http://www.perlmonks.org?node_id=941881

svenXY has asked for the wisdom of the Perl Monks concerning the following question:

Hi,

I'm confused with a negative lookahead. Here's some example code. I want it to match on $needs and to not match on $has. The examples 3 and 4 are only as a reference to a fixed string, but I basically want ^($start(.*?))(?!$end)$. Any ideas?

#!/usr/bin/perl
use strict;
use warnings;

my $start = "log4j.rootLogger=";
my $startlong = "log4j.rootLogger=INFO, FILE";
my $end = ", SYSLOG";

my $needs = 'log4j.rootLogger=INFO, FILE';
my $has = $needs . ', SYSLOG';

print "needs_matched $needs\n" if $needs =~ /^($start(\w+(,\s)?)+)(?!$
+end)$/;
print "has_matched $has\n" if $has =~ /^($start(\w+(,\s)?)+)(?!$end)$/
+;
print "needs_matched $needs\n" if $needs =~ /^(log4j.rootLogger=INFO, 
+FILE)(?!, SYSLOG)$/;
print "has_matched $has\n" if $has =~ /^(log4j.rootLogger=INFO, FILE)(
+?!, SYSLOG)$/;

[download]

Regards,
svenXY

Replies are listed 'Best First'.
Re: regex: negative lookahead
by kennethk (Abbot) on Dec 05, 2011 at 15:54 UTC
    If you take a look at what you have in $1 after your failed miss for 2, the issue will become more apparent, I think; print $1; yields log4j.rootLogger=INFO, FILE, SYSLOG. Essentially, you have already consumed ", SYSLOG" by the time you get to the end of the string, so clearly that bit does not follow. A negative look-behind will actually yield what you intended:

    print "has_matched $has\n" if $has =~ /^($start(\w+(,\s)?)+)(?<!$end)$/;

[reply]
[d/l]
[select]
      Hi,
      ++kennethk - thanks for clarifying. I'm still a little bit confused by the difference between (?!...) and (?<!...) though.
      Regards,
      svenXY
[reply]
        The difference is a lookahead versus a lookbehind. If you ignore the lookahead/lookbehind because they are zero width (don't consume characters), you would expect /^($start(\w+(,\s)?)+)$/ to grab an entire string between start and finish. When you get to the part of the regular expression where you have your assertion, the cursor is here: 
        log4j.rootLogger=INFO, FILE, SYSLOG
                                   ^

        [download]
        If you look ahead of this position, there is nothing, which clearly passes the negative lookahead check. The offending string is behind your position, thus you need a negative lookbehind.
[reply]
[d/l]
[select]
Re: regex: negative lookahead
by Khen1950fx (Canon) on Dec 05, 2011 at 20:50 UTC
    I've never had a the need to use a lookaround, but your question piqued my curiosity. As an experiment, I used Regexp::Assemble to rework your example because it just does the right thing with assertions, lookaheads, and lookbehinds. 
    #! /usr/bin/perl -slw

use strict;
use Regexp::Assemble;

my $needs = Regexp::Assemble->new->add( qw[ FILE INFO ] );
my $has   = Regexp::Assemble->new->add( qw[ FILE INFO SYSLOG ] );

while( defined( $_ = <DATA> )) {
    chomp;
    if( /($needs)/ ) {
        print $needs->as_string;
    }
}
while( defined( $_ = <DATA> )) {
    chomp;
    if( /($has)/ ) {
        print $has->as_string
    }
}

__DATA__
FILE
INFO
SYSLOG

    [download]
    I believe, if I"ve understood you correctly, that it works the way that it should. What do you think?
[reply]
[d/l]
Re: regex: negative lookahead
by vinian (Beadle) on Dec 06, 2011 at 05:01 UTC

    i don't think there is something to do with regex negative lookahead, the problem lies in the
    (\w+(,\s)?)+
    it will match "INFO, FILE, SYSLOG", but not "INFO, FILE".
    "+" is greedy, it eaten whatever match by "( ... )" . I don't how to make the above regex match less, but i use this
    print "has_matched $has\n"    if $has   =~ /^($start(\w+(,\s)?)+?)$(?<!$end)/;
    and it will not match $has.

[reply]
[d/l]
[select]

Back to Seekers of Perl Wisdom