3

I'm using jison (a javascript equivalent of Bison) and I'm having the following precedence problem. I'll illustrate it using the calculator demo http://zaach.github.com/jison/try/

It works fine as is. The precedence is

%left '+' '-'
%left '*' '/'
%left '^'
%left UMINUS

and the grammar is

e
    : e '+' e
        {$$ = $1+$3;}
    | e '-' e
        {$$ = $1-$3;}
    | e '*' e
        {$$ = $1*$3;}
    | e '/' e
        {$$ = $1/$3;}
    | e '^' e
        {$$ = Math.pow($1, $3);}
    | '-' e %prec UMINUS
        {$$ = -$2;}

If I change the '*' line to be 

    | e '*' e %prec TIMESPREC
        {$$ = $1*$3;}

and change the precedence to 

%left '+' '-'
%left TIMESPREC '/'
%left '^'
%left UMINUS

it doesn't work any more. Isn't it supposed to work the same? This could be useful, e.g., if one wanted to eventually have an algebra syntax where 2 x + 3 is parsed as (2x)+3.

Thanks!

user1213898
  • 173
  • 7

1 Answers1

2

The reason for this is because %prec will only set the precedence of the rule, not on all the individual tokens. So, the individual precedence of the tokens on the RHS of the rule still matters.

So, setting %prec on your multiplication rule will not alter the precedence of the '*' symbol. When precedence tries to resolve conflicts it will compare the reduce actions priority with the priority of the token that is to be shifted. Since * has no priority, the grammar becomes ambiguous with your change.

Dervall
  • 5,736
  • 3
  • 25
  • 48
  • Thanks, that explains why! Now, what is the right solution if I want to allow things like 2 x + 3 instead of 2*x+3, where there is no *? Thanks – user1213898 Feb 18 '12 at 22:42
  • You'll basically have to rewrite the grammar to allow this. It's a nontrivial matter, you can't simply remove the `*` in the rule and set a precedence since that will cause conflicts. In essence, `%prec` won't solve your problems with this. By the way, there's no better way to say thanks than to upvote or accept answers :) – Dervall Feb 18 '12 at 22:53