[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [RFC] %language declaration

From: Paolo Bonzini
Subject: Re: [RFC] %language declaration
Date: Fri, 15 Sep 2006 10:23:45 +0200
User-agent: Thunderbird (Macintosh/20060719)

Akim Demaille wrote:
"Paolo" == Paolo Bonzini <address@hidden> writes:

 > Convenient abbreviations, or convenient forms.  In this case, in C you
 > only need to add "%glr-parser" for example.  In C++, you have to change

 >    %skeleton ""

 > to

 >    %glr-parser
 >    %skeleton ""

Because that's really *two* changes!  You both change the computation
of the automaton, and the skeleton you use!
I think 99% users could not care less of what skeleton is used. They just want a GLR parser, a C++ parser, a Java parser, a push parser, a pure parser, a combination of these, whatever.
 > With %language, the change would be the same -- add "%glr-parser".

What do you do about people who want to use/work on a different
implementation?  Will they be able to use a simple name such as
"push.c"?  Will they have to patch bison to have it accept this name?
Why not?  %skeleton is going to remain, but mostly for developers.

And I will modify my patch so that the list of skeletons that implement the automata is *not* in Bison; the list of languages might be though.
 > I'm also not going to delve into naming schemes for skeletons, and so
 > on.  It's not worth for sure while we have only two grammar kinds, and
 > if anything I'd rather have a config file in /usr/share/bison than
 > doing file system scanning.

This is not nice for user toying with their skeletons.  If you walk
that way, please don't touch the semantics of %skeleton at all.
Provide some other directives.
Nobody is thinking of removing %skeleton.


reply via email to

[Prev in Thread] Current Thread [Next in Thread]