Re: push parser

[Bob Rossi]

On Fri, Sep 29, 2006 at 12:01:01PM -0700, Paul Eggert wrote:
> Bob Rossi <address@hidden> writes:
> 
> >> What should I do with yynerrs?
> 
> I'd leave it out.  It's not documented (and not that useful), so no
> harm done.

OK.
> > One issue I had is, now the yyparse function (which calls yypushparse) 
> > depends on yylex to take 2 parameters. This breaks the small test I have
> > that is not in the test suite because my yylex takes nothing. Should we
> > force the user to have yylex take these parameters? or should we somehow
> > modify yyparse ()?
> 
> How about if we call the push-parser lexer "yypushlex" rather than
> "yylex"?  That would be consistent with renaming yyparse to
> yypushparse.

OK, that still doesn't completly help me understand what would be 
appropriate. In terms of the testsuite, we use yylex (...) with
2 parameters. This works, but either way I can change it to work if 
things change. Outside of the testsuite is a different story.

If I have yypushparse call yylex or yypushlex, should we define those in
the generated parser since bison will require them in order to compile?
or will bison force the user to define them? 

If we force the user to define that function, it add's an unnecessary 
complication to the push parser. That is, the user is mostly likely 
going to call yypushparse, and not care about yyparse at all. However,
yyparse is the function causing the user to define the yypushlex
function. What do you think I should do to resolve this?

Thanks,
Bob Rossi