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

I thought I was versed in pattern matching until I came across this code. After deconstructing a Perl Golf solution from mtve, I found myself unable to understand why and how 2 lines of code work.

http://terje.perlgolf.org/

#!/usr/bin/perl

for(@a=glob"{\321,.,\321}"x2) {
    $x = $_;
    for (@a) {
        $h = ~$x;
        print '*' if $_ =~ m/$h/; #Why does this line match?
        print ' ' if  $_ !~ m/$h/; #Why does this line not?

        #The next line is just for match tracking.
        $matches{"$_ =~ $h"} = $_ =~ m/$h/ ? '*' : ' ';
    }
    print "\n";
}

foreach $key(sort keys %matches) {
        print "$key    $matches{$key}\n";
}

[download]

Here's the output:

*********
* ** ** *
*********
***   ***
* *   * *
***   ***
*********
* ** ** *
*********
.. =~ ..    *
.. =~ .Ñ
.. =~ Ñ.
.. =~ ÑÑ
.Ñ =~ ..    *
.Ñ =~ .Ñ    *
.Ñ =~ Ñ.
.Ñ =~ ÑÑ
Ñ. =~ ..    *
Ñ. =~ .Ñ
Ñ. =~ Ñ.    *
Ñ. =~ ÑÑ
ÑÑ =~ ..    *
ÑÑ =~ .Ñ    *
ÑÑ =~ Ñ.    *
ÑÑ =~ ÑÑ    *

[download]

It seems odd to me that ÑÑ =~ Ñ. would match but Ñ. =~ ÑÑ would not. Here's and even simpler version that shows that Ñ =~ . matches and . =~ Ñ does not. 

#!/usr/bin/perl

for(@a=glob"{\321,.,\321}"x1) {
    $x = $_;
    for (@a) {
        $h = ~$x;
        print '*' if $_ =~ m/$h/; #Why does this line match?
        print ' ' if  $_ !~ m/$h/; #Why does this line not?

        #The next line is just for match tracking.
        $matches{"$_ =~ $h"} = $_ =~ m/$h/ ? '*' : ' ';
    }
    print "\n";
}

foreach $key(sort keys %matches) {
        print "$key    $matches{$key}\n";
}

[download]

Here's the output:

***
* *
***
. =~ .    *
. =~ Ñ
Ñ =~ .    *
Ñ =~ Ñ    *

[download]

Replies are listed 'Best First'.
Re: Cantor's Revenge Matching
by Paladin (Vicar) on May 01, 2003 at 16:45 UTC
    It seems odd to me that ÑÑ =~ Ñ. would match but Ñ. =~ ÑÑ would not. Here's and even simpler version that shows that Ñ =~ . matches and . =~ Ñ does not.

    . matches any character (except \n by default) so m/Ñ./ matches a Ñ followed by any character, which ÑÑ is. m/ÑÑ/ matches only ÑÑ, which Ñ. is not. The difference is that in a regex, . has special meaning, while in a string, it doesn't.

[reply]
[d/l]
[select]

      So the value in $h is being used as regex code. Somehow I thought the pattern needed to pre-compiled using qr//.

      $h =  qr/../;

      Thank you so much for the answer.

[reply]
[d/l]
        Correct. Regexen go through double-quotish interpolation (with a few exceptions, like using single quotes as the delimiters) before being passed on to the regex engine, so you can say stuff like: 
        my $foo = 'first part';
my $anything = '.*';
my $bar = 'last part';

$_ =~ /$foo$anything$bar/;

        [download]
        and have it mean the same as: 
        $_ =~ /first part.*last part/;

        [download]
[reply]
[d/l]
[select]
Re: Cantor's Revenge Matching
by queue (Beadle) on May 01, 2003 at 16:38 UTC
    I might be missing something here, but the dot on the LHS is a literal dot character. The dot on the RHS is the dot metacharacter, which matches any single character. So a dot on the LHS isn't going to be matched by anything on the RHS except a literal dot or the dot metacharacter.
[reply]
Re: Cantor's Revenge Matching
by terrencebrown (Acolyte) on May 01, 2003 at 17:40 UTC

    Very good for both answers. Thank you.

[reply]

Back to Seekers of Perl Wisdom