Re: prologue alternatives

[Joel E. Denny]

On Sun, 17 Sep 2006, Paul Eggert wrote:

> "Joel E. Denny" <address@hidden> writes:
> 
> > Are you suggesting YYSTYPE shouldn't be a single union?  
> 
> Yes, if that's a better way to handle multiple uses of %union.

I'm not sure how that would go.

> > I see the Bison grammar file language as something of a meta language that 
> > can do better than C... and %type-code{...} seems to make that possible.
> 
> I guess that's where we disagree.  I think Bison's grammar language is
> inherently linked to the underlying implementation language (C or Java
> or whatnot), and that people can't make sensible use of one without
> understanding the other.

I don't mean to say that people can use one without understanding the 
other.  I just think Bison can sometimes help alleviate some of the 
limitations in C (or whatever) in order to improve the maintainability of 
parser specifications.

> It's possible to design things differently, but that would be a major
> redesign, and arguably it wouldn't be Yacc-like (or Bison-like :-) any
> more.

Of course, I haven't used Bison as long as you have, so I don't have the 
same sense of what's Bison-like.  However, I think we're discussing 
reasonably small steps towards a better Bison.

> > If I suddenly realized that I needed to add a header that contained 
> > YYSTYPE,
> 
> This is not a normal small change that one would make to a grammar.
> It's all right if a disruptive change like that causes, well,
> disruption.

I agree that it's a minor point.

> > if we're going for maximum flexibility in all areas, we could 
> > take your proposal and extend your $declare{YYSTYPE} to a general variable 
> > language.  Where "stype-code" is a user variable and "YYSTYPE" is a Bison 
> > variable:
> >
> >   %header {
> >     $var{stype-code}
> >     $var{YYSTYPE}
> >   }
> 
> Something like that might be reasonable, yes.
> 
> Also, it might make sense to let the user specify the name of
> the header, or even to generate multiple (different) headers.
>
> More generally, perhaps we should have a sublanguage to let the user
> generate arbitrary files containing information of interest.

So, replace %header with:

  %file "my_parser.h" {
    $var{stype-code}
    $var{YYSTYPE}
  }

In this case, the user might wish not to use %defines.  However, some of 
the skeletons require %defines.  In order to get the declaration order 
right in that required header file, it might be necessary to have:

  %defines "my_parser.h" {
    $var{stype-code}
    $var{YYSTYPE}
  }

to tell Bison that "my_parser.h" is the header file it requires.