Knowing which regular expression matched 1

[MacTommy]

I know how to solve this in a not so elegant way, but somehow I have the feeling I am missing something obvious.

What I want is this: I've got a couple (possibly 1000's) of regexps. I want to go through a string, each time finding the first regexp that matches, and returning some relevant bit that matched, plus a string of which regexp it was that matched.

So, let's say if regexp 1 matched, I want to return $4 (relevant to that first regexp that is), and the string 'FIRST'. If the second one matched I want to return $2, and 'SECOND', e.g.

And preferrable of course I want to go through the string just once, and do the matching just once...

 

[MacTommy]

Well, I figured it out and just in case someone has the some problem and manages to stumble upon this page, I post the solution.
It lies in using the (?{ Perl code }) construct in your regular expressions and using $^N in this code.

This next bit works and does what I described above:

use strict;

my $sString = "blurb pppqqqqrrr and some weabcdeeff and some ";

my ($sType, $sValue);
while( (($sType, $sValue) = matcher(\$sString)) && defined $sType ) {
	print "type: $sType, value: $sValue\n";
}

sub matcher {
	my ($srString) = @_;

	if( $$srString =~
			/# First
			 (a(b(c(d(eee?ff))(?{$sType = 'First'; $sValue = $^N;}))))\ and\ some|
			 # Second one
			 ppp(q+(r+)(?{$sType = 'Second'; $sValue = $^N;}))\ and\ some
			 /gsx ) { # "
		return($sType, $sValue);
	}

	return(undef, undef);
}

Output:

type: Second, value: rrr
type: First, value: deeff

Read more on all this on perlre

 

[KevinADC]

Very cool and I hope I never have to use it. Makes my brain hurt a little just looking at it.

------------------------------------------
- Kevin, perl coder unexceptional!

 

[KevinADC]

You might want to make that a perl FAQ too.

------------------------------------------
- Kevin, perl coder unexceptional!

 

[MacTommy]

The reason I want to use it is because I am making kind of a lexicaliser like this. It uses regular expressions in stead of just plain strings (like a normal dictionary does) and it returns lexical category, plus the actual match. And it skips any part that doesn't match anything.
Does that clarify anything... ?!?

And just what would that FAQ be..?!?
Can you tell me something obscure of which the manual says it is expirimential and can be altered/omitted any release now, that I hope I never have to use..?!?
;-)

 

[MacTommy]

experimental that is...
Too bad you can't edit these posts once you clicked 'submit' (too soon...).

 

[ishnid]

I could be very wrong ... but ...

I can't help thinking there's a neater way of achieving what you're looking to do (of which I'm not entirely sure either ;-))

The main reason I'm thinking this is because you have a lot of parentheses in your regexp which don't appear to be doing anything.

$^N is the text from the last capturing group that matched. For the first part of the regexp, that'll always be either (d(eee?ff)) or (eee?ff). For the second, it'll always be (r+), right?

Can you expand a little (perhaps with a real example) of what your lexicaliser is supposed to do. Maybe Parse::RecDescent (a very powerful module that's a total pain to learn but well worth it) might be useful for this kind of task.

 

[MacTommy]

Yes, I am sorry. The parentheses were just for the sake of the example (as I mentioned $2 and $4 in the first post). They don't make much sense apart from that in the example I admit...

I am actually using this together with Parse::Yapp, but Parse::RecDescent might be worth a look as well. Thanks!