Re: push parser

[Bob Rossi]

On Sat, Dec 16, 2006 at 07:39:12AM -0500, Bob Rossi wrote:
> On Fri, Dec 15, 2006 at 09:32:29PM -0500, Joel E. Denny wrote:
> > On Fri, 15 Dec 2006, Joel E. Denny wrote:
> > 
> > > Bob and Paul, I can't seem to communicate this adequately in English, so 
> > > I 
> > > think the best way to explain my view is to write the patch that 
> > > implements it.  It should be very short, but it may be a few hours before 
> > > I can get some time.  Can we pick up the discussion again after that?
> > 
> > The patch is below, but I'll wait for both of your reviews before 
> > committing it.  It's very short.  Notice that for push parsers now:
> 
> Well, it looks fine to me. I have the same hesitations now that I had
> before you coded the patch. I think you did a reasonable job explaining
> your position previously. I want to make it clear, I don't have enough
> experience with bison to give this patch a proper review. I'm not sure
> of the long term effects of such a change.
> 
> Below was my main issue with the patch which I meant to respond
> publically,
> 
> > > How are we going to handle the case when the push parser does not want    
> > >                             
> > > to generate code that is in the pure_if check? That was my primary        
> > >                             
> > > concern from the beginning. Setting the pure option whenever push-parser  
> > >                             
> > > is set forces the rest of bison to loose the fact that it's generating a  
> > >                             
> > > push parser, and not a pure parser. (Now I understand that push parsers   
> > >                             
> > > are pure, but I'm talking about the fact that there are pure_if           
> > >                             
> > > statements that may not actually want to be true in the push mode).       
> > >                             
>                                                                               
>                           
> > b4_pure_if([                                                                
> >                             
> >    code for pure and push                                                   
> >                             
> >    b4_push_if([], [[code for pure but not push]])])                         
> >                             
> > 
> > So far, this hasn't happened.  If it does, I don't think it will be often.  
> >                             
> 
> So, if you are happy with this issue, and Paul is also, then so am I!
> I do want to mention that I enjoy seeing the problems that this patch
> solves. 
> 
> BTW, after knowing all of this, if I were to write the patch, I would do
> it this way. I would leave the pure-parser option set to false. But I
> would write a new macro, b4_pure_or_push (or something), and replace all
> calls to b4_pure that are supposed to be b4_pure_or_push. That would
> leave bison with the ability to perform all combinations of checks,
> which I don't think it can currently do now.
> 
> 
> Thanks,
> Bob Rossi