[antlr-interest] Skipping or ignoring whole subtrees in a tree grammar

[Frank Smith]

The wildcard is an interesting beast.

Only use it on the bare minimum amount of stuff in your grammar (e.g. only
the rule you need skipped!), because I had some seemingly unrelated errors
crop up in my grammar that I ended up tracing back to overusing the wildcard
in some rules.

in my case it was:

^('fun' TYPE ID param* block)

worked perfectly with:
^('fun' TYPE ID param* .)

and caused all kinds of heartache with:
^('fun' TYPE ID . .)

so even though i didn't use param in the action that followed i left it in
there for the sake of it actually working :)

good luck with yours

On 8/13/07, Hardy, Stephen <Stephen_Hardy at rabbit.com> wrote:
>
> Frank,
>
> thanks for the response!  It seems that '.' is a bit more magic than I
> thought.  That is, the semantics seem to be that '.' matches the token
> plus any subtree that it may have, not just the 'root' token itself
> (although I stand open to correction on this interpretation).
>
> Regards,
> SJH
>
>
> > -----Original Message-----
> > From: Frank Smith [mailto:fasmith718 at gmail.com]
> > Sent: Monday, August 13, 2007 8:44 AM
> > To: Hardy, Stephen
> > Cc: antlr-interest at antlr.org
> > Subject: Re: [antlr-interest] Skipping or ignoring whole
> > subtrees in a tree grammar
> >
> > my mistake: you can access bar by doing foo.start.getChild(0);
> >
> > Sorry for the mixup
> >
> > --Frank
> >
> >
> > On 8/13/07, Frank Smith <fasmith718 at gmail.com> wrote:
> >
> >       Use the wild card.  That causes a tree grammar to skip
> > a token e.g.:
> >
> >       ...
> >       | ^(DELETED_FOO .)
> >       ...
> >
> >       you can still access bar if you want by using
> > DELETED_FOO.start.getChild(0); if you want to use it in an
> > action somehow, but not be stepped through automatically by
> > the walker.  I used this a lot- works great.
> >
> >       Best regards,
> >            Frank
> >
> >
> >
> >       On 8/13/07, Hardy, Stephen < Stephen_Hardy at rabbit.com
> > <mailto:Stephen_Hardy at rabbit.com> > wrote:
> >
> >               Hi all,
> >
> >               supposing I have a tree grammar which includes
> > something like
> >
> >               foo
> >                  : ^(FOO          bar  { some action... } )
> >                  | ^(DELETED_FOO  bar)
> >                  ;
> >
> >               Suppose bar is a complex, deeply recursive,
> > subtree; and I wish to avoid
> >               as much as possible the overhead of parsing it
> > in the DELETED_FOO case.
> >               (Imagine that DELETED_FOO was once a FOO, but
> > the parser or application
> >               decided after the fact that it was an
> > irrelevant part of the input, and
> >               changed the token to mark it as deleted, but
> > without actually removing
> >               the subtree.  Why not delete the tree?  Well, a
> > different tree grammar
> >               may still want to see the contents, just not this one.)
> >
> >               Is there a way of telling ANTLR to basically
> > 'skip' any analysis of the
> >               subtree under DELETED_FOO?  For example, a
> > routine which simply consumes
> >               tokens from the DOWN which immediately follows
> > DELETED_FOO until the
> >               matching (balanced) UP token is seen?
> >
> >               I seem to recall a mention of something like
> > this in TDAR, but darned if
> >               I can locate it.
> >
> >               Regards,
> >               SJH
> >
> >
> >
> >
> >
> >
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.antlr.org/pipermail/antlr-interest/attachments/20070813/f864bd55/attachment-0001.html