Merge branch '10.0' into merge-wip

2 files changed, 146 insertions, 9 deletions

diff --git a/storage/perfschema/gen_pfs_lex_token.cc b/storage/perfschema/gen_pfs_lex_token.cc
index 7581255b284..926982dd25d 100644
--- a/storage/perfschema/gen_pfs_lex_token.cc
+++ b/storage/perfschema/gen_pfs_lex_token.cc
@@ -1,5 +1,5 @@
 /*
-   Copyright (c) 2011, 2012, Oracle and/or its affiliates. All rights reserved.
+   Copyright (c) 2011, 2014, Oracle and/or its affiliates. All rights reserved.
 
    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
@@ -36,10 +36,13 @@
   See also YYMAXUTOK.
 */
 #define MY_MAX_TOKEN 1000
+/** Generated token. */
 struct gen_lex_token_string
 {
   const char *m_token_string;
   int m_token_length;
+  bool m_append_space;
+  bool m_start_expr;
 };
 
 gen_lex_token_string compiled_token_array[MY_MAX_TOKEN];
@@ -76,6 +79,13 @@ void set_token(int tok, const char *str)
 
   compiled_token_array[tok].m_token_string= str;
   compiled_token_array[tok].m_token_length= strlen(str);
+  compiled_token_array[tok].m_append_space= true;
+  compiled_token_array[tok].m_start_expr= false;
+}
+
+void set_start_expr_token(int tok)
+{
+  compiled_token_array[tok].m_start_expr= true;
 }
 
 void compute_tokens()
@@ -91,6 +101,8 @@ void compute_tokens()
   {
     compiled_token_array[tok].m_token_string= "(unknown)";
     compiled_token_array[tok].m_token_length= 9;
+    compiled_token_array[tok].m_append_space= true;
+    compiled_token_array[tok].m_start_expr= false;
   }
 
   /*
@@ -102,6 +114,7 @@ void compute_tokens()
     str[0]= (char) tok;
     compiled_token_array[tok].m_token_string= str;
     compiled_token_array[tok].m_token_length= 1;
+    compiled_token_array[tok].m_append_space= true;
   }
 
   max_token_seen= 255;
@@ -202,6 +215,71 @@ void compute_tokens()
   max_token_seen++;
   tok_pfs_unused= max_token_seen;
   set_token(tok_pfs_unused, "UNUSED");
+
+  /*
+    Fix whitespace for some special tokens.
+  */
+
+  /*
+    The lexer parses "@@variable" as '@', '@', 'variable',
+    returning a token for '@' alone.
+
+    This is incorrect, '@' is not really a token,
+    because the syntax "@ @ variable" (with spaces) is not accepted:
+    The lexer keeps some internal state after the '@' fake token.
+
+    To work around this, digest text are printed as "@@variable".
+  */
+  compiled_token_array[(int) '@'].m_append_space= false;
+
+  /*
+    Define additional properties for tokens.
+
+    List all the token that are followed by an expression.
+    This is needed to differentiate unary from binary
+    '+' and '-' operators, because we want to:
+    - reduce <unary +> <NUM> to <?>,
+    - preserve <...> <binary +> <NUM> as is.
+  */
+  set_start_expr_token('(');
+  set_start_expr_token(',');
+  set_start_expr_token(EVERY_SYM);
+  set_start_expr_token(AT_SYM);
+  set_start_expr_token(STARTS_SYM);
+  set_start_expr_token(ENDS_SYM);
+  set_start_expr_token(DEFAULT);
+  set_start_expr_token(RETURN_SYM);
+  set_start_expr_token(IF_SYM);
+  set_start_expr_token(ELSEIF_SYM);
+  set_start_expr_token(CASE_SYM);
+  set_start_expr_token(WHEN_SYM);
+  set_start_expr_token(WHILE_SYM);
+  set_start_expr_token(UNTIL_SYM);
+  set_start_expr_token(SELECT_SYM);
+
+  set_start_expr_token(OR_SYM);
+  set_start_expr_token(OR2_SYM);
+  set_start_expr_token(XOR);
+  set_start_expr_token(AND_SYM);
+  set_start_expr_token(AND_AND_SYM);
+  set_start_expr_token(NOT_SYM);
+  set_start_expr_token(BETWEEN_SYM);
+  set_start_expr_token(LIKE);
+  set_start_expr_token(REGEXP);
+
+  set_start_expr_token('|');
+  set_start_expr_token('&');
+  set_start_expr_token(SHIFT_LEFT);
+  set_start_expr_token(SHIFT_RIGHT);
+  set_start_expr_token('+');
+  set_start_expr_token('-');
+  set_start_expr_token(INTERVAL_SYM);
+  set_start_expr_token('*');
+  set_start_expr_token('/');
+  set_start_expr_token('%');
+  set_start_expr_token(DIV_SYM);
+  set_start_expr_token(MOD_SYM);
+  set_start_expr_token('^');
 }
 
 void print_tokens()
@@ -214,20 +292,26 @@ void print_tokens()
 
   for (tok= 0; tok<256; tok++)
   {
-    printf("/* %03d */  { \"\\x%02x\", 1},\n", tok, tok);
+    printf("/* %03d */  { \"\\x%02x\", 1, %s, %s},\n",
+           tok,
+           tok,
+           compiled_token_array[tok].m_append_space ? "true" : "false",
+           compiled_token_array[tok].m_start_expr ? "true" : "false");
   }
 
   printf("/* PART 2: named tokens. */\n");
 
   for (tok= 256; tok<= max_token_seen; tok++)
   {
-    printf("/* %03d */  { \"%s\", %d},\n",
+    printf("/* %03d */  { \"%s\", %d, %s, %s},\n",
            tok,
            compiled_token_array[tok].m_token_string,
-           compiled_token_array[tok].m_token_length);
+           compiled_token_array[tok].m_token_length,
+           compiled_token_array[tok].m_append_space ? "true" : "false",
+           compiled_token_array[tok].m_start_expr ? "true" : "false");
   }
 
-  printf("/* DUMMY */ { \"\", 0}\n");
+  printf("/* DUMMY */ { \"\", 0, false, false}\n");
   printf("};\n");
 
   printf("/* PFS specific tokens. */\n");
@@ -254,6 +338,8 @@ int main(int argc,char **argv)
   printf("{\n");
   printf("  const char *m_token_string;\n");
   printf("  int m_token_length;\n");
+  printf("  bool m_append_space;\n");
+  printf("  bool m_start_expr;\n");
   printf("};\n");
   printf("typedef struct lex_token_string lex_token_string;\n");
 
diff --git a/storage/perfschema/pfs_digest.cc b/storage/perfschema/pfs_digest.cc
index 1fbf72c19c8..31fffcb570b 100644
--- a/storage/perfschema/pfs_digest.cc
+++ b/storage/perfschema/pfs_digest.cc
@@ -603,16 +603,67 @@ PSI_digest_locker* pfs_digest_add_token_v1(PSI_digest_locker *locker,
 
   switch (token)
   {
-    case BIN_NUM:
+    case NUM:
+    case LONG_NUM:
+    case ULONGLONG_NUM:
     case DECIMAL_NUM:
     case FLOAT_NUM:
+    case BIN_NUM:
     case HEX_NUM:
+    {
+      bool found_unary;
+      do
+      {
+        found_unary= false;
+        peek_last_two_tokens(digest_storage, state->m_last_id_index,
+                             &last_token, &last_token2);
+
+        if ((last_token == '-') || (last_token == '+'))
+        {
+          /*
+            We need to differentiate:
+            - a <unary minus> operator
+            - a <unary plus> operator
+            from
+            - a <binary minus> operator
+            - a <binary plus> operator
+            to only reduce "a = -1" to "a = ?", and not change "b - 1" to "b ?"
+
+            Binary operators are found inside an expression,
+            while unary operators are found at the beginning of an expression, or after operators.
+
+            To achieve this, every token that is followed by an <expr> expression
+            in the SQL grammar is flagged.
+            See sql/sql_yacc.yy
+            See sql/gen_lex_token.cc
+
+            For example,
+            "(-1)" is parsed as "(", "-", NUM, ")", and lex_token_array["("].m_start_expr is true,
+            so reduction of the "-" NUM is done, the result is "(?)".
+            "(a-1)" is parsed as "(", ID, "-", NUM, ")", and lex_token_array[ID].m_start_expr is false,
+            so the operator is binary, no reduction is done, and the result is "(a-?)".
+          */
+          if (lex_token_array[last_token2].m_start_expr)
+          {
+            /*
+              REDUCE:
+              TOK_PFS_GENERIC_VALUE := (UNARY_PLUS | UNARY_MINUS) (NUM | LOG_NUM | ... | FLOAT_NUM)
+
+              REDUCE:
+              TOK_PFS_GENERIC_VALUE := (UNARY_PLUS | UNARY_MINUS) TOK_PFS_GENERIC_VALUE
+            */
+            token= TOK_PFS_GENERIC_VALUE;
+            digest_storage->m_byte_count-= PFS_SIZE_OF_A_TOKEN;
+            found_unary= true;
+          }
+        }
+      } while (found_unary);
+    }
+    /* fall through, for case NULL_SYM below */
     case LEX_HOSTNAME:
-    case LONG_NUM:
-    case NUM:
     case TEXT_STRING:
     case NCHAR_STRING:
-    case ULONGLONG_NUM:
+    case PARAM_MARKER:
     {
       /*
         REDUCE: