A new grammar in combination with %glr-parser worked.
%{
#include <stdio.h>
#include <stdlib.h>
extern int yylex();
extern int yyparse();
extern FILE *yyin;
void yyerror(const char *s);
%}
%union {
char cval;
int ival;
float fval;
char *sval;
}
/* define the "terminal symbol" token types */
...
%start syntax
%glr-parser
%%
syntax:
| syntax rule { printf("Success!\n"); }
rule: LEFT_ANGLEB rule_name RIGHT_ANGLEB { printf("found LHS rule
name\n"); }EXPANDS_TO expression ;
expression : list | list OR expression ;
list : term | term list ;
term : literal | LEFT_ANGLEB rule_name RIGHT_ANGLEB ;
rule_name: IDENTIFIER ;
/* actually, the original BNF did not use quotes */
literal : LITERAL_STRING | QUOTED_STRING ;
%%
This grammar has 1 shift/reduce conflict, as opposed to 6 in the
previous grammar, and %glr-parser works this time.
Question:
(1)I understand that if I have a file of the form:
<lhs1> ::= <rhs1> <lhs2> ::= <rhs2>, bison parses my rules by default
as (<lhs1> ::= <rhs1> <lhs2>) ::= <rhs2> and generates a syntax error
when it reaches the ::= before <rhs2>. In Bison terminology (I hope),
after recognising <rhs1>, I want Bison to reduce <lhs1> ::= <rhs1> to
<rule>, but the next token '<' does not give it a reason to do so. It would
have to skip '<', 'lhs2', '>' and finally arrive at '::=' to know that
(<lhs1> ::= <rhs1> <lhs2>) will not work. Therefore, as written, a 2-token
lookahead as suggested by Ron Burk would not help me. Am I
understanding correctly?
However, if I defined a token RULE_NAME as, for example
\< [A-Za-z_]+\>
then after matching the longest possible sequence of tokens in:
NAME ::= NAME ...<end-of-rule> NAME ::= <whatever> so that I am at
<end-of-rule>, where I want to reduce, then a 2-token lookahead *would*
help in this case. Is that right?
Regards,
Test User.
_______________________________________________
help-bison@gnu.org https://lists.gnu.org/mailman/listinfo/help-bison
