Redundancy/conflicts in expression rules.

[Rainer Schuetze]

On 13.02.2014 01:00, Iain Buclaw wrote:
> This might be because the rules I have in place are not complete (I'm
> not implementing a complete D expression parser), but I have my doubts
> on this, and currently leaning on a possible problem in the documentation.
>
>
> I've built up rules in yacc based on what's documented here:
> http://dlang.org/expression.html
>
> However I seem to be getting shift/reduce conflicts around:
>
> AndAndExpression:
>          OrExpression
> |       AndAndExpression && OrExpression
>          CmpExpression
> |       AndAndExpression && CmpExpression
> ;
>
> OrExpression:
>          XorExpression
> |       OrExpression | XorExpression
> ;
>
> XorExpression:
>          AndExpression
> |       XorExpression ^ AndExpression
> ;
>
> AndExpression:
>          ShiftExpression
> |       AndExpression & ShiftExpression
> ;
>
> CmpExpression:
>          ShiftExpression
> |       EqualExpression
> |       IdentityExpression
> |       RelExpression
> ;
>
> CmpExpression:
>          ShiftExpression
> |       EqualExpression
> |       IdentityExpression
> |       RelExpression
> ;

I think this is by design to disallow comparison operators and binary 
operators in the same expression without paranthesis:

int x = a & b < c;

op.d(2): Error: b < c must be parenthesized when next to operator &

The grammar in the spec doesn't play nice with generators and isn't 
always correct, but in this case, I think it is.