Re: [PATCH] Use variants to support objects as semantic values.

[Akim Demaille]

Le 4 nov. 08 à 11:50, Paolo Bonzini a écrit :

> > I just say that the java skeleton now uses %define stype, which is

s/say/saw/

> > correlated.  I regret that the names were not discussed first, now we
> > face a lack of structure, although Joel had been careful up to now to
> > propose something consistent.
>
> "%define stype" goes with "#define YYSTYPE" in C/C++; it was  
> introduced
> before dotted names were introduced for %defines.  Note that it is in
> principle possible to move from "#define YYXYZ" to "%define xyz" for
> this and other preprocessor macros in the C/C++ skeletons.

I really want to get rid of the CPP directives in favor of Bison  
directives.  And I'd like that we continue getting rid of specific  
directives such as %debug to move into a more consistent %define debug  
style.  It has the added value that then it then automatically benefit  
from --define (which is somewhere waiting to the pulled from the  
candidates/make-symbol branch) which saves us from inventing a  
bazillion of new options (--locations, --debug etc.), documenting them  
twice (once as a directive, and then as an option) etc.

Of course backward and yacc compatibilities have to be considered.

> What about doing like this:
>
> 1) %union => use unions

Makes sense :)

> 2) %define stype => use given type as ::semantic_type

I really dislike stype, it looks like (verbose :) assembly language.

> 3) no %union, no %define stype, b4_tag_seen_flag=1,
> %define global_tokens_and_yystype => do not define YYSTYPE at global

I guess this is not what you meant here.

> scope, and use YYSTYPE as ::semantic_type
>
> 4) no %union, no %define stype, b4_tag_seen_flag=1,
> no %define global_tokens_and_yystype => use variants
>
> 5) no %union, no %define stype, b4_tag_seen_flag=0
> => use int as ::semantic_type
>
> ?

I have once been fond of such cascading rules, but it turned out it  
was too complex to be easy to teach.  I guess I would rather have  
something like

%define semantic.type.kind union|variant|type
// in the latter case
%define semantic.type.value  USER-DEFINED (defaults to int or YYSTYPE)

%union would %define semantic.type.kind union

Or maybe use impossible values ?

%define semantic.type "%union" or "%variant" or "USER-TYPE".