Adaptive comment

src/sql/parser/ob_parser.cpp

@@ -250,6 +250,14 @@ int ObParser::parse(
       ObCharset::is_valid_collation(connection_collation_)
           ? (ObCharset::charset_type_by_coll(connection_collation_) != CHARSET_UTF8MB4)
           : false;
+  parse_result.malloc_pool_ = allocator_;
+  parse_result.sql_mode_ = sql_mode_;
+  parse_result.need_parameterize_ = (FP_MODE == parse_mode || FP_PARAMERIZE_AND_FILTER_HINT_MODE == parse_mode);
+  parse_result.minus_ctx_.pos_ = -1;
+  parse_result.minus_ctx_.raw_sql_offset_ = -1;
+  parse_result.charset_info_ = ObCharset::get_charset(connection_collation_);
+  parse_result.connection_collation_ = connection_collation_;
+  parse_result.mysql_compatible_comment_ = false;
 
   if (parse_result.is_fp_ || parse_result.is_dynamic_sql_) {
     int64_t new_length = parse_result.is_fp_ ? stmt.length() + 1 : stmt.length() * 2;

src/sql/parser/parse_node.h

@@ -249,6 +249,8 @@ typedef struct {
   const struct ObCharsetInfo* charset_info_;
   int last_well_formed_len_;
   bool may_bool_value_;  // used for true/false in sql parser
+  int connection_collation_;       // connection collation
+  bool mysql_compatible_comment_;  // whether the parser is parsing "/*! xxxx */"
 
 #ifdef SQL_PARSER_COMPILATION
   TokenPosInfo* comment_list_;

src/sql/parser/sql_parser_mysql_mode.l

@@ -10,6 +10,9 @@
  * See the Mulan PubL v2 for more details.
  */
 
+/* unput() change the yyin but it doesn't change ParserResult->input_sql_.
+// use unput() function may have unexpected result while copy string.
+*/
 %option noyywrap nounput noinput nodefault case-insensitive
 %option noyyalloc noyyrealloc noyyfree
 %option reentrant bison-bridge bison-locations
@@ -84,6 +87,10 @@ btend         {backtick}
 btdouble      {backtick}{backtick}
 btcontent     [^`]+
 
+mysql_compatible_comment_with_version \/\*\![0-9]{5}
+mysql_compatible_comment_without_version \/\*\!
+mysql_compatible_comment_end \*\/
+
 rowidPattern  (WITH{whitespace}ROWID)
 
 %%
@@ -1139,6 +1146,14 @@ Timestamp{whitespace}?\"[^\"]*\" {
   }
 }
 
+{mysql_compatible_comment_without_version} {
+  // if is a mysql comment without version. For example, /*!any sql str*/
+  // mysql_comment without version, processed as common sql str;
+  // place before `c_cmt_begin` to avoid (the '/*!') being hidden by '/*')
+  ParseResult *p = (ParseResult *)yyextra;
+  p->mysql_compatible_comment_ = true;
+}
+
 {c_cmt_begin} {
 BEGIN(in_c_comment);
 #ifdef SQL_PARSER_COMPILATION
@@ -1171,6 +1186,25 @@ BEGIN(in_c_comment);
 #endif
 }
 
+{mysql_compatible_comment_end} {
+  //for mysql compatible comment:
+  // only "*/" should be matched, duplicated '*' (e.g., "***/") will report a error.
+  ParseResult *p = (ParseResult *)yyextra;
+  if (p->mysql_compatible_comment_){
+    p->mysql_compatible_comment_ = false;
+    BEGIN(INITIAL);
+  } else {
+    // The sql could be "select */*!xxx*/ from t1;". We can't directly raise a syntax
+    // error here. We should treat the "*/" as '*' and '/' by return '*' and unput '/';
+
+    // yyless will change the yytext and yyleng.
+    char c_ret = yytext[0];
+    yyless(1);
+    p->yycolumn_ = p->yycolumn_ - 1;
+    return c_ret;
+  }
+}
+
 <in_c_comment><<EOF>>  {
   yyerror(yylloc, yyextra, "unterminated log_level string\n");
   return PARSER_SYNTAX_ERROR;
@@ -1401,6 +1435,15 @@ BEGIN(in_c_comment);
   }
 }
 
+{mysql_compatible_comment_with_version} {
+  // comment with version: /*!50600 any sql str*/
+  // comment without version: /*!any sql str*/
+  // we do not add a start_condition, since some sql string need to be processed in INITIAL state.
+  // instead of a new start_condition, we use a extra field (mysql_compatible_comment_) to mark the adaptive comment.
+  ParseResult *p = (ParseResult *)yyextra;
+  p->mysql_compatible_comment_ = true;
+}
+
 [\n] {
   yylineno ++;
   if (IS_FAST_PARAMETERIZE) {

src/sql/parser/type_name.c

@@ -320,7 +320,6 @@ const char* get_type_name(int type)
 	case T_FUN_SYS_TIME_TO_SEC : return "T_FUN_SYS_TIME_TO_SEC";
 	case T_FUN_SYS_SEC_TO_TIME : return "T_FUN_SYS_SEC_TO_TIME";
 	case T_FUN_SYS_INTERVAL : return "T_FUN_SYS_INTERVAL";
-	case T_FUN_SYS_CONVERT_TZ : return "T_FUN_SYS_CONVERT_TZ";
 	case T_FUN_UDF : return "T_FUN_UDF";
 	case T_FUN_SYS_SUBTIME : return "T_FUN_SYS_SUBTIME";
 	case T_FUN_SYS_SQRT : return "T_FUN_SYS_SQRT";
@@ -520,6 +519,7 @@ const char* get_type_name(int type)
 	case T_FUN_SYS_UTC_TIME : return "T_FUN_SYS_UTC_TIME";
 	case T_FUN_SYS_UTC_DATE : return "T_FUN_SYS_UTC_DATE";
 	case T_FUN_SYS_TIME_FORMAT : return "T_FUN_SYS_TIME_FORMAT";
+	case T_FUN_SYS_CONVERT_TZ : return "T_FUN_SYS_CONVERT_TZ";
 	case T_FUN_SYS_END : return "T_FUN_SYS_END";
 	case T_MAX_OP : return "T_MAX_OP";
 	case T_FUN_MATCH_AGAINST : return "T_FUN_MATCH_AGAINST";