Re: counting overlapping patterns

[Limbic~Region]

Anonymous Monk,
Here is some code that covers your example, but it admittedly only works on fixed strings since it doesn't use regular expressions at all.

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

print str_count('AAAA', 'AA'), "\n";

sub str_count {
    my ($str, $pat) = @_;
    my $tot;
    for ( 0 .. length( $str ) - length( $pat ) ) {
        $tot++ if index($str, $pat, $_) - $_ == 0;;
    }
    return $tot;
}

[Limbic~Region]

While discussing other methods that avoided regular expressions in the CB with bobf and nothingmuch, I mentioned an unpack/hash solution. nothinmuch asked to see it, so here is a highly untested alternative.

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

print str_count('ABAABAAAA', 'AA'), "\n";

sub str_count {
    my ($str, $pat) = @_;
    my %substr;
    my ($p_len, $s_len) = (length $pat, length $str);

    my $template = ("A$p_len" . 'X' . ($p_len - 1)) x ($s_len - $p_len
+ + 1);
    $substr{$_}++ for unpack $template, $str;
    return $substr{$pat};
}