Hi gavin
Am 2007-03-31 um 22:35 schrieb Gavin Lambert:
> At 04:25 1/04/2007, Tobias Pape wrote:
> >unaryPattern returns [ pASTPattern p ]
> > : s=unarySelector
> > { $p = (pASTPattern) ASTUnaryPattern_new((char *) $s.text-
> > >chars); }
> > ;
> >
> >binaryPattern returns [ pASTPattern p ]
> > : s=binarySelector v=argument
> > { $p = (pASTPattern) ASTBinaryPattern_new((char *) $s.text-
> > >chars, v.v); }
> > ;
>
> I don't think you're allowed to upcast like that in the C-language
> parsers. At least not if 'ASTBinaryPattern' inherits from
> ASTPattern, anyway, rather than simply being an ASTPattern with a
> different constructor (in which case you probably don't need the
> cast). Inheritance is usually faked via containment and you're not
> going to get the compiler's help to convert pointers, so unless the
> structure is laid out in a very specific way then you're going to
> end up with an invalid pointer.
leave my ASTs away ;)
yes, exactly, my AST* layout is special, its kinf of fake-OOP.
This will explain more to you:
for my Patterns, I got the rules as:
==============================================
pattern returns [ pASTPattern p ]
: up=unaryPattern { $p = up.p; }
| bp=binaryPattern { $p = bp.p;}
| kp=keywordPattern { $p = (pASTPattern) kp.p; }
;
unaryPattern returns [ pASTPattern p ]
: s=unarySelector
{ $p = (pASTPattern) ASTUnaryPattern_new((char *) $s.text-
>chars); }
;
binaryPattern returns [ pASTPattern p ]
: s=binarySelector v=argument
{ $p = (pASTPattern) ASTBinaryPattern_new((char *) $s.text-
>chars, v.v); }
;
keywordPattern returns [ pASTKeywordPattern p ]
: (s=keyword v=argument {
$p = ASTKeywordPattern_new();
$p->add_argument($p,(char *) $s.text->chars, v.v);
} )+
;
==============================================
As you can see, "pattern" returns an pASTPattern p, but if
assigned is in the subrules as pASTBinaryPattern or pASTUnaryPattern,
cc would complain,
when i assign "$p = up.b;" or "$p = bp.p;".
and, yes, the cast is intended, here the layouts:
==============================================
[from ASTPattern.h]
// ASTPattern
#define NODE_ASTPattern 6
#define ASTPATTERN_FORMAT \
ASTNODE_FORMAT;\
/*abstract */ p_string (*get_selector)(pASTPattern self);\
/*abstract */ pList (*get_arguments)(pASTPattern self)
struct _ASTPattern {
// ASTPattern
ASTNODE_FORMAT;
/*abstract */ p_string (*get_selector)(pASTPattern self);
/*abstract */ pList (*get_arguments)(pASTPattern self);
};
==============================================
==============================================
[from ASTUnaryPattern.h]
// ASTUnaryPattern
#define NODE_ASTUnaryPattern 7
#define ASTUNARYPATTERN_FORMAT \
ASTPATTERN_FORMAT; \
/* private */ char * _selector
struct _ASTUnaryPattern {
// ASTUnaryPattern
ASTPATTERN_FORMAT;
/* private */ char * _selector;
};
==============================================
==============================================
[from ASTBinaryPattern.h]
// ASTBinaryPattern
#define NODE_ASTBinaryPattern 8
#define ASTBINARYPATTERN_FORMAT \
ASTPATTERN_FORMAT;\
/* private */ char * _selector;\
/* private */ struct _ASTVariable* _parameter
struct _ASTBinaryPattern {
// ASTBinaryPattern
ASTPATTERN_FORMAT;
/* private */ char * _selector;
/* private */ struct _ASTVariable* _parameter;
};
==============================================
as you can see, only the "visible" fields differ, and the constructors
are there for specific asignmenst.
A constructed ASTPattern would differ significantly (?) from any
other AST*Pattern.
you know what I mean?
>>These casts, fact, arent my problems at all. I'm merely messing
with libantlr3c ;)<<
But thanks for your notice. If you want, I'll get you a copy of my
speciffic "C Class Layout" (without late binding, thoug)
thanks,
-Tobias
-------------- next part --------------
A non-text attachment was scrubbed...
Name: PGP.sig
Type: application/pgp-signature
Size: 186 bytes
Desc: Signierter Teil der Nachricht
Url : http://www.antlr.org/pipermail/antlr-interest/attachments/20070331/4fabdfcb/attachment.bin