如何使用Bison / Yacc和递归规则构建数组

Question

With Bison, I figured out how to get everything into one long string as follows:

arg_list:
    WORD arg_list { strcat( $1, "IFS" ); $$ = strcat($1, $2); }  |
    WORD
    ;

and:

WORD arg_list { printf("%s, %s\n", $1, $2); }

But the problem is that I will then have to split up $2 in the second rule again to parse it. Is there a way to populate an array instead of just using concatenation? Am I going about this the wrong way?

If I need to build something like a linked list that could make sense, just not sure what would be the proper way to bind to arg_list, and then clean up the memory.

Answer 1

If you have an array type with a push_front operation, this is trivially:

arg_list:
    WORD arg_list { $$ = $2.push_front($1); }
    WORD { $$ = new Array<string>($1); }

without that, it requires more work. You can use a vector and add the strings on the end (which will be in the reversed order). Or you can use a linked list (which is easier if you're using straight C):

arg_list:
    WORD arg_list { $$ = malloc(sizeof(struct list_elem));
                    $$->next = $2;
                    $$->val = $1; }
    WORD          { $$ = malloc(sizeof(struct list_elem));
                    $$->next = 0;
                    $$->val = $1; }

Answer 2

%union {
  char *char_ptr;
}
%token STRING
%type <char_ptr> STRING string
%%
...
string:
    STRING        /* Lexic analyzer return STRING and set yylval = yytext; */
  | string STRING
    { char *str = (char*) malloc(strlen($1) + strlen($2) + 1);
      strcpy(str, $1);
      strcat(str, $2);
      free($2);
      free($1);
      $$ = str;
    }
  ;
%%