Re: Some seeming context sensitivity, and shift/reduce errors

[Hans Aberg]

On 8 Dec 2007, at 20:20, Jack T Mudge III wrote:

[Please keep the cc to the Help-Bison list, as other may want to join  
in.]

> > As for resolving shift/reduce conflicts, one can try looking in
> > the .output parser description file that Bison can generate, at the
> > conflicting states, and their parsing position "."s. Then apply token
> > precedences (%left etc) to the tokens immediately and after those  
> > "."s.

> It's worth a try. The only problem I see with using either left or  
> right
> association is that this seems to change according to context, e.g.
>
> **  *some* *bold* text
>
> where the two **s at the beginning don't have association, the  
> third and 5th *
> would have right-association, and the 4th and 6th * would have
> left-association. That's where I run into a problem. Depending  
> where the *
> is, and whether a * has come before it, changes the meaning of the *.

One can push the context sensitivities into the actions. This way,  
one can even handle language like C++. An alternate approach is to  
use the GLR parser (see the manual).

> As I said, i'll try the comp.comilers mailing list. I think you're  
> right in
> that they may have a suggestion or two.

Look around for some similar languages, which helps avoiding  
pitfalls, and may give ideas on grammar design.

> I went ahead and purchased a copy of
> the Dragon Book's second edition (just hasn't shipped yet....),  
> when that
> arrives, I'll thumb through that before I try the mailing list. The  
> book may
> answer my question :).

Also look into the FAQ of comp.compilers, which has more suggestions  
- it is posted there monthly.

  Hans Aberg