Question

I am constructing a MySQL grammar validator with ANTLR. I started with the sql_yacc.yy from the MySQL source code, but I have some difficulties converting the following grammar. I tried many times, but it doesn’t work. Can anyone help me?

expr
  : expr or expr 
  | expr XOR expr
  | expr and expr
  | NOT_SYM expr 
  | bool_pri IS TRUE_SYM 
  | bool_pri IS not TRUE_SYM 
  | bool_pri IS FALSE_SYM
  | bool_pri IS not FALSE_SYM 
  | bool_pri IS UNKNOWN_SYM
  | bool_pri IS not UNKNOWN_SYM 
  | bool_pri
  ;

bool_pri
  : bool_pri IS NULL_SYM 
  | bool_pri IS not NULL_SYM 
  | bool_pri EQUAL_SYM predicate 
  | bool_pri comp_op predicate
  | bool_pri comp_op all_or_any '(' subselect ')' 
  | predicate
  ;

predicate
  : bit_expr IN_SYM '(' subselect ')'
  | bit_expr not IN_SYM '(' subselect ')'
  | bit_expr IN_SYM '(' expr ')'
  | bit_expr IN_SYM '(' expr ',' expr_list ')'
  | bit_expr not IN_SYM '(' expr ')'
  | bit_expr not IN_SYM '(' expr ',' expr_list ')'
  | bit_expr BETWEEN_SYM bit_expr AND_SYM predicate
  | bit_expr not BETWEEN_SYM bit_expr AND_SYM predicate
  | bit_expr SOUNDS_SYM LIKE bit_expr
  | bit_expr LIKE simple_expr opt_escape
  | bit_expr not LIKE simple_expr opt_escape
  | bit_expr REGEXP bit_expr
  | bit_expr not REGEXP bit_expr
  | bit_expr
  ;

bit_expr
  : bit_expr '|' bit_expr 
  | bit_expr '&' bit_expr
  | bit_expr SHIFT_LEFT bit_expr 
  | bit_expr SHIFT_RIGHT bit_expr 
  | bit_expr '+' bit_expr
  | bit_expr '-' bit_expr 
  | bit_expr '+' INTERVAL_SYM expr interval 
  | bit_expr '-' INTERVAL_SYM expr interval
  | bit_expr '*' bit_expr
  | bit_expr '/' bit_expr 
  | bit_expr '%' bit_expr 
  | bit_expr DIV_SYM bit_expr 
  | bit_expr MOD_SYM bit_expr 
  | bit_expr '^' bit_expr
  | simple_expr
  ;

Answer 1

ANTLR cannot cope with left-recursion, so there’s no trivial way to convert sql_yacc.yy into the ANTLR equivalent. You might want to have a look at the following resources from the ANTLR Wiki:

• Left-Recursion Removal
• SQL grammar for MySQL dialect (ANTLR v3)

Note that the MySQL grammar is incomplete, but might give you a starting point.