I wrote the following before finding the delightful Sort::Naturally library function. I reccommend using that over what I've done, but maybe people are curious how it might be done. Or maybe someone needs to fine-tune or customize it.

The following code is NOT efficient. In particular, the calling of naturalSortInner with $a and $b as arguments completely eliminates the efficiency of having $a and $b. However, it was necessary for the recursive nature of naturalSortInner.

#!/usr/bin/perl

sub naturalSortInner {
    $x = uc( shift );
    $y = uc( shift );
    if( !($x =~ /\d+(\.\d+)?/) ) {
        return $x cmp $y;
    }
    $xBefore = $`;
    $xMatch = $&;
    $xAfter = $';
    if( !($y =~ /\d+(\.\d+)?/) ) {
        return $x cmp $y;
    }
    if( $xBefore eq $` ) {
        if( $xMatch == $& ) {
            return naturalSortInner( $xAfter, $' );
        } else {
            return $xMatch <=> $&;
        }
    } else {
        return $x cmp $y;
    }
    print "\n<before: '$xBefore', match: '$xMatch', after: '$xAfter'>\
+n";
}
sub naturalSort {
    naturalSortInner( $a, $b );
}


@arr = (
    'beta',
    'Alpha',
    'Gamma1',
    'Gamma',
    '23',
    '5',
    'Version1',
    'Version1.1',
    'Version1.2',
    'Version11.1-Sub1',
    'Version11.1-Sub10',
    'Version11.1-Sub3',
    'x23sub5',
    'Version2',
    'Version2.1',
    'GammaGlobulin',
    'Gamma10',
    'c',
    'Gamma2',
    'Gamma3',
);

print join( "\n", sort naturalSort @arr ) . "\n";
[download]