If you run the code with  -Mre=debug you can see it first matching two digits, then matching the lookahead string (thus failing), backtracking, trying to match a single digit, then succeeding:

  Setting an EVAL scope, savestack=5
   0 <> <2005-12->        |  1:  CURLY {4,4}
                           DIGIT can match 4 times out of 4...
  Setting an EVAL scope, savestack=5
   4 <2005> <-12->        |  4:    EXACT <->
   5 <2005-> <12->        |  6:    CURLY {1,3}
                           DIGIT can match 2 times out of 3...
  Setting an EVAL scope, savestack=5
   7 <2005-12> <->        |  9:      UNLESSM[-0]
   7 <2005-12> <->        | 11:        EXACT <->
   8 <2005-12-> <>        | 13:        SUCCEED
                                  could match...
                                failed...
   6 <2005-1> <2->        |  9:      UNLESSM[-0]
   6 <2005-1> <2->        | 11:        EXACT <->
                                  failed...
   6 <2005-1> <2->        | 15:      END
Match successful!
[download]

Dave.

Cheers,

JohnGG

Look-aheads and look-behinds are IMO advanced constructions we don't need most of the time for simple problems. They demand more thought and many times are not what we wanted at the end. For example, from your description ("I expected it not to match because of the '-' at the end.") and your two solutions, I would suggest the use of simple /\d{4}-\d{1-3}([^-\d]|$)/, where [^-\d] prevents the pattern to match after 1, 2 or 3 digits and encountering yet another digit or dash, and $ in the alternation makes it succeed at the end of the line.

#!/usr/bin/perl -w

use strict ;
use warnings;

for my $d qw(2005-12 2005-100 2005-1- 2005-12- 2005-123- 2005-1000) {
    printf "%-10s: ", $d;
    if ( $d =~ /\d{4}-\d{1,3}([^-\d]|$)/ ) {
      print "yep\n" ;
    } else {
      print "nope\n";
    }
}
[download]

2005-12   : yep
2005-100  : yep
2005-1-   : nope
2005-12-  : nope
2005-123- : nope
2005-1000 : nope
[download]

It's actually possible to avoid matching the extra character by using (?>...).

for my $d qw(2005-12 2005-100 2005-1- 2005-12- 2005-123- 2005-1000) {
    printf "%-10s: ", $d;
    if ( $d =~ /(?>\d{4}-\d{1,3})(?!-)/ ) {
      print "yep\n" ;
    } else {
      print "nope\n";
    }
}
[download]

2005-12   : yep
2005-100  : yep
2005-1-   : nope
2005-12-  : nope
2005-123- : nope
2005-1000 : yep
[download]

Use (?![-\d]) if you don't want the last to match.

/^\d{4}\-\d{2}$/ or die; # match on end of string after \d{2}, or ...
/^\d{4}\-\d{2}\s+/ or die; # match on whitespace delimiter
# etc.
[download]

I always try to keep my questions as short as possible!
So they often do not describing the true issue
Anyway, I get your point, but in my case, in which I've written a 'generic date parser/converter' I don't really want to use ^ and $, it would limit the number of possible date formats, for example


"2005-031"
" 2005-31"
"2005031 "
"|2005-031 12:11:22| "
"Some time ago 1776-07-04 ....."
[download]

And I'm not even started to scratch the surface of what my 'generic date parser/converter' can do more :)

Thnx
LuCa

In that case I would be inclined to maintain a list of regexps - one for each allowable format - rather than (I predict) torturing one into handling successive new requirements until it finally dies in an agony of unmaintainability. I might even put them in a configuration file rather than code for easy update in production environments, load and chop them them into an array and then try them out successively on the data until a match is found or the possible formats exhausted.

-M

Free your mind

You probably already know about Date::Manip. It has a function ParseDateString that should do some of what you want.

   /\d{4}-(?>\d{1,3})(?!-)/
[download]

for (qw(2005-12 2005-100 2005-1- 2005-12- 2005-123- 2005-1000)) {
    printf "%-10s: ", $_;
    if ( /\d{4}-(?>\d{1,3})(?!-)/ ) {
      print "yep\n" ;
    } else {
      print "nope\n";
    }
}
[download]

2005-12   : yep
2005-100  : yep
2005-1-   : nope
2005-12-  : nope
2005-123- : nope
2005-1000 : yep

   /\d{4}-(?>\d{1,3})(?![\-\d])/
[download]

Update I shouldn't post in a hurry. I now see ikegami has posted a post very similar to mine. Duh.

\d{4}-(\d{3}|\d{2}|\d)(?!-)
[download]

You might want to consider a negative lookahead on \d as well:

\d{4}-\d{1,3}(?![-\d])
[download]

Of course, this means you won't match a string like "2005-1222" either. The original would match that, and if that's intentional, you need an even trickier version:

\d{4}-\d{1,3}(?!\d*-)
[download]

Here's betting one of these should be what you want. :-)

print "Just another Perl ${\(trickster and hacker)},"
The Sidhekin proves Sidhe did it!