in reply to Re: Matching Exact Word
in thread Matching a Word Exactly

sleepy i guess :) not anchors, but match an optional word before Guinea, then check 
my( $word, $guinea ) = /(\w+)?\s*\b(Guinea)\b/;
if( $word eq ucfirst $word ){
   warn "Not the Guinea I want ($word $guinea)";
}

[download]

Replies are listed 'Best First'.
Re: Matching Exact Word
by jonadab (Parson) on Oct 09, 2014 at 11:01 UTC

    This will fail in some cases.

    Geographically, Guinea is thousands of miles from here. (This fails immediately because of the comma; if the comma were removed, it would still fail.)

    If what you want is to match Guinea but not New Guinea or Equatorial Guinea, then what you probably really want is a negative lookbehind assertion that specifically rules out being preceded by "New " or "Equatorial ". Similarly, a negative lookahead assertion at the end can preclude Guinea Pig and Guinnea-Bisseau.

[reply]
      If what you want is to match Guinea but not New Guinea or Equatorial Guinea, then what you probably really want is a negative lookbehind assertion that specifically rules out being preceded by "New " or "Equatorial "

      One caveat:  You can't use alternation in the look-behind assertion because variable-length negative look-behind assertion isn't supported. Instead, you must list the alternatives separately. You can, of course, use alternation in the look-ahead assertion.

      use strict;
use warnings;

my $pattern = qr{
    (?<!New\s)
    (?<!Equatorial\s)
    Guinea
    (?![\s-](?:Bissau|pig))
}ix;

while (my $text = <DATA>) {
    my $match = $text =~ m/$pattern/ ? 1 : 0;

    print "$match $text";

    # This prints...
    # 0 Papua New Guinea
    # 1 I live in Guinea.
    # 1 i live in guinea, but i don't have a shift key.
    # 0 Guinea-Bissau
    # 0 Guinea Bissau
    # 0 Equatorial Guinea
    # 0 I love guinea pigs!
}

__DATA__
Papua New Guinea
I live in Guinea.
i live in guinea, but i don't have a shift key.
Guinea-Bissau
Guinea Bissau
Equatorial Guinea
I love guinea pigs!

      [download]
[reply]
[d/l]

      You are right. However, even this code may fail, if somebody misspells the country names.

      It is more a linguistic problem than a pattern recognition one, and, as such, seems extraordinary difficult to tackle in a failproof way (which would require an AI, a syntaxic and contextual analysis, etc.)

      However, as you mentioned, using negative look-ahead and negative look-behind assertions should allow him to avoid the most common other words.

[reply]

In Section Seekers of Perl Wisdom