/* Copyright (c) 2000, 2016, Oracle and/or its affiliates. Copyright (c) 2010, 2016, MariaDB This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; version 2 of the License. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program; if not, write to the Free Software Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA */ /* Copy data from a textfile to table */ /* 2006-12 Erik Wetterberg : LOAD XML added */ #include #include "sql_priv.h" #include "unireg.h" #include "sql_load.h" #include "sql_load.h" #include "sql_cache.h" // query_cache_* #include "sql_base.h" // fill_record_n_invoke_before_triggers #include #include "sql_view.h" // check_key_in_view #include "sql_insert.h" // check_that_all_fields_are_given_values, // write_record #include "sql_acl.h" // INSERT_ACL, UPDATE_ACL #include "log_event.h" // Delete_file_log_event, // Execute_load_query_log_event, // LOG_EVENT_UPDATE_TABLE_MAP_VERSION_F #include #include "rpl_mi.h" #include "sql_repl.h" #include "sp_head.h" #include "sql_trigger.h" #include "sql_derived.h" #include "sql_show.h" extern "C" int _my_b_net_read(IO_CACHE *info, uchar *Buffer, size_t Count); class XML_TAG { public: int level; String field; String value; XML_TAG(int l, String f, String v); }; XML_TAG::XML_TAG(int l, String f, String v) { level= l; field.append(f); value.append(v); } #define GET (stack_pos != stack ? *--stack_pos : my_b_get(&cache)) #define PUSH(A) *(stack_pos++)=(A) class READ_INFO { File file; uchar *buffer, /* Buffer for read text */ *end_of_buff; /* Data in bufferts ends here */ uint buff_length, /* Length of buffert */ max_length; /* Max length of row */ const uchar *field_term_ptr,*line_term_ptr; const char *line_start_ptr,*line_start_end; uint field_term_length,line_term_length,enclosed_length; int field_term_char,line_term_char,enclosed_char,escape_char; int *stack,*stack_pos; bool found_end_of_line,start_of_line,eof; NET *io_net; int level; /* for load xml */ #if MYSQL_VERSION_ID >= 100200 #error This 10.0 and 10.1 specific fix should be removed in 10.2. #error Fix read_mbtail() to use my_charlen() instead of my_charlen_tmp() #else int my_charlen_tmp(CHARSET_INFO *cs, const char *str, const char *end) { my_wc_t wc; return cs->cset->mb_wc(cs, &wc, (const uchar *) str, (const uchar *) end); } /** Read a tail of a multi-byte character. The first byte of the character is assumed to be already read from the file and appended to "str". @returns true - if EOF happened unexpectedly @returns false - no EOF happened: found a good multi-byte character, or a bad byte sequence Note: The return value depends only on EOF: - read_mbtail() returns "false" is a good character was read, but also - read_mbtail() returns "false" if an incomplete byte sequence was found and no EOF happened. For example, suppose we have an ujis file with bytes 0x8FA10A, where: - 0x8FA1 is an incomplete prefix of a 3-byte character (it should be [8F][A1-FE][A1-FE] to make a full 3-byte character) - 0x0A is a line demiliter This file has some broken data, the trailing [A1-FE] is missing. In this example it works as follows: - 0x8F is read from the file and put into "data" before the call for read_mbtail() - 0xA1 is read from the file and put into "data" by read_mbtail() - 0x0A is kept in the read queue, so the next read iteration after the current read_mbtail() call will normally find it and recognize as a line delimiter - the current call for read_mbtail() returns "false", because no EOF happened */ bool read_mbtail(String *str) { int chlen; if ((chlen= my_charlen_tmp(read_charset, str->end() - 1, str->end())) == 1) return false; // Single byte character found for (uint32 length0= str->length() - 1 ; MY_CS_IS_TOOSMALL(chlen); ) { int chr= GET; if (chr == my_b_EOF) { DBUG_PRINT("info", ("read_mbtail: chlen=%d; unexpected EOF", chlen)); return true; // EOF } str->append(chr); chlen= my_charlen_tmp(read_charset, str->ptr() + length0, str->end()); if (chlen == MY_CS_ILSEQ) { /** It has been an incomplete (but a valid) sequence so far, but the last byte turned it into a bad byte sequence. Unget the very last byte. */ str->length(str->length() - 1); PUSH(chr); DBUG_PRINT("info", ("read_mbtail: ILSEQ")); return false; // Bad byte sequence } } DBUG_PRINT("info", ("read_mbtail: chlen=%d", chlen)); return false; // Good multi-byte character } #endif public: bool error,line_cuted,found_null,enclosed; uchar *row_start, /* Found row starts here */ *row_end; /* Found row ends here */ CHARSET_INFO *read_charset; LOAD_FILE_IO_CACHE cache; READ_INFO(File file,uint tot_length,CHARSET_INFO *cs, String &field_term,String &line_start,String &line_term, String &enclosed,int escape,bool get_it_from_net, bool is_fifo); ~READ_INFO(); int read_field(); int read_fixed_length(void); int next_line(void); char unescape(char chr); int terminator(const uchar *ptr, uint length); bool find_start_of_fields(); /* load xml */ List taglist; int read_value(int delim, String *val); int read_xml(THD *thd); int clear_level(int level); my_off_t file_length() { return cache.end_of_file; } my_off_t position() { return my_b_tell(&cache); } /** skip all data till the eof. */ void skip_data_till_eof() { while (GET != my_b_EOF) ; } }; static int read_fixed_length(THD *thd, COPY_INFO &info, TABLE_LIST *table_list, List &fields_vars, List &set_fields, List &set_values, READ_INFO &read_info, ulong skip_lines, bool ignore_check_option_errors); static int read_sep_field(THD *thd, COPY_INFO &info, TABLE_LIST *table_list, List &fields_vars, List &set_fields, List &set_values, READ_INFO &read_info, String &enclosed, ulong skip_lines, bool ignore_check_option_errors); static int read_xml_field(THD *thd, COPY_INFO &info, TABLE_LIST *table_list, List &fields_vars, List &set_fields, List &set_values, READ_INFO &read_info, String &enclosed, ulong skip_lines, bool ignore_check_option_errors); #ifndef EMBEDDED_LIBRARY static bool write_execute_load_query_log_event(THD *, sql_exchange*, const char*, const char*, bool, enum enum_duplicates, bool, bool, int); #endif /* EMBEDDED_LIBRARY */ /* Execute LOAD DATA query SYNOPSYS mysql_load() thd - current thread ex - sql_exchange object representing source file and its parsing rules table_list - list of tables to which we are loading data fields_vars - list of fields and variables to which we read data from file set_fields - list of fields mentioned in set clause set_values - expressions to assign to fields in previous list handle_duplicates - indicates whenever we should emit error or replace row if we will meet duplicates. ignore - - indicates whenever we should ignore duplicates read_file_from_client - is this LOAD DATA LOCAL ? RETURN VALUES TRUE - error / FALSE - success */ int mysql_load(THD *thd,sql_exchange *ex,TABLE_LIST *table_list, List &fields_vars, List &set_fields, List &set_values, enum enum_duplicates handle_duplicates, bool ignore, bool read_file_from_client) { char name[FN_REFLEN]; File file; TABLE *table= NULL; int error= 0; String *field_term=ex->field_term,*escaped=ex->escaped; String *enclosed=ex->enclosed; bool is_fifo=0; #ifndef EMBEDDED_LIBRARY killed_state killed_status; bool is_concurrent; #endif char *db = table_list->db; // This is never null /* If path for file is not defined, we will use the current database. If this is not set, we will use the directory where the table to be loaded is located */ char *tdb= thd->db ? thd->db : db; // Result is never null ulong skip_lines= ex->skip_lines; bool transactional_table __attribute__((unused)); DBUG_ENTER("mysql_load"); /* Bug #34283 mysqlbinlog leaves tmpfile after termination if binlog contains load data infile, so in mixed mode we go to row-based for avoiding the problem. */ thd->set_current_stmt_binlog_format_row_if_mixed(); #ifdef EMBEDDED_LIBRARY read_file_from_client = 0; //server is always in the same process #endif if (escaped->length() > 1 || enclosed->length() > 1) { my_message(ER_WRONG_FIELD_TERMINATORS, ER_THD(thd, ER_WRONG_FIELD_TERMINATORS), MYF(0)); DBUG_RETURN(TRUE); } /* Report problems with non-ascii separators */ if (!escaped->is_ascii() || !enclosed->is_ascii() || !field_term->is_ascii() || !ex->line_term->is_ascii() || !ex->line_start->is_ascii()) { push_warning(thd, Sql_condition::WARN_LEVEL_WARN, WARN_NON_ASCII_SEPARATOR_NOT_IMPLEMENTED, ER_THD(thd, WARN_NON_ASCII_SEPARATOR_NOT_IMPLEMENTED)); } if (open_and_lock_tables(thd, table_list, TRUE, 0)) DBUG_RETURN(TRUE); if (mysql_handle_single_derived(thd->lex, table_list, DT_MERGE_FOR_INSERT) || mysql_handle_single_derived(thd->lex, table_list, DT_PREPARE)) DBUG_RETURN(TRUE); if (setup_tables_and_check_access(thd, &thd->lex->select_lex.context, &thd->lex->select_lex.top_join_list, table_list, thd->lex->select_lex.leaf_tables, FALSE, INSERT_ACL | UPDATE_ACL, INSERT_ACL | UPDATE_ACL, FALSE)) DBUG_RETURN(-1); if (!table_list->table || // do not suport join view !table_list->single_table_updatable() || // and derived tables check_key_in_view(thd, table_list)) { my_error(ER_NON_UPDATABLE_TABLE, MYF(0), table_list->alias, "LOAD"); DBUG_RETURN(TRUE); } if (table_list->prepare_where(thd, 0, TRUE) || table_list->prepare_check_option(thd)) { DBUG_RETURN(TRUE); } thd_proc_info(thd, "executing"); /* Let us emit an error if we are loading data to table which is used in subselect in SET clause like we do it for INSERT. The main thing to fix to remove this restriction is to ensure that the table is marked to be 'used for insert' in which case we should never mark this table as 'const table' (ie, one that has only one row). */ if (unique_table(thd, table_list, table_list->next_global, 0)) { my_error(ER_UPDATE_TABLE_USED, MYF(0), table_list->table_name, "LOAD DATA"); DBUG_RETURN(TRUE); } table= table_list->table; transactional_table= table->file->has_transactions(); #ifndef EMBEDDED_LIBRARY is_concurrent= (table_list->lock_type == TL_WRITE_CONCURRENT_INSERT); #endif if (!fields_vars.elements) { Field_iterator_table_ref field_iterator; field_iterator.set(table_list); for (; !field_iterator.end_of_fields(); field_iterator.next()) { Item *item; if (!(item= field_iterator.create_item(thd))) DBUG_RETURN(TRUE); fields_vars.push_back(item->real_item(), thd->mem_root); } bitmap_set_all(table->write_set); /* Let us also prepare SET clause, altough it is probably empty in this case. */ if (setup_fields(thd, 0, set_fields, MARK_COLUMNS_WRITE, 0, 0) || setup_fields(thd, 0, set_values, MARK_COLUMNS_READ, 0, 0)) DBUG_RETURN(TRUE); } else { // Part field list /* TODO: use this conds for 'WITH CHECK OPTIONS' */ if (setup_fields(thd, 0, fields_vars, MARK_COLUMNS_WRITE, 0, 0) || setup_fields(thd, 0, set_fields, MARK_COLUMNS_WRITE, 0, 0) || check_that_all_fields_are_given_values(thd, table, table_list)) DBUG_RETURN(TRUE); /* Add all fields with default functions to table->write_set. */ if (table->default_field) table->mark_default_fields_for_write(); /* Fix the expressions in SET clause */ if (setup_fields(thd, 0, set_values, MARK_COLUMNS_READ, 0, 0)) DBUG_RETURN(TRUE); } switch_to_nullable_trigger_fields(fields_vars, table); switch_to_nullable_trigger_fields(set_fields, table); switch_to_nullable_trigger_fields(set_values, table); table->prepare_triggers_for_insert_stmt_or_event(); table->mark_columns_needed_for_insert(); if (table->vfield) { for (Field **vfield_ptr= table->vfield; *vfield_ptr; vfield_ptr++) { if ((*vfield_ptr)->stored_in_db) { thd->lex->unit.insert_table_with_stored_vcol= table; break; } } } uint tot_length=0; bool use_blobs= 0, use_vars= 0; List_iterator_fast it(fields_vars); Item *item; while ((item= it++)) { Item *real_item= item->real_item(); if (real_item->type() == Item::FIELD_ITEM) { Field *field= ((Item_field*)real_item)->field; if (field->flags & BLOB_FLAG) { use_blobs= 1; tot_length+= 256; // Will be extended if needed } else tot_length+= field->field_length; } else if (item->type() == Item::STRING_ITEM) use_vars= 1; } if (use_blobs && !ex->line_term->length() && !field_term->length()) { my_message(ER_BLOBS_AND_NO_TERMINATED, ER_THD(thd, ER_BLOBS_AND_NO_TERMINATED), MYF(0)); DBUG_RETURN(TRUE); } if (use_vars && !field_term->length() && !enclosed->length()) { my_error(ER_LOAD_FROM_FIXED_SIZE_ROWS_TO_VAR, MYF(0)); DBUG_RETURN(TRUE); } /* We can't give an error in the middle when using LOCAL files */ if (read_file_from_client && handle_duplicates == DUP_ERROR) ignore= 1; #ifndef EMBEDDED_LIBRARY if (read_file_from_client) { (void)net_request_file(&thd->net,ex->file_name); file = -1; } else #endif { #ifdef DONT_ALLOW_FULL_LOAD_DATA_PATHS ex->file_name+=dirname_length(ex->file_name); #endif if (!dirname_length(ex->file_name)) { strxnmov(name, FN_REFLEN-1, mysql_real_data_home, tdb, NullS); (void) fn_format(name, ex->file_name, name, "", MY_RELATIVE_PATH | MY_UNPACK_FILENAME); } else { (void) fn_format(name, ex->file_name, mysql_real_data_home, "", MY_RELATIVE_PATH | MY_UNPACK_FILENAME | MY_RETURN_REAL_PATH); } if (thd->rgi_slave) { #if defined(HAVE_REPLICATION) && !defined(MYSQL_CLIENT) if (strncmp(thd->rgi_slave->rli->slave_patternload_file, name, thd->rgi_slave->rli->slave_patternload_file_size)) { /* LOAD DATA INFILE in the slave SQL Thread can only read from --slave-load-tmpdir". This should never happen. Please, report a bug. */ sql_print_error("LOAD DATA INFILE in the slave SQL Thread can only read from --slave-load-tmpdir. " \ "Please, report a bug."); my_error(ER_OPTION_PREVENTS_STATEMENT, MYF(0), "--slave-load-tmpdir"); DBUG_RETURN(TRUE); } #else /* This is impossible and should never happen. */ DBUG_ASSERT(FALSE); #endif } else if (!is_secure_file_path(name)) { /* Read only allowed from within dir specified by secure_file_priv */ my_error(ER_OPTION_PREVENTS_STATEMENT, MYF(0), "--secure-file-priv"); DBUG_RETURN(TRUE); } #if !defined(__WIN__) && ! defined(__NETWARE__) MY_STAT stat_info; if (!my_stat(name, &stat_info, MYF(MY_WME))) DBUG_RETURN(TRUE); // if we are not in slave thread, the file must be: if (!thd->slave_thread && !((stat_info.st_mode & S_IFLNK) != S_IFLNK && // symlink ((stat_info.st_mode & S_IFREG) == S_IFREG || // regular file (stat_info.st_mode & S_IFIFO) == S_IFIFO))) // named pipe { my_error(ER_TEXTFILE_NOT_READABLE, MYF(0), name); DBUG_RETURN(TRUE); } if ((stat_info.st_mode & S_IFIFO) == S_IFIFO) is_fifo= 1; #endif if ((file= mysql_file_open(key_file_load, name, O_RDONLY, MYF(MY_WME))) < 0) DBUG_RETURN(TRUE); } COPY_INFO info; bzero((char*) &info,sizeof(info)); info.ignore= ignore; info.handle_duplicates=handle_duplicates; info.escape_char= (escaped->length() && (ex->escaped_given() || !(thd->variables.sql_mode & MODE_NO_BACKSLASH_ESCAPES))) ? (*escaped)[0] : INT_MAX; READ_INFO read_info(file,tot_length, ex->cs ? ex->cs : thd->variables.collation_database, *field_term,*ex->line_start, *ex->line_term, *enclosed, info.escape_char, read_file_from_client, is_fifo); if (read_info.error) { if (file >= 0) mysql_file_close(file, MYF(0)); // no files in net reading DBUG_RETURN(TRUE); // Can't allocate buffers } #ifndef EMBEDDED_LIBRARY if (mysql_bin_log.is_open()) { read_info.cache.thd = thd; read_info.cache.wrote_create_file = 0; read_info.cache.last_pos_in_file = HA_POS_ERROR; read_info.cache.log_delayed= transactional_table; } #endif /*!EMBEDDED_LIBRARY*/ thd->count_cuted_fields= CHECK_FIELD_WARN; /* calc cuted fields */ thd->cuted_fields=0L; /* Skip lines if there is a line terminator */ if (ex->line_term->length() && ex->filetype != FILETYPE_XML) { /* ex->skip_lines needs to be preserved for logging */ while (skip_lines > 0) { skip_lines--; if (read_info.next_line()) break; } } thd_proc_info(thd, "reading file"); if (!(error= MY_TEST(read_info.error))) { table->reset_default_fields(); table->next_number_field=table->found_next_number_field; if (ignore || handle_duplicates == DUP_REPLACE) table->file->extra(HA_EXTRA_IGNORE_DUP_KEY); if (handle_duplicates == DUP_REPLACE && (!table->triggers || !table->triggers->has_delete_triggers())) table->file->extra(HA_EXTRA_WRITE_CAN_REPLACE); if (thd->locked_tables_mode <= LTM_LOCK_TABLES) table->file->ha_start_bulk_insert((ha_rows) 0); table->copy_blobs=1; thd->abort_on_warning= !ignore && thd->is_strict_mode(); thd_progress_init(thd, 2); if (table_list->table->validate_default_values_of_unset_fields(thd)) { read_info.error= true; error= 1; } else if (ex->filetype == FILETYPE_XML) /* load xml */ error= read_xml_field(thd, info, table_list, fields_vars, set_fields, set_values, read_info, *(ex->line_term), skip_lines, ignore); else if (!field_term->length() && !enclosed->length()) error= read_fixed_length(thd, info, table_list, fields_vars, set_fields, set_values, read_info, skip_lines, ignore); else error= read_sep_field(thd, info, table_list, fields_vars, set_fields, set_values, read_info, *enclosed, skip_lines, ignore); thd_proc_info(thd, "End bulk insert"); if (!error) thd_progress_next_stage(thd); if (thd->locked_tables_mode <= LTM_LOCK_TABLES && table->file->ha_end_bulk_insert() && !error) { table->file->print_error(my_errno, MYF(0)); error= 1; } table->file->extra(HA_EXTRA_NO_IGNORE_DUP_KEY); table->file->extra(HA_EXTRA_WRITE_CANNOT_REPLACE); table->next_number_field=0; } if (file >= 0) mysql_file_close(file, MYF(0)); free_blobs(table); /* if pack_blob was used */ table->copy_blobs=0; thd->count_cuted_fields= CHECK_FIELD_IGNORE; /* simulated killing in the middle of per-row loop must be effective for binlogging */ DBUG_EXECUTE_IF("simulate_kill_bug27571", { error=1; thd->killed= KILL_QUERY; };); #ifndef EMBEDDED_LIBRARY killed_status= (error == 0) ? NOT_KILLED : thd->killed; #endif /* We must invalidate the table in query cache before binlog writing and ha_autocommit_... */ query_cache_invalidate3(thd, table_list, 0); if (error) { if (read_file_from_client) read_info.skip_data_till_eof(); #ifndef EMBEDDED_LIBRARY if (mysql_bin_log.is_open()) { { /* Make sure last block (the one which caused the error) gets logged. */ log_loaded_block(&read_info.cache, 0, 0); /* If the file was not empty, wrote_create_file is true */ if (read_info.cache.wrote_create_file) { int errcode= query_error_code(thd, killed_status == NOT_KILLED); /* since there is already an error, the possible error of writing binary log will be ignored */ if (thd->transaction.stmt.modified_non_trans_table) (void) write_execute_load_query_log_event(thd, ex, table_list->db, table_list->table_name, is_concurrent, handle_duplicates, ignore, transactional_table, errcode); else { Delete_file_log_event d(thd, db, transactional_table); (void) mysql_bin_log.write(&d); } } } } #endif /*!EMBEDDED_LIBRARY*/ error= -1; // Error on read goto err; } sprintf(name, ER_THD(thd, ER_LOAD_INFO), (ulong) info.records, (ulong) info.deleted, (ulong) (info.records - info.copied), (long) thd->get_stmt_da()->current_statement_warn_count()); if (thd->transaction.stmt.modified_non_trans_table) thd->transaction.all.modified_non_trans_table= TRUE; thd->transaction.all.m_unsafe_rollback_flags|= (thd->transaction.stmt.m_unsafe_rollback_flags & THD_TRANS::DID_WAIT); #ifndef EMBEDDED_LIBRARY if (mysql_bin_log.is_open()) { /* We need to do the job that is normally done inside binlog_query() here, which is to ensure that the pending event is written before tables are unlocked and before any other events are written. We also need to update the table map version for the binary log to mark that table maps are invalid after this point. */ if (thd->is_current_stmt_binlog_format_row()) error= thd->binlog_flush_pending_rows_event(TRUE, transactional_table); else { /* As already explained above, we need to call log_loaded_block() to have the last block logged */ log_loaded_block(&read_info.cache, 0, 0); if (read_info.cache.wrote_create_file) { int errcode= query_error_code(thd, killed_status == NOT_KILLED); error= write_execute_load_query_log_event(thd, ex, table_list->db, table_list->table_name, is_concurrent, handle_duplicates, ignore, transactional_table, errcode); } /* Flushing the IO CACHE while writing the execute load query log event may result in error (for instance, because the max_binlog_size has been reached, and rotation of the binary log failed). */ error= error || mysql_bin_log.get_log_file()->error; } if (error) goto err; } #endif /*!EMBEDDED_LIBRARY*/ /* ok to client sent only after binlog write and engine commit */ my_ok(thd, info.copied + info.deleted, 0L, name); err: DBUG_ASSERT(transactional_table || !(info.copied || info.deleted) || thd->transaction.stmt.modified_non_trans_table); table->file->ha_release_auto_increment(); table->auto_increment_field_not_null= FALSE; thd->abort_on_warning= 0; DBUG_RETURN(error); } #ifndef EMBEDDED_LIBRARY /* Not a very useful function; just to avoid duplication of code */ static bool write_execute_load_query_log_event(THD *thd, sql_exchange* ex, const char* db_arg, /* table's database */ const char* table_name_arg, bool is_concurrent, enum enum_duplicates duplicates, bool ignore, bool transactional_table, int errcode) { char *load_data_query; my_off_t fname_start, fname_end; List fv; Item *item, *val; int n; const char *tdb= (thd->db != NULL ? thd->db : db_arg); const char *qualify_db= NULL; char command_buffer[1024]; String query_str(command_buffer, sizeof(command_buffer), system_charset_info); Load_log_event lle(thd, ex, tdb, table_name_arg, fv, is_concurrent, duplicates, ignore, transactional_table); /* force in a LOCAL if there was one in the original. */ if (thd->lex->local_file) lle.set_fname_outside_temp_buf(ex->file_name, strlen(ex->file_name)); query_str.length(0); if (!thd->db || strcmp(db_arg, thd->db)) { /* If used database differs from table's database, prefix table name with database name so that it becomes a FQ name. */ qualify_db= db_arg; } lle.print_query(thd, FALSE, (const char *) ex->cs?ex->cs->csname:NULL, &query_str, &fname_start, &fname_end, qualify_db); /* prepare fields-list and SET if needed; print_query won't do that for us. */ if (!thd->lex->field_list.is_empty()) { List_iterator li(thd->lex->field_list); query_str.append(" ("); n= 0; while ((item= li++)) { if (n++) query_str.append(", "); if (item->real_type() == Item::FIELD_ITEM) append_identifier(thd, &query_str, item->name, strlen(item->name)); else { /* Actually Item_user_var_as_out_param despite claiming STRING_ITEM. */ DBUG_ASSERT(item->type() == Item::STRING_ITEM); ((Item_user_var_as_out_param *)item)->print_for_load(thd, &query_str); } } query_str.append(")"); } if (!thd->lex->update_list.is_empty()) { List_iterator lu(thd->lex->update_list); List_iterator lv(thd->lex->value_list); query_str.append(STRING_WITH_LEN(" SET ")); n= 0; while ((item= lu++)) { val= lv++; if (n++) query_str.append(STRING_WITH_LEN(", ")); append_identifier(thd, &query_str, item->name, strlen(item->name)); query_str.append(val->name); } } if (!(load_data_query= (char *)thd->strmake(query_str.ptr(), query_str.length()))) return TRUE; Execute_load_query_log_event e(thd, load_data_query, query_str.length(), (uint) (fname_start - 1), (uint) fname_end, (duplicates == DUP_REPLACE) ? LOAD_DUP_REPLACE : (ignore ? LOAD_DUP_IGNORE : LOAD_DUP_ERROR), transactional_table, FALSE, FALSE, errcode); return mysql_bin_log.write(&e); } #endif /**************************************************************************** ** Read of rows of fixed size + optional garage + optonal newline ****************************************************************************/ static int read_fixed_length(THD *thd, COPY_INFO &info, TABLE_LIST *table_list, List &fields_vars, List &set_fields, List &set_values, READ_INFO &read_info, ulong skip_lines, bool ignore_check_option_errors) { List_iterator_fast it(fields_vars); Item_field *sql_field; TABLE *table= table_list->table; bool err, progress_reports, auto_increment_field_not_null=false; ulonglong counter, time_to_report_progress; DBUG_ENTER("read_fixed_length"); counter= 0; time_to_report_progress= MY_HOW_OFTEN_TO_WRITE/10; progress_reports= 1; if ((thd->progress.max_counter= read_info.file_length()) == ~(my_off_t) 0) progress_reports= 0; while ((sql_field= (Item_field*) it++)) { if (table->field[sql_field->field->field_index] == table->next_number_field) auto_increment_field_not_null= true; } while (!read_info.read_fixed_length()) { if (thd->killed) { thd->send_kill_message(); DBUG_RETURN(1); } if (progress_reports) { thd->progress.counter= read_info.position(); if (++counter >= time_to_report_progress) { time_to_report_progress+= MY_HOW_OFTEN_TO_WRITE/10; thd_progress_report(thd, thd->progress.counter, thd->progress.max_counter); } } if (skip_lines) { /* We could implement this with a simple seek if: - We are not using DATA INFILE LOCAL - escape character is "" - line starting prefix is "" */ skip_lines--; continue; } it.rewind(); uchar *pos=read_info.row_start; #ifdef HAVE_valgrind read_info.row_end[0]=0; #endif restore_record(table, s->default_values); /* There is no variables in fields_vars list in this format so this conversion is safe. */ while ((sql_field= (Item_field*) it++)) { Field *field= sql_field->field; table->auto_increment_field_not_null= auto_increment_field_not_null; /* No fields specified in fields_vars list can be null in this format. Mark field as not null, we should do this for each row because of restore_record... */ field->set_notnull(); if (pos == read_info.row_end) { thd->cuted_fields++; /* Not enough fields */ push_warning_printf(thd, Sql_condition::WARN_LEVEL_WARN, ER_WARN_TOO_FEW_RECORDS, ER_THD(thd, ER_WARN_TOO_FEW_RECORDS), thd->get_stmt_da()->current_row_for_warning()); /* Timestamp fields that are NOT NULL are autoupdated if there is no corresponding value in the data file. */ if (!field->maybe_null() && field->type() == FIELD_TYPE_TIMESTAMP) field->set_time(); } else { uint length; uchar save_chr; if ((length=(uint) (read_info.row_end-pos)) > field->field_length) length=field->field_length; save_chr=pos[length]; pos[length]='\0'; // Safeguard aganst malloc field->store((char*) pos,length,read_info.read_charset); pos[length]=save_chr; if ((pos+=length) > read_info.row_end) pos= read_info.row_end; /* Fills rest with space */ } /* Do not auto-update this field. */ field->set_has_explicit_value(); } if (pos != read_info.row_end) { thd->cuted_fields++; /* To long row */ push_warning_printf(thd, Sql_condition::WARN_LEVEL_WARN, ER_WARN_TOO_MANY_RECORDS, ER_THD(thd, ER_WARN_TOO_MANY_RECORDS), thd->get_stmt_da()->current_row_for_warning()); } if (thd->killed || fill_record_n_invoke_before_triggers(thd, table, set_fields, set_values, ignore_check_option_errors, TRG_EVENT_INSERT) || (table->default_field && table->update_default_fields())) DBUG_RETURN(1); switch (table_list->view_check_option(thd, ignore_check_option_errors)) { case VIEW_CHECK_SKIP: read_info.next_line(); goto continue_loop; case VIEW_CHECK_ERROR: DBUG_RETURN(-1); } err= write_record(thd, table, &info); table->auto_increment_field_not_null= FALSE; if (err) DBUG_RETURN(1); /* We don't need to reset auto-increment field since we are restoring its default value at the beginning of each loop iteration. */ if (read_info.next_line()) // Skip to next line break; if (read_info.line_cuted) { thd->cuted_fields++; /* To long row */ push_warning_printf(thd, Sql_condition::WARN_LEVEL_WARN, ER_WARN_TOO_MANY_RECORDS, ER_THD(thd, ER_WARN_TOO_MANY_RECORDS), thd->get_stmt_da()->current_row_for_warning()); } thd->get_stmt_da()->inc_current_row_for_warning(); continue_loop:; } DBUG_RETURN(MY_TEST(read_info.error)); } static int read_sep_field(THD *thd, COPY_INFO &info, TABLE_LIST *table_list, List &fields_vars, List &set_fields, List &set_values, READ_INFO &read_info, String &enclosed, ulong skip_lines, bool ignore_check_option_errors) { List_iterator_fast it(fields_vars); Item *item; TABLE *table= table_list->table; uint enclosed_length; bool err, progress_reports; ulonglong counter, time_to_report_progress; DBUG_ENTER("read_sep_field"); enclosed_length=enclosed.length(); counter= 0; time_to_report_progress= MY_HOW_OFTEN_TO_WRITE/10; progress_reports= 1; if ((thd->progress.max_counter= read_info.file_length()) == ~(my_off_t) 0) progress_reports= 0; for (;;it.rewind()) { if (thd->killed) { thd->send_kill_message(); DBUG_RETURN(1); } if (progress_reports) { thd->progress.counter= read_info.position(); if (++counter >= time_to_report_progress) { time_to_report_progress+= MY_HOW_OFTEN_TO_WRITE/10; thd_progress_report(thd, thd->progress.counter, thd->progress.max_counter); } } restore_record(table, s->default_values); while ((item= it++)) { uint length; uchar *pos; Item_field *real_item; if (read_info.read_field()) break; /* If this line is to be skipped we don't want to fill field or var */ if (skip_lines) continue; pos=read_info.row_start; length=(uint) (read_info.row_end-pos); real_item= item->field_for_view_update(); if ((!read_info.enclosed && (enclosed_length && length == 4 && !memcmp(pos, STRING_WITH_LEN("NULL")))) || (length == 1 && read_info.found_null)) { if (item->type() == Item::STRING_ITEM) { ((Item_user_var_as_out_param *)item)->set_null_value( read_info.read_charset); } else if (!real_item) { my_error(ER_NONUPDATEABLE_COLUMN, MYF(0), item->name); DBUG_RETURN(1); } else { Field *field= real_item->field; if (field->reset()) { my_error(ER_WARN_NULL_TO_NOTNULL, MYF(0), field->field_name, thd->get_stmt_da()->current_row_for_warning()); DBUG_RETURN(1); } field->set_null(); if (!field->maybe_null()) { /* Timestamp fields that are NOT NULL are autoupdated if there is no corresponding value in the data file. */ if (field->type() == MYSQL_TYPE_TIMESTAMP) field->set_time(); else if (field != table->next_number_field) field->set_warning(Sql_condition::WARN_LEVEL_WARN, ER_WARN_NULL_TO_NOTNULL, 1); } /* Do not auto-update this field. */ field->set_has_explicit_value(); } continue; } if (item->type() == Item::STRING_ITEM) { ((Item_user_var_as_out_param *)item)->set_value((char*) pos, length, read_info.read_charset); } else if (!real_item) { my_error(ER_NONUPDATEABLE_COLUMN, MYF(0), item->name); DBUG_RETURN(1); } else { Field *field= real_item->field; field->set_notnull(); read_info.row_end[0]=0; // Safe to change end marker if (field == table->next_number_field) table->auto_increment_field_not_null= TRUE; field->store((char*) pos, length, read_info.read_charset); field->set_has_explicit_value(); } } if (thd->is_error()) read_info.error= 1; if (read_info.error) break; if (skip_lines) { skip_lines--; continue; } if (item) { /* Have not read any field, thus input file is simply ended */ if (item == fields_vars.head()) break; for (; item ; item= it++) { Item_field *real_item= item->field_for_view_update(); if (item->type() == Item::STRING_ITEM) { ((Item_user_var_as_out_param *)item)->set_null_value( read_info.read_charset); } else if (!real_item) { my_error(ER_NONUPDATEABLE_COLUMN, MYF(0), item->name); DBUG_RETURN(1); } else { Field *field= real_item->field; if (field->reset()) { my_error(ER_WARN_NULL_TO_NOTNULL, MYF(0),field->field_name, thd->get_stmt_da()->current_row_for_warning()); DBUG_RETURN(1); } if (!field->maybe_null() && field->type() == FIELD_TYPE_TIMESTAMP) field->set_time(); field->set_has_explicit_value(); /* TODO: We probably should not throw warning for each field. But how about intention to always have the same number of warnings in THD::cuted_fields (and get rid of cuted_fields in the end ?) */ thd->cuted_fields++; push_warning_printf(thd, Sql_condition::WARN_LEVEL_WARN, ER_WARN_TOO_FEW_RECORDS, ER_THD(thd, ER_WARN_TOO_FEW_RECORDS), thd->get_stmt_da()->current_row_for_warning()); } } } if (thd->killed || fill_record_n_invoke_before_triggers(thd, table, set_fields, set_values, ignore_check_option_errors, TRG_EVENT_INSERT) || (table->default_field && table->update_default_fields())) DBUG_RETURN(1); switch (table_list->view_check_option(thd, ignore_check_option_errors)) { case VIEW_CHECK_SKIP: read_info.next_line(); goto continue_loop; case VIEW_CHECK_ERROR: DBUG_RETURN(-1); } err= write_record(thd, table, &info); table->auto_increment_field_not_null= FALSE; if (err) DBUG_RETURN(1); /* We don't need to reset auto-increment field since we are restoring its default value at the beginning of each loop iteration. */ if (read_info.next_line()) // Skip to next line break; if (read_info.line_cuted) { thd->cuted_fields++; /* To long row */ push_warning_printf(thd, Sql_condition::WARN_LEVEL_WARN, ER_WARN_TOO_MANY_RECORDS, ER_THD(thd, ER_WARN_TOO_MANY_RECORDS), thd->get_stmt_da()->current_row_for_warning()); if (thd->killed) DBUG_RETURN(1); } thd->get_stmt_da()->inc_current_row_for_warning(); continue_loop:; } DBUG_RETURN(MY_TEST(read_info.error)); } /**************************************************************************** ** Read rows in xml format ****************************************************************************/ static int read_xml_field(THD *thd, COPY_INFO &info, TABLE_LIST *table_list, List &fields_vars, List &set_fields, List &set_values, READ_INFO &read_info, String &row_tag, ulong skip_lines, bool ignore_check_option_errors) { List_iterator_fast it(fields_vars); Item *item; TABLE *table= table_list->table; bool no_trans_update_stmt; CHARSET_INFO *cs= read_info.read_charset; DBUG_ENTER("read_xml_field"); no_trans_update_stmt= !table->file->has_transactions(); for ( ; ; it.rewind()) { if (thd->killed) { thd->send_kill_message(); DBUG_RETURN(1); } // read row tag and save values into tag list if (read_info.read_xml(thd)) break; List_iterator_fast xmlit(read_info.taglist); xmlit.rewind(); XML_TAG *tag= NULL; #ifndef DBUG_OFF DBUG_PRINT("read_xml_field", ("skip_lines=%d", (int) skip_lines)); while ((tag= xmlit++)) { DBUG_PRINT("read_xml_field", ("got tag:%i '%s' '%s'", tag->level, tag->field.c_ptr(), tag->value.c_ptr())); } #endif restore_record(table, s->default_values); while ((item= it++)) { /* If this line is to be skipped we don't want to fill field or var */ if (skip_lines) continue; /* find field in tag list */ xmlit.rewind(); tag= xmlit++; while(tag && strcmp(tag->field.c_ptr(), item->name) != 0) tag= xmlit++; Item_field *real_item= item->field_for_view_update(); if (!tag) // found null { if (item->type() == Item::STRING_ITEM) ((Item_user_var_as_out_param *) item)->set_null_value(cs); else if (!real_item) { my_error(ER_NONUPDATEABLE_COLUMN, MYF(0), item->name); DBUG_RETURN(1); } else { Field *field= real_item->field; field->reset(); field->set_null(); if (field == table->next_number_field) table->auto_increment_field_not_null= TRUE; if (!field->maybe_null()) { if (field->type() == FIELD_TYPE_TIMESTAMP) field->set_time(); else if (field != table->next_number_field) field->set_warning(Sql_condition::WARN_LEVEL_WARN, ER_WARN_NULL_TO_NOTNULL, 1); } /* Do not auto-update this field. */ field->set_has_explicit_value(); } continue; } if (item->type() == Item::STRING_ITEM) ((Item_user_var_as_out_param *) item)->set_value( (char *) tag->value.ptr(), tag->value.length(), cs); else if (!real_item) { my_error(ER_NONUPDATEABLE_COLUMN, MYF(0), item->name); DBUG_RETURN(1); } else { Field *field= ((Item_field *)item)->field; field->set_notnull(); if (field == table->next_number_field) table->auto_increment_field_not_null= TRUE; field->store((char *) tag->value.ptr(), tag->value.length(), cs); field->set_has_explicit_value(); } } if (read_info.error) break; if (skip_lines) { skip_lines--; continue; } if (item) { /* Have not read any field, thus input file is simply ended */ if (item == fields_vars.head()) break; for ( ; item; item= it++) { Item_field *real_item= item->field_for_view_update(); if (item->type() == Item::STRING_ITEM) ((Item_user_var_as_out_param *)item)->set_null_value(cs); else if (!real_item) { my_error(ER_NONUPDATEABLE_COLUMN, MYF(0), item->name); DBUG_RETURN(1); } else { /* QQ: We probably should not throw warning for each field. But how about intention to always have the same number of warnings in THD::cuted_fields (and get rid of cuted_fields in the end ?) */ thd->cuted_fields++; push_warning_printf(thd, Sql_condition::WARN_LEVEL_WARN, ER_WARN_TOO_FEW_RECORDS, ER_THD(thd, ER_WARN_TOO_FEW_RECORDS), thd->get_stmt_da()->current_row_for_warning()); } } } if (thd->killed || fill_record_n_invoke_before_triggers(thd, table, set_fields, set_values, ignore_check_option_errors, TRG_EVENT_INSERT) || (table->default_field && table->update_default_fields())) DBUG_RETURN(1); switch (table_list->view_check_option(thd, ignore_check_option_errors)) { case VIEW_CHECK_SKIP: read_info.next_line(); goto continue_loop; case VIEW_CHECK_ERROR: DBUG_RETURN(-1); } if (write_record(thd, table, &info)) DBUG_RETURN(1); /* We don't need to reset auto-increment field since we are restoring its default value at the beginning of each loop iteration. */ thd->transaction.stmt.modified_non_trans_table= no_trans_update_stmt; thd->get_stmt_da()->inc_current_row_for_warning(); continue_loop:; } DBUG_RETURN(MY_TEST(read_info.error) || thd->is_error()); } /* load xml end */ /* Unescape all escape characters, mark \N as null */ char READ_INFO::unescape(char chr) { /* keep this switch synchornous with the ESCAPE_CHARS macro */ switch(chr) { case 'n': return '\n'; case 't': return '\t'; case 'r': return '\r'; case 'b': return '\b'; case '0': return 0; // Ascii null case 'Z': return '\032'; // Win32 end of file case 'N': found_null=1; /* fall through */ default: return chr; } } /* Read a line using buffering If last line is empty (in line mode) then it isn't outputed */ READ_INFO::READ_INFO(File file_par, uint tot_length, CHARSET_INFO *cs, String &field_term, String &line_start, String &line_term, String &enclosed_par, int escape, bool get_it_from_net, bool is_fifo) :file(file_par), buffer(NULL), buff_length(tot_length), escape_char(escape), found_end_of_line(false), eof(false), error(false), line_cuted(false), found_null(false), read_charset(cs) { /* Field and line terminators must be interpreted as sequence of unsigned char. Otherwise, non-ascii terminators will be negative on some platforms, and positive on others (depending on the implementation of char). */ field_term_ptr= static_cast(static_cast(field_term.ptr())); field_term_length= field_term.length(); line_term_ptr= static_cast(static_cast(line_term.ptr())); line_term_length= line_term.length(); level= 0; /* for load xml */ if (line_start.length() == 0) { line_start_ptr=0; start_of_line= 0; } else { line_start_ptr= line_start.ptr(); line_start_end=line_start_ptr+line_start.length(); start_of_line= 1; } /* If field_terminator == line_terminator, don't use line_terminator */ if (field_term_length == line_term_length && !memcmp(field_term_ptr,line_term_ptr,field_term_length)) { line_term_length=0; line_term_ptr= NULL; } enclosed_char= (enclosed_length=enclosed_par.length()) ? (uchar) enclosed_par[0] : INT_MAX; field_term_char= field_term_length ? field_term_ptr[0] : INT_MAX; line_term_char= line_term_length ? line_term_ptr[0] : INT_MAX; /* Set of a stack for unget if long terminators */ uint length= MY_MAX(cs->mbmaxlen, MY_MAX(field_term_length, line_term_length)) + 1; set_if_bigger(length,line_start.length()); stack=stack_pos=(int*) sql_alloc(sizeof(int)*length); if (!(buffer=(uchar*) my_malloc(buff_length+1,MYF(MY_WME | MY_THREAD_SPECIFIC)))) error= 1; /* purecov: inspected */ else { end_of_buff=buffer+buff_length; if (init_io_cache(&cache,(get_it_from_net) ? -1 : file, 0, (get_it_from_net) ? READ_NET : (is_fifo ? READ_FIFO : READ_CACHE),0L,1, MYF(MY_WME | MY_THREAD_SPECIFIC))) { my_free(buffer); /* purecov: inspected */ buffer= NULL; error=1; } else { #ifndef EMBEDDED_LIBRARY if (get_it_from_net) cache.read_function = _my_b_net_read; if (mysql_bin_log.is_open()) { cache.real_read_function= cache.read_function; cache.read_function= log_loaded_block; } #endif } } } READ_INFO::~READ_INFO() { ::end_io_cache(&cache); my_free(buffer); List_iterator xmlit(taglist); XML_TAG *t; while ((t= xmlit++)) delete(t); } inline int READ_INFO::terminator(const uchar *ptr,uint length) { int chr=0; // Keep gcc happy uint i; for (i=1 ; i < length ; i++) { if ((chr=GET) != *(uchar*)++ptr) { break; } } if (i == length) return 1; PUSH(chr); while (i-- > 1) PUSH(*--ptr); return 0; } /** Read a field. The data in the loaded file was presumably escaped using - either select_export::send_data() OUTFILE - or mysql_real_escape_string() using the same character set with the one specified in the current "LOAD DATA INFILE ... CHARACTER SET ..." (or the default LOAD character set). Note, non-escaped multi-byte characters are scanned as a single entity. This is needed to correctly distinguish between: - 0x5C as an escape character versus - 0x5C as the second byte in a multi-byte sequence (big5, cp932, gbk, sjis) Parts of escaped multi-byte characters are scanned on different loop iterations. See the comment about 0x5C handling in select_export::send_data() in sql_class.cc. READ_INFO::read_field() does not check wellformedness. Raising wellformedness errors or warnings in READ_INFO::read_field() would be wrong, as the data after unescaping can go into a BLOB field, or into a TEXT/VARCHAR field of a different character set. The loop below only makes sure to revert escaping made by select_export::send_data() or mysql_real_escape_string(). Wellformedness is checked later, during Field::store(str,length,cs) time. Note, in some cases users can supply data which did not go through escaping properly. For example, utf8 "\" (backslash followed by LATIN SMALL LETTER A WITH DIAERESIS) is improperly escaped data that could not be generated by select_export::send_data() / mysql_real_escape_string(): - either there should be two backslashes: "\\" - or there should be no backslashes at all: "" "\" and " are scanned on two different loop iterations and store "" into the field. Note, adding useless escapes before multi-byte characters like in the example above is safe in case of utf8, but is not safe in case of character sets that have escape_with_backslash_is_dangerous==TRUE, such as big5, cp932, gbk, sjis. This can lead to mis-interpretation of the data. Suppose we have a big5 character "<5C>" followed by <30> (digit 0). If we add an extra escape before this sequence, then we'll get <5C><5C><30>. The first loop iteration will turn <5C> into . The second loop iteration will turn <5C><30> into <30>. So the program that generates a dump file for further use with LOAD DATA must make sure to use escapes properly. */ int READ_INFO::read_field() { int chr,found_enclosed_char; uchar *to,*new_buffer; found_null=0; if (found_end_of_line) return 1; // One have to call next_line /* Skip until we find 'line_start' */ if (start_of_line) { // Skip until line_start start_of_line=0; if (find_start_of_fields()) return 1; } if ((chr=GET) == my_b_EOF) { found_end_of_line=eof=1; return 1; } to=buffer; if (chr == enclosed_char) { found_enclosed_char=enclosed_char; *to++=(uchar) chr; // If error } else { found_enclosed_char= INT_MAX; PUSH(chr); } for (;;) { // Make sure we have enough space for the longest multi-byte character. while ( to + read_charset->mbmaxlen < end_of_buff) { chr = GET; if (chr == my_b_EOF) goto found_eof; if (chr == escape_char) { if ((chr=GET) == my_b_EOF) { *to++= (uchar) escape_char; goto found_eof; } /* When escape_char == enclosed_char, we treat it like we do for handling quotes in SQL parsing -- you can double-up the escape_char to include it literally, but it doesn't do escapes like \n. This allows: LOAD DATA ... ENCLOSED BY '"' ESCAPED BY '"' with data like: "fie""ld1", "field2" */ if (escape_char != enclosed_char || chr == escape_char) { *to++ = (uchar) unescape((char) chr); continue; } PUSH(chr); chr= escape_char; } #ifdef ALLOW_LINESEPARATOR_IN_STRINGS if (chr == line_term_char) #else if (chr == line_term_char && found_enclosed_char == INT_MAX) #endif { if (terminator(line_term_ptr,line_term_length)) { // Maybe unexpected linefeed enclosed=0; found_end_of_line=1; row_start=buffer; row_end= to; return 0; } } if (chr == found_enclosed_char) { if ((chr=GET) == found_enclosed_char) { // Remove dupplicated *to++ = (uchar) chr; continue; } // End of enclosed field if followed by field_term or line_term if (chr == my_b_EOF || (chr == line_term_char && terminator(line_term_ptr, line_term_length))) { /* Maybe unexpected linefeed */ enclosed=1; found_end_of_line=1; row_start=buffer+1; row_end= to; return 0; } if (chr == field_term_char && terminator(field_term_ptr,field_term_length)) { enclosed=1; row_start=buffer+1; row_end= to; return 0; } /* The string didn't terminate yet. Store back next character for the loop */ PUSH(chr); /* copy the found term character to 'to' */ chr= found_enclosed_char; } else if (chr == field_term_char && found_enclosed_char == INT_MAX) { if (terminator(field_term_ptr,field_term_length)) { enclosed=0; row_start=buffer; row_end= to; return 0; } } #ifdef USE_MB #endif *to++ = (uchar) chr; #if MYSQL_VERSION_ID >= 100200 #error This 10.0 and 10.1 specific fix should be removed in 10.2 #else if (my_mbcharlen(read_charset, (uchar) chr) > 1) { /* A known MBHEAD found. Try to scan the full multi-byte character. Otherwise, a possible following second byte 0x5C would be mis-interpreted as an escape on the next iteration. (Important for big5, gbk, sjis, cp932). */ String tmp((char *) to - 1, read_charset->mbmaxlen, read_charset); tmp.length(1); bool eof= read_mbtail(&tmp); to+= tmp.length() - 1; if (eof) goto found_eof; } #endif } /* ** We come here if buffer is too small. Enlarge it and continue */ if (!(new_buffer=(uchar*) my_realloc((char*) buffer,buff_length+1+IO_SIZE, MYF(MY_WME | MY_THREAD_SPECIFIC)))) return (error=1); to=new_buffer + (to-buffer); buffer=new_buffer; buff_length+=IO_SIZE; end_of_buff=buffer+buff_length; } found_eof: enclosed=0; found_end_of_line=eof=1; row_start=buffer; row_end=to; return 0; } /* Read a row with fixed length. NOTES The row may not be fixed size on disk if there are escape characters in the file. IMPLEMENTATION NOTE One can't use fixed length with multi-byte charset ** RETURN 0 ok 1 error */ int READ_INFO::read_fixed_length() { int chr; uchar *to; if (found_end_of_line) return 1; // One have to call next_line if (start_of_line) { // Skip until line_start start_of_line=0; if (find_start_of_fields()) return 1; } to=row_start=buffer; while (to < end_of_buff) { if ((chr=GET) == my_b_EOF) goto found_eof; if (chr == escape_char) { if ((chr=GET) == my_b_EOF) { *to++= (uchar) escape_char; goto found_eof; } *to++ =(uchar) unescape((char) chr); continue; } if (chr == line_term_char) { if (terminator(line_term_ptr,line_term_length)) { // Maybe unexpected linefeed found_end_of_line=1; row_end= to; return 0; } } *to++ = (uchar) chr; } row_end=to; // Found full line return 0; found_eof: found_end_of_line=eof=1; row_start=buffer; row_end=to; return to == buffer ? 1 : 0; } int READ_INFO::next_line() { line_cuted=0; start_of_line= line_start_ptr != 0; if (found_end_of_line || eof) { found_end_of_line=0; return eof; } found_end_of_line=0; if (!line_term_length) return 0; // No lines for (;;) { int chr = GET; #ifdef USE_MB if (my_mbcharlen(read_charset, chr) > 1) { for (uint i=1; chr != my_b_EOF && i xmlit(taglist); xmlit.rewind(); XML_TAG *tag; while ((tag= xmlit++)) { if(tag->level >= level_arg) { xmlit.remove(); delete tag; } } DBUG_RETURN(0); } /* Convert an XML entity to Unicode value. Return -1 on error; */ static int my_xml_entity_to_char(const char *name, uint length) { if (length == 2) { if (!memcmp(name, "gt", length)) return '>'; if (!memcmp(name, "lt", length)) return '<'; } else if (length == 3) { if (!memcmp(name, "amp", length)) return '&'; } else if (length == 4) { if (!memcmp(name, "quot", length)) return '"'; if (!memcmp(name, "apos", length)) return '\''; } return -1; } /** @brief Convert newline, linefeed, tab to space @param chr character @details According to the "XML 1.0" standard, only space (#x20) characters, carriage returns, line feeds or tabs are considered as spaces. Convert all of them to space (#x20) for parsing simplicity. */ static int my_tospace(int chr) { return (chr == '\t' || chr == '\r' || chr == '\n') ? ' ' : chr; } /* Read an xml value: handle multibyte and xml escape */ int READ_INFO::read_value(int delim, String *val) { int chr; String tmp; for (chr= GET; my_tospace(chr) != delim && chr != my_b_EOF;) { #ifdef USE_MB if (my_mbcharlen(read_charset, chr) > 1) { DBUG_PRINT("read_xml",("multi byte")); int i, ml= my_mbcharlen(read_charset, chr); for (i= 1; i < ml; i++) { val->append(chr); /* Don't use my_tospace() in the middle of a multi-byte character TODO: check that the multi-byte sequence is valid. */ chr= GET; if (chr == my_b_EOF) return chr; } } #endif if(chr == '&') { tmp.length(0); for (chr= my_tospace(GET) ; chr != ';' ; chr= my_tospace(GET)) { if (chr == my_b_EOF) return chr; tmp.append(chr); } if ((chr= my_xml_entity_to_char(tmp.ptr(), tmp.length())) >= 0) val->append(chr); else { val->append('&'); val->append(tmp); val->append(';'); } } else val->append(chr); chr= GET; } return my_tospace(chr); } /* Read a record in xml format tags and attributes are stored in taglist when tag set in ROWS IDENTIFIED BY is closed, we are ready and return */ int READ_INFO::read_xml(THD *thd) { DBUG_ENTER("READ_INFO::read_xml"); int chr, chr2, chr3; int delim= 0; String tag, attribute, value; bool in_tag= false; tag.length(0); attribute.length(0); value.length(0); for (chr= my_tospace(GET); chr != my_b_EOF ; ) { switch(chr){ case '<': /* read tag */ /* TODO: check if this is a comment */ chr= my_tospace(GET); if(chr == '!') { chr2= GET; chr3= GET; if(chr2 == '-' && chr3 == '-') { chr2= 0; chr3= 0; chr= my_tospace(GET); while(chr != '>' || chr2 != '-' || chr3 != '-') { if(chr == '-') { chr3= chr2; chr2= chr; } else if (chr2 == '-') { chr2= 0; chr3= 0; } chr= my_tospace(GET); if (chr == my_b_EOF) goto found_eof; } break; } } tag.length(0); while(chr != '>' && chr != ' ' && chr != '/' && chr != my_b_EOF) { if(chr != delim) /* fix for the '' - stored in line_term if((tag.length() == line_term_length -2) && (memcmp(tag.ptr(), line_term_ptr + 1, tag.length()) == 0)) { DBUG_PRINT("read_xml", ("start-of-row: %i %s %s", level,tag.c_ptr_safe(), line_term_ptr)); } if(chr == ' ' || chr == '>') { level++; clear_level(level + 1); } if (chr == ' ') in_tag= true; else in_tag= false; break; case ' ': /* read attribute */ while(chr == ' ') /* skip blanks */ chr= my_tospace(GET); if(!in_tag) break; while(chr != '=' && chr != '/' && chr != '>' && chr != my_b_EOF) { attribute.append(chr); chr= my_tospace(GET); } break; case '>': /* end tag - read tag value */ in_tag= false; chr= read_value('<', &value); if(chr == my_b_EOF) goto found_eof; /* save value to list */ if (tag.length() > 0 && value.length() > 0) { DBUG_PRINT("read_xml", ("lev:%i tag:%s val:%s", level,tag.c_ptr_safe(), value.c_ptr_safe())); XML_TAG *tmp= new XML_TAG(level, tag, value); if (!tmp || taglist.push_front(tmp, thd->mem_root)) DBUG_RETURN(1); // End of memory } tag.length(0); value.length(0); attribute.length(0); break; case '/': /* close tag */ chr= my_tospace(GET); /* Decrease the 'level' only when (i) It's not an */ /* (without space) empty tag i.e. or, (ii) */ /* It is of format */ if(chr != '>' || in_tag) { level--; in_tag= false; } if(chr != '>') /* if this is an empty tag */ tag.length(0); /* we should keep tag value */ while(chr != '>' && chr != my_b_EOF) { tag.append(chr); chr= my_tospace(GET); } if((tag.length() == line_term_length -2) && (memcmp(tag.ptr(), line_term_ptr + 1, tag.length()) == 0)) { DBUG_PRINT("read_xml", ("found end-of-row %i %s", level, tag.c_ptr_safe())); DBUG_RETURN(0); //normal return } chr= my_tospace(GET); break; case '=': /* attribute name end - read the value */ //check for tag field and attribute name if(!memcmp(tag.c_ptr_safe(), STRING_WITH_LEN("field")) && !memcmp(attribute.c_ptr_safe(), STRING_WITH_LEN("name"))) { /* this is format xx where actual fieldname is in attribute */ delim= my_tospace(GET); tag.length(0); attribute.length(0); chr= '<'; /* we pretend that it is a tag */ level--; break; } //check for " or ' chr= GET; if (chr == my_b_EOF) goto found_eof; if(chr == '"' || chr == '\'') { delim= chr; } else { delim= ' '; /* no delimiter, use space */ PUSH(chr); } chr= read_value(delim, &value); if (attribute.length() > 0 && value.length() > 0) { DBUG_PRINT("read_xml", ("lev:%i att:%s val:%s\n", level + 1, attribute.c_ptr_safe(), value.c_ptr_safe())); XML_TAG *tmp= new XML_TAG(level + 1, attribute, value); if (!tmp || taglist.push_front(tmp, thd->mem_root)) DBUG_RETURN(1); // End of memory } attribute.length(0); value.length(0); if (chr != ' ') chr= my_tospace(GET); break; default: chr= my_tospace(GET); } /* end switch */ } /* end while */ found_eof: DBUG_PRINT("read_xml",("Found eof")); eof= 1; DBUG_RETURN(1); }