[antlr-interest] Token position 0, -1

[A Z]

This is a bug in the C code generator. The parser code should call
createTypeToken() but instead calls createTypeText().  You can fix it by
modifying
antlr-3.4/tool/src/main/resources/org/antlr/codegen/templates/C/AST.stg
towards the end of the file and recompiling the jar file.


On Tue, Jul 10, 2012 at 3:55 AM, Mike Lischke <mike at lischke-online.de>wrote:

>
> Stefan, Jim,
>
> thanks for the hint about the start/stop tokens.
>
> >> Is there a similar construct if you have a rule instead of a token? For
> >> instance:
> >>
> >> target_table:
> >>      qualified_identifier -> ^(TABLE_REF_ID qualified_identifier)
> >> ;
> >
> > I use in some case the start attribute of the subrule:
> >
> > target_table:
> >       qualified_identifier -> ^(TABLE_REF_ID[$qualified_identifier.start]
> > qualified_identifier)
>
> This doesn't work with my grammar (using the C target). Instead of taking
> the location only the token text of my virtual token is set to NULL. The
> generated C code is:
>
>             if ( BACKTRACKING==0 )
>
>             {
>                 pANTLR3_REWRITE_RULE_SUBTREE_STREAM stream_retval;
>
>
> stream_retval=antlr3RewriteRuleSubtreeStreamNewAEE(ADAPTOR,  RECOGNIZER,
> (pANTLR3_UINT8)"token retval", retval.tree != NULL ? retval.tree : NULL);
>
>                 root_0                      =
> (pANTLR3_BASE_TREE)(ADAPTOR->nilNode(ADAPTOR));
>                 retval.tree    = root_0;
>                 // 580:23: -> ^( TABLE_REF_ID[$qualified_identifier.start]
> qualified_identifier )
>                 {
>                     // MySQL55.g:580:26: ^(
> TABLE_REF_ID[$qualified_identifier.start] qualified_identifier )
>                     {
>                         pANTLR3_BASE_TREE root_1 =
> (pANTLR3_BASE_TREE)(ADAPTOR->nilNode(ADAPTOR));
>                         root_1 =
> (pANTLR3_BASE_TREE)(ADAPTOR->becomeRoot(ADAPTOR,
>
>                         #if 1 == 2
>
> (pANTLR3_BASE_TREE)ADAPTOR->createTypeTokenText(ADAPTOR, TABLE_REF_ID,
> TOKTEXT((qualified_identifier695.start)))
>                         #else
>
> (pANTLR3_BASE_TREE)ADAPTOR->createTypeText(ADAPTOR, TABLE_REF_ID,
> (pANTLR3_UINT8)(qualified_identifier695.start))
>                         #endif
>
>                         , root_1));
>
>                         ADAPTOR->addChild(ADAPTOR, root_1,
> stream_qualified_identifier == NULL ? NULL :
> stream_qualified_identifier->nextTree(stream_qualified_identifier));
>
>                         ADAPTOR->addChild(ADAPTOR, root_0, root_1);
>                     }
>
>                 }
>
>
>                 retval.tree = root_0; // set result root
>
> which shows that the start token is used only to create the token text for
> root_1, which is clearly not what we want. Also, this 1 == 2 part seems
> quite useless and is just bloating the already huge file (almost 134K lines
> already).
>
> Mike
> --
> www.soft-gems.net
>
>
>
> List: http://www.antlr.org/mailman/listinfo/antlr-interest
> Unsubscribe:
> http://www.antlr.org/mailman/options/antlr-interest/your-email-address
>