Re: Regex niceties for the parser (was Re: To get things started...)

Text::Trie is an amazing little module.  I've abused it in a macro processing
source filter to build optimum regexps out of lists of tokens.

I've attached a proof-of-concept/benchmark program using the interesting bits
from POE::Preprocessor to build, test, and time regexps.  Results of a sample
run are included after __END__.

-- Rocco Caputo / [email protected] / poe.perl.org


On Mon, 27 Nov 2000 12:37:52 -0500, Dan Sugalski wrote:

>At 05:17 PM 11/27/00 +0000, Nicholas Clark wrote:
>> > >>>>> "ST" == Sam Tregar <[email protected]> writes:
>> > ST> Perhaps we really need a new kind of regex that works by-design against
>> > ST> streams of bytes?
>>
>>I don't think any change is needed in the regex syntax. I think just being
>>careful to use minimal matching whenever possible should be sufficient.
>
>What we could use, though, is a way to specify weighted alternatives in 
>regexes, so we could have a single regex look for multiple things at once 
>and match the one we'd prefer if there are alternatives. I don't even 
>really care if we can't do it inside a regex specification itself and need 
>to do odd things with regex objects.
>
>Yeah, I know we can do this with alternations and such, but that can get 
>really messy when we're pasting a half-zillion regexes together.
>
>					Dan