Re: Multiple identifiers for same terminal and rule

On Sat, Aug 30, 2003 at 08:34:11AM -0700, Randal L. Schwartz wrote:
> >>>>> "Colin" == Colin Kuskie <[email protected]> writes:
> 
> Colin> I often find myself giving identical productions different
> Colin> identifiers to make my grammar clearer and to get around the limitation
> Colin> of named access only giving the last match to an identifier, like
> Colin> the following:
> 
> What I've done is define the details once, then just had a simple
> rule that renamed it:
> 
> cell: cell_or_inst
> inst: cell_or_inst
> cell_or_inst: /\w+/

That's a good work-around.  Thanks!

> Oh, and leave off the { $item[1] } for those.  That's the default. :)

I know, but you have to know that most of the crew that I work with
would learn a lot from reading PORM.  I try to write the more complex
stuff as clearly as I can so that they can follow it (and I don't
forget it later as well!).

That's why I suggested the syntax.  To me, it seems more clear.

It's very analogous to what I think Perl6 would allow:

$cell = $inst = m:/\w+/;

Maybe I'll try hacking P::RD and see if I can make a patch and
supporting tests.  It shouldn't be too hard, since I don't want to
change the tokenizer, only the lexer.  But it's going to break the
current behavior where that particular syntax creates an empty pattern
that always matches.

Colin