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

use List::Util qw(first);
use Benchmark;

my %test;
for my $j (1970..2020) {
    for my $m (1..12) {
        for my $t (1..31) {
            my $d = sprintf '%02d.%02d.%d', $t, $m, $j;
            $test{$d}++;
        }
    }
}

my %found;

my $m_grep = sub {
    foreach my $m (1..12) {
        if (grep { /\.0?$m\./ } keys %test) {
            $found{$m} ||= 1;
        }
    }
};

my $m_first = sub {
    foreach my $m (1..12) {
        if (first { /\.0?$m\./ } keys %test) {
            $found{$m} ||= 1;
        }
    }
};

my $m_pq = sub {
    my %notfound;
    @notfound{1..12} = ();
    my $re = join "|", keys %notfound;
    for my $key (keys %test) {
        if ($key =~ m/\.0?($re)\./) {
            my $month = $1 + 0;
            $found{$month} ||= 1;
            delete $notfound{$month};
            $re = join "|", keys %notfound;
            # wenn alle gefunden, braucht man gar nicht mehr suchen
            last unless keys %notfound;
        }
    }
};

timethese(100, {
    'grep'  => $m_grep,
    'first' => $m_first,
    'pq'    => $m_pq
});

Benchmark: timing 100 iterations of first, grep, pq...
     first: 13 wallclock secs (13.55 usr +  0.00 sys = 13.55 CPU) @ 7.38/s (n=100)
      grep: 59 wallclock secs (58.09 usr +  0.17 sys = 58.27 CPU) @ 1.72/s (n=100)
        pq:  1 wallclock secs ( 0.86 usr +  0.00 sys =  0.86 CPU) @ 116.41/s (n=100)