISTS %Q.'%q_idx';" + "DROP TABLE IF EXISTS %Q.'%q_config';", + pConfig->zDb, pConfig->zName, + pConfig->zDb, pConfig->zName, + pConfig->zDb, pConfig->zName + ); + if( rc==SQLITE_OK && pConfig->bColumnsize ){ + rc = fts5ExecPrintf(pConfig->db, 0, + "DROP TABLE IF EXISTS %Q.'%q_docsize';", + pConfig->zDb, pConfig->zName + ); + } + if( rc==SQLITE_OK && pConfig->eContent==FTS5_CONTENT_NORMAL ){ + rc = fts5ExecPrintf(pConfig->db, 0, + "DROP TABLE IF EXISTS %Q.'%q_content';", + pConfig->zDb, pConfig->zName + ); + } + return rc; +} + +static void fts5StorageRenameOne( + Fts5Config *pConfig, /* Current FTS5 configuration */ + int *pRc, /* IN/OUT: Error code */ + const char *zTail, /* Tail of table name e.g. "data", "config" */ + const char *zName /* New name of FTS5 table */ +){ + if( *pRc==SQLITE_OK ){ + *pRc = fts5ExecPrintf(pConfig->db, 0, + "ALTER TABLE %Q.'%q_%s' RENAME TO '%q_%s';", + pConfig->zDb, pConfig->zName, zTail, zName, zTail + ); + } +} + +static int sqlite3Fts5StorageRename(Fts5Storage *pStorage, const char *zName){ + Fts5Config *pConfig = pStorage->pConfig; + int rc = sqlite3Fts5StorageSync(pStorage, 1); + + fts5StorageRenameOne(pConfig, &rc, "data", zName); + fts5StorageRenameOne(pConfig, &rc, "idx", zName); + fts5StorageRenameOne(pConfig, &rc, "config", zName); + if( pConfig->bColumnsize ){ + fts5StorageRenameOne(pConfig, &rc, "docsize", zName); + } + if( pConfig->eContent==FTS5_CONTENT_NORMAL ){ + fts5StorageRenameOne(pConfig, &rc, "content", zName); + } + return rc; +} + +/* +** Create the shadow table named zPost, with definition zDefn. Return +** SQLITE_OK if successful, or an SQLite error code otherwise. +*/ +static int sqlite3Fts5CreateTable( + Fts5Config *pConfig, /* FTS5 configuration */ + const char *zPost, /* Shadow table to create (e.g. "content") */ + const char *zDefn, /* Columns etc. for shadow table */ + int bWithout, /* True for without rowid */ + char **pzErr /* OUT: Error message */ +){ + int rc; + char *zErr = 0; + + rc = fts5ExecPrintf(pConfig->db, &zErr, "CREATE TABLE %Q.'%q_%q'(%s)%s", + pConfig->zDb, pConfig->zName, zPost, zDefn, bWithout?" WITHOUT ROWID":"" + ); + if( zErr ){ + *pzErr = sqlite3_mprintf( + "fts5: error creating shadow table %q_%s: %s", + pConfig->zName, zPost, zErr + ); + sqlite3_free(zErr); + } + + return rc; +} + +/* +** Open a new Fts5Index handle. If the bCreate argument is true, create +** and initialize the underlying tables +** +** If successful, set *pp to point to the new object and return SQLITE_OK. +** Otherwise, set *pp to NULL and return an SQLite error code. +*/ +static int sqlite3Fts5StorageOpen( + Fts5Config *pConfig, + Fts5Index *pIndex, + int bCreate, + Fts5Storage **pp, + char **pzErr /* OUT: Error message */ +){ + int rc = SQLITE_OK; + Fts5Storage *p; /* New object */ + int nByte; /* Bytes of space to allocate */ + + nByte = sizeof(Fts5Storage) /* Fts5Storage object */ + + pConfig->nCol * sizeof(i64); /* Fts5Storage.aTotalSize[] */ + *pp = p = (Fts5Storage*)sqlite3_malloc(nByte); + if( !p ) return SQLITE_NOMEM; + + memset(p, 0, nByte); + p->aTotalSize = (i64*)&p[1]; + p->pConfig = pConfig; + p->pIndex = pIndex; + + if( bCreate ){ + if( pConfig->eContent==FTS5_CONTENT_NORMAL ){ + int nDefn = 32 + pConfig->nCol*10; + char *zDefn = sqlite3_malloc(32 + pConfig->nCol * 10); + if( zDefn==0 ){ + rc = SQLITE_NOMEM; + }else{ + int i; + int iOff; + sqlite3_snprintf(nDefn, zDefn, "id INTEGER PRIMARY KEY"); + iOff = (int)strlen(zDefn); + for(i=0; i<pConfig->nCol; i++){ + sqlite3_snprintf(nDefn-iOff, &zDefn[iOff], ", c%d", i); + iOff += (int)strlen(&zDefn[iOff]); + } + rc = sqlite3Fts5CreateTable(pConfig, "content", zDefn, 0, pzErr); + } + sqlite3_free(zDefn); + } + + if( rc==SQLITE_OK && pConfig->bColumnsize ){ + rc = sqlite3Fts5CreateTable( + pConfig, "docsize", "id INTEGER PRIMARY KEY, sz BLOB", 0, pzErr + ); + } + if( rc==SQLITE_OK ){ + rc = sqlite3Fts5CreateTable( + pConfig, "config", "k PRIMARY KEY, v", 1, pzErr + ); + } + if( rc==SQLITE_OK ){ + rc = sqlite3Fts5StorageConfigValue(p, "version", 0, FTS5_CURRENT_VERSION); + } + } + + if( rc ){ + sqlite3Fts5StorageClose(p); + *pp = 0; + } + return rc; +} + +/* +** Close a handle opened by an earlier call to sqlite3Fts5StorageOpen(). +*/ +static int sqlite3Fts5StorageClose(Fts5Storage *p){ + int rc = SQLITE_OK; + if( p ){ + int i; + + /* Finalize all SQL statements */ + for(i=0; i<ArraySize(p->aStmt); i++){ + sqlite3_finalize(p->aStmt[i]); + } + + sqlite3_free(p); + } + return rc; +} + +typedef struct Fts5InsertCtx Fts5InsertCtx; +struct Fts5InsertCtx { + Fts5Storage *pStorage; + int iCol; + int szCol; /* Size of column value in tokens */ +}; + +/* +** Tokenization callback used when inserting tokens into the FTS index. +*/ +static int fts5StorageInsertCallback( + void *pContext, /* Pointer to Fts5InsertCtx object */ + int tflags, + const char *pToken, /* Buffer containing token */ + int nToken, /* Size of token in bytes */ + int iUnused1, /* Start offset of token */ + int iUnused2 /* End offset of token */ +){ + Fts5InsertCtx *pCtx = (Fts5InsertCtx*)pContext; + Fts5Index *pIdx = pCtx->pStorage->pIndex; + UNUSED_PARAM2(iUnused1, iUnused2); + if( (tflags & FTS5_TOKEN_COLOCATED)==0 || pCtx->szCol==0 ){ + pCtx->szCol++; + } + return sqlite3Fts5IndexWrite(pIdx, pCtx->iCol, pCtx->szCol-1, pToken, nToken); +} + +/* +** If a row with rowid iDel is present in the %_content table, add the +** delete-markers to the FTS index necessary to delete it. Do not actually +** remove the %_content row at this time though. +*/ +static int fts5StorageDeleteFromIndex( + Fts5Storage *p, + i64 iDel, + sqlite3_value **apVal +){ + Fts5Config *pConfig = p->pConfig; + sqlite3_stmt *pSeek = 0; /* SELECT to read row iDel from %_data */ + int rc; /* Return code */ + int rc2; /* sqlite3_reset() return code */ + int iCol; + Fts5InsertCtx ctx; + + if( apVal==0 ){ + rc = fts5StorageGetStmt(p, FTS5_STMT_LOOKUP, &pSeek, 0); + if( rc!=SQLITE_OK ) return rc; + sqlite3_bind_int64(pSeek, 1, iDel); + if( sqlite3_step(pSeek)!=SQLITE_ROW ){ + return sqlite3_reset(pSeek); + } + } + + ctx.pStorage = p; + ctx.iCol = -1; + rc = sqlite3Fts5IndexBeginWrite(p->pIndex, 1, iDel); + for(iCol=1; rc==SQLITE_OK && iCol<=pConfig->nCol; iCol++){ + if( pConfig->abUnindexed[iCol-1]==0 ){ + const char *zText; + int nText; + if( pSeek ){ + zText = (const char*)sqlite3_column_text(pSeek, iCol); + nText = sqlite3_column_bytes(pSeek, iCol); + }else{ + zText = (const char*)sqlite3_value_text(apVal[iCol-1]); + nText = sqlite3_value_bytes(apVal[iCol-1]); + } + ctx.szCol = 0; + rc = sqlite3Fts5Tokenize(pConfig, FTS5_TOKENIZE_DOCUMENT, + zText, nText, (void*)&ctx, fts5StorageInsertCallback + ); + p->aTotalSize[iCol-1] -= (i64)ctx.szCol; + } + } + p->nTotalRow--; + + rc2 = sqlite3_reset(pSeek); + if( rc==SQLITE_OK ) rc = rc2; + return rc; +} + + +/* +** Insert a record into the %_docsize table. Specifically, do: +** +** INSERT OR REPLACE INTO %_docsize(id, sz) VALUES(iRowid, pBuf); +** +** If there is no %_docsize table (as happens if the columnsize=0 option +** is specified when the FTS5 table is created), this function is a no-op. +*/ +static int fts5StorageInsertDocsize( + Fts5Storage *p, /* Storage module to write to */ + i64 iRowid, /* id value */ + Fts5Buffer *pBuf /* sz value */ +){ + int rc = SQLITE_OK; + if( p->pConfig->bColumnsize ){ + sqlite3_stmt *pReplace = 0; + rc = fts5StorageGetStmt(p, FTS5_STMT_REPLACE_DOCSIZE, &pReplace, 0); + if( rc==SQLITE_OK ){ + sqlite3_bind_int64(pReplace, 1, iRowid); + sqlite3_bind_blob(pReplace, 2, pBuf->p, pBuf->n, SQLITE_STATIC); + sqlite3_step(pReplace); + rc = sqlite3_reset(pReplace); + } + } + return rc; +} + +/* +** Load the contents of the "averages" record from disk into the +** p->nTotalRow and p->aTotalSize[] variables. If successful, and if +** argument bCache is true, set the p->bTotalsValid flag to indicate +** that the contents of aTotalSize[] and nTotalRow are valid until +** further notice. +** +** Return SQLITE_OK if successful, or an SQLite error code if an error +** occurs. +*/ +static int fts5StorageLoadTotals(Fts5Storage *p, int bCache){ + int rc = SQLITE_OK; + if( p->bTotalsValid==0 ){ + rc = sqlite3Fts5IndexGetAverages(p->pIndex, &p->nTotalRow, p->aTotalSize); + p->bTotalsValid = bCache; + } + return rc; +} + +/* +** Store the current contents of the p->nTotalRow and p->aTotalSize[] +** variables in the "averages" record on disk. +** +** Return SQLITE_OK if successful, or an SQLite error code if an error +** occurs. +*/ +static int fts5StorageSaveTotals(Fts5Storage *p){ + int nCol = p->pConfig->nCol; + int i; + Fts5Buffer buf; + int rc = SQLITE_OK; + memset(&buf, 0, sizeof(buf)); + + sqlite3Fts5BufferAppendVarint(&rc, &buf, p->nTotalRow); + for(i=0; i<nCol; i++){ + sqlite3Fts5BufferAppendVarint(&rc, &buf, p->aTotalSize[i]); + } + if( rc==SQLITE_OK ){ + rc = sqlite3Fts5IndexSetAverages(p->pIndex, buf.p, buf.n); + } + sqlite3_free(buf.p); + + return rc; +} + +/* +** Remove a row from the FTS table. +*/ +static int sqlite3Fts5StorageDelete(Fts5Storage *p, i64 iDel, sqlite3_value **apVal){ + Fts5Config *pConfig = p->pConfig; + int rc; + sqlite3_stmt *pDel = 0; + + assert( pConfig->eContent!=FTS5_CONTENT_NORMAL || apVal==0 ); + rc = fts5StorageLoadTotals(p, 1); + + /* Delete the index records */ + if( rc==SQLITE_OK ){ + rc = fts5StorageDeleteFromIndex(p, iDel, apVal); + } + + /* Delete the %_docsize record */ + if( rc==SQLITE_OK && pConfig->bColumnsize ){ + rc = fts5StorageGetStmt(p, FTS5_STMT_DELETE_DOCSIZE, &pDel, 0); + if( rc==SQLITE_OK ){ + sqlite3_bind_int64(pDel, 1, iDel); + sqlite3_step(pDel); + rc = sqlite3_reset(pDel); + } + } + + /* Delete the %_content record */ + if( pConfig->eContent==FTS5_CONTENT_NORMAL ){ + if( rc==SQLITE_OK ){ + rc = fts5StorageGetStmt(p, FTS5_STMT_DELETE_CONTENT, &pDel, 0); + } + if( rc==SQLITE_OK ){ + sqlite3_bind_int64(pDel, 1, iDel); + sqlite3_step(pDel); + rc = sqlite3_reset(pDel); + } + } + + /* Write the averages record */ + if( rc==SQLITE_OK ){ + rc = fts5StorageSaveTotals(p); + } + + return rc; +} + +/* +** Delete all entries in the FTS5 index. +*/ +static int sqlite3Fts5StorageDeleteAll(Fts5Storage *p){ + Fts5Config *pConfig = p->pConfig; + int rc; + + /* Delete the contents of the %_data and %_docsize tables. */ + rc = fts5ExecPrintf(pConfig->db, 0, + "DELETE FROM %Q.'%q_data';" + "DELETE FROM %Q.'%q_idx';", + pConfig->zDb, pConfig->zName, + pConfig->zDb, pConfig->zName + ); + if( rc==SQLITE_OK && pConfig->bColumnsize ){ + rc = fts5ExecPrintf(pConfig->db, 0, + "DELETE FROM %Q.'%q_docsize';", + pConfig->zDb, pConfig->zName + ); + } + + /* Reinitialize the %_data table. This call creates the initial structure + ** and averages records. */ + if( rc==SQLITE_OK ){ + rc = sqlite3Fts5IndexReinit(p->pIndex); + } + if( rc==SQLITE_OK ){ + rc = sqlite3Fts5StorageConfigValue(p, "version", 0, FTS5_CURRENT_VERSION); + } + return rc; +} + +static int sqlite3Fts5StorageRebuild(Fts5Storage *p){ + Fts5Buffer buf = {0,0,0}; + Fts5Config *pConfig = p->pConfig; + sqlite3_stmt *pScan = 0; + Fts5InsertCtx ctx; + int rc; + + memset(&ctx, 0, sizeof(Fts5InsertCtx)); + ctx.pStorage = p; + rc = sqlite3Fts5StorageDeleteAll(p); + if( rc==SQLITE_OK ){ + rc = fts5StorageLoadTotals(p, 1); + } + + if( rc==SQLITE_OK ){ + rc = fts5StorageGetStmt(p, FTS5_STMT_SCAN, &pScan, 0); + } + + while( rc==SQLITE_OK && SQLITE_ROW==sqlite3_step(pScan) ){ + i64 iRowid = sqlite3_column_int64(pScan, 0); + + sqlite3Fts5BufferZero(&buf); + rc = sqlite3Fts5IndexBeginWrite(p->pIndex, 0, iRowid); + for(ctx.iCol=0; rc==SQLITE_OK && ctx.iCol<pConfig->nCol; ctx.iCol++){ + ctx.szCol = 0; + if( pConfig->abUnindexed[ctx.iCol]==0 ){ + rc = sqlite3Fts5Tokenize(pConfig, + FTS5_TOKENIZE_DOCUMENT, + (const char*)sqlite3_column_text(pScan, ctx.iCol+1), + sqlite3_column_bytes(pScan, ctx.iCol+1), + (void*)&ctx, + fts5StorageInsertCallback + ); + } + sqlite3Fts5BufferAppendVarint(&rc, &buf, ctx.szCol); + p->aTotalSize[ctx.iCol] += (i64)ctx.szCol; + } + p->nTotalRow++; + + if( rc==SQLITE_OK ){ + rc = fts5StorageInsertDocsize(p, iRowid, &buf); + } + } + sqlite3_free(buf.p); + + /* Write the averages record */ + if( rc==SQLITE_OK ){ + rc = fts5StorageSaveTotals(p); + } + return rc; +} + +static int sqlite3Fts5StorageOptimize(Fts5Storage *p){ + return sqlite3Fts5IndexOptimize(p->pIndex); +} + +static int sqlite3Fts5StorageMerge(Fts5Storage *p, int nMerge){ + return sqlite3Fts5IndexMerge(p->pIndex, nMerge); +} + +/* +** Allocate a new rowid. This is used for "external content" tables when +** a NULL value is inserted into the rowid column. The new rowid is allocated +** by inserting a dummy row into the %_docsize table. The dummy will be +** overwritten later. +** +** If the %_docsize table does not exist, SQLITE_MISMATCH is returned. In +** this case the user is required to provide a rowid explicitly. +*/ +static int fts5StorageNewRowid(Fts5Storage *p, i64 *piRowid){ + int rc = SQLITE_MISMATCH; + if( p->pConfig->bColumnsize ){ + sqlite3_stmt *pReplace = 0; + rc = fts5StorageGetStmt(p, FTS5_STMT_REPLACE_DOCSIZE, &pReplace, 0); + if( rc==SQLITE_OK ){ + sqlite3_bind_null(pReplace, 1); + sqlite3_bind_null(pReplace, 2); + sqlite3_step(pReplace); + rc = sqlite3_reset(pReplace); + } + if( rc==SQLITE_OK ){ + *piRowid = sqlite3_last_insert_rowid(p->pConfig->db); + } + } + return rc; +} + +/* +** Insert a new row into the FTS content table. +*/ +static int sqlite3Fts5StorageContentInsert( + Fts5Storage *p, + sqlite3_value **apVal, + i64 *piRowid +){ + Fts5Config *pConfig = p->pConfig; + int rc = SQLITE_OK; + + /* Insert the new row into the %_content table. */ + if( pConfig->eContent!=FTS5_CONTENT_NORMAL ){ + if( sqlite3_value_type(apVal[1])==SQLITE_INTEGER ){ + *piRowid = sqlite3_value_int64(apVal[1]); + }else{ + rc = fts5StorageNewRowid(p, piRowid); + } + }else{ + sqlite3_stmt *pInsert = 0; /* Statement to write %_content table */ + int i; /* Counter variable */ + rc = fts5StorageGetStmt(p, FTS5_STMT_INSERT_CONTENT, &pInsert, 0); + for(i=1; rc==SQLITE_OK && i<=pConfig->nCol+1; i++){ + rc = sqlite3_bind_value(pInsert, i, apVal[i]); + } + if( rc==SQLITE_OK ){ + sqlite3_step(pInsert); + rc = sqlite3_reset(pInsert); + } + *piRowid = sqlite3_last_insert_rowid(pConfig->db); + } + + return rc; +} + +/* +** Insert new entries into the FTS index and %_docsize table. +*/ +static int sqlite3Fts5StorageIndexInsert( + Fts5Storage *p, + sqlite3_value **apVal, + i64 iRowid +){ + Fts5Config *pConfig = p->pConfig; + int rc = SQLITE_OK; /* Return code */ + Fts5InsertCtx ctx; /* Tokenization callback context object */ + Fts5Buffer buf; /* Buffer used to build up %_docsize blob */ + + memset(&buf, 0, sizeof(Fts5Buffer)); + ctx.pStorage = p; + rc = fts5StorageLoadTotals(p, 1); + + if( rc==SQLITE_OK ){ + rc = sqlite3Fts5IndexBeginWrite(p->pIndex, 0, iRowid); + } + for(ctx.iCol=0; rc==SQLITE_OK && ctx.iCol<pConfig->nCol; ctx.iCol++){ + ctx.szCol = 0; + if( pConfig->abUnindexed[ctx.iCol]==0 ){ + rc = sqlite3Fts5Tokenize(pConfig, + FTS5_TOKENIZE_DOCUMENT, + (const char*)sqlite3_value_text(apVal[ctx.iCol+2]), + sqlite3_value_bytes(apVal[ctx.iCol+2]), + (void*)&ctx, + fts5StorageInsertCallback + ); + } + sqlite3Fts5BufferAppendVarint(&rc, &buf, ctx.szCol); + p->aTotalSize[ctx.iCol] += (i64)ctx.szCol; + } + p->nTotalRow++; + + /* Write the %_docsize record */ + if( rc==SQLITE_OK ){ + rc = fts5StorageInsertDocsize(p, iRowid, &buf); + } + sqlite3_free(buf.p); + + /* Write the averages record */ + if( rc==SQLITE_OK ){ + rc = fts5StorageSaveTotals(p); + } + + return rc; +} + +static int fts5StorageCount(Fts5Storage *p, const char *zSuffix, i64 *pnRow){ + Fts5Config *pConfig = p->pConfig; + char *zSql; + int rc; + + zSql = sqlite3_mprintf("SELECT count(*) FROM %Q.'%q_%s'", + pConfig->zDb, pConfig->zName, zSuffix + ); + if( zSql==0 ){ + rc = SQLITE_NOMEM; + }else{ + sqlite3_stmt *pCnt = 0; + rc = sqlite3_prepare_v2(pConfig->db, zSql, -1, &pCnt, 0); + if( rc==SQLITE_OK ){ + if( SQLITE_ROW==sqlite3_step(pCnt) ){ + *pnRow = sqlite3_column_int64(pCnt, 0); + } + rc = sqlite3_finalize(pCnt); + } + } + + sqlite3_free(zSql); + return rc; +} + +/* +** Context object used by sqlite3Fts5StorageIntegrity(). +*/ +typedef struct Fts5IntegrityCtx Fts5IntegrityCtx; +struct Fts5IntegrityCtx { + i64 iRowid; + int iCol; + int szCol; + u64 cksum; + Fts5Termset *pTermset; + Fts5Config *pConfig; +}; + + +/* +** Tokenization callback used by integrity check. +*/ +static int fts5StorageIntegrityCallback( + void *pContext, /* Pointer to Fts5IntegrityCtx object */ + int tflags, + const char *pToken, /* Buffer containing token */ + int nToken, /* Size of token in bytes */ + int iUnused1, /* Start offset of token */ + int iUnused2 /* End offset of token */ +){ + Fts5IntegrityCtx *pCtx = (Fts5IntegrityCtx*)pContext; + Fts5Termset *pTermset = pCtx->pTermset; + int bPresent; + int ii; + int rc = SQLITE_OK; + int iPos; + int iCol; + + UNUSED_PARAM2(iUnused1, iUnused2); + + if( (tflags & FTS5_TOKEN_COLOCATED)==0 || pCtx->szCol==0 ){ + pCtx->szCol++; + } + + switch( pCtx->pConfig->eDetail ){ + case FTS5_DETAIL_FULL: + iPos = pCtx->szCol-1; + iCol = pCtx->iCol; + break; + + case FTS5_DETAIL_COLUMNS: + iPos = pCtx->iCol; + iCol = 0; + break; + + default: + assert( pCtx->pConfig->eDetail==FTS5_DETAIL_NONE ); + iPos = 0; + iCol = 0; + break; + } + + rc = sqlite3Fts5TermsetAdd(pTermset, 0, pToken, nToken, &bPresent); + if( rc==SQLITE_OK && bPresent==0 ){ + pCtx->cksum ^= sqlite3Fts5IndexEntryCksum( + pCtx->iRowid, iCol, iPos, 0, pToken, nToken + ); + } + + for(ii=0; rc==SQLITE_OK && ii<pCtx->pConfig->nPrefix; ii++){ + const int nChar = pCtx->pConfig->aPrefix[ii]; + int nByte = sqlite3Fts5IndexCharlenToBytelen(pToken, nToken, nChar); + if( nByte ){ + rc = sqlite3Fts5TermsetAdd(pTermset, ii+1, pToken, nByte, &bPresent); + if( bPresent==0 ){ + pCtx->cksum ^= sqlite3Fts5IndexEntryCksum( + pCtx->iRowid, iCol, iPos, ii+1, pToken, nByte + ); + } + } + } + + return rc; +} + +/* +** Check that the contents of the FTS index match that of the %_content +** table. Return SQLITE_OK if they do, or SQLITE_CORRUPT if not. Return +** some other SQLite error code if an error occurs while attempting to +** determine this. +*/ +static int sqlite3Fts5StorageIntegrity(Fts5Storage *p){ + Fts5Config *pConfig = p->pConfig; + int rc; /* Return code */ + int *aColSize; /* Array of size pConfig->nCol */ + i64 *aTotalSize; /* Array of size pConfig->nCol */ + Fts5IntegrityCtx ctx; + sqlite3_stmt *pScan; + + memset(&ctx, 0, sizeof(Fts5IntegrityCtx)); + ctx.pConfig = p->pConfig; + aTotalSize = (i64*)sqlite3_malloc(pConfig->nCol * (sizeof(int)+sizeof(i64))); + if( !aTotalSize ) return SQLITE_NOMEM; + aColSize = (int*)&aTotalSize[pConfig->nCol]; + memset(aTotalSize, 0, sizeof(i64) * pConfig->nCol); + + /* Generate the expected index checksum based on the contents of the + ** %_content table. This block stores the checksum in ctx.cksum. */ + rc = fts5StorageGetStmt(p, FTS5_STMT_SCAN, &pScan, 0); + if( rc==SQLITE_OK ){ + int rc2; + while( SQLITE_ROW==sqlite3_step(pScan) ){ + int i; + ctx.iRowid = sqlite3_column_int64(pScan, 0); + ctx.szCol = 0; + if( pConfig->bColumnsize ){ + rc = sqlite3Fts5StorageDocsize(p, ctx.iRowid, aColSize); + } + if( rc==SQLITE_OK && pConfig->eDetail==FTS5_DETAIL_NONE ){ + rc = sqlite3Fts5TermsetNew(&ctx.pTermset); + } + for(i=0; rc==SQLITE_OK && i<pConfig->nCol; i++){ + if( pConfig->abUnindexed[i] ) continue; + ctx.iCol = i; + ctx.szCol = 0; + if( pConfig->eDetail==FTS5_DETAIL_COLUMNS ){ + rc = sqlite3Fts5TermsetNew(&ctx.pTermset); + } + if( rc==SQLITE_OK ){ + rc = sqlite3Fts5Tokenize(pConfig, + FTS5_TOKENIZE_DOCUMENT, + (const char*)sqlite3_column_text(pScan, i+1), + sqlite3_column_bytes(pScan, i+1), + (void*)&ctx, + fts5StorageIntegrityCallback + ); + } + if( rc==SQLITE_OK && pConfig->bColumnsize && ctx.szCol!=aColSize[i] ){ + rc = FTS5_CORRUPT; + } + aTotalSize[i] += ctx.szCol; + if( pConfig->eDetail==FTS5_DETAIL_COLUMNS ){ + sqlite3Fts5TermsetFree(ctx.pTermset); + ctx.pTermset = 0; + } + } + sqlite3Fts5TermsetFree(ctx.pTermset); + ctx.pTermset = 0; + + if( rc!=SQLITE_OK ) break; + } + rc2 = sqlite3_reset(pScan); + if( rc==SQLITE_OK ) rc = rc2; + } + + /* Test that the "totals" (sometimes called "averages") record looks Ok */ + if( rc==SQLITE_OK ){ + int i; + rc = fts5StorageLoadTotals(p, 0); + for(i=0; rc==SQLITE_OK && i<pConfig->nCol; i++){ + if( p->aTotalSize[i]!=aTotalSize[i] ) rc = FTS5_CORRUPT; + } + } + + /* Check that the %_docsize and %_content tables contain the expected + ** number of rows. */ + if( rc==SQLITE_OK && pConfig->eContent==FTS5_CONTENT_NORMAL ){ + i64 nRow = 0; + rc = fts5StorageCount(p, "content", &nRow); + if( rc==SQLITE_OK && nRow!=p->nTotalRow ) rc = FTS5_CORRUPT; + } + if( rc==SQLITE_OK && pConfig->bColumnsize ){ + i64 nRow = 0; + rc = fts5StorageCount(p, "docsize", &nRow); + if( rc==SQLITE_OK && nRow!=p->nTotalRow ) rc = FTS5_CORRUPT; + } + + /* Pass the expected checksum down to the FTS index module. It will + ** verify, amongst other things, that it matches the checksum generated by + ** inspecting the index itself. */ + if( rc==SQLITE_OK ){ + rc = sqlite3Fts5IndexIntegrityCheck(p->pIndex, ctx.cksum); + } + + sqlite3_free(aTotalSize); + return rc; +} + +/* +** Obtain an SQLite statement handle that may be used to read data from the +** %_content table. +*/ +static int sqlite3Fts5StorageStmt( + Fts5Storage *p, + int eStmt, + sqlite3_stmt **pp, + char **pzErrMsg +){ + int rc; + assert( eStmt==FTS5_STMT_SCAN_ASC + || eStmt==FTS5_STMT_SCAN_DESC + || eStmt==FTS5_STMT_LOOKUP + ); + rc = fts5StorageGetStmt(p, eStmt, pp, pzErrMsg); + if( rc==SQLITE_OK ){ + assert( p->aStmt[eStmt]==*pp ); + p->aStmt[eStmt] = 0; + } + return rc; +} + +/* +** Release an SQLite statement handle obtained via an earlier call to +** sqlite3Fts5StorageStmt(). The eStmt parameter passed to this function +** must match that passed to the sqlite3Fts5StorageStmt() call. +*/ +static void sqlite3Fts5StorageStmtRelease( + Fts5Storage *p, + int eStmt, + sqlite3_stmt *pStmt +){ + assert( eStmt==FTS5_STMT_SCAN_ASC + || eStmt==FTS5_STMT_SCAN_DESC + || eStmt==FTS5_STMT_LOOKUP + ); + if( p->aStmt[eStmt]==0 ){ + sqlite3_reset(pStmt); + p->aStmt[eStmt] = pStmt; + }else{ + sqlite3_finalize(pStmt); + } +} + +static int fts5StorageDecodeSizeArray( + int *aCol, int nCol, /* Array to populate */ + const u8 *aBlob, int nBlob /* Record to read varints from */ +){ + int i; + int iOff = 0; + for(i=0; i<nCol; i++){ + if( iOff>=nBlob ) return 1; + iOff += fts5GetVarint32(&aBlob[iOff], aCol[i]); + } + return (iOff!=nBlob); +} + +/* +** Argument aCol points to an array of integers containing one entry for +** each table column. This function reads the %_docsize record for the +** specified rowid and populates aCol[] with the results. +** +** An SQLite error code is returned if an error occurs, or SQLITE_OK +** otherwise. +*/ +static int sqlite3Fts5StorageDocsize(Fts5Storage *p, i64 iRowid, int *aCol){ + int nCol = p->pConfig->nCol; /* Number of user columns in table */ + sqlite3_stmt *pLookup = 0; /* Statement to query %_docsize */ + int rc; /* Return Code */ + + assert( p->pConfig->bColumnsize ); + rc = fts5StorageGetStmt(p, FTS5_STMT_LOOKUP_DOCSIZE, &pLookup, 0); + if( rc==SQLITE_OK ){ + int bCorrupt = 1; + sqlite3_bind_int64(pLookup, 1, iRowid); + if( SQLITE_ROW==sqlite3_step(pLookup) ){ + const u8 *aBlob = sqlite3_column_blob(pLookup, 0); + int nBlob = sqlite3_column_bytes(pLookup, 0); + if( 0==fts5StorageDecodeSizeArray(aCol, nCol, aBlob, nBlob) ){ + bCorrupt = 0; + } + } + rc = sqlite3_reset(pLookup); + if( bCorrupt && rc==SQLITE_OK ){ + rc = FTS5_CORRUPT; + } + } + + return rc; +} + +static int sqlite3Fts5StorageSize(Fts5Storage *p, int iCol, i64 *pnToken){ + int rc = fts5StorageLoadTotals(p, 0); + if( rc==SQLITE_OK ){ + *pnToken = 0; + if( iCol<0 ){ + int i; + for(i=0; i<p->pConfig->nCol; i++){ + *pnToken += p->aTotalSize[i]; + } + }else if( iCol<p->pConfig->nCol ){ + *pnToken = p->aTotalSize[iCol]; + }else{ + rc = SQLITE_RANGE; + } + } + return rc; +} + +static int sqlite3Fts5StorageRowCount(Fts5Storage *p, i64 *pnRow){ + int rc = fts5StorageLoadTotals(p, 0); + if( rc==SQLITE_OK ){ + *pnRow = p->nTotalRow; + } + return rc; +} + +/* +** Flush any data currently held in-memory to disk. +*/ +static int sqlite3Fts5StorageSync(Fts5Storage *p, int bCommit){ + if( bCommit && p->bTotalsValid ){ + int rc = fts5StorageSaveTotals(p); + p->bTotalsValid = 0; + if( rc!=SQLITE_OK ) return rc; + } + return sqlite3Fts5IndexSync(p->pIndex, bCommit); +} + +static int sqlite3Fts5StorageRollback(Fts5Storage *p){ + p->bTotalsValid = 0; + return sqlite3Fts5IndexRollback(p->pIndex); +} + +static int sqlite3Fts5StorageConfigValue( + Fts5Storage *p, + const char *z, + sqlite3_value *pVal, + int iVal +){ + sqlite3_stmt *pReplace = 0; + int rc = fts5StorageGetStmt(p, FTS5_STMT_REPLACE_CONFIG, &pReplace, 0); + if( rc==SQLITE_OK ){ + sqlite3_bind_text(pReplace, 1, z, -1, SQLITE_STATIC); + if( pVal ){ + sqlite3_bind_value(pReplace, 2, pVal); + }else{ + sqlite3_bind_int(pReplace, 2, iVal); + } + sqlite3_step(pReplace); + rc = sqlite3_reset(pReplace); + } + if( rc==SQLITE_OK && pVal ){ + int iNew = p->pConfig->iCookie + 1; + rc = sqlite3Fts5IndexSetCookie(p->pIndex, iNew); + if( rc==SQLITE_OK ){ + p->pConfig->iCookie = iNew; + } + } + return rc; +} + + + +#line 1 "fts5_tokenize.c" +/* +** 2014 May 31 +** +** The author disclaims copyright to this source code. In place of +** a legal notice, here is a blessing: +** +** May you do good and not evil. +** May you find forgiveness for yourself and forgive others. +** May you share freely, never taking more than you give. +** +****************************************************************************** +*/ + + +/* #include "fts5Int.h" */ + +/************************************************************************** +** Start of ascii tokenizer implementation. +*/ + +/* +** For tokenizers with no "unicode" modifier, the set of token characters +** is the same as the set of ASCII range alphanumeric characters. +*/ +static unsigned char aAsciiTokenChar[128] = { + 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 0x00..0x0F */ + 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 0x10..0x1F */ + 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 0x20..0x2F */ + 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, /* 0x30..0x3F */ + 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, /* 0x40..0x4F */ + 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, /* 0x50..0x5F */ + 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, /* 0x60..0x6F */ + 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, /* 0x70..0x7F */ +}; + +typedef struct AsciiTokenizer AsciiTokenizer; +struct AsciiTokenizer { + unsigned char aTokenChar[128]; +}; + +static void fts5AsciiAddExceptions( + AsciiTokenizer *p, + const char *zArg, + int bTokenChars +){ + int i; + for(i=0; zArg[i]; i++){ + if( (zArg[i] & 0x80)==0 ){ + p->aTokenChar[(int)zArg[i]] = (unsigned char)bTokenChars; + } + } +} + +/* +** Delete a "ascii" tokenizer. +*/ +static void fts5AsciiDelete(Fts5Tokenizer *p){ + sqlite3_free(p); +} + +/* +** Create an "ascii" tokenizer. +*/ +static int fts5AsciiCreate( + void *pUnused, + const char **azArg, int nArg, + Fts5Tokenizer **ppOut +){ + int rc = SQLITE_OK; + AsciiTokenizer *p = 0; + UNUSED_PARAM(pUnused); + if( nArg%2 ){ + rc = SQLITE_ERROR; + }else{ + p = sqlite3_malloc(sizeof(AsciiTokenizer)); + if( p==0 ){ + rc = SQLITE_NOMEM; + }else{ + int i; + memset(p, 0, sizeof(AsciiTokenizer)); + memcpy(p->aTokenChar, aAsciiTokenChar, sizeof(aAsciiTokenChar)); + for(i=0; rc==SQLITE_OK && i<nArg; i+=2){ + const char *zArg = azArg[i+1]; + if( 0==sqlite3_stricmp(azArg[i], "tokenchars") ){ + fts5AsciiAddExceptions(p, zArg, 1); + }else + if( 0==sqlite3_stricmp(azArg[i], "separators") ){ + fts5AsciiAddExceptions(p, zArg, 0); + }else{ + rc = SQLITE_ERROR; + } + } + if( rc!=SQLITE_OK ){ + fts5AsciiDelete((Fts5Tokenizer*)p); + p = 0; + } + } + } + + *ppOut = (Fts5Tokenizer*)p; + return rc; +} + + +static void asciiFold(char *aOut, const char *aIn, int nByte){ + int i; + for(i=0; i<nByte; i++){ + char c = aIn[i]; + if( c>='A' && c<='Z' ) c += 32; + aOut[i] = c; + } +} + +/* +** Tokenize some text using the ascii tokenizer. +*/ +static int fts5AsciiTokenize( + Fts5Tokenizer *pTokenizer, + void *pCtx, + int iUnused, + const char *pText, int nText, + int (*xToken)(void*, int, const char*, int nToken, int iStart, int iEnd) +){ + AsciiTokenizer *p = (AsciiTokenizer*)pTokenizer; + int rc = SQLITE_OK; + int ie; + int is = 0; + + char aFold[64]; + int nFold = sizeof(aFold); + char *pFold = aFold; + unsigned char *a = p->aTokenChar; + + UNUSED_PARAM(iUnused); + + while( is<nText && rc==SQLITE_OK ){ + int nByte; + + /* Skip any leading divider characters. */ + while( is<nText && ((pText[is]&0x80)==0 && a[(int)pText[is]]==0) ){ + is++; + } + if( is==nText ) break; + + /* Count the token characters */ + ie = is+1; + while( ie<nText && ((pText[ie]&0x80) || a[(int)pText[ie]] ) ){ + ie++; + } + + /* Fold to lower case */ + nByte = ie-is; + if( nByte>nFold ){ + if( pFold!=aFold ) sqlite3_free(pFold); + pFold = sqlite3_malloc(nByte*2); + if( pFold==0 ){ + rc = SQLITE_NOMEM; + break; + } + nFold = nByte*2; + } + asciiFold(pFold, &pText[is], nByte); + + /* Invoke the token callback */ + rc = xToken(pCtx, 0, pFold, nByte, is, ie); + is = ie+1; + } + + if( pFold!=aFold ) sqlite3_free(pFold); + if( rc==SQLITE_DONE ) rc = SQLITE_OK; + return rc; +} + +/************************************************************************** +** Start of unicode61 tokenizer implementation. +*/ + + +/* +** The following two macros - READ_UTF8 and WRITE_UTF8 - have been copied +** from the sqlite3 source file utf.c. If this file is compiled as part +** of the amalgamation, they are not required. +*/ +#ifndef SQLITE_AMALGAMATION + +static const unsigned char sqlite3Utf8Trans1[] = { + 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, + 0x08, 0x09, 0x0a, 0x0b, 0x0c, 0x0d, 0x0e, 0x0f, + 0x10, 0x11, 0x12, 0x13, 0x14, 0x15, 0x16, 0x17, + 0x18, 0x19, 0x1a, 0x1b, 0x1c, 0x1d, 0x1e, 0x1f, + 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, + 0x08, 0x09, 0x0a, 0x0b, 0x0c, 0x0d, 0x0e, 0x0f, + 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, + 0x00, 0x01, 0x02, 0x03, 0x00, 0x01, 0x00, 0x00, +}; + +#define READ_UTF8(zIn, zTerm, c) \ + c = *(zIn++); \ + if( c>=0xc0 ){ \ + c = sqlite3Utf8Trans1[c-0xc0]; \ + while( zIn!=zTerm && (*zIn & 0xc0)==0x80 ){ \ + c = (c<<6) + (0x3f & *(zIn++)); \ + } \ + if( c<0x80 \ + || (c&0xFFFFF800)==0xD800 \ + || (c&0xFFFFFFFE)==0xFFFE ){ c = 0xFFFD; } \ + } + + +#define WRITE_UTF8(zOut, c) { \ + if( c<0x00080 ){ \ + *zOut++ = (unsigned char)(c&0xFF); \ + } \ + else if( c<0x00800 ){ \ + *zOut++ = 0xC0 + (unsigned char)((c>>6)&0x1F); \ + *zOut++ = 0x80 + (unsigned char)(c & 0x3F); \ + } \ + else if( c<0x10000 ){ \ + *zOut++ = 0xE0 + (unsigned char)((c>>12)&0x0F); \ + *zOut++ = 0x80 + (unsigned char)((c>>6) & 0x3F); \ + *zOut++ = 0x80 + (unsigned char)(c & 0x3F); \ + }else{ \ + *zOut++ = 0xF0 + (unsigned char)((c>>18) & 0x07); \ + *zOut++ = 0x80 + (unsigned char)((c>>12) & 0x3F); \ + *zOut++ = 0x80 + (unsigned char)((c>>6) & 0x3F); \ + *zOut++ = 0x80 + (unsigned char)(c & 0x3F); \ + } \ +} + +#endif /* ifndef SQLITE_AMALGAMATION */ + +typedef struct Unicode61Tokenizer Unicode61Tokenizer; +struct Unicode61Tokenizer { + unsigned char aTokenChar[128]; /* ASCII range token characters */ + char *aFold; /* Buffer to fold text into */ + int nFold; /* Size of aFold[] in bytes */ + int bRemoveDiacritic; /* True if remove_diacritics=1 is set */ + int nException; + int *aiException; +}; + +static int fts5UnicodeAddExceptions( + Unicode61Tokenizer *p, /* Tokenizer object */ + const char *z, /* Characters to treat as exceptions */ + int bTokenChars /* 1 for 'tokenchars', 0 for 'separators' */ +){ + int rc = SQLITE_OK; + int n = (int)strlen(z); + int *aNew; + + if( n>0 ){ + aNew = (int*)sqlite3_realloc(p->aiException, (n+p->nException)*sizeof(int)); + if( aNew ){ + int nNew = p->nException; + const unsigned char *zCsr = (const unsigned char*)z; + const unsigned char *zTerm = (const unsigned char*)&z[n]; + while( zCsr<zTerm ){ + int iCode; + int bToken; + READ_UTF8(zCsr, zTerm, iCode); + if( iCode<128 ){ + p->aTokenChar[iCode] = (unsigned char)bTokenChars; + }else{ + bToken = sqlite3Fts5UnicodeIsalnum(iCode); + assert( (bToken==0 || bToken==1) ); + assert( (bTokenChars==0 || bTokenChars==1) ); + if( bToken!=bTokenChars && sqlite3Fts5UnicodeIsdiacritic(iCode)==0 ){ + int i; + for(i=0; i<nNew; i++){ + if( aNew[i]>iCode ) break; + } + memmove(&aNew[i+1], &aNew[i], (nNew-i)*sizeof(int)); + aNew[i] = iCode; + nNew++; + } + } + } + p->aiException = aNew; + p->nException = nNew; + }else{ + rc = SQLITE_NOMEM; + } + } + + return rc; +} + +/* +** Return true if the p->aiException[] array contains the value iCode. +*/ +static int fts5UnicodeIsException(Unicode61Tokenizer *p, int iCode){ + if( p->nException>0 ){ + int *a = p->aiException; + int iLo = 0; + int iHi = p->nException-1; + + while( iHi>=iLo ){ + int iTest = (iHi + iLo) / 2; + if( iCode==a[iTest] ){ + return 1; + }else if( iCode>a[iTest] ){ + iLo = iTest+1; + }else{ + iHi = iTest-1; + } + } + } + + return 0; +} + +/* +** Delete a "unicode61" tokenizer. +*/ +static void fts5UnicodeDelete(Fts5Tokenizer *pTok){ + if( pTok ){ + Unicode61Tokenizer *p = (Unicode61Tokenizer*)pTok; + sqlite3_free(p->aiException); + sqlite3_free(p->aFold); + sqlite3_free(p); + } + return; +} + +/* +** Create a "unicode61" tokenizer. +*/ +static int fts5UnicodeCreate( + void *pUnused, + const char **azArg, int nArg, + Fts5Tokenizer **ppOut +){ + int rc = SQLITE_OK; /* Return code */ + Unicode61Tokenizer *p = 0; /* New tokenizer object */ + + UNUSED_PARAM(pUnused); + + if( nArg%2 ){ + rc = SQLITE_ERROR; + }else{ + p = (Unicode61Tokenizer*)sqlite3_malloc(sizeof(Unicode61Tokenizer)); + if( p ){ + int i; + memset(p, 0, sizeof(Unicode61Tokenizer)); + memcpy(p->aTokenChar, aAsciiTokenChar, sizeof(aAsciiTokenChar)); + p->bRemoveDiacritic = 1; + p->nFold = 64; + p->aFold = sqlite3_malloc(p->nFold * sizeof(char)); + if( p->aFold==0 ){ + rc = SQLITE_NOMEM; + } + for(i=0; rc==SQLITE_OK && i<nArg; i+=2){ + const char *zArg = azArg[i+1]; + if( 0==sqlite3_stricmp(azArg[i], "remove_diacritics") ){ + if( (zArg[0]!='0' && zArg[0]!='1') || zArg[1] ){ + rc = SQLITE_ERROR; + } + p->bRemoveDiacritic = (zArg[0]=='1'); + }else + if( 0==sqlite3_stricmp(azArg[i], "tokenchars") ){ + rc = fts5UnicodeAddExceptions(p, zArg, 1); + }else + if( 0==sqlite3_stricmp(azArg[i], "separators") ){ + rc = fts5UnicodeAddExceptions(p, zArg, 0); + }else{ + rc = SQLITE_ERROR; + } + } + }else{ + rc = SQLITE_NOMEM; + } + if( rc!=SQLITE_OK ){ + fts5UnicodeDelete((Fts5Tokenizer*)p); + p = 0; + } + *ppOut = (Fts5Tokenizer*)p; + } + return rc; +} + +/* +** Return true if, for the purposes of tokenizing with the tokenizer +** passed as the first argument, codepoint iCode is considered a token +** character (not a separator). +*/ +static int fts5UnicodeIsAlnum(Unicode61Tokenizer *p, int iCode){ + assert( (sqlite3Fts5UnicodeIsalnum(iCode) & 0xFFFFFFFE)==0 ); + return sqlite3Fts5UnicodeIsalnum(iCode) ^ fts5UnicodeIsException(p, iCode); +} + +static int fts5UnicodeTokenize( + Fts5Tokenizer *pTokenizer, + void *pCtx, + int iUnused, + const char *pText, int nText, + int (*xToken)(void*, int, const char*, int nToken, int iStart, int iEnd) +){ + Unicode61Tokenizer *p = (Unicode61Tokenizer*)pTokenizer; + int rc = SQLITE_OK; + unsigned char *a = p->aTokenChar; + + unsigned char *zTerm = (unsigned char*)&pText[nText]; + unsigned char *zCsr = (unsigned char *)pText; + + /* Output buffer */ + char *aFold = p->aFold; + int nFold = p->nFold; + const char *pEnd = &aFold[nFold-6]; + + UNUSED_PARAM(iUnused); + + /* Each iteration of this loop gobbles up a contiguous run of separators, + ** then the next token. */ + while( rc==SQLITE_OK ){ + int iCode; /* non-ASCII codepoint read from input */ + char *zOut = aFold; + int is; + int ie; + + /* Skip any separator characters. */ + while( 1 ){ + if( zCsr>=zTerm ) goto tokenize_done; + if( *zCsr & 0x80 ) { + /* A character outside of the ascii range. Skip past it if it is + ** a separator character. Or break out of the loop if it is not. */ + is = zCsr - (unsigned char*)pText; + READ_UTF8(zCsr, zTerm, iCode); + if( fts5UnicodeIsAlnum(p, iCode) ){ + goto non_ascii_tokenchar; + } + }else{ + if( a[*zCsr] ){ + is = zCsr - (unsigned char*)pText; + goto ascii_tokenchar; + } + zCsr++; + } + } + + /* Run through the tokenchars. Fold them into the output buffer along + ** the way. */ + while( zCsr<zTerm ){ + + /* Grow the output buffer so that there is sufficient space to fit the + ** largest possible utf-8 character. */ + if( zOut>pEnd ){ + aFold = sqlite3_malloc(nFold*2); + if( aFold==0 ){ + rc = SQLITE_NOMEM; + goto tokenize_done; + } + zOut = &aFold[zOut - p->aFold]; + memcpy(aFold, p->aFold, nFold); + sqlite3_free(p->aFold); + p->aFold = aFold; + p->nFold = nFold = nFold*2; + pEnd = &aFold[nFold-6]; + } + + if( *zCsr & 0x80 ){ + /* An non-ascii-range character. Fold it into the output buffer if + ** it is a token character, or break out of the loop if it is not. */ + READ_UTF8(zCsr, zTerm, iCode); + if( fts5UnicodeIsAlnum(p,iCode)||sqlite3Fts5UnicodeIsdiacritic(iCode) ){ + non_ascii_tokenchar: + iCode = sqlite3Fts5UnicodeFold(iCode, p->bRemoveDiacritic); + if( iCode ) WRITE_UTF8(zOut, iCode); + }else{ + break; + } + }else if( a[*zCsr]==0 ){ + /* An ascii-range separator character. End of token. */ + break; + }else{ + ascii_tokenchar: + if( *zCsr>='A' && *zCsr<='Z' ){ + *zOut++ = *zCsr + 32; + }else{ + *zOut++ = *zCsr; + } + zCsr++; + } + ie = zCsr - (unsigned char*)pText; + } + + /* Invoke the token callback */ + rc = xToken(pCtx, 0, aFold, zOut-aFold, is, ie); + } + + tokenize_done: + if( rc==SQLITE_DONE ) rc = SQLITE_OK; + return rc; +} + +/************************************************************************** +** Start of porter stemmer implementation. +*/ + +/* Any tokens larger than this (in bytes) are passed through without +** stemming. */ +#define FTS5_PORTER_MAX_TOKEN 64 + +typedef struct PorterTokenizer PorterTokenizer; +struct PorterTokenizer { + fts5_tokenizer tokenizer; /* Parent tokenizer module */ + Fts5Tokenizer *pTokenizer; /* Parent tokenizer instance */ + char aBuf[FTS5_PORTER_MAX_TOKEN + 64]; +}; + +/* +** Delete a "porter" tokenizer. +*/ +static void fts5PorterDelete(Fts5Tokenizer *pTok){ + if( pTok ){ + PorterTokenizer *p = (PorterTokenizer*)pTok; + if( p->pTokenizer ){ + p->tokenizer.xDelete(p->pTokenizer); + } + sqlite3_free(p); + } +} + +/* +** Create a "porter" tokenizer. +*/ +static int fts5PorterCreate( + void *pCtx, + const char **azArg, int nArg, + Fts5Tokenizer **ppOut +){ + fts5_api *pApi = (fts5_api*)pCtx; + int rc = SQLITE_OK; + PorterTokenizer *pRet; + void *pUserdata = 0; + const char *zBase = "unicode61"; + + if( nArg>0 ){ + zBase = azArg[0]; + } + + pRet = (PorterTokenizer*)sqlite3_malloc(sizeof(PorterTokenizer)); + if( pRet ){ + memset(pRet, 0, sizeof(PorterTokenizer)); + rc = pApi->xFindTokenizer(pApi, zBase, &pUserdata, &pRet->tokenizer); + }else{ + rc = SQLITE_NOMEM; + } + if( rc==SQLITE_OK ){ + int nArg2 = (nArg>0 ? nArg-1 : 0); + const char **azArg2 = (nArg2 ? &azArg[1] : 0); + rc = pRet->tokenizer.xCreate(pUserdata, azArg2, nArg2, &pRet->pTokenizer); + } + + if( rc!=SQLITE_OK ){ + fts5PorterDelete((Fts5Tokenizer*)pRet); + pRet = 0; + } + *ppOut = (Fts5Tokenizer*)pRet; + return rc; +} + +typedef struct PorterContext PorterContext; +struct PorterContext { + void *pCtx; + int (*xToken)(void*, int, const char*, int, int, int); + char *aBuf; +}; + +typedef struct PorterRule PorterRule; +struct PorterRule { + const char *zSuffix; + int nSuffix; + int (*xCond)(char *zStem, int nStem); + const char *zOutput; + int nOutput; +}; + +#if 0 +static int fts5PorterApply(char *aBuf, int *pnBuf, PorterRule *aRule){ + int ret = -1; + int nBuf = *pnBuf; + PorterRule *p; + + for(p=aRule; p->zSuffix; p++){ + assert( strlen(p->zSuffix)==p->nSuffix ); + assert( strlen(p->zOutput)==p->nOutput ); + if( nBuf<p->nSuffix ) continue; + if( 0==memcmp(&aBuf[nBuf - p->nSuffix], p->zSuffix, p->nSuffix) ) break; + } + + if( p->zSuffix ){ + int nStem = nBuf - p->nSuffix; + if( p->xCond==0 || p->xCond(aBuf, nStem) ){ + memcpy(&aBuf[nStem], p->zOutput, p->nOutput); + *pnBuf = nStem + p->nOutput; + ret = p - aRule; + } + } + + return ret; +} +#endif + +static int fts5PorterIsVowel(char c, int bYIsVowel){ + return ( + c=='a' || c=='e' || c=='i' || c=='o' || c=='u' || (bYIsVowel && c=='y') + ); +} + +static int fts5PorterGobbleVC(char *zStem, int nStem, int bPrevCons){ + int i; + int bCons = bPrevCons; + + /* Scan for a vowel */ + for(i=0; i<nStem; i++){ + if( 0==(bCons = !fts5PorterIsVowel(zStem[i], bCons)) ) break; + } + + /* Scan for a consonent */ + for(i++; i<nStem; i++){ + if( (bCons = !fts5PorterIsVowel(zStem[i], bCons)) ) return i+1; + } + return 0; +} + +/* porter rule condition: (m > 0) */ +static int fts5Porter_MGt0(char *zStem, int nStem){ + return !!fts5PorterGobbleVC(zStem, nStem, 0); +} + +/* porter rule condition: (m > 1) */ +static int fts5Porter_MGt1(char *zStem, int nStem){ + int n; + n = fts5PorterGobbleVC(zStem, nStem, 0); + if( n && fts5PorterGobbleVC(&zStem[n], nStem-n, 1) ){ + return 1; + } + return 0; +} + +/* porter rule condition: (m = 1) */ +static int fts5Porter_MEq1(char *zStem, int nStem){ + int n; + n = fts5PorterGobbleVC(zStem, nStem, 0); + if( n && 0==fts5PorterGobbleVC(&zStem[n], nStem-n, 1) ){ + return 1; + } + return 0; +} + +/* porter rule condition: (*o) */ +static int fts5Porter_Ostar(char *zStem, int nStem){ + if( zStem[nStem-1]=='w' || zStem[nStem-1]=='x' || zStem[nStem-1]=='y' ){ + return 0; + }else{ + int i; + int mask = 0; + int bCons = 0; + for(i=0; i<nStem; i++){ + bCons = !fts5PorterIsVowel(zStem[i], bCons); + assert( bCons==0 || bCons==1 ); + mask = (mask << 1) + bCons; + } + return ((mask & 0x0007)==0x0005); + } +} + +/* porter rule condition: (m > 1 and (*S or *T)) */ +static int fts5Porter_MGt1_and_S_or_T(char *zStem, int nStem){ + assert( nStem>0 ); + return (zStem[nStem-1]=='s' || zStem[nStem-1]=='t') + && fts5Porter_MGt1(zStem, nStem); +} + +/* porter rule condition: (*v*) */ +static int fts5Porter_Vowel(char *zStem, int nStem){ + int i; + for(i=0; i<nStem; i++){ + if( fts5PorterIsVowel(zStem[i], i>0) ){ + return 1; + } + } + return 0; +} + + +/************************************************************************** +*************************************************************************** +** GENERATED CODE STARTS HERE (mkportersteps.tcl) +*/ + +static int fts5PorterStep4(char *aBuf, int *pnBuf){ + int ret = 0; + int nBuf = *pnBuf; + switch( aBuf[nBuf-2] ){ + + case 'a': + if( nBuf>2 && 0==memcmp("al", &aBuf[nBuf-2], 2) ){ + if( fts5Porter_MGt1(aBuf, nBuf-2) ){ + *pnBuf = nBuf - 2; + } + } + break; + + case 'c': + if( nBuf>4 && 0==memcmp("ance", &aBuf[nBuf-4], 4) ){ + if( fts5Porter_MGt1(aBuf, nBuf-4) ){ + *pnBuf = nBuf - 4; + } + }else if( nBuf>4 && 0==memcmp("ence", &aBuf[nBuf-4], 4) ){ + if( fts5Porter_MGt1(aBuf, nBuf-4) ){ + *pnBuf = nBuf - 4; + } + } + break; + + case 'e': + if( nBuf>2 && 0==memcmp("er", &aBuf[nBuf-2], 2) ){ + if( fts5Porter_MGt1(aBuf, nBuf-2) ){ + *pnBuf = nBuf - 2; + } + } + break; + + case 'i': + if( nBuf>2 && 0==memcmp("ic", &aBuf[nBuf-2], 2) ){ + if( fts5Porter_MGt1(aBuf, nBuf-2) ){ + *pnBuf = nBuf - 2; + } + } + break; + + case 'l': + if( nBuf>4 && 0==memcmp("able", &aBuf[nBuf-4], 4) ){ + if( fts5Porter_MGt1(aBuf, nBuf-4) ){ + *pnBuf = nBuf - 4; + } + }else if( nBuf>4 && 0==memcmp("ible", &aBuf[nBuf-4], 4) ){ + if( fts5Porter_MGt1(aBuf, nBuf-4) ){ + *pnBuf = nBuf - 4; + } + } + break; + + case 'n': + if( nBuf>3 && 0==memcmp("ant", &aBuf[nBuf-3], 3) ){ + if( fts5Porter_MGt1(aBuf, nBuf-3) ){ + *pnBuf = nBuf - 3; + } + }else if( nBuf>5 && 0==memcmp("ement", &aBuf[nBuf-5], 5) ){ + if( fts5Porter_MGt1(aBuf, nBuf-5) ){ + *pnBuf = nBuf - 5; + } + }else if( nBuf>4 && 0==memcmp("ment", &aBuf[nBuf-4], 4) ){ + if( fts5Porter_MGt1(aBuf, nBuf-4) ){ + *pnBuf = nBuf - 4; + } + }else if( nBuf>3 && 0==memcmp("ent", &aBuf[nBuf-3], 3) ){ + if( fts5Porter_MGt1(aBuf, nBuf-3) ){ + *pnBuf = nBuf - 3; + } + } + break; + + case 'o': + if( nBuf>3 && 0==memcmp("ion", &aBuf[nBuf-3], 3) ){ + if( fts5Porter_MGt1_and_S_or_T(aBuf, nBuf-3) ){ + *pnBuf = nBuf - 3; + } + }else if( nBuf>2 && 0==memcmp("ou", &aBuf[nBuf-2], 2) ){ + if( fts5Porter_MGt1(aBuf, nBuf-2) ){ + *pnBuf = nBuf - 2; + } + } + break; + + case 's': + if( nBuf>3 && 0==memcmp("ism", &aBuf[nBuf-3], 3) ){ + if( fts5Porter_MGt1(aBuf, nBuf-3) ){ + *pnBuf = nBuf - 3; + } + } + break; + + case 't': + if( nBuf>3 && 0==memcmp("ate", &aBuf[nBuf-3], 3) ){ + if( fts5Porter_MGt1(aBuf, nBuf-3) ){ + *pnBuf = nBuf - 3; + } + }else if( nBuf>3 && 0==memcmp("iti", &aBuf[nBuf-3], 3) ){ + if( fts5Porter_MGt1(aBuf, nBuf-3) ){ + *pnBuf = nBuf - 3; + } + } + break; + + case 'u': + if( nBuf>3 && 0==memcmp("ous", &aBuf[nBuf-3], 3) ){ + if( fts5Porter_MGt1(aBuf, nBuf-3) ){ + *pnBuf = nBuf - 3; + } + } + break; + + case 'v': + if( nBuf>3 && 0==memcmp("ive", &aBuf[nBuf-3], 3) ){ + if( fts5Porter_MGt1(aBuf, nBuf-3) ){ + *pnBuf = nBuf - 3; + } + } + break; + + case 'z': + if( nBuf>3 && 0==memcmp("ize", &aBuf[nBuf-3], 3) ){ + if( fts5Porter_MGt1(aBuf, nBuf-3) ){ + *pnBuf = nBuf - 3; + } + } + break; + + } + return ret; +} + + +static int fts5PorterStep1B2(char *aBuf, int *pnBuf){ + int ret = 0; + int nBuf = *pnBuf; + switch( aBuf[nBuf-2] ){ + + case 'a': + if( nBuf>2 && 0==memcmp("at", &aBuf[nBuf-2], 2) ){ + memcpy(&aBuf[nBuf-2], "ate", 3); + *pnBuf = nBuf - 2 + 3; + ret = 1; + } + break; + + case 'b': + if( nBuf>2 && 0==memcmp("bl", &aBuf[nBuf-2], 2) ){ + memcpy(&aBuf[nBuf-2], "ble", 3); + *pnBuf = nBuf - 2 + 3; + ret = 1; + } + break; + + case 'i': + if( nBuf>2 && 0==memcmp("iz", &aBuf[nBuf-2], 2) ){ + memcpy(&aBuf[nBuf-2], "ize", 3); + *pnBuf = nBuf - 2 + 3; + ret = 1; + } + break; + + } + return ret; +} + + +static int fts5PorterStep2(char *aBuf, int *pnBuf){ + int ret = 0; + int nBuf = *pnBuf; + switch( aBuf[nBuf-2] ){ + + case 'a': + if( nBuf>7 && 0==memcmp("ational", &aBuf[nBuf-7], 7) ){ + if( fts5Porter_MGt0(aBuf, nBuf-7) ){ + memcpy(&aBuf[nBuf-7], "ate", 3); + *pnBuf = nBuf - 7 + 3; + } + }else if( nBuf>6 && 0==memcmp("tional", &aBuf[nBuf-6], 6) ){ + if( fts5Porter_MGt0(aBuf, nBuf-6) ){ + memcpy(&aBuf[nBuf-6], "tion", 4); + *pnBuf = nBuf - 6 + 4; + } + } + break; + + case 'c': + if( nBuf>4 && 0==memcmp("enci", &aBuf[nBuf-4], 4) ){ + if( fts5Porter_MGt0(aBuf, nBuf-4) ){ + memcpy(&aBuf[nBuf-4], "ence", 4); + *pnBuf = nBuf - 4 + 4; + } + }else if( nBuf>4 && 0==memcmp("anci", &aBuf[nBuf-4], 4) ){ + if( fts5Porter_MGt0(aBuf, nBuf-4) ){ + memcpy(&aBuf[nBuf-4], "ance", 4); + *pnBuf = nBuf - 4 + 4; + } + } + break; + + case 'e': + if( nBuf>4 && 0==memcmp("izer", &aBuf[nBuf-4], 4) ){ + if( fts5Porter_MGt0(aBuf, nBuf-4) ){ + memcpy(&aBuf[nBuf-4], "ize", 3); + *pnBuf = nBuf - 4 + 3; + } + } + break; + + case 'g': + if( nBuf>4 && 0==memcmp("logi", &aBuf[nBuf-4], 4) ){ + if( fts5Porter_MGt0(aBuf, nBuf-4) ){ + memcpy(&aBuf[nBuf-4], "log", 3); + *pnBuf = nBuf - 4 + 3; + } + } + break; + + case 'l': + if( nBuf>3 && 0==memcmp("bli", &aBuf[nBuf-3], 3) ){ + if( fts5Porter_MGt0(aBuf, nBuf-3) ){ + memcpy(&aBuf[nBuf-3], "ble", 3); + *pnBuf = nBuf - 3 + 3; + } + }else if( nBuf>4 && 0==memcmp("alli", &aBuf[nBuf-4], 4) ){ + if( fts5Porter_MGt0(aBuf, nBuf-4) ){ + memcpy(&aBuf[nBuf-4], "al", 2); + *pnBuf = nBuf - 4 + 2; + } + }else if( nBuf>5 && 0==memcmp("entli", &aBuf[nBuf-5], 5) ){ + if( fts5Porter_MGt0(aBuf, nBuf-5) ){ + memcpy(&aBuf[nBuf-5], "ent", 3); + *pnBuf = nBuf - 5 + 3; + } + }else if( nBuf>3 && 0==memcmp("eli", &aBuf[nBuf-3], 3) ){ + if( fts5Porter_MGt0(aBuf, nBuf-3) ){ + memcpy(&aBuf[nBuf-3], "e", 1); + *pnBuf = nBuf - 3 + 1; + } + }else if( nBuf>5 && 0==memcmp("ousli", &aBuf[nBuf-5], 5) ){ + if( fts5Porter_MGt0(aBuf, nBuf-5) ){ + memcpy(&aBuf[nBuf-5], "ous", 3); + *pnBuf = nBuf - 5 + 3; + } + } + break; + + case 'o': + if( nBuf>7 && 0==memcmp("ization", &aBuf[nBuf-7], 7) ){ + if( fts5Porter_MGt0(aBuf, nBuf-7) ){ + memcpy(&aBuf[nBuf-7], "ize", 3); + *pnBuf = nBuf - 7 + 3; + } + }else if( nBuf>5 && 0==memcmp("ation", &aBuf[nBuf-5], 5) ){ + if( fts5Porter_MGt0(aBuf, nBuf-5) ){ + memcpy(&aBuf[nBuf-5], "ate", 3); + *pnBuf = nBuf - 5 + 3; + } + }else if( nBuf>4 && 0==memcmp("ator", &aBuf[nBuf-4], 4) ){ + if( fts5Porter_MGt0(aBuf, nBuf-4) ){ + memcpy(&aBuf[nBuf-4], "ate", 3); + *pnBuf = nBuf - 4 + 3; + } + } + break; + + case 's': + if( nBuf>5 && 0==memcmp("alism", &aBuf[nBuf-5], 5) ){ + if( fts5Porter_MGt0(aBuf, nBuf-5) ){ + memcpy(&aBuf[nBuf-5], "al", 2); + *pnBuf = nBuf - 5 + 2; + } + }else if( nBuf>7 && 0==memcmp("iveness", &aBuf[nBuf-7], 7) ){ + if( fts5Porter_MGt0(aBuf, nBuf-7) ){ + memcpy(&aBuf[nBuf-7], "ive", 3); + *pnBuf = nBuf - 7 + 3; + } + }else if( nBuf>7 && 0==memcmp("fulness", &aBuf[nBuf-7], 7) ){ + if( fts5Porter_MGt0(aBuf, nBuf-7) ){ + memcpy(&aBuf[nBuf-7], "ful", 3); + *pnBuf = nBuf - 7 + 3; + } + }else if( nBuf>7 && 0==memcmp("ousness", &aBuf[nBuf-7], 7) ){ + if( fts5Porter_MGt0(aBuf, nBuf-7) ){ + memcpy(&aBuf[nBuf-7], "ous", 3); + *pnBuf = nBuf - 7 + 3; + } + } + break; + + case 't': + if( nBuf>5 && 0==memcmp("aliti", &aBuf[nBuf-5], 5) ){ + if( fts5Porter_MGt0(aBuf, nBuf-5) ){ + memcpy(&aBuf[nBuf-5], "al", 2); + *pnBuf = nBuf - 5 + 2; + } + }else if( nBuf>5 && 0==memcmp("iviti", &aBuf[nBuf-5], 5) ){ + if( fts5Porter_MGt0(aBuf, nBuf-5) ){ + memcpy(&aBuf[nBuf-5], "ive", 3); + *pnBuf = nBuf - 5 + 3; + } + }else if( nBuf>6 && 0==memcmp("biliti", &aBuf[nBuf-6], 6) ){ + if( fts5Porter_MGt0(aBuf, nBuf-6) ){ + memcpy(&aBuf[nBuf-6], "ble", 3); + *pnBuf = nBuf - 6 + 3; + } + } + break; + + } + return ret; +} + + +static int fts5PorterStep3(char *aBuf, int *pnBuf){ + int ret = 0; + int nBuf = *pnBuf; + switch( aBuf[nBuf-2] ){ + + case 'a': + if( nBuf>4 && 0==memcmp("ical", &aBuf[nBuf-4], 4) ){ + if( fts5Porter_MGt0(aBuf, nBuf-4) ){ + memcpy(&aBuf[nBuf-4], "ic", 2); + *pnBuf = nBuf - 4 + 2; + } + } + break; + + case 's': + if( nBuf>4 && 0==memcmp("ness", &aBuf[nBuf-4], 4) ){ + if( fts5Porter_MGt0(aBuf, nBuf-4) ){ + *pnBuf = nBuf - 4; + } + } + break; + + case 't': + if( nBuf>5 && 0==memcmp("icate", &aBuf[nBuf-5], 5) ){ + if( fts5Porter_MGt0(aBuf, nBuf-5) ){ + memcpy(&aBuf[nBuf-5], "ic", 2); + *pnBuf = nBuf - 5 + 2; + } + }else if( nBuf>5 && 0==memcmp("iciti", &aBuf[nBuf-5], 5) ){ + if( fts5Porter_MGt0(aBuf, nBuf-5) ){ + memcpy(&aBuf[nBuf-5], "ic", 2); + *pnBuf = nBuf - 5 + 2; + } + } + break; + + case 'u': + if( nBuf>3 && 0==memcmp("ful", &aBuf[nBuf-3], 3) ){ + if( fts5Porter_MGt0(aBuf, nBuf-3) ){ + *pnBuf = nBuf - 3; + } + } + break; + + case 'v': + if( nBuf>5 && 0==memcmp("ative", &aBuf[nBuf-5], 5) ){ + if( fts5Porter_MGt0(aBuf, nBuf-5) ){ + *pnBuf = nBuf - 5; + } + } + break; + + case 'z': + if( nBuf>5 && 0==memcmp("alize", &aBuf[nBuf-5], 5) ){ + if( fts5Porter_MGt0(aBuf, nBuf-5) ){ + memcpy(&aBuf[nBuf-5], "al", 2); + *pnBuf = nBuf - 5 + 2; + } + } + break; + + } + return ret; +} + + +static int fts5PorterStep1B(char *aBuf, int *pnBuf){ + int ret = 0; + int nBuf = *pnBuf; + switch( aBuf[nBuf-2] ){ + + case 'e': + if( nBuf>3 && 0==memcmp("eed", &aBuf[nBuf-3], 3) ){ + if( fts5Porter_MGt0(aBuf, nBuf-3) ){ + memcpy(&aBuf[nBuf-3], "ee", 2); + *pnBuf = nBuf - 3 + 2; + } + }else if( nBuf>2 && 0==memcmp("ed", &aBuf[nBuf-2], 2) ){ + if( fts5Porter_Vowel(aBuf, nBuf-2) ){ + *pnBuf = nBuf - 2; + ret = 1; + } + } + break; + + case 'n': + if( nBuf>3 && 0==memcmp("ing", &aBuf[nBuf-3], 3) ){ + if( fts5Porter_Vowel(aBuf, nBuf-3) ){ + *pnBuf = nBuf - 3; + ret = 1; + } + } + break; + + } + return ret; +} + +/* +** GENERATED CODE ENDS HERE (mkportersteps.tcl) +*************************************************************************** +**************************************************************************/ + +static void fts5PorterStep1A(char *aBuf, int *pnBuf){ + int nBuf = *pnBuf; + if( aBuf[nBuf-1]=='s' ){ + if( aBuf[nBuf-2]=='e' ){ + if( (nBuf>4 && aBuf[nBuf-4]=='s' && aBuf[nBuf-3]=='s') + || (nBuf>3 && aBuf[nBuf-3]=='i' ) + ){ + *pnBuf = nBuf-2; + }else{ + *pnBuf = nBuf-1; + } + } + else if( aBuf[nBuf-2]!='s' ){ + *pnBuf = nBuf-1; + } + } +} + +static int fts5PorterCb( + void *pCtx, + int tflags, + const char *pToken, + int nToken, + int iStart, + int iEnd +){ + PorterContext *p = (PorterContext*)pCtx; + + char *aBuf; + int nBuf; + + if( nToken>FTS5_PORTER_MAX_TOKEN || nToken<3 ) goto pass_through; + aBuf = p->aBuf; + nBuf = nToken; + memcpy(aBuf, pToken, nBuf); + + /* Step 1. */ + fts5PorterStep1A(aBuf, &nBuf); + if( fts5PorterStep1B(aBuf, &nBuf) ){ + if( fts5PorterStep1B2(aBuf, &nBuf)==0 ){ + char c = aBuf[nBuf-1]; + if( fts5PorterIsVowel(c, 0)==0 + && c!='l' && c!='s' && c!='z' && c==aBuf[nBuf-2] + ){ + nBuf--; + }else if( fts5Porter_MEq1(aBuf, nBuf) && fts5Porter_Ostar(aBuf, nBuf) ){ + aBuf[nBuf++] = 'e'; + } + } + } + + /* Step 1C. */ + if( aBuf[nBuf-1]=='y' && fts5Porter_Vowel(aBuf, nBuf-1) ){ + aBuf[nBuf-1] = 'i'; + } + + /* Steps 2 through 4. */ + fts5PorterStep2(aBuf, &nBuf); + fts5PorterStep3(aBuf, &nBuf); + fts5PorterStep4(aBuf, &nBuf); + + /* Step 5a. */ + assert( nBuf>0 ); + if( aBuf[nBuf-1]=='e' ){ + if( fts5Porter_MGt1(aBuf, nBuf-1) + || (fts5Porter_MEq1(aBuf, nBuf-1) && !fts5Porter_Ostar(aBuf, nBuf-1)) + ){ + nBuf--; + } + } + + /* Step 5b. */ + if( nBuf>1 && aBuf[nBuf-1]=='l' + && aBuf[nBuf-2]=='l' && fts5Porter_MGt1(aBuf, nBuf-1) + ){ + nBuf--; + } + + return p->xToken(p->pCtx, tflags, aBuf, nBuf, iStart, iEnd); + + pass_through: + return p->xToken(p->pCtx, tflags, pToken, nToken, iStart, iEnd); +} + +/* +** Tokenize using the porter tokenizer. +*/ +static int fts5PorterTokenize( + Fts5Tokenizer *pTokenizer, + void *pCtx, + int flags, + const char *pText, int nText, + int (*xToken)(void*, int, const char*, int nToken, int iStart, int iEnd) +){ + PorterTokenizer *p = (PorterTokenizer*)pTokenizer; + PorterContext sCtx; + sCtx.xToken = xToken; + sCtx.pCtx = pCtx; + sCtx.aBuf = p->aBuf; + return p->tokenizer.xTokenize( + p->pTokenizer, (void*)&sCtx, flags, pText, nText, fts5PorterCb + ); +} + +/* +** Register all built-in tokenizers with FTS5. +*/ +static int sqlite3Fts5TokenizerInit(fts5_api *pApi){ + struct BuiltinTokenizer { + const char *zName; + fts5_tokenizer x; + } aBuiltin[] = { + { "unicode61", {fts5UnicodeCreate, fts5UnicodeDelete, fts5UnicodeTokenize}}, + { "ascii", {fts5AsciiCreate, fts5AsciiDelete, fts5AsciiTokenize }}, + { "porter", {fts5PorterCreate, fts5PorterDelete, fts5PorterTokenize }}, + }; + + int rc = SQLITE_OK; /* Return code */ + int i; /* To iterate through builtin functions */ + + for(i=0; rc==SQLITE_OK && i<ArraySize(aBuiltin); i++){ + rc = pApi->xCreateTokenizer(pApi, + aBuiltin[i].zName, + (void*)pApi, + &aBuiltin[i].x, + 0 + ); + } + + return rc; +} + + + +#line 1 "fts5_unicode2.c" +/* +** 2012 May 25 +** +** The author disclaims copyright to this source code. In place of +** a legal notice, here is a blessing: +** +** May you do good and not evil. +** May you find forgiveness for yourself and forgive others. +** May you share freely, never taking more than you give. +** +****************************************************************************** +*/ + +/* +** DO NOT EDIT THIS MACHINE GENERATED FILE. +*/ + + +#include <assert.h> + +/* +** Return true if the argument corresponds to a unicode codepoint +** classified as either a letter or a number. Otherwise false. +** +** The results are undefined if the value passed to this function +** is less than zero. +*/ +static int sqlite3Fts5UnicodeIsalnum(int c){ + /* Each unsigned integer in the following array corresponds to a contiguous + ** range of unicode codepoints that are not either letters or numbers (i.e. + ** codepoints for which this function should return 0). + ** + ** The most significant 22 bits in each 32-bit value contain the first + ** codepoint in the range. The least significant 10 bits are used to store + ** the size of the range (always at least 1). In other words, the value + ** ((C<<22) + N) represents a range of N codepoints starting with codepoint + ** C. It is not possible to represent a range larger than 1023 codepoints + ** using this format. + */ + static const unsigned int aEntry[] = { + 0x00000030, 0x0000E807, 0x00016C06, 0x0001EC2F, 0x0002AC07, + 0x0002D001, 0x0002D803, 0x0002EC01, 0x0002FC01, 0x00035C01, + 0x0003DC01, 0x000B0804, 0x000B480E, 0x000B9407, 0x000BB401, + 0x000BBC81, 0x000DD401, 0x000DF801, 0x000E1002, 0x000E1C01, + 0x000FD801, 0x00120808, 0x00156806, 0x00162402, 0x00163C01, + 0x00164437, 0x0017CC02, 0x00180005, 0x00181816, 0x00187802, + 0x00192C15, 0x0019A804, 0x0019C001, 0x001B5001, 0x001B580F, + 0x001B9C07, 0x001BF402, 0x001C000E, 0x001C3C01, 0x001C4401, + 0x001CC01B, 0x001E980B, 0x001FAC09, 0x001FD804, 0x00205804, + 0x00206C09, 0x00209403, 0x0020A405, 0x0020C00F, 0x00216403, + 0x00217801, 0x0023901B, 0x00240004, 0x0024E803, 0x0024F812, + 0x00254407, 0x00258804, 0x0025C001, 0x00260403, 0x0026F001, + 0x0026F807, 0x00271C02, 0x00272C03, 0x00275C01, 0x00278802, + 0x0027C802, 0x0027E802, 0x00280403, 0x0028F001, 0x0028F805, + 0x00291C02, 0x00292C03, 0x00294401, 0x0029C002, 0x0029D401, + 0x002A0403, 0x002AF001, 0x002AF808, 0x002B1C03, 0x002B2C03, + 0x002B8802, 0x002BC002, 0x002C0403, 0x002CF001, 0x002CF807, + 0x002D1C02, 0x002D2C03, 0x002D5802, 0x002D8802, 0x002DC001, + 0x002E0801, 0x002EF805, 0x002F1803, 0x002F2804, 0x002F5C01, + 0x002FCC08, 0x00300403, 0x0030F807, 0x00311803, 0x00312804, + 0x00315402, 0x00318802, 0x0031FC01, 0x00320802, 0x0032F001, + 0x0032F807, 0x00331803, 0x00332804, 0x00335402, 0x00338802, + 0x00340802, 0x0034F807, 0x00351803, 0x00352804, 0x00355C01, + 0x00358802, 0x0035E401, 0x00360802, 0x00372801, 0x00373C06, + 0x00375801, 0x00376008, 0x0037C803, 0x0038C401, 0x0038D007, + 0x0038FC01, 0x00391C09, 0x00396802, 0x003AC401, 0x003AD006, + 0x003AEC02, 0x003B2006, 0x003C041F, 0x003CD00C, 0x003DC417, + 0x003E340B, 0x003E6424, 0x003EF80F, 0x003F380D, 0x0040AC14, + 0x00412806, 0x00415804, 0x00417803, 0x00418803, 0x00419C07, + 0x0041C404, 0x0042080C, 0x00423C01, 0x00426806, 0x0043EC01, + 0x004D740C, 0x004E400A, 0x00500001, 0x0059B402, 0x005A0001, + 0x005A6C02, 0x005BAC03, 0x005C4803, 0x005CC805, 0x005D4802, + 0x005DC802, 0x005ED023, 0x005F6004, 0x005F7401, 0x0060000F, + 0x0062A401, 0x0064800C, 0x0064C00C, 0x00650001, 0x00651002, + 0x0066C011, 0x00672002, 0x00677822, 0x00685C05, 0x00687802, + 0x0069540A, 0x0069801D, 0x0069FC01, 0x006A8007, 0x006AA006, + 0x006C0005, 0x006CD011, 0x006D6823, 0x006E0003, 0x006E840D, + 0x006F980E, 0x006FF004, 0x00709014, 0x0070EC05, 0x0071F802, + 0x00730008, 0x00734019, 0x0073B401, 0x0073C803, 0x00770027, + 0x0077F004, 0x007EF401, 0x007EFC03, 0x007F3403, 0x007F7403, + 0x007FB403, 0x007FF402, 0x00800065, 0x0081A806, 0x0081E805, + 0x00822805, 0x0082801A, 0x00834021, 0x00840002, 0x00840C04, + 0x00842002, 0x00845001, 0x00845803, 0x00847806, 0x00849401, + 0x00849C01, 0x0084A401, 0x0084B801, 0x0084E802, 0x00850005, + 0x00852804, 0x00853C01, 0x00864264, 0x00900027, 0x0091000B, + 0x0092704E, 0x00940200, 0x009C0475, 0x009E53B9, 0x00AD400A, + 0x00B39406, 0x00B3BC03, 0x00B3E404, 0x00B3F802, 0x00B5C001, + 0x00B5FC01, 0x00B7804F, 0x00B8C00C, 0x00BA001A, 0x00BA6C59, + 0x00BC00D6, 0x00BFC00C, 0x00C00005, 0x00C02019, 0x00C0A807, + 0x00C0D802, 0x00C0F403, 0x00C26404, 0x00C28001, 0x00C3EC01, + 0x00C64002, 0x00C6580A, 0x00C70024, 0x00C8001F, 0x00C8A81E, + 0x00C94001, 0x00C98020, 0x00CA2827, 0x00CB003F, 0x00CC0100, + 0x01370040, 0x02924037, 0x0293F802, 0x02983403, 0x0299BC10, + 0x029A7C01, 0x029BC008, 0x029C0017, 0x029C8002, 0x029E2402, + 0x02A00801, 0x02A01801, 0x02A02C01, 0x02A08C09, 0x02A0D804, + 0x02A1D004, 0x02A20002, 0x02A2D011, 0x02A33802, 0x02A38012, + 0x02A3E003, 0x02A4980A, 0x02A51C0D, 0x02A57C01, 0x02A60004, + 0x02A6CC1B, 0x02A77802, 0x02A8A40E, 0x02A90C01, 0x02A93002, + 0x02A97004, 0x02A9DC03, 0x02A9EC01, 0x02AAC001, 0x02AAC803, + 0x02AADC02, 0x02AAF802, 0x02AB0401, 0x02AB7802, 0x02ABAC07, + 0x02ABD402, 0x02AF8C0B, 0x03600001, 0x036DFC02, 0x036FFC02, + 0x037FFC01, 0x03EC7801, 0x03ECA401, 0x03EEC810, 0x03F4F802, + 0x03F7F002, 0x03F8001A, 0x03F88007, 0x03F8C023, 0x03F95013, + 0x03F9A004, 0x03FBFC01, 0x03FC040F, 0x03FC6807, 0x03FCEC06, + 0x03FD6C0B, 0x03FF8007, 0x03FFA007, 0x03FFE405, 0x04040003, + 0x0404DC09, 0x0405E411, 0x0406400C, 0x0407402E, 0x040E7C01, + 0x040F4001, 0x04215C01, 0x04247C01, 0x0424FC01, 0x04280403, + 0x04281402, 0x04283004, 0x0428E003, 0x0428FC01, 0x04294009, + 0x0429FC01, 0x042CE407, 0x04400003, 0x0440E016, 0x04420003, + 0x0442C012, 0x04440003, 0x04449C0E, 0x04450004, 0x04460003, + 0x0446CC0E, 0x04471404, 0x045AAC0D, 0x0491C004, 0x05BD442E, + 0x05BE3C04, 0x074000F6, 0x07440027, 0x0744A4B5, 0x07480046, + 0x074C0057, 0x075B0401, 0x075B6C01, 0x075BEC01, 0x075C5401, + 0x075CD401, 0x075D3C01, 0x075DBC01, 0x075E2401, 0x075EA401, + 0x075F0C01, 0x07BBC002, 0x07C0002C, 0x07C0C064, 0x07C2800F, + 0x07C2C40E, 0x07C3040F, 0x07C3440F, 0x07C4401F, 0x07C4C03C, + 0x07C5C02B, 0x07C7981D, 0x07C8402B, 0x07C90009, 0x07C94002, + 0x07CC0021, 0x07CCC006, 0x07CCDC46, 0x07CE0014, 0x07CE8025, + 0x07CF1805, 0x07CF8011, 0x07D0003F, 0x07D10001, 0x07D108B6, + 0x07D3E404, 0x07D4003E, 0x07D50004, 0x07D54018, 0x07D7EC46, + 0x07D9140B, 0x07DA0046, 0x07DC0074, 0x38000401, 0x38008060, + 0x380400F0, + }; + static const unsigned int aAscii[4] = { + 0xFFFFFFFF, 0xFC00FFFF, 0xF8000001, 0xF8000001, + }; + + if( (unsigned int)c<128 ){ + return ( (aAscii[c >> 5] & (1 << (c & 0x001F)))==0 ); + }else if( (unsigned int)c<(1<<22) ){ + unsigned int key = (((unsigned int)c)<<10) | 0x000003FF; + int iRes = 0; + int iHi = sizeof(aEntry)/sizeof(aEntry[0]) - 1; + int iLo = 0; + while( iHi>=iLo ){ + int iTest = (iHi + iLo) / 2; + if( key >= aEntry[iTest] ){ + iRes = iTest; + iLo = iTest+1; + }else{ + iHi = iTest-1; + } + } + assert( aEntry[0]<key ); + assert( key>=aEntry[iRes] ); + return (((unsigned int)c) >= ((aEntry[iRes]>>10) + (aEntry[iRes]&0x3FF))); + } + return 1; +} + + +/* +** If the argument is a codepoint corresponding to a lowercase letter +** in the ASCII range with a diacritic added, return the codepoint +** of the ASCII letter only. For example, if passed 235 - "LATIN +** SMALL LETTER E WITH DIAERESIS" - return 65 ("LATIN SMALL LETTER +** E"). The resuls of passing a codepoint that corresponds to an +** uppercase letter are undefined. +*/ +static int fts5_remove_diacritic(int c){ + unsigned short aDia[] = { + 0, 1797, 1848, 1859, 1891, 1928, 1940, 1995, + 2024, 2040, 2060, 2110, 2168, 2206, 2264, 2286, + 2344, 2383, 2472, 2488, 2516, 2596, 2668, 2732, + 2782, 2842, 2894, 2954, 2984, 3000, 3028, 3336, + 3456, 3696, 3712, 3728, 3744, 3896, 3912, 3928, + 3968, 4008, 4040, 4106, 4138, 4170, 4202, 4234, + 4266, 4296, 4312, 4344, 4408, 4424, 4472, 4504, + 6148, 6198, 6264, 6280, 6360, 6429, 6505, 6529, + 61448, 61468, 61534, 61592, 61642, 61688, 61704, 61726, + 61784, 61800, 61836, 61880, 61914, 61948, 61998, 62122, + 62154, 62200, 62218, 62302, 62364, 62442, 62478, 62536, + 62554, 62584, 62604, 62640, 62648, 62656, 62664, 62730, + 62924, 63050, 63082, 63274, 63390, + }; + char aChar[] = { + '\0', 'a', 'c', 'e', 'i', 'n', 'o', 'u', 'y', 'y', 'a', 'c', + 'd', 'e', 'e', 'g', 'h', 'i', 'j', 'k', 'l', 'n', 'o', 'r', + 's', 't', 'u', 'u', 'w', 'y', 'z', 'o', 'u', 'a', 'i', 'o', + 'u', 'g', 'k', 'o', 'j', 'g', 'n', 'a', 'e', 'i', 'o', 'r', + 'u', 's', 't', 'h', 'a', 'e', 'o', 'y', '\0', '\0', '\0', '\0', + '\0', '\0', '\0', '\0', 'a', 'b', 'd', 'd', 'e', 'f', 'g', 'h', + 'h', 'i', 'k', 'l', 'l', 'm', 'n', 'p', 'r', 'r', 's', 't', + 'u', 'v', 'w', 'w', 'x', 'y', 'z', 'h', 't', 'w', 'y', 'a', + 'e', 'i', 'o', 'u', 'y', + }; + + unsigned int key = (((unsigned int)c)<<3) | 0x00000007; + int iRes = 0; + int iHi = sizeof(aDia)/sizeof(aDia[0]) - 1; + int iLo = 0; + while( iHi>=iLo ){ + int iTest = (iHi + iLo) / 2; + if( key >= aDia[iTest] ){ + iRes = iTest; + iLo = iTest+1; + }else{ + iHi = iTest-1; + } + } + assert( key>=aDia[iRes] ); + return ((c > (aDia[iRes]>>3) + (aDia[iRes]&0x07)) ? c : (int)aChar[iRes]); +} + + +/* +** Return true if the argument interpreted as a unicode codepoint +** is a diacritical modifier character. +*/ +static int sqlite3Fts5UnicodeIsdiacritic(int c){ + unsigned int mask0 = 0x08029FDF; + unsigned int mask1 = 0x000361F8; + if( c<768 || c>817 ) return 0; + return (c < 768+32) ? + (mask0 & (1 << (c-768))) : + (mask1 & (1 << (c-768-32))); +} + + +/* +** Interpret the argument as a unicode codepoint. If the codepoint +** is an upper case character that has a lower case equivalent, +** return the codepoint corresponding to the lower case version. +** Otherwise, return a copy of the argument. +** +** The results are undefined if the value passed to this function +** is less than zero. +*/ +static int sqlite3Fts5UnicodeFold(int c, int bRemoveDiacritic){ + /* Each entry in the following array defines a rule for folding a range + ** of codepoints to lower case. The rule applies to a range of nRange + ** codepoints starting at codepoint iCode. + ** + ** If the least significant bit in flags is clear, then the rule applies + ** to all nRange codepoints (i.e. all nRange codepoints are upper case and + ** need to be folded). Or, if it is set, then the rule only applies to + ** every second codepoint in the range, starting with codepoint C. + ** + ** The 7 most significant bits in flags are an index into the aiOff[] + ** array. If a specific codepoint C does require folding, then its lower + ** case equivalent is ((C + aiOff[flags>>1]) & 0xFFFF). + ** + ** The contents of this array are generated by parsing the CaseFolding.txt + ** file distributed as part of the "Unicode Character Database". See + ** http://www.unicode.org for details. + */ + static const struct TableEntry { + unsigned short iCode; + unsigned char flags; + unsigned char nRange; + } aEntry[] = { + {65, 14, 26}, {181, 64, 1}, {192, 14, 23}, + {216, 14, 7}, {256, 1, 48}, {306, 1, 6}, + {313, 1, 16}, {330, 1, 46}, {376, 116, 1}, + {377, 1, 6}, {383, 104, 1}, {385, 50, 1}, + {386, 1, 4}, {390, 44, 1}, {391, 0, 1}, + {393, 42, 2}, {395, 0, 1}, {398, 32, 1}, + {399, 38, 1}, {400, 40, 1}, {401, 0, 1}, + {403, 42, 1}, {404, 46, 1}, {406, 52, 1}, + {407, 48, 1}, {408, 0, 1}, {412, 52, 1}, + {413, 54, 1}, {415, 56, 1}, {416, 1, 6}, + {422, 60, 1}, {423, 0, 1}, {425, 60, 1}, + {428, 0, 1}, {430, 60, 1}, {431, 0, 1}, + {433, 58, 2}, {435, 1, 4}, {439, 62, 1}, + {440, 0, 1}, {444, 0, 1}, {452, 2, 1}, + {453, 0, 1}, {455, 2, 1}, {456, 0, 1}, + {458, 2, 1}, {459, 1, 18}, {478, 1, 18}, + {497, 2, 1}, {498, 1, 4}, {502, 122, 1}, + {503, 134, 1}, {504, 1, 40}, {544, 110, 1}, + {546, 1, 18}, {570, 70, 1}, {571, 0, 1}, + {573, 108, 1}, {574, 68, 1}, {577, 0, 1}, + {579, 106, 1}, {580, 28, 1}, {581, 30, 1}, + {582, 1, 10}, {837, 36, 1}, {880, 1, 4}, + {886, 0, 1}, {902, 18, 1}, {904, 16, 3}, + {908, 26, 1}, {910, 24, 2}, {913, 14, 17}, + {931, 14, 9}, {962, 0, 1}, {975, 4, 1}, + {976, 140, 1}, {977, 142, 1}, {981, 146, 1}, + {982, 144, 1}, {984, 1, 24}, {1008, 136, 1}, + {1009, 138, 1}, {1012, 130, 1}, {1013, 128, 1}, + {1015, 0, 1}, {1017, 152, 1}, {1018, 0, 1}, + {1021, 110, 3}, {1024, 34, 16}, {1040, 14, 32}, + {1120, 1, 34}, {1162, 1, 54}, {1216, 6, 1}, + {1217, 1, 14}, {1232, 1, 88}, {1329, 22, 38}, + {4256, 66, 38}, {4295, 66, 1}, {4301, 66, 1}, + {7680, 1, 150}, {7835, 132, 1}, {7838, 96, 1}, + {7840, 1, 96}, {7944, 150, 8}, {7960, 150, 6}, + {7976, 150, 8}, {7992, 150, 8}, {8008, 150, 6}, + {8025, 151, 8}, {8040, 150, 8}, {8072, 150, 8}, + {8088, 150, 8}, {8104, 150, 8}, {8120, 150, 2}, + {8122, 126, 2}, {8124, 148, 1}, {8126, 100, 1}, + {8136, 124, 4}, {8140, 148, 1}, {8152, 150, 2}, + {8154, 120, 2}, {8168, 150, 2}, {8170, 118, 2}, + {8172, 152, 1}, {8184, 112, 2}, {8186, 114, 2}, + {8188, 148, 1}, {8486, 98, 1}, {8490, 92, 1}, + {8491, 94, 1}, {8498, 12, 1}, {8544, 8, 16}, + {8579, 0, 1}, {9398, 10, 26}, {11264, 22, 47}, + {11360, 0, 1}, {11362, 88, 1}, {11363, 102, 1}, + {11364, 90, 1}, {11367, 1, 6}, {11373, 84, 1}, + {11374, 86, 1}, {11375, 80, 1}, {11376, 82, 1}, + {11378, 0, 1}, {11381, 0, 1}, {11390, 78, 2}, + {11392, 1, 100}, {11499, 1, 4}, {11506, 0, 1}, + {42560, 1, 46}, {42624, 1, 24}, {42786, 1, 14}, + {42802, 1, 62}, {42873, 1, 4}, {42877, 76, 1}, + {42878, 1, 10}, {42891, 0, 1}, {42893, 74, 1}, + {42896, 1, 4}, {42912, 1, 10}, {42922, 72, 1}, + {65313, 14, 26}, + }; + static const unsigned short aiOff[] = { + 1, 2, 8, 15, 16, 26, 28, 32, + 37, 38, 40, 48, 63, 64, 69, 71, + 79, 80, 116, 202, 203, 205, 206, 207, + 209, 210, 211, 213, 214, 217, 218, 219, + 775, 7264, 10792, 10795, 23228, 23256, 30204, 54721, + 54753, 54754, 54756, 54787, 54793, 54809, 57153, 57274, + 57921, 58019, 58363, 61722, 65268, 65341, 65373, 65406, + 65408, 65410, 65415, 65424, 65436, 65439, 65450, 65462, + 65472, 65476, 65478, 65480, 65482, 65488, 65506, 65511, + 65514, 65521, 65527, 65528, 65529, + }; + + int ret = c; + + assert( sizeof(unsigned short)==2 && sizeof(unsigned char)==1 ); + + if( c<128 ){ + if( c>='A' && c<='Z' ) ret = c + ('a' - 'A'); + }else if( c<65536 ){ + const struct TableEntry *p; + int iHi = sizeof(aEntry)/sizeof(aEntry[0]) - 1; + int iLo = 0; + int iRes = -1; + + assert( c>aEntry[0].iCode ); + while( iHi>=iLo ){ + int iTest = (iHi + iLo) / 2; + int cmp = (c - aEntry[iTest].iCode); + if( cmp>=0 ){ + iRes = iTest; + iLo = iTest+1; + }else{ + iHi = iTest-1; + } + } + + assert( iRes>=0 && c>=aEntry[iRes].iCode ); + p = &aEntry[iRes]; + if( c<(p->iCode + p->nRange) && 0==(0x01 & p->flags & (p->iCode ^ c)) ){ + ret = (c + (aiOff[p->flags>>1])) & 0x0000FFFF; + assert( ret>0 ); + } + + if( bRemoveDiacritic ) ret = fts5_remove_diacritic(ret); + } + + else if( c>=66560 && c<66600 ){ + ret = c + 40; + } + + return ret; +} + +#line 1 "fts5_varint.c" +/* +** 2015 May 30 +** +** The author disclaims copyright to this source code. In place of +** a legal notice, here is a blessing: +** +** May you do good and not evil. +** May you find forgiveness for yourself and forgive others. +** May you share freely, never taking more than you give. +** +****************************************************************************** +** +** Routines for varint serialization and deserialization. +*/ + + +/* #include "fts5Int.h" */ + +/* +** This is a copy of the sqlite3GetVarint32() routine from the SQLite core. +** Except, this version does handle the single byte case that the core +** version depends on being handled before its function is called. +*/ +static int sqlite3Fts5GetVarint32(const unsigned char *p, u32 *v){ + u32 a,b; + + /* The 1-byte case. Overwhelmingly the most common. */ + a = *p; + /* a: p0 (unmasked) */ + if (!(a&0x80)) + { + /* Values between 0 and 127 */ + *v = a; + return 1; + } + + /* The 2-byte case */ + p++; + b = *p; + /* b: p1 (unmasked) */ + if (!(b&0x80)) + { + /* Values between 128 and 16383 */ + a &= 0x7f; + a = a<<7; + *v = a | b; + return 2; + } + + /* The 3-byte case */ + p++; + a = a<<14; + a |= *p; + /* a: p0<<14 | p2 (unmasked) */ + if (!(a&0x80)) + { + /* Values between 16384 and 2097151 */ + a &= (0x7f<<14)|(0x7f); + b &= 0x7f; + b = b<<7; + *v = a | b; + return 3; + } + + /* A 32-bit varint is used to store size information in btrees. + ** Objects are rarely larger than 2MiB limit of a 3-byte varint. + ** A 3-byte varint is sufficient, for example, to record the size + ** of a 1048569-byte BLOB or string. + ** + ** We only unroll the first 1-, 2-, and 3- byte cases. The very + ** rare larger cases can be handled by the slower 64-bit varint + ** routine. + */ + { + u64 v64; + u8 n; + p -= 2; + n = sqlite3Fts5GetVarint(p, &v64); + *v = (u32)v64; + assert( n>3 && n<=9 ); + return n; + } +} + + +/* +** Bitmasks used by sqlite3GetVarint(). These precomputed constants +** are defined here rather than simply putting the constant expressions +** inline in order to work around bugs in the RVT compiler. +** +** SLOT_2_0 A mask for (0x7f<<14) | 0x7f +** +** SLOT_4_2_0 A mask for (0x7f<<28) | SLOT_2_0 +*/ +#define SLOT_2_0 0x001fc07f +#define SLOT_4_2_0 0xf01fc07f + +/* +** Read a 64-bit variable-length integer from memory starting at p[0]. +** Return the number of bytes read. The value is stored in *v. +*/ +static u8 sqlite3Fts5GetVarint(const unsigned char *p, u64 *v){ + u32 a,b,s; + + a = *p; + /* a: p0 (unmasked) */ + if (!(a&0x80)) + { + *v = a; + return 1; + } + + p++; + b = *p; + /* b: p1 (unmasked) */ + if (!(b&0x80)) + { + a &= 0x7f; + a = a<<7; + a |= b; + *v = a; + return 2; + } + + /* Verify that constants are precomputed correctly */ + assert( SLOT_2_0 == ((0x7f<<14) | (0x7f)) ); + assert( SLOT_4_2_0 == ((0xfU<<28) | (0x7f<<14) | (0x7f)) ); + + p++; + a = a<<14; + a |= *p; + /* a: p0<<14 | p2 (unmasked) */ + if (!(a&0x80)) + { + a &= SLOT_2_0; + b &= 0x7f; + b = b<<7; + a |= b; + *v = a; + return 3; + } + + /* CSE1 from below */ + a &= SLOT_2_0; + p++; + b = b<<14; + b |= *p; + /* b: p1<<14 | p3 (unmasked) */ + if (!(b&0x80)) + { + b &= SLOT_2_0; + /* moved CSE1 up */ + /* a &= (0x7f<<14)|(0x7f); */ + a = a<<7; + a |= b; + *v = a; + return 4; + } + + /* a: p0<<14 | p2 (masked) */ + /* b: p1<<14 | p3 (unmasked) */ + /* 1:save off p0<<21 | p1<<14 | p2<<7 | p3 (masked) */ + /* moved CSE1 up */ + /* a &= (0x7f<<14)|(0x7f); */ + b &= SLOT_2_0; + s = a; + /* s: p0<<14 | p2 (masked) */ + + p++; + a = a<<14; + a |= *p; + /* a: p0<<28 | p2<<14 | p4 (unmasked) */ + if (!(a&0x80)) + { + /* we can skip these cause they were (effectively) done above in calc'ing s */ + /* a &= (0x7f<<28)|(0x7f<<14)|(0x7f); */ + /* b &= (0x7f<<14)|(0x7f); */ + b = b<<7; + a |= b; + s = s>>18; + *v = ((u64)s)<<32 | a; + return 5; + } + + /* 2:save off p0<<21 | p1<<14 | p2<<7 | p3 (masked) */ + s = s<<7; + s |= b; + /* s: p0<<21 | p1<<14 | p2<<7 | p3 (masked) */ + + p++; + b = b<<14; + b |= *p; + /* b: p1<<28 | p3<<14 | p5 (unmasked) */ + if (!(b&0x80)) + { + /* we can skip this cause it was (effectively) done above in calc'ing s */ + /* b &= (0x7f<<28)|(0x7f<<14)|(0x7f); */ + a &= SLOT_2_0; + a = a<<7; + a |= b; + s = s>>18; + *v = ((u64)s)<<32 | a; + return 6; + } + + p++; + a = a<<14; + a |= *p; + /* a: p2<<28 | p4<<14 | p6 (unmasked) */ + if (!(a&0x80)) + { + a &= SLOT_4_2_0; + b &= SLOT_2_0; + b = b<<7; + a |= b; + s = s>>11; + *v = ((u64)s)<<32 | a; + return 7; + } + + /* CSE2 from below */ + a &= SLOT_2_0; + p++; + b = b<<14; + b |= *p; + /* b: p3<<28 | p5<<14 | p7 (unmasked) */ + if (!(b&0x80)) + { + b &= SLOT_4_2_0; + /* moved CSE2 up */ + /* a &= (0x7f<<14)|(0x7f); */ + a = a<<7; + a |= b; + s = s>>4; + *v = ((u64)s)<<32 | a; + return 8; + } + + p++; + a = a<<15; + a |= *p; + /* a: p4<<29 | p6<<15 | p8 (unmasked) */ + + /* moved CSE2 up */ + /* a &= (0x7f<<29)|(0x7f<<15)|(0xff); */ + b &= SLOT_2_0; + b = b<<8; + a |= b; + + s = s<<4; + b = p[-4]; + b &= 0x7f; + b = b>>3; + s |= b; + + *v = ((u64)s)<<32 | a; + + return 9; +} + +/* +** The variable-length integer encoding is as follows: +** +** KEY: +** A = 0xxxxxxx 7 bits of data and one flag bit +** B = 1xxxxxxx 7 bits of data and one flag bit +** C = xxxxxxxx 8 bits of data +** +** 7 bits - A +** 14 bits - BA +** 21 bits - BBA +** 28 bits - BBBA +** 35 bits - BBBBA +** 42 bits - BBBBBA +** 49 bits - BBBBBBA +** 56 bits - BBBBBBBA +** 64 bits - BBBBBBBBC +*/ + +#ifdef SQLITE_NOINLINE +# define FTS5_NOINLINE SQLITE_NOINLINE +#else +# define FTS5_NOINLINE +#endif + +/* +** Write a 64-bit variable-length integer to memory starting at p[0]. +** The length of data write will be between 1 and 9 bytes. The number +** of bytes written is returned. +** +** A variable-length integer consists of the lower 7 bits of each byte +** for all bytes that have the 8th bit set and one byte with the 8th +** bit clear. Except, if we get to the 9th byte, it stores the full +** 8 bits and is the last byte. +*/ +static int FTS5_NOINLINE fts5PutVarint64(unsigned char *p, u64 v){ + int i, j, n; + u8 buf[10]; + if( v & (((u64)0xff000000)<<32) ){ + p[8] = (u8)v; + v >>= 8; + for(i=7; i>=0; i--){ + p[i] = (u8)((v & 0x7f) | 0x80); + v >>= 7; + } + return 9; + } + n = 0; + do{ + buf[n++] = (u8)((v & 0x7f) | 0x80); + v >>= 7; + }while( v!=0 ); + buf[0] &= 0x7f; + assert( n<=9 ); + for(i=0, j=n-1; j>=0; j--, i++){ + p[i] = buf[j]; + } + return n; +} + +static int sqlite3Fts5PutVarint(unsigned char *p, u64 v){ + if( v<=0x7f ){ + p[0] = v&0x7f; + return 1; + } + if( v<=0x3fff ){ + p[0] = ((v>>7)&0x7f)|0x80; + p[1] = v&0x7f; + return 2; + } + return fts5PutVarint64(p,v); +} + + +static int sqlite3Fts5GetVarintLen(u32 iVal){ +#if 0 + if( iVal<(1 << 7 ) ) return 1; +#endif + assert( iVal>=(1 << 7) ); + if( iVal<(1 << 14) ) return 2; + if( iVal<(1 << 21) ) return 3; + if( iVal<(1 << 28) ) return 4; + return 5; +} + + +#line 1 "fts5_vocab.c" +/* +** 2015 May 08 +** +** The author disclaims copyright to this source code. In place of +** a legal notice, here is a blessing: +** +** May you do good and not evil. +** May you find forgiveness for yourself and forgive others. +** May you share freely, never taking more than you give. +** +****************************************************************************** +** +** This is an SQLite virtual table module implementing direct access to an +** existing FTS5 index. The module may create several different types of +** tables: +** +** col: +** CREATE TABLE vocab(term, col, doc, cnt, PRIMARY KEY(term, col)); +** +** One row for each term/column combination. The value of $doc is set to +** the number of fts5 rows that contain at least one instance of term +** $term within column $col. Field $cnt is set to the total number of +** instances of term $term in column $col (in any row of the fts5 table). +** +** row: +** CREATE TABLE vocab(term, doc, cnt, PRIMARY KEY(term)); +** +** One row for each term in the database. The value of $doc is set to +** the number of fts5 rows that contain at least one instance of term +** $term. Field $cnt is set to the total number of instances of term +** $term in the database. +*/ + + +/* #include "fts5Int.h" */ + + +typedef struct Fts5VocabTable Fts5VocabTable; +typedef struct Fts5VocabCursor Fts5VocabCursor; + +struct Fts5VocabTable { + sqlite3_vtab base; + char *zFts5Tbl; /* Name of fts5 table */ + char *zFts5Db; /* Db containing fts5 table */ + sqlite3 *db; /* Database handle */ + Fts5Global *pGlobal; /* FTS5 global object for this database */ + int eType; /* FTS5_VOCAB_COL or ROW */ +}; + +struct Fts5VocabCursor { + sqlite3_vtab_cursor base; + sqlite3_stmt *pStmt; /* Statement holding lock on pIndex */ + Fts5Index *pIndex; /* Associated FTS5 index */ + + int bEof; /* True if this cursor is at EOF */ + Fts5IndexIter *pIter; /* Term/rowid iterator object */ + + int nLeTerm; /* Size of zLeTerm in bytes */ + char *zLeTerm; /* (term <= $zLeTerm) paramater, or NULL */ + + /* These are used by 'col' tables only */ + Fts5Config *pConfig; /* Fts5 table configuration */ + int iCol; + i64 *aCnt; + i64 *aDoc; + + /* Output values used by 'row' and 'col' tables */ + i64 rowid; /* This table's current rowid value */ + Fts5Buffer term; /* Current value of 'term' column */ +}; + +#define FTS5_VOCAB_COL 0 +#define FTS5_VOCAB_ROW 1 + +#define FTS5_VOCAB_COL_SCHEMA "term, col, doc, cnt" +#define FTS5_VOCAB_ROW_SCHEMA "term, doc, cnt" + +/* +** Bits for the mask used as the idxNum value by xBestIndex/xFilter. +*/ +#define FTS5_VOCAB_TERM_EQ 0x01 +#define FTS5_VOCAB_TERM_GE 0x02 +#define FTS5_VOCAB_TERM_LE 0x04 + + +/* +** Translate a string containing an fts5vocab table type to an +** FTS5_VOCAB_XXX constant. If successful, set *peType to the output +** value and return SQLITE_OK. Otherwise, set *pzErr to an error message +** and return SQLITE_ERROR. +*/ +static int fts5VocabTableType(const char *zType, char **pzErr, int *peType){ + int rc = SQLITE_OK; + char *zCopy = sqlite3Fts5Strndup(&rc, zType, -1); + if( rc==SQLITE_OK ){ + sqlite3Fts5Dequote(zCopy); + if( sqlite3_stricmp(zCopy, "col")==0 ){ + *peType = FTS5_VOCAB_COL; + }else + + if( sqlite3_stricmp(zCopy, "row")==0 ){ + *peType = FTS5_VOCAB_ROW; + }else + { + *pzErr = sqlite3_mprintf("fts5vocab: unknown table type: %Q", zCopy); + rc = SQLITE_ERROR; + } + sqlite3_free(zCopy); + } + + return rc; +} + + +/* +** The xDisconnect() virtual table method. +*/ +static int fts5VocabDisconnectMethod(sqlite3_vtab *pVtab){ + Fts5VocabTable *pTab = (Fts5VocabTable*)pVtab; + sqlite3_free(pTab); + return SQLITE_OK; +} + +/* +** The xDestroy() virtual table method. +*/ +static int fts5VocabDestroyMethod(sqlite3_vtab *pVtab){ + Fts5VocabTable *pTab = (Fts5VocabTable*)pVtab; + sqlite3_free(pTab); + return SQLITE_OK; +} + +/* +** This function is the implementation of both the xConnect and xCreate +** methods of the FTS3 virtual table. +** +** The argv[] array contains the following: +** +** argv[0] -> module name ("fts5vocab") +** argv[1] -> database name +** argv[2] -> table name +** +** then: +** +** argv[3] -> name of fts5 table +** argv[4] -> type of fts5vocab table +** +** or, for tables in the TEMP schema only. +** +** argv[3] -> name of fts5 tables database +** argv[4] -> name of fts5 table +** argv[5] -> type of fts5vocab table +*/ +static int fts5VocabInitVtab( + sqlite3 *db, /* The SQLite database connection */ + void *pAux, /* Pointer to Fts5Global object */ + int argc, /* Number of elements in argv array */ + const char * const *argv, /* xCreate/xConnect argument array */ + sqlite3_vtab **ppVTab, /* Write the resulting vtab structure here */ + char **pzErr /* Write any error message here */ +){ + const char *azSchema[] = { + "CREATE TABlE vocab(" FTS5_VOCAB_COL_SCHEMA ")", + "CREATE TABlE vocab(" FTS5_VOCAB_ROW_SCHEMA ")" + }; + + Fts5VocabTable *pRet = 0; + int rc = SQLITE_OK; /* Return code */ + int bDb; + + bDb = (argc==6 && strlen(argv[1])==4 && memcmp("temp", argv[1], 4)==0); + + if( argc!=5 && bDb==0 ){ + *pzErr = sqlite3_mprintf("wrong number of vtable arguments"); + rc = SQLITE_ERROR; + }else{ + int nByte; /* Bytes of space to allocate */ + const char *zDb = bDb ? argv[3] : argv[1]; + const char *zTab = bDb ? argv[4] : argv[3]; + const char *zType = bDb ? argv[5] : argv[4]; + int nDb = (int)strlen(zDb)+1; + int nTab = (int)strlen(zTab)+1; + int eType = 0; + + rc = fts5VocabTableType(zType, pzErr, &eType); + if( rc==SQLITE_OK ){ + assert( eType>=0 && eType<ArraySize(azSchema) ); + rc = sqlite3_declare_vtab(db, azSchema[eType]); + } + + nByte = sizeof(Fts5VocabTable) + nDb + nTab; + pRet = sqlite3Fts5MallocZero(&rc, nByte); + if( pRet ){ + pRet->pGlobal = (Fts5Global*)pAux; + pRet->eType = eType; + pRet->db = db; + pRet->zFts5Tbl = (char*)&pRet[1]; + pRet->zFts5Db = &pRet->zFts5Tbl[nTab]; + memcpy(pRet->zFts5Tbl, zTab, nTab); + memcpy(pRet->zFts5Db, zDb, nDb); + sqlite3Fts5Dequote(pRet->zFts5Tbl); + sqlite3Fts5Dequote(pRet->zFts5Db); + } + } + + *ppVTab = (sqlite3_vtab*)pRet; + return rc; +} + + +/* +** The xConnect() and xCreate() methods for the virtual table. All the +** work is done in function fts5VocabInitVtab(). +*/ +static int fts5VocabConnectMethod( + sqlite3 *db, /* Database connection */ + void *pAux, /* Pointer to tokenizer hash table */ + int argc, /* Number of elements in argv array */ + const char * const *argv, /* xCreate/xConnect argument array */ + sqlite3_vtab **ppVtab, /* OUT: New sqlite3_vtab object */ + char **pzErr /* OUT: sqlite3_malloc'd error message */ +){ + return fts5VocabInitVtab(db, pAux, argc, argv, ppVtab, pzErr); +} +static int fts5VocabCreateMethod( + sqlite3 *db, /* Database connection */ + void *pAux, /* Pointer to tokenizer hash table */ + int argc, /* Number of elements in argv array */ + const char * const *argv, /* xCreate/xConnect argument array */ + sqlite3_vtab **ppVtab, /* OUT: New sqlite3_vtab object */ + char **pzErr /* OUT: sqlite3_malloc'd error message */ +){ + return fts5VocabInitVtab(db, pAux, argc, argv, ppVtab, pzErr); +} + +/* +** Implementation of the xBestIndex method. +*/ +static int fts5VocabBestIndexMethod( + sqlite3_vtab *pUnused, + sqlite3_index_info *pInfo +){ + int i; + int iTermEq = -1; + int iTermGe = -1; + int iTermLe = -1; + int idxNum = 0; + int nArg = 0; + + UNUSED_PARAM(pUnused); + + for(i=0; i<pInfo->nConstraint; i++){ + struct sqlite3_index_constraint *p = &pInfo->aConstraint[i]; + if( p->usable==0 ) continue; + if( p->iColumn==0 ){ /* term column */ + if( p->op==SQLITE_INDEX_CONSTRAINT_EQ ) iTermEq = i; + if( p->op==SQLITE_INDEX_CONSTRAINT_LE ) iTermLe = i; + if( p->op==SQLITE_INDEX_CONSTRAINT_LT ) iTermLe = i; + if( p->op==SQLITE_INDEX_CONSTRAINT_GE ) iTermGe = i; + if( p->op==SQLITE_INDEX_CONSTRAINT_GT ) iTermGe = i; + } + } + + if( iTermEq>=0 ){ + idxNum |= FTS5_VOCAB_TERM_EQ; + pInfo->aConstraintUsage[iTermEq].argvIndex = ++nArg; + pInfo->estimatedCost = 100; + }else{ + pInfo->estimatedCost = 1000000; + if( iTermGe>=0 ){ + idxNum |= FTS5_VOCAB_TERM_GE; + pInfo->aConstraintUsage[iTermGe].argvIndex = ++nArg; + pInfo->estimatedCost = pInfo->estimatedCost / 2; + } + if( iTermLe>=0 ){ + idxNum |= FTS5_VOCAB_TERM_LE; + pInfo->aConstraintUsage[iTermLe].argvIndex = ++nArg; + pInfo->estimatedCost = pInfo->estimatedCost / 2; + } + } + + pInfo->idxNum = idxNum; + + return SQLITE_OK; +} + +/* +** Implementation of xOpen method. +*/ +static int fts5VocabOpenMethod( + sqlite3_vtab *pVTab, + sqlite3_vtab_cursor **ppCsr +){ + Fts5VocabTable *pTab = (Fts5VocabTable*)pVTab; + Fts5Index *pIndex = 0; + Fts5Config *pConfig = 0; + Fts5VocabCursor *pCsr = 0; + int rc = SQLITE_OK; + sqlite3_stmt *pStmt = 0; + char *zSql = 0; + + zSql = sqlite3Fts5Mprintf(&rc, + "SELECT t.%Q FROM %Q.%Q AS t WHERE t.%Q MATCH '*id'", + pTab->zFts5Tbl, pTab->zFts5Db, pTab->zFts5Tbl, pTab->zFts5Tbl + ); + if( zSql ){ + rc = sqlite3_prepare_v2(pTab->db, zSql, -1, &pStmt, 0); + } + sqlite3_free(zSql); + assert( rc==SQLITE_OK || pStmt==0 ); + if( rc==SQLITE_ERROR ) rc = SQLITE_OK; + + if( pStmt && sqlite3_step(pStmt)==SQLITE_ROW ){ + i64 iId = sqlite3_column_int64(pStmt, 0); + pIndex = sqlite3Fts5IndexFromCsrid(pTab->pGlobal, iId, &pConfig); + } + + if( rc==SQLITE_OK && pIndex==0 ){ + rc = sqlite3_finalize(pStmt); + pStmt = 0; + if( rc==SQLITE_OK ){ + pVTab->zErrMsg = sqlite3_mprintf( + "no such fts5 table: %s.%s", pTab->zFts5Db, pTab->zFts5Tbl + ); + rc = SQLITE_ERROR; + } + } + + if( rc==SQLITE_OK ){ + int nByte = pConfig->nCol * sizeof(i64) * 2 + sizeof(Fts5VocabCursor); + pCsr = (Fts5VocabCursor*)sqlite3Fts5MallocZero(&rc, nByte); + } + + if( pCsr ){ + pCsr->pIndex = pIndex; + pCsr->pStmt = pStmt; + pCsr->pConfig = pConfig; + pCsr->aCnt = (i64*)&pCsr[1]; + pCsr->aDoc = &pCsr->aCnt[pConfig->nCol]; + }else{ + sqlite3_finalize(pStmt); + } + + *ppCsr = (sqlite3_vtab_cursor*)pCsr; + return rc; +} + +static void fts5VocabResetCursor(Fts5VocabCursor *pCsr){ + pCsr->rowid = 0; + sqlite3Fts5IterClose(pCsr->pIter); + pCsr->pIter = 0; + sqlite3_free(pCsr->zLeTerm); + pCsr->nLeTerm = -1; + pCsr->zLeTerm = 0; +} + +/* +** Close the cursor. For additional information see the documentation +** on the xClose method of the virtual table interface. +*/ +static int fts5VocabCloseMethod(sqlite3_vtab_cursor *pCursor){ + Fts5VocabCursor *pCsr = (Fts5VocabCursor*)pCursor; + fts5VocabResetCursor(pCsr); + sqlite3Fts5BufferFree(&pCsr->term); + sqlite3_finalize(pCsr->pStmt); + sqlite3_free(pCsr); + return SQLITE_OK; +} + + +/* +** Advance the cursor to the next row in the table. +*/ +static int fts5VocabNextMethod(sqlite3_vtab_cursor *pCursor){ + Fts5VocabCursor *pCsr = (Fts5VocabCursor*)pCursor; + Fts5VocabTable *pTab = (Fts5VocabTable*)pCursor->pVtab; + int rc = SQLITE_OK; + int nCol = pCsr->pConfig->nCol; + + pCsr->rowid++; + + if( pTab->eType==FTS5_VOCAB_COL ){ + for(pCsr->iCol++; pCsr->iCol<nCol; pCsr->iCol++){ + if( pCsr->aDoc[pCsr->iCol] ) break; + } + } + + if( pTab->eType==FTS5_VOCAB_ROW || pCsr->iCol>=nCol ){ + if( sqlite3Fts5IterEof(pCsr->pIter) ){ + pCsr->bEof = 1; + }else{ + const char *zTerm; + int nTerm; + + zTerm = sqlite3Fts5IterTerm(pCsr->pIter, &nTerm); + if( pCsr->nLeTerm>=0 ){ + int nCmp = MIN(nTerm, pCsr->nLeTerm); + int bCmp = memcmp(pCsr->zLeTerm, zTerm, nCmp); + if( bCmp<0 || (bCmp==0 && pCsr->nLeTerm<nTerm) ){ + pCsr->bEof = 1; + return SQLITE_OK; + } + } + + sqlite3Fts5BufferSet(&rc, &pCsr->term, nTerm, (const u8*)zTerm); + memset(pCsr->aCnt, 0, nCol * sizeof(i64)); + memset(pCsr->aDoc, 0, nCol * sizeof(i64)); + pCsr->iCol = 0; + + assert( pTab->eType==FTS5_VOCAB_COL || pTab->eType==FTS5_VOCAB_ROW ); + while( rc==SQLITE_OK ){ + const u8 *pPos; int nPos; /* Position list */ + i64 iPos = 0; /* 64-bit position read from poslist */ + int iOff = 0; /* Current offset within position list */ + + pPos = pCsr->pIter->pData; + nPos = pCsr->pIter->nData; + switch( pCsr->pConfig->eDetail ){ + case FTS5_DETAIL_FULL: + pPos = pCsr->pIter->pData; + nPos = pCsr->pIter->nData; + if( pTab->eType==FTS5_VOCAB_ROW ){ + while( 0==sqlite3Fts5PoslistNext64(pPos, nPos, &iOff, &iPos) ){ + pCsr->aCnt[0]++; + } + pCsr->aDoc[0]++; + }else{ + int iCol = -1; + while( 0==sqlite3Fts5PoslistNext64(pPos, nPos, &iOff, &iPos) ){ + int ii = FTS5_POS2COLUMN(iPos); + pCsr->aCnt[ii]++; + if( iCol!=ii ){ + if( ii>=nCol ){ + rc = FTS5_CORRUPT; + break; + } + pCsr->aDoc[ii]++; + iCol = ii; + } + } + } + break; + + case FTS5_DETAIL_COLUMNS: + if( pTab->eType==FTS5_VOCAB_ROW ){ + pCsr->aDoc[0]++; + }else{ + while( 0==sqlite3Fts5PoslistNext64(pPos, nPos, &iOff,&iPos) ){ + assert_nc( iPos>=0 && iPos<nCol ); + if( iPos>=nCol ){ + rc = FTS5_CORRUPT; + break; + } + pCsr->aDoc[iPos]++; + } + } + break; + + default: + assert( pCsr->pConfig->eDetail==FTS5_DETAIL_NONE ); + pCsr->aDoc[0]++; + break; + } + + if( rc==SQLITE_OK ){ + rc = sqlite3Fts5IterNextScan(pCsr->pIter); + } + + if( rc==SQLITE_OK ){ + zTerm = sqlite3Fts5IterTerm(pCsr->pIter, &nTerm); + if( nTerm!=pCsr->term.n || memcmp(zTerm, pCsr->term.p, nTerm) ){ + break; + } + if( sqlite3Fts5IterEof(pCsr->pIter) ) break; + } + } + } + } + + if( rc==SQLITE_OK && pCsr->bEof==0 && pTab->eType==FTS5_VOCAB_COL ){ + while( pCsr->aDoc[pCsr->iCol]==0 ) pCsr->iCol++; + assert( pCsr->iCol<pCsr->pConfig->nCol ); + } + return rc; +} + +/* +** This is the xFilter implementation for the virtual table. +*/ +static int fts5VocabFilterMethod( + sqlite3_vtab_cursor *pCursor, /* The cursor used for this query */ + int idxNum, /* Strategy index */ + const char *zUnused, /* Unused */ + int nUnused, /* Number of elements in apVal */ + sqlite3_value **apVal /* Arguments for the indexing scheme */ +){ + Fts5VocabCursor *pCsr = (Fts5VocabCursor*)pCursor; + int rc = SQLITE_OK; + + int iVal = 0; + int f = FTS5INDEX_QUERY_SCAN; + const char *zTerm = 0; + int nTerm = 0; + + sqlite3_value *pEq = 0; + sqlite3_value *pGe = 0; + sqlite3_value *pLe = 0; + + UNUSED_PARAM2(zUnused, nUnused); + + fts5VocabResetCursor(pCsr); + if( idxNum & FTS5_VOCAB_TERM_EQ ) pEq = apVal[iVal++]; + if( idxNum & FTS5_VOCAB_TERM_GE ) pGe = apVal[iVal++]; + if( idxNum & FTS5_VOCAB_TERM_LE ) pLe = apVal[iVal++]; + + if( pEq ){ + zTerm = (const char *)sqlite3_value_text(pEq); + nTerm = sqlite3_value_bytes(pEq); + f = 0; + }else{ + if( pGe ){ + zTerm = (const char *)sqlite3_value_text(pGe); + nTerm = sqlite3_value_bytes(pGe); + } + if( pLe ){ + const char *zCopy = (const char *)sqlite3_value_text(pLe); + pCsr->nLeTerm = sqlite3_value_bytes(pLe); + pCsr->zLeTerm = sqlite3_malloc(pCsr->nLeTerm+1); + if( pCsr->zLeTerm==0 ){ + rc = SQLITE_NOMEM; + }else{ + memcpy(pCsr->zLeTerm, zCopy, pCsr->nLeTerm+1); + } + } + } + + + if( rc==SQLITE_OK ){ + rc = sqlite3Fts5IndexQuery(pCsr->pIndex, zTerm, nTerm, f, 0, &pCsr->pIter); + } + if( rc==SQLITE_OK ){ + rc = fts5VocabNextMethod(pCursor); + } + + return rc; +} + +/* +** This is the xEof method of the virtual table. SQLite calls this +** routine to find out if it has reached the end of a result set. +*/ +static int fts5VocabEofMethod(sqlite3_vtab_cursor *pCursor){ + Fts5VocabCursor *pCsr = (Fts5VocabCursor*)pCursor; + return pCsr->bEof; +} + +static int fts5VocabColumnMethod( + sqlite3_vtab_cursor *pCursor, /* Cursor to retrieve value from */ + sqlite3_context *pCtx, /* Context for sqlite3_result_xxx() calls */ + int iCol /* Index of column to read value from */ +){ + Fts5VocabCursor *pCsr = (Fts5VocabCursor*)pCursor; + int eDetail = pCsr->pConfig->eDetail; + int eType = ((Fts5VocabTable*)(pCursor->pVtab))->eType; + i64 iVal = 0; + + if( iCol==0 ){ + sqlite3_result_text( + pCtx, (const char*)pCsr->term.p, pCsr->term.n, SQLITE_TRANSIENT + ); + }else if( eType==FTS5_VOCAB_COL ){ + assert( iCol==1 || iCol==2 || iCol==3 ); + if( iCol==1 ){ + if( eDetail!=FTS5_DETAIL_NONE ){ + const char *z = pCsr->pConfig->azCol[pCsr->iCol]; + sqlite3_result_text(pCtx, z, -1, SQLITE_STATIC); + } + }else if( iCol==2 ){ + iVal = pCsr->aDoc[pCsr->iCol]; + }else{ + iVal = pCsr->aCnt[pCsr->iCol]; + } + }else{ + assert( iCol==1 || iCol==2 ); + if( iCol==1 ){ + iVal = pCsr->aDoc[0]; + }else{ + iVal = pCsr->aCnt[0]; + } + } + + if( iVal>0 ) sqlite3_result_int64(pCtx, iVal); + return SQLITE_OK; +} + +/* +** This is the xRowid method. The SQLite core calls this routine to +** retrieve the rowid for the current row of the result set. The +** rowid should be written to *pRowid. +*/ +static int fts5VocabRowidMethod( + sqlite3_vtab_cursor *pCursor, + sqlite_int64 *pRowid +){ + Fts5VocabCursor *pCsr = (Fts5VocabCursor*)pCursor; + *pRowid = pCsr->rowid; + return SQLITE_OK; +} + +static int sqlite3Fts5VocabInit(Fts5Global *pGlobal, sqlite3 *db){ + static const sqlite3_module fts5Vocab = { + /* iVersion */ 2, + /* xCreate */ fts5VocabCreateMethod, + /* xConnect */ fts5VocabConnectMethod, + /* xBestIndex */ fts5VocabBestIndexMethod, + /* xDisconnect */ fts5VocabDisconnectMethod, + /* xDestroy */ fts5VocabDestroyMethod, + /* xOpen */ fts5VocabOpenMethod, + /* xClose */ fts5VocabCloseMethod, + /* xFilter */ fts5VocabFilterMethod, + /* xNext */ fts5VocabNextMethod, + /* xEof */ fts5VocabEofMethod, + /* xColumn */ fts5VocabColumnMethod, + /* xRowid */ fts5VocabRowidMethod, + /* xUpdate */ 0, + /* xBegin */ 0, + /* xSync */ 0, + /* xCommit */ 0, + /* xRollback */ 0, + /* xFindFunction */ 0, + /* xRename */ 0, + /* xSavepoint */ 0, + /* xRelease */ 0, + /* xRollbackTo */ 0, + }; + void *p = (void*)pGlobal; + + return sqlite3_create_module_v2(db, "fts5vocab", &fts5Vocab, p, 0); +} + + + + + +#endif /* !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS5) */ diff --git a/src/libtracker-fts/fts5.h b/src/libtracker-fts/fts5.h new file mode 100644 index 000000000..96ecb38e3 --- /dev/null +++ b/src/libtracker-fts/fts5.h @@ -0,0 +1,578 @@ +/* +** 2014 May 31 +** +** The author disclaims copyright to this source code. In place of +** a legal notice, here is a blessing: +** +** May you do good and not evil. +** May you find forgiveness for yourself and forgive others. +** May you share freely, never taking more than you give. +** +****************************************************************************** +** +** Interfaces to extend FTS5. Using the interfaces defined in this file, +** FTS5 may be extended with: +** +** * custom tokenizers, and +** * custom auxiliary functions. +*/ + + +#ifndef _FTS5_H +#define _FTS5_H + +#include "sqlite3.h" + +#ifdef __cplusplus +extern "C" { +#endif + +/************************************************************************* +** CUSTOM AUXILIARY FUNCTIONS +** +** Virtual table implementations may overload SQL functions by implementing +** the sqlite3_module.xFindFunction() method. +*/ + +typedef struct Fts5ExtensionApi Fts5ExtensionApi; +typedef struct Fts5Context Fts5Context; +typedef struct Fts5PhraseIter Fts5PhraseIter; + +typedef void (*fts5_extension_function)( + const Fts5ExtensionApi *pApi, /* API offered by current FTS version */ + Fts5Context *pFts, /* First arg to pass to pApi functions */ + sqlite3_context *pCtx, /* Context for returning result/error */ + int nVal, /* Number of values in apVal[] array */ + sqlite3_value **apVal /* Array of trailing arguments */ +); + +struct Fts5PhraseIter { + const unsigned char *a; + const unsigned char *b; +}; + +/* +** EXTENSION API FUNCTIONS +** +** xUserData(pFts): +** Return a copy of the context pointer the extension function was +** registered with. +** +** xColumnTotalSize(pFts, iCol, pnToken): +** If parameter iCol is less than zero, set output variable *pnToken +** to the total number of tokens in the FTS5 table. Or, if iCol is +** non-negative but less than the number of columns in the table, return +** the total number of tokens in column iCol, considering all rows in +** the FTS5 table. +** +** If parameter iCol is greater than or equal to the number of columns +** in the table, SQLITE_RANGE is returned. Or, if an error occurs (e.g. +** an OOM condition or IO error), an appropriate SQLite error code is +** returned. +** +** xColumnCount(pFts): +** Return the number of columns in the table. +** +** xColumnSize(pFts, iCol, pnToken): +** If parameter iCol is less than zero, set output variable *pnToken +** to the total number of tokens in the current row. Or, if iCol is +** non-negative but less than the number of columns in the table, set +** *pnToken to the number of tokens in column iCol of the current row. +** +** If parameter iCol is greater than or equal to the number of columns +** in the table, SQLITE_RANGE is returned. Or, if an error occurs (e.g. +** an OOM condition or IO error), an appropriate SQLite error code is +** returned. +** +** This function may be quite inefficient if used with an FTS5 table +** created with the "columnsize=0" option. +** +** xColumnText: +** This function attempts to retrieve the text of column iCol of the +** current document. If successful, (*pz) is set to point to a buffer +** containing the text in utf-8 encoding, (*pn) is set to the size in bytes +** (not characters) of the buffer and SQLITE_OK is returned. Otherwise, +** if an error occurs, an SQLite error code is returned and the final values +** of (*pz) and (*pn) are undefined. +** +** xPhraseCount: +** Returns the number of phrases in the current query expression. +** +** xPhraseSize: +** Returns the number of tokens in phrase iPhrase of the query. Phrases +** are numbered starting from zero. +** +** xInstCount: +** Set *pnInst to the total number of occurrences of all phrases within +** the query within the current row. Return SQLITE_OK if successful, or +** an error code (i.e. SQLITE_NOMEM) if an error occurs. +** +** This API can be quite slow if used with an FTS5 table created with the +** "detail=none" or "detail=column" option. If the FTS5 table is created +** with either "detail=none" or "detail=column" and "content=" option +** (i.e. if it is a contentless table), then this API always returns 0. +** +** xInst: +** Query for the details of phrase match iIdx within the current row. +** Phrase matches are numbered starting from zero, so the iIdx argument +** should be greater than or equal to zero and smaller than the value +** output by xInstCount(). +** +** Usually, output parameter *piPhrase is set to the phrase number, *piCol +** to the column in which it occurs and *piOff the token offset of the +** first token of the phrase. The exception is if the table was created +** with the offsets=0 option specified. In this case *piOff is always +** set to -1. +** +** Returns SQLITE_OK if successful, or an error code (i.e. SQLITE_NOMEM) +** if an error occurs. +** +** This API can be quite slow if used with an FTS5 table created with the +** "detail=none" or "detail=column" option. +** +** xRowid: +** Returns the rowid of the current row. +** +** xTokenize: +** Tokenize text using the tokenizer belonging to the FTS5 table. +** +** xQueryPhrase(pFts5, iPhrase, pUserData, xCallback): +** This API function is used to query the FTS table for phrase iPhrase +** of the current query. Specifically, a query equivalent to: +** +** ... FROM ftstable WHERE ftstable MATCH $p ORDER BY rowid +** +** with $p set to a phrase equivalent to the phrase iPhrase of the +** current query is executed. For each row visited, the callback function +** passed as the fourth argument is invoked. The context and API objects +** passed to the callback function may be used to access the properties of +** each matched row. Invoking Api.xUserData() returns a copy of the pointer +** passed as the third argument to pUserData. +** +** If the callback function returns any value other than SQLITE_OK, the +** query is abandoned and the xQueryPhrase function returns immediately. +** If the returned value is SQLITE_DONE, xQueryPhrase returns SQLITE_OK. +** Otherwise, the error code is propagated upwards. +** +** If the query runs to completion without incident, SQLITE_OK is returned. +** Or, if some error occurs before the query completes or is aborted by +** the callback, an SQLite error code is returned. +** +** +** xSetAuxdata(pFts5, pAux, xDelete) +** +** Save the pointer passed as the second argument as the extension functions +** "auxiliary data". The pointer may then be retrieved by the current or any +** future invocation of the same fts5 extension function made as part of +** of the same MATCH query using the xGetAuxdata() API. +** +** Each extension function is allocated a single auxiliary data slot for +** each FTS query (MATCH expression). If the extension function is invoked +** more than once for a single FTS query, then all invocations share a +** single auxiliary data context. +** +** If there is already an auxiliary data pointer when this function is +** invoked, then it is replaced by the new pointer. If an xDelete callback +** was specified along with the original pointer, it is invoked at this +** point. +** +** The xDelete callback, if one is specified, is also invoked on the +** auxiliary data pointer after the FTS5 query has finished. +** +** If an error (e.g. an OOM condition) occurs within this function, an +** the auxiliary data is set to NULL and an error code returned. If the +** xDelete parameter was not NULL, it is invoked on the auxiliary data +** pointer before returning. +** +** +** xGetAuxdata(pFts5, bClear) +** +** Returns the current auxiliary data pointer for the fts5 extension +** function. See the xSetAuxdata() method for details. +** +** If the bClear argument is non-zero, then the auxiliary data is cleared +** (set to NULL) before this function returns. In this case the xDelete, +** if any, is not invoked. +** +** +** xRowCount(pFts5, pnRow) +** +** This function is used to retrieve the total number of rows in the table. +** In other words, the same value that would be returned by: +** +** SELECT count(*) FROM ftstable; +** +** xPhraseFirst() +** This function is used, along with type Fts5PhraseIter and the xPhraseNext +** method, to iterate through all instances of a single query phrase within +** the current row. This is the same information as is accessible via the +** xInstCount/xInst APIs. While the xInstCount/xInst APIs are more convenient +** to use, this API may be faster under some circumstances. To iterate +** through instances of phrase iPhrase, use the following code: +** +** Fts5PhraseIter iter; +** int iCol, iOff; +** for(pApi->xPhraseFirst(pFts, iPhrase, &iter, &iCol, &iOff); +** iCol>=0; +** pApi->xPhraseNext(pFts, &iter, &iCol, &iOff) +** ){ +** // An instance of phrase iPhrase at offset iOff of column iCol +** } +** +** The Fts5PhraseIter structure is defined above. Applications should not +** modify this structure directly - it should only be used as shown above +** with the xPhraseFirst() and xPhraseNext() API methods (and by +** xPhraseFirstColumn() and xPhraseNextColumn() as illustrated below). +** +** This API can be quite slow if used with an FTS5 table created with the +** "detail=none" or "detail=column" option. If the FTS5 table is created +** with either "detail=none" or "detail=column" and "content=" option +** (i.e. if it is a contentless table), then this API always iterates +** through an empty set (all calls to xPhraseFirst() set iCol to -1). +** +** xPhraseNext() +** See xPhraseFirst above. +** +** xPhraseFirstColumn() +** This function and xPhraseNextColumn() are similar to the xPhraseFirst() +** and xPhraseNext() APIs described above. The difference is that instead +** of iterating through all instances of a phrase in the current row, these +** APIs are used to iterate through the set of columns in the current row +** that contain one or more instances of a specified phrase. For example: +** +** Fts5PhraseIter iter; +** int iCol; +** for(pApi->xPhraseFirstColumn(pFts, iPhrase, &iter, &iCol); +** iCol>=0; +** pApi->xPhraseNextColumn(pFts, &iter, &iCol) +** ){ +** // Column iCol contains at least one instance of phrase iPhrase +** } +** +** This API can be quite slow if used with an FTS5 table created with the +** "detail=none" option. If the FTS5 table is created with either +** "detail=none" "content=" option (i.e. if it is a contentless table), +** then this API always iterates through an empty set (all calls to +** xPhraseFirstColumn() set iCol to -1). +** +** The information accessed using this API and its companion +** xPhraseFirstColumn() may also be obtained using xPhraseFirst/xPhraseNext +** (or xInst/xInstCount). The chief advantage of this API is that it is +** significantly more efficient than those alternatives when used with +** "detail=column" tables. +** +** xPhraseNextColumn() +** See xPhraseFirstColumn above. +*/ +struct Fts5ExtensionApi { + int iVersion; /* Currently always set to 3 */ + + void *(*xUserData)(Fts5Context*); + + int (*xColumnCount)(Fts5Context*); + int (*xRowCount)(Fts5Context*, sqlite3_int64 *pnRow); + int (*xColumnTotalSize)(Fts5Context*, int iCol, sqlite3_int64 *pnToken); + + int (*xTokenize)(Fts5Context*, + const char *pText, int nText, /* Text to tokenize */ + void *pCtx, /* Context passed to xToken() */ + int (*xToken)(void*, int, const char*, int, int, int) /* Callback */ + ); + + int (*xPhraseCount)(Fts5Context*); + int (*xPhraseSize)(Fts5Context*, int iPhrase); + + int (*xInstCount)(Fts5Context*, int *pnInst); + int (*xInst)(Fts5Context*, int iIdx, int *piPhrase, int *piCol, int *piOff); + + sqlite3_int64 (*xRowid)(Fts5Context*); + int (*xColumnText)(Fts5Context*, int iCol, const char **pz, int *pn); + int (*xColumnSize)(Fts5Context*, int iCol, int *pnToken); + + int (*xQueryPhrase)(Fts5Context*, int iPhrase, void *pUserData, + int(*)(const Fts5ExtensionApi*,Fts5Context*,void*) + ); + int (*xSetAuxdata)(Fts5Context*, void *pAux, void(*xDelete)(void*)); + void *(*xGetAuxdata)(Fts5Context*, int bClear); + + int (*xPhraseFirst)(Fts5Context*, int iPhrase, Fts5PhraseIter*, int*, int*); + void (*xPhraseNext)(Fts5Context*, Fts5PhraseIter*, int *piCol, int *piOff); + + int (*xPhraseFirstColumn)(Fts5Context*, int iPhrase, Fts5PhraseIter*, int*); + void (*xPhraseNextColumn)(Fts5Context*, Fts5PhraseIter*, int *piCol); +}; + +/* +** CUSTOM AUXILIARY FUNCTIONS +*************************************************************************/ + +/************************************************************************* +** CUSTOM TOKENIZERS +** +** Applications may also register custom tokenizer types. A tokenizer +** is registered by providing fts5 with a populated instance of the +** following structure. All structure methods must be defined, setting +** any member of the fts5_tokenizer struct to NULL leads to undefined +** behaviour. The structure methods are expected to function as follows: +** +** xCreate: +** This function is used to allocate and inititalize a tokenizer instance. +** A tokenizer instance is required to actually tokenize text. +** +** The first argument passed to this function is a copy of the (void*) +** pointer provided by the application when the fts5_tokenizer object +** was registered with FTS5 (the third argument to xCreateTokenizer()). +** The second and third arguments are an array of nul-terminated strings +** containing the tokenizer arguments, if any, specified following the +** tokenizer name as part of the CREATE VIRTUAL TABLE statement used +** to create the FTS5 table. +** +** The final argument is an output variable. If successful, (*ppOut) +** should be set to point to the new tokenizer handle and SQLITE_OK +** returned. If an error occurs, some value other than SQLITE_OK should +** be returned. In this case, fts5 assumes that the final value of *ppOut +** is undefined. +** +** xDelete: +** This function is invoked to delete a tokenizer handle previously +** allocated using xCreate(). Fts5 guarantees that this function will +** be invoked exactly once for each successful call to xCreate(). +** +** xTokenize: +** This function is expected to tokenize the nText byte string indicated +** by argument pText. pText may or may not be nul-terminated. The first +** argument passed to this function is a pointer to an Fts5Tokenizer object +** returned by an earlier call to xCreate(). +** +** The second argument indicates the reason that FTS5 is requesting +** tokenization of the supplied text. This is always one of the following +** four values: +** +** <ul><li> <b>FTS5_TOKENIZE_DOCUMENT</b> - A document is being inserted into +** or removed from the FTS table. The tokenizer is being invoked to +** determine the set of tokens to add to (or delete from) the +** FTS index. +** +** <li> <b>FTS5_TOKENIZE_QUERY</b> - A MATCH query is being executed +** against the FTS index. The tokenizer is being called to tokenize +** a bareword or quoted string specified as part of the query. +** +** <li> <b>(FTS5_TOKENIZE_QUERY | FTS5_TOKENIZE_PREFIX)</b> - Same as +** FTS5_TOKENIZE_QUERY, except that the bareword or quoted string is +** followed by a "*" character, indicating that the last token +** returned by the tokenizer will be treated as a token prefix. +** +** <li> <b>FTS5_TOKENIZE_AUX</b> - The tokenizer is being invoked to +** satisfy an fts5_api.xTokenize() request made by an auxiliary +** function. Or an fts5_api.xColumnSize() request made by the same +** on a columnsize=0 database. +** </ul> +** +** For each token in the input string, the supplied callback xToken() must +** be invoked. The first argument to it should be a copy of the pointer +** passed as the second argument to xTokenize(). The third and fourth +** arguments are a pointer to a buffer containing the token text, and the +** size of the token in bytes. The 4th and 5th arguments are the byte offsets +** of the first byte of and first byte immediately following the text from +** which the token is derived within the input. +** +** The second argument passed to the xToken() callback ("tflags") should +** normally be set to 0. The exception is if the tokenizer supports +** synonyms. In this case see the discussion below for details. +** +** FTS5 assumes the xToken() callback is invoked for each token in the +** order that they occur within the input text. +** +** If an xToken() callback returns any value other than SQLITE_OK, then +** the tokenization should be abandoned and the xTokenize() method should +** immediately return a copy of the xToken() return value. Or, if the +** input buffer is exhausted, xTokenize() should return SQLITE_OK. Finally, +** if an error occurs with the xTokenize() implementation itself, it +** may abandon the tokenization and return any error code other than +** SQLITE_OK or SQLITE_DONE. +** +** SYNONYM SUPPORT +** +** Custom tokenizers may also support synonyms. Consider a case in which a +** user wishes to query for a phrase such as "first place". Using the +** built-in tokenizers, the FTS5 query 'first + place' will match instances +** of "first place" within the document set, but not alternative forms +** such as "1st place". In some applications, it would be better to match +** all instances of "first place" or "1st place" regardless of which form +** the user specified in the MATCH query text. +** +** There are several ways to approach this in FTS5: +** +** <ol><li> By mapping all synonyms to a single token. In this case, the +** In the above example, this means that the tokenizer returns the +** same token for inputs "first" and "1st". Say that token is in +** fact "first", so that when the user inserts the document "I won +** 1st place" entries are added to the index for tokens "i", "won", +** "first" and "place". If the user then queries for '1st + place', +** the tokenizer substitutes "first" for "1st" and the query works +** as expected. +** +** <li> By adding multiple synonyms for a single term to the FTS index. +** In this case, when tokenizing query text, the tokenizer may +** provide multiple synonyms for a single term within the document. +** FTS5 then queries the index for each synonym individually. For +** example, faced with the query: +** +** <codeblock> +** ... MATCH 'first place'</codeblock> +** +** the tokenizer offers both "1st" and "first" as synonyms for the +** first token in the MATCH query and FTS5 effectively runs a query +** similar to: +** +** <codeblock> +** ... MATCH '(first OR 1st) place'</codeblock> +** +** except that, for the purposes of auxiliary functions, the query +** still appears to contain just two phrases - "(first OR 1st)" +** being treated as a single phrase. +** +** <li> By adding multiple synonyms for a single term to the FTS index. +** Using this method, when tokenizing document text, the tokenizer +** provides multiple synonyms for each token. So that when a +** document such as "I won first place" is tokenized, entries are +** added to the FTS index for "i", "won", "first", "1st" and +** "place". +** +** This way, even if the tokenizer does not provide synonyms +** when tokenizing query text (it should not - to do would be +** inefficient), it doesn't matter if the user queries for +** 'first + place' or '1st + place', as there are entires in the +** FTS index corresponding to both forms of the first token. +** </ol> +** +** Whether it is parsing document or query text, any call to xToken that +** specifies a <i>tflags</i> argument with the FTS5_TOKEN_COLOCATED bit +** is considered to supply a synonym for the previous token. For example, +** when parsing the document "I won first place", a tokenizer that supports +** synonyms would call xToken() 5 times, as follows: +** +** <codeblock> +** xToken(pCtx, 0, "i", 1, 0, 1); +** xToken(pCtx, 0, "won", 3, 2, 5); +** xToken(pCtx, 0, "first", 5, 6, 11); +** xToken(pCtx, FTS5_TOKEN_COLOCATED, "1st", 3, 6, 11); +** xToken(pCtx, 0, "place", 5, 12, 17); +**</codeblock> +** +** It is an error to specify the FTS5_TOKEN_COLOCATED flag the first time +** xToken() is called. Multiple synonyms may be specified for a single token +** by making multiple calls to xToken(FTS5_TOKEN_COLOCATED) in sequence. +** There is no limit to the number of synonyms that may be provided for a +** single token. +** +** In many cases, method (1) above is the best approach. It does not add +** extra data to the FTS index or require FTS5 to query for multiple terms, +** so it is efficient in terms of disk space and query speed. However, it +** does not support prefix queries very well. If, as suggested above, the +** token "first" is subsituted for "1st" by the tokenizer, then the query: +** +** <codeblock> +** ... MATCH '1s*'</codeblock> +** +** will not match documents that contain the token "1st" (as the tokenizer +** will probably not map "1s" to any prefix of "first"). +** +** For full prefix support, method (3) may be preferred. In this case, +** because the index contains entries for both "first" and "1st", prefix +** queries such as 'fi*' or '1s*' will match correctly. However, because +** extra entries are added to the FTS index, this method uses more space +** within the database. +** +** Method (2) offers a midpoint between (1) and (3). Using this method, +** a query such as '1s*' will match documents that contain the literal +** token "1st", but not "first" (assuming the tokenizer is not able to +** provide synonyms for prefixes). However, a non-prefix query like '1st' +** will match against "1st" and "first". This method does not require +** extra disk space, as no extra entries are added to the FTS index. +** On the other hand, it may require more CPU cycles to run MATCH queries, +** as separate queries of the FTS index are required for each synonym. +** +** When using methods (2) or (3), it is important that the tokenizer only +** provide synonyms when tokenizing document text (method (2)) or query +** text (method (3)), not both. Doing so will not cause any errors, but is +** inefficient. +*/ +typedef struct Fts5Tokenizer Fts5Tokenizer; +typedef struct fts5_tokenizer fts5_tokenizer; +struct fts5_tokenizer { + int (*xCreate)(void*, const char **azArg, int nArg, Fts5Tokenizer **ppOut); + void (*xDelete)(Fts5Tokenizer*); + int (*xTokenize)(Fts5Tokenizer*, + void *pCtx, + int flags, /* Mask of FTS5_TOKENIZE_* flags */ + const char *pText, int nText, + int (*xToken)( + void *pCtx, /* Copy of 2nd argument to xTokenize() */ + int tflags, /* Mask of FTS5_TOKEN_* flags */ + const char *pToken, /* Pointer to buffer containing token */ + int nToken, /* Size of token in bytes */ + int iStart, /* Byte offset of token within input text */ + int iEnd /* Byte offset of end of token within input text */ + ) + ); +}; + +/* Flags that may be passed as the third argument to xTokenize() */ +#define FTS5_TOKENIZE_QUERY 0x0001 +#define FTS5_TOKENIZE_PREFIX 0x0002 +#define FTS5_TOKENIZE_DOCUMENT 0x0004 +#define FTS5_TOKENIZE_AUX 0x0008 + +/* Flags that may be passed by the tokenizer implementation back to FTS5 +** as the third argument to the supplied xToken callback. */ +#define FTS5_TOKEN_COLOCATED 0x0001 /* Same position as prev. token */ + +/* +** END OF CUSTOM TOKENIZERS +*************************************************************************/ + +/************************************************************************* +** FTS5 EXTENSION REGISTRATION API +*/ +typedef struct fts5_api fts5_api; +struct fts5_api { + int iVersion; /* Currently always set to 2 */ + + /* Create a new tokenizer */ + int (*xCreateTokenizer)( + fts5_api *pApi, + const char *zName, + void *pContext, + fts5_tokenizer *pTokenizer, + void (*xDestroy)(void*) + ); + + /* Find an existing tokenizer */ + int (*xFindTokenizer)( + fts5_api *pApi, + const char *zName, + void **ppContext, + fts5_tokenizer *pTokenizer + ); + + /* Create a new auxiliary function */ + int (*xCreateFunction)( + fts5_api *pApi, + const char *zName, + void *pContext, + fts5_extension_function xFunction, + void (*xDestroy)(void*) + ); +}; + +/* +** END OF REGISTRATION API +*************************************************************************/ + +#ifdef __cplusplus +} /* end of the 'extern "C"' block */ +#endif + +#endif /* _FTS5_H */ + diff --git a/src/libtracker-fts/tracker-fts-tokenizer.c b/src/libtracker-fts/tracker-fts-tokenizer.c index c45d73d9a..26764aaf3 100644 --- a/src/libtracker-fts/tracker-fts-tokenizer.c +++ b/src/libtracker-fts/tracker-fts-tokenizer.c @@ -19,7 +19,7 @@ * 02110-1301 USA */ -/* FTS3/4 Tokenizer using TrackerParser */ +/* FTS5 Tokenizer using TrackerParser */ #include "config.h" @@ -27,209 +27,392 @@ #include <string.h> #include <libtracker-common/tracker-parser.h> +#include <libtracker-data/tracker-ontologies.h> #include "tracker-fts-tokenizer.h" #include "tracker-fts-config.h" -#include "fts3_tokenizer.h" +#include "fts5.h" +typedef struct TrackerTokenizerData TrackerTokenizerData; typedef struct TrackerTokenizer TrackerTokenizer; -typedef struct TrackerCursor TrackerCursor; + +struct TrackerTokenizerData { + TrackerLanguage *language; + int max_word_length; + int max_words; + gboolean enable_stemmer; + gboolean enable_unaccent; + gboolean ignore_numbers; + gboolean ignore_stop_words; +}; struct TrackerTokenizer { - sqlite3_tokenizer base; - TrackerLanguage *language; - int max_word_length; - int max_words; - gboolean enable_stemmer; - gboolean enable_unaccent; - gboolean ignore_numbers; - gboolean ignore_stop_words; + TrackerTokenizerData *data; + TrackerParser *parser; }; -struct TrackerCursor { - sqlite3_tokenizer_cursor base; +static int +tracker_tokenizer_create (void *data, + const char **argv, + int argc, + Fts5Tokenizer **tokenizer_out) +{ + TrackerTokenizer *tokenizer; + + tokenizer = g_new0 (TrackerTokenizer, 1); + tokenizer->data = data; + tokenizer->parser = tracker_parser_new (tokenizer->data->language); + + *tokenizer_out = (Fts5Tokenizer *) tokenizer; + + return SQLITE_OK; +} + +static void +tracker_tokenizer_destroy (Fts5Tokenizer *fts5_tokenizer) +{ + TrackerTokenizer *tokenizer = (TrackerTokenizer *) fts5_tokenizer; + + tracker_parser_free (tokenizer->parser); + g_free (tokenizer); +} - TrackerTokenizer *tokenizer; - TrackerParser *parser; - guint n_words; +typedef int (*TokenFunc) (void *pCtx, /* Copy of 2nd argument to xTokenize() */ + int flags, /* Mask of FTS5_TOKEN_* flags */ + const char *token, /* Pointer to buffer containing token */ + int n_token, /* Size of token in bytes */ + int start, /* Byte offset of token within input text */ + int end); /* Byte offset of end of token within input text */ + +static int +tracker_tokenizer_tokenize (Fts5Tokenizer *fts5_tokenizer, + void *ctx, + int flags, /* Mask of FTS5_TOKENIZE_* flags */ + const char *text, + int length, + TokenFunc token_func) +{ + TrackerTokenizer *tokenizer = (TrackerTokenizer *) fts5_tokenizer; + TrackerTokenizerData *data = tokenizer->data; + const gchar *token; + gboolean stop_word; + int n_tokens = 0, pos, start, end, len; + int rc = SQLITE_OK; + + if (length <= 0) + return rc; + + tracker_parser_reset (tokenizer->parser, text, length, + data->max_word_length, + data->enable_stemmer, + data->enable_unaccent, + data->ignore_stop_words, + TRUE, + data->ignore_numbers); + + while (n_tokens < data->max_words) { + token = tracker_parser_next (tokenizer->parser, + &pos, + &start, &end, + &stop_word, + &len); + + if (!token) + break; + + if (stop_word && data->ignore_stop_words) + continue; + + rc = token_func (ctx, 0, token, len, start, end); + + if (rc != SQLITE_OK) + break; + + n_tokens++; + } + + return rc; +} + +/* Our custom tokenizer: */ +static const fts5_tokenizer tracker_tokenizer_module = { + tracker_tokenizer_create, /* xCreate */ + tracker_tokenizer_destroy, /* xDelete */ + tracker_tokenizer_tokenize, /* xTokenize */ }; -/* -** Create a new tokenizer instance. -*/ -static int trackerCreate( - int argc, /* Number of entries in argv[] */ - const char * const *argv, /* Tokenizer creation arguments */ - sqlite3_tokenizer **ppTokenizer /* OUT: Created tokenizer */ -){ - TrackerTokenizer *p; - TrackerFTSConfig *config; - - p = (TrackerTokenizer *)sqlite3_malloc(sizeof(TrackerTokenizer)); - if( !p ){ - return SQLITE_NOMEM; - } - memset(p, 0, sizeof(TrackerTokenizer)); - p->language = tracker_language_new (NULL); - - config = tracker_fts_config_new (); - - p->max_word_length = tracker_fts_config_get_max_word_length (config); - p->enable_stemmer = tracker_fts_config_get_enable_stemmer (config); - p->enable_unaccent = tracker_fts_config_get_enable_unaccent (config); - p->ignore_numbers = tracker_fts_config_get_ignore_numbers (config); - - /* disable stop words if TRACKER_FTS_STOP_WORDS is set to 0 - used by tests - * otherwise, get value from the conf file */ - p->ignore_stop_words = (g_strcmp0 (g_getenv ("TRACKER_FTS_STOP_WORDS"), "0") == 0 ? - FALSE : tracker_fts_config_get_ignore_stop_words (config)); - - p->max_words = tracker_fts_config_get_max_words_to_index (config); - - g_object_unref (config); - - *ppTokenizer = (sqlite3_tokenizer *)p; - - return SQLITE_OK; +static TrackerTokenizerData * +tracker_tokenizer_data_new (void) +{ + TrackerTokenizerData *p; + TrackerFTSConfig *config; + + config = tracker_fts_config_new (); + + p = g_new0 (TrackerTokenizerData, 1); + p->language = tracker_language_new (NULL); + p->max_word_length = tracker_fts_config_get_max_word_length (config); + p->enable_stemmer = tracker_fts_config_get_enable_stemmer (config); + p->enable_unaccent = tracker_fts_config_get_enable_unaccent (config); + p->ignore_numbers = tracker_fts_config_get_ignore_numbers (config); + p->max_words = tracker_fts_config_get_max_words_to_index (config); + + /* disable stop words if TRACKER_FTS_STOP_WORDS is set to 0 - used by tests + * otherwise, get value from the conf file */ + p->ignore_stop_words = (g_strcmp0 (g_getenv ("TRACKER_FTS_STOP_WORDS"), "0") == 0 ? + FALSE : tracker_fts_config_get_ignore_stop_words (config)); + + g_object_unref (config); + + return p; } -/* -** Destroy a tokenizer -*/ -static int trackerDestroy(sqlite3_tokenizer *pTokenizer){ - TrackerTokenizer *p = (TrackerTokenizer *)pTokenizer; - g_object_unref (p->language); - sqlite3_free(p); - return SQLITE_OK; +static void +tracker_tokenizer_data_free (gpointer user_data) +{ + TrackerTokenizerData *data = user_data; + + g_object_unref (data->language); + g_free (data); } -/* -** Prepare to begin tokenizing a particular string. The input -** string to be tokenized is pInput[0..nBytes-1]. A cursor -** used to incrementally tokenize this string is returned in -** *ppCursor. -*/ -static int trackerOpen( - sqlite3_tokenizer *pTokenizer, /* The tokenizer */ - const char *zInput, /* Input string */ - int nInput, /* Length of zInput in bytes */ - sqlite3_tokenizer_cursor **ppCursor /* OUT: Tokenization cursor */ -){ - TrackerTokenizer *p = (TrackerTokenizer *)pTokenizer; - TrackerParser *parser; - TrackerCursor *pCsr; - - if ( nInput<0 ){ - nInput = strlen(zInput); - } - - parser = tracker_parser_new (p->language); - tracker_parser_reset (parser, zInput, nInput, - p->max_word_length, - p->enable_stemmer, - p->enable_unaccent, - p->ignore_stop_words, - TRUE, - p->ignore_numbers); - - pCsr = (TrackerCursor *)sqlite3_malloc(sizeof(TrackerCursor)); - memset(pCsr, 0, sizeof(TrackerCursor)); - pCsr->tokenizer = p; - pCsr->parser = parser; - - *ppCursor = (sqlite3_tokenizer_cursor *)pCsr; - return SQLITE_OK; +static int +offsets_tokenizer_func (void *data, + int flags, + const char *token, + int n_token, + int start, + int end) +{ + GArray *offsets = data; + g_array_append_val (offsets, start); + return SQLITE_OK; } -/* -** Close a tokenization cursor. -*/ -static int trackerClose(sqlite3_tokenizer_cursor *pCursor){ - TrackerCursor *pCsr = (TrackerCursor *)pCursor; - - tracker_parser_free (pCsr->parser); - sqlite3_free(pCsr); - return SQLITE_OK; +static void +tracker_offsets_function (const Fts5ExtensionApi *api, + Fts5Context *fts_ctx, + sqlite3_context *ctx, + int n_args, + sqlite3_value **args) +{ + GString *str; + int rc, n_hits, i; + GArray *offsets = NULL; + const gchar * const *property_names; + gint cur_col = -1; + + if (n_args > 0) { + sqlite3_result_error (ctx, "Invalid argument count", -1); + return; + } + + property_names = api->xUserData (fts_ctx); + rc = api->xInstCount (fts_ctx, &n_hits); + + if (rc != SQLITE_OK) { + sqlite3_result_null (ctx); + return; + } + + str = g_string_new (NULL); + + for (i = 0; i < n_hits; i++) { + int phrase, col, n_token; + + rc = api->xInst (fts_ctx, i, &phrase, &col, &n_token); + + if (cur_col != col) { + const char *text; + int length; + + if (offsets) + g_array_free (offsets, TRUE); + + offsets = g_array_new (FALSE, FALSE, sizeof (gint)); + api->xColumnText (fts_ctx, col, &text, &length); + api->xTokenize (fts_ctx, text, length, + offsets, &offsets_tokenizer_func); + cur_col = col; + } + + if (str->len != 0) + g_string_append_c (str, ','); + + if (rc != SQLITE_OK) + break; + + g_string_append_printf (str, "%s,%d", + property_names[col], + g_array_index (offsets, gint, n_token)); + } + + if (offsets) + g_array_free (offsets, TRUE); + + if (rc == SQLITE_OK) { + sqlite3_result_text (ctx, str->str, str->len, g_free); + g_string_free (str, FALSE); + } else { + sqlite3_result_error_code (ctx, rc); + g_string_free (str, TRUE); + } } -/* -** Extract the next token from a tokenization cursor. -*/ -static int trackerNext( - sqlite3_tokenizer_cursor *pCursor, /* Cursor returned by simpleOpen */ - const char **ppToken, /* OUT: *ppToken is the token text */ - int *pnBytes, /* OUT: Number of bytes in token */ - int *piStartOffset, /* OUT: Starting offset of token */ - int *piEndOffset, /* OUT: Ending offset of token */ - int *piPosition /* OUT: Position integer of token */ -){ - TrackerCursor *cursor = (TrackerCursor *) pCursor; - TrackerTokenizer *p; - const gchar *pToken; - gboolean stop_word; - int pos, start, end, len; - - p = cursor->tokenizer; - - if (cursor->n_words > p->max_words){ - return SQLITE_DONE; - } - - do { - pToken = tracker_parser_next (cursor->parser, - &pos, - &start, &end, - &stop_word, - &len); - - if (!pToken){ - return SQLITE_DONE; - } - } while (stop_word && p->ignore_stop_words); - - *ppToken = pToken; - *piStartOffset = start; - *piEndOffset = end; - *piPosition = pos; - *pnBytes = len; - - cursor->n_words++; - - return SQLITE_OK; +static GHashTable * +get_fts_weights (sqlite3_context *context) +{ + static GHashTable *weights = NULL; + static GMutex mutex; + int rc = SQLITE_DONE; + + g_mutex_lock (&mutex); + + if (G_UNLIKELY (weights == NULL)) { + sqlite3_stmt *stmt; + sqlite3 *db; + const gchar *uri; + + weights = g_hash_table_new (g_str_hash, g_str_equal); + db = sqlite3_context_db_handle (context); + rc = sqlite3_prepare_v2 (db, + "SELECT \"rdf:Property\".\"tracker:weight\", " + "(SELECT Uri FROM Resource where Resource.ID=\"rdf:Property\".ID) " + "FROM \"rdf:Property\" " + "WHERE \"rdf:Property\".\"tracker:fulltextIndexed\" = 1 ", + -1, &stmt, NULL); + + while ((rc = sqlite3_step (stmt)) != SQLITE_DONE) { + if (rc == SQLITE_ROW) { + TrackerProperty *property; + guint weight; + + weight = sqlite3_column_int (stmt, 0); + uri = sqlite3_column_text (stmt, 1); + + property = tracker_ontologies_get_property_by_uri (uri); + g_hash_table_insert (weights, + (gpointer) tracker_property_get_name (property), + GUINT_TO_POINTER (weight)); + } else if (rc != SQLITE_BUSY) { + break; + } + } + + sqlite3_finalize (stmt); + + if (rc != SQLITE_DONE) { + g_hash_table_destroy (weights); + weights = NULL; + } + } + + g_mutex_unlock (&mutex); + + return weights; } -/* -** The set of routines that implement the simple tokenizer -*/ -static const sqlite3_tokenizer_module trackerTokenizerModule = { - 0, /* iVersion */ - trackerCreate, /* xCreate */ - trackerDestroy, /* xDestroy */ - trackerOpen, /* xOpen */ - trackerClose, /* xClose */ - trackerNext, /* xNext */ -}; +static void +tracker_rank_function (const Fts5ExtensionApi *api, + Fts5Context *fts_ctx, + sqlite3_context *ctx, + int n_args, + sqlite3_value **args) +{ + const gchar * const *property_names; + int i, rc, n_columns, n_tokens; + GHashTable *weights; + gdouble rank = 0; + + if (n_args != 0) { + sqlite3_result_error (ctx, "Invalid argument count", -1); + return; + } + + n_columns = api->xColumnCount (fts_ctx); + property_names = api->xUserData (fts_ctx); + weights = get_fts_weights (ctx); + + if (!weights) { + sqlite3_result_error (ctx, "Could not read FTS weights", -1); + return; + } + + for (i = 0; i < n_columns; i++) { + const gchar *property; + guint weight; + + rc = api->xColumnSize (fts_ctx, i, &n_tokens); + if (rc != SQLITE_OK) + break; + + if (n_tokens <= 0) + continue; + + property = property_names[i]; + weight = GPOINTER_TO_UINT (g_hash_table_lookup (weights, property)); + rank += weight; + } + + if (rc == SQLITE_OK) { + sqlite3_result_double (ctx, rank); + } else { + sqlite3_result_error_code (ctx, rc); + } +} -/* -** Set *ppModule to point at the implementation of the tracker tokenizer. -*/ -gboolean tracker_tokenizer_initialize (sqlite3 *db) { - const sqlite3_tokenizer_module *pTokenizer; - int rc = SQLITE_OK; - sqlite3_stmt *stmt; - - pTokenizer = &trackerTokenizerModule; - rc = sqlite3_prepare_v2(db, "SELECT fts3_tokenizer(?, ?)", - -1, &stmt, 0); - - if (rc != SQLITE_OK) { - return FALSE; - } - - sqlite3_bind_text(stmt, 1, "TrackerTokenizer", -1, SQLITE_STATIC); - sqlite3_bind_blob(stmt, 2, &pTokenizer, sizeof(pTokenizer), SQLITE_STATIC); - sqlite3_step(stmt); - rc = sqlite3_finalize(stmt); - - return (rc == SQLITE_OK); +static fts5_api * +get_fts5_api (sqlite3 *db) { + int rc = SQLITE_OK; + sqlite3_stmt *stmt; + fts5_api *api; + + rc = sqlite3_prepare_v2(db, "SELECT fts5()", + -1, &stmt, 0); + + if (rc != SQLITE_OK) { + return NULL; + } + + if (sqlite3_step (stmt) != SQLITE_ROW) { + return NULL; + } + + memcpy (&api, sqlite3_column_blob (stmt, 0), sizeof (api)); + sqlite3_finalize (stmt); + + return api; +} + +gboolean +tracker_tokenizer_initialize (sqlite3 *db, + const gchar **property_names) +{ + TrackerTokenizerData *data; + fts5_tokenizer *tokenizer; + fts5_api *api; + + api = get_fts5_api (db); + + if (!api) + return FALSE; + + data = tracker_tokenizer_data_new (); + tokenizer = (fts5_tokenizer *) &tracker_tokenizer_module; + api->xCreateTokenizer (api, "TrackerTokenizer", data, tokenizer, + tracker_tokenizer_data_free); + + /* Offsets */ + api->xCreateFunction (api, "tracker_offsets", + g_strdupv ((gchar **) property_names), + &tracker_offsets_function, + (GDestroyNotify) g_strfreev); + + /* Rank */ + api->xCreateFunction (api, "tracker_rank", + g_strdupv ((gchar **) property_names), + &tracker_rank_function, + (GDestroyNotify) g_strfreev); + + return TRUE; } diff --git a/src/libtracker-fts/tracker-fts-tokenizer.h b/src/libtracker-fts/tracker-fts-tokenizer.h index 3e86295a5..30843d23e 100644 --- a/src/libtracker-fts/tracker-fts-tokenizer.h +++ b/src/libtracker-fts/tracker-fts-tokenizer.h @@ -21,11 +21,11 @@ #include <sqlite3.h> #include <glib.h> -#include "fts3_tokenizer.h" #ifndef __TRACKER_FTS_TOKENIZER_H__ #define __TRACKER_FTS_TOKENIZER_H__ -gboolean tracker_tokenizer_initialize (sqlite3 *db); +gboolean tracker_tokenizer_initialize (sqlite3 *db, + const gchar **property_names); #endif /* __TRACKER_FTS_TOKENIZER_H__ */ diff --git a/src/libtracker-fts/tracker-fts.c b/src/libtracker-fts/tracker-fts.c index 0571fb0f4..deb951a23 100644 --- a/src/libtracker-fts/tracker-fts.c +++ b/src/libtracker-fts/tracker-fts.c @@ -28,13 +28,16 @@ #ifndef HAVE_BUILTIN_FTS -#include "fts3.h" +#include "sqlite3.h" +#include "fts5.h" + static gsize module_initialized = 0; +int sqlite3_fts5_init (); + #endif static gboolean initialized = FALSE; -static GPrivate property_names_key = G_PRIVATE_INIT ((GDestroyNotify) g_strfreev); gboolean @@ -47,13 +50,13 @@ tracker_fts_init (void) #ifdef HAVE_BUILTIN_FTS initialized = TRUE; - /* SQLite has all needed FTS4 features compiled in */ + /* SQLite has all needed FTS5 features compiled in */ return TRUE; #else int rc = SQLITE_OK; if (g_once_init_enter (&module_initialized)) { - rc = sqlite3_auto_extension ((void (*) (void)) fts4_extension_init); + rc = sqlite3_auto_extension ((void (*) (void)) sqlite3_fts5_init); g_once_init_leave (&module_initialized, (rc == SQLITE_OK)); } @@ -70,202 +73,45 @@ tracker_fts_shutdown (void) return TRUE; } - /* Nothing to do, there is no fts4_extension_shutdown() */ initialized = FALSE; return TRUE; } -static void -function_rank (sqlite3_context *context, - int argc, - sqlite3_value *argv[]) -{ - guint *matchinfo, *weights; - gdouble rank = 0; - gint i, n_columns; - - if (argc != 2) { - sqlite3_result_error(context, - "wrong number of arguments to function rank()", - -1); - return; - } - - matchinfo = (unsigned int *) sqlite3_value_blob (argv[0]); - weights = (unsigned int *) sqlite3_value_blob (argv[1]); - n_columns = matchinfo[0]; - - for (i = 0; i < n_columns; i++) { - if (matchinfo[i + 1] != 0) { - rank += (gdouble) weights[i]; - } - } - - sqlite3_result_double(context, rank); -} - -static void -function_offsets (sqlite3_context *context, - int argc, - sqlite3_value *argv[]) +static gchar ** +get_fts_properties (GHashTable *tables) { - gchar *offsets; - const gchar * const * names; - gint offset_values[4]; - GString *result = NULL; - gint i = 0; - - if (argc != 2) { - sqlite3_result_error(context, - "wrong number of arguments to function tracker_offsets()", - -1); - return; - } - - offsets = (gchar *) sqlite3_value_text (argv[0]); - names = sqlite3_value_blob (argv[1]); - - while (offsets && *offsets) { - offset_values[i] = g_strtod (offsets, &offsets); - - /* All 4 values from the quartet have been gathered */ - if (i == 3) { - if (!result) { - result = g_string_new (""); - } else { - g_string_append_c (result, ','); - } - - g_string_append_printf (result, - "%s,%d", - names[offset_values[0]], - offset_values[2]); - - } - - i = (i + 1) % 4; - } - - sqlite3_result_text (context, - (result) ? g_string_free (result, FALSE) : NULL, - -1, g_free); -} - -static void -function_weights (sqlite3_context *context, - int argc, - sqlite3_value *argv[]) -{ - static guint *weights = NULL; - static GMutex mutex; - static gint length; - int rc = SQLITE_DONE; - - g_mutex_lock (&mutex); - - if (G_UNLIKELY (weights == NULL)) { - GArray *weight_array; - sqlite3_stmt *stmt; - sqlite3 *db; - - weight_array = g_array_new (FALSE, FALSE, sizeof (guint)); - db = sqlite3_context_db_handle (context); - rc = sqlite3_prepare_v2 (db, - "SELECT \"rdf:Property\".\"tracker:weight\" " - "FROM \"rdf:Property\" " - "WHERE \"rdf:Property\".\"tracker:fulltextIndexed\" = 1 " - "ORDER BY \"rdf:Property\".ID ", - -1, &stmt, NULL); - - while ((rc = sqlite3_step (stmt)) != SQLITE_DONE) { - if (rc == SQLITE_ROW) { - guint weight; - weight = sqlite3_column_int (stmt, 0); - g_array_append_val (weight_array, weight); - } else if (rc != SQLITE_BUSY) { - break; - } - } + GList *table_columns, *columns; + gchar **property_names; + GHashTableIter iter; - sqlite3_finalize (stmt); + columns = NULL; + g_hash_table_iter_init (&iter, tables); - if (rc == SQLITE_DONE) { - length = weight_array->len * g_array_get_element_size (weight_array); - weights = (guint *) g_array_free (weight_array, FALSE); - } else { - g_array_free (weight_array, TRUE); - } + while (g_hash_table_iter_next (&iter, NULL, (gpointer *) &table_columns)) { + columns = g_list_concat (columns, g_list_copy (table_columns)); } - g_mutex_unlock (&mutex); - - if (rc == SQLITE_DONE) - sqlite3_result_blob (context, weights, length, NULL); - else - sqlite3_result_error_code (context, rc); -} - -static void -function_property_names (sqlite3_context *context, - int argc, - sqlite3_value *argv[]) -{ - gchar **property_names; - gint length; - - property_names = g_private_get (&property_names_key); - length = g_strv_length (property_names) * sizeof (gchar *); - sqlite3_result_blob (context, property_names, length, NULL); -} + property_names = tracker_glist_to_string_list (columns); + g_list_free (columns); -static void -fts_register_functions (sqlite3 *db) -{ - sqlite3_create_function (db, "tracker_rank", 2, SQLITE_ANY, - NULL, &function_rank, - NULL, NULL); - sqlite3_create_function (db, "tracker_offsets", 2, SQLITE_ANY, - NULL, &function_offsets, - NULL, NULL); - sqlite3_create_function (db, "fts_column_weights", 0, SQLITE_ANY, - NULL, &function_weights, - NULL, NULL); - sqlite3_create_function (db, "fts_property_names", 0, SQLITE_ANY, - NULL, &function_property_names, - NULL, NULL); + return property_names; } gboolean tracker_fts_init_db (sqlite3 *db, GHashTable *tables) { - GHashTableIter iter; - GList *columns; - GList *table_columns; gchar **property_names; + gboolean retval; g_return_val_if_fail (initialized == TRUE, FALSE); - if (!tracker_tokenizer_initialize (db)) { - return FALSE; - } - - /* Set up GStrv 'property_names' */ - columns = NULL; - g_hash_table_iter_init (&iter, tables); - while (g_hash_table_iter_next (&iter, NULL, (gpointer *) &table_columns)) { - columns = g_list_concat (columns, g_list_copy (table_columns)); - } - - property_names = tracker_glist_to_string_list (columns); - g_private_replace (&property_names_key, property_names); - g_list_free (columns); - - /* Register functions with the database, including one to get property names */ - fts_register_functions (db); + property_names = get_fts_properties (tables); + retval = tracker_tokenizer_initialize (db, (const gchar **) property_names); + g_strfreev (property_names); - return TRUE; + return retval; } gboolean @@ -282,13 +128,16 @@ tracker_fts_create_table (sqlite3 *db, g_return_val_if_fail (initialized == TRUE, FALSE); + if (g_hash_table_size (tables) == 0) + return TRUE; + /* Create view on tables/columns marked as FTS-indexed */ g_hash_table_iter_init (&iter, tables); str = g_string_new ("CREATE VIEW fts_view AS SELECT Resource.ID as rowid "); from = g_string_new ("FROM Resource "); fts = g_string_new ("CREATE VIRTUAL TABLE "); - g_string_append_printf (fts, "%s USING fts4(content=\"fts_view\", ", + g_string_append_printf (fts, "%s USING fts5(content=\"fts_view\", ", table_name); while (g_hash_table_iter_next (&iter, (gpointer *) &index_table, @@ -321,17 +170,28 @@ tracker_fts_create_table (sqlite3 *db, g_string_append (str, from->str); g_string_free (from, TRUE); - rc = sqlite3_exec(db, str->str, NULL, 0, NULL); + rc = sqlite3_exec(db, str->str, NULL, NULL, NULL); g_string_free (str, TRUE); if (rc != SQLITE_OK) { + g_assert_not_reached(); return FALSE; } g_string_append (fts, "tokenize=TrackerTokenizer)"); - rc = sqlite3_exec(db, fts->str, NULL, 0, NULL); + rc = sqlite3_exec(db, fts->str, NULL, NULL, NULL); g_string_free (fts, TRUE); + if (rc != SQLITE_OK) + return FALSE; + + str = g_string_new (NULL); + g_string_append_printf (str, + "INSERT INTO %s(%s, rank) VALUES('rank', 'tracker_rank()')", + table_name, table_name); + rc = sqlite3_exec (db, str->str, NULL, NULL, NULL); + g_string_free (str, TRUE); + return (rc == SQLITE_OK); } @@ -349,17 +209,21 @@ tracker_fts_alter_table (sqlite3 *db, tmp_name = g_strdup_printf ("%s_TMP", table_name); query = g_strdup_printf ("DROP VIEW fts_view"); - rc = sqlite3_prepare_v2 (db, query, -1, NULL, NULL); + rc = sqlite3_exec (db, query, NULL, NULL, NULL); + g_free (query); + + query = g_strdup_printf ("DROP TABLE fts5"); + rc = sqlite3_exec (db, query, NULL, NULL, NULL); + g_free (query); if (!tracker_fts_create_table (db, tmp_name, tables, grouped_columns)) { g_free (tmp_name); - g_free (query); return FALSE; } - query = g_strdup_printf ("INSERT INTO %s (docid) SELECT docid FROM %s", - tmp_name, table_name); - rc = sqlite3_prepare_v2 (db, query, -1, NULL, NULL); + query = g_strdup_printf ("INSERT INTO %s (rowid) SELECT rowid FROM fts_view", + tmp_name); + rc = sqlite3_exec (db, query, NULL, NULL, NULL); g_free (query); if (rc != SQLITE_OK) { @@ -369,7 +233,7 @@ tracker_fts_alter_table (sqlite3 *db, query = g_strdup_printf ("INSERT INTO %s(%s) VALUES('rebuild')", tmp_name, tmp_name); - rc = sqlite3_prepare_v2 (db, query, -1, NULL, NULL); + rc = sqlite3_exec (db, query, NULL, NULL, NULL); g_free (query); if (rc != SQLITE_OK) { @@ -379,17 +243,11 @@ tracker_fts_alter_table (sqlite3 *db, query = g_strdup_printf ("ALTER TABLE %s RENAME TO %s", tmp_name, table_name); - rc = sqlite3_prepare_v2 (db, query, -1, NULL, NULL); + rc = sqlite3_exec (db, query, NULL, NULL, NULL); g_free (query); g_free (tmp_name); - - if (rc != SQLITE_OK) { - g_free (tmp_name); - return FALSE; - } - - return TRUE; + return rc == SQLITE_OK; } void @@ -401,6 +259,6 @@ tracker_fts_rebuild_tokens (sqlite3 *db, /* This special query rebuilds the tokens in the given FTS table */ query = g_strdup_printf ("INSERT INTO %s(%s) VALUES('rebuild')", table_name, table_name); - sqlite3_exec(db, query, NULL, 0, NULL); + sqlite3_exec(db, query, NULL, NULL, NULL); g_free (query); } diff --git a/src/ontologies/32-nco.ontology b/src/ontologies/32-nco.ontology index a7c80f5ba..6e7abb5a9 100644 --- a/src/ontologies/32-nco.ontology +++ b/src/ontologies/32-nco.ontology @@ -11,7 +11,7 @@ nco: a tracker:Namespace, tracker:Ontology ; tracker:prefix "nco" ; - nao:lastModified "2011-09-29T12:20:00Z" . + nao:lastModified "2016-02-28T21:30:00Z" . nco:Role a rdfs:Class ; rdfs:label "Role" ; @@ -599,7 +599,6 @@ nco:hobby a rdf:Property ; nrl:maxCardinality 1 ; rdfs:domain nco:PersonContact ; rdfs:range xsd:string ; - tracker:fulltextIndexed true ; tracker:weight 2 . # Same remarks as in url |