I have a list of IP addresses that I want sorted in a human readable fashion. A simple sort() on a list of IPs will not work because the octets may be: one, two, or three digits long which confuses sort(). Here is a simple sorting function for a list of IP addresses:

my @ips    = qw(198.15.0.20 4.2.2.1 10.11.1.1 10.100.1.1 65.182.224.40);
my @sorted = sort by_ip_address @ips;

sub by_ip_address {
    return ip2long($a) <=> ip2long($b);
}

Note: You will need my ip2long() function for this to work.

If you want to sort an array naturally (the way a human would) you can use Perl's sort() function, but use a custom sort method:

my @input  = qw(foo foo250 foo12 foo23 bar999 bar7 bar17 bar99 18);
my @sorted = sort { natural(); } @input;

print join(", ", @sorted);

sub natural {
    # Separate the word and numeric parts
    my ($word_a, $num_a) = $a =~ /(.*?)(\d+|$)/;
    my ($word_b, $num_b) = $b =~ /(.*?)(\d+|$)/;

    #print "$a / $b: $word_a, $num_a, $word_b, $num_b\n";

    # If the words are diff it's an alpha sort on the words
    if ($word_a ne $word_b) {
        return $word_a cmp $word_b;
    # Words are the same, numeric sort the number part
    } else {
        return ($num_a || 0) <=> ($num_b || 0);
    }
}

See also: Natural Sort

I have a array with a bunch of names like vlan-1, vlan100, vlan34 which do not sort appropriately using Perl's standard sort() function. Sort::Naturally to the rescue! I didn't want to install an entire module for one sort operation, and require a dependency, so I ripped out just the natural sort function and included that in my script.

sub nsort {
    my($cmp, $lc);
    return @_ if @_ < 2;   # Just to be CLEVER.

    my($x, $i);  # scratch vars

    map
        $_->[0],

    sort {
        # Uses $i as the index variable, $x as the result.
        $x = 0;
        $i = 1;

        while($i < @$a and $i < @$b) {
            last if ($x = ($a->[$i] cmp $b->[$i])); # lexicographic
            ++$i;

            last if ($x = ($a->[$i] <=> $b->[$i])); # numeric
            ++$i;
        }

        $x || (@$a <=> @$b) || ($a->[0] cmp $b->[0]);
    }

    map {
        my @bit = ($x = defined($_) ? $_ : '');

        if($x =~ m/^[+-]?(?=\d|\.\d)\d*(?:\.\d*)?(?:[Ee](?:[+-]?\d+))?\z/s) {
            # It's entirely purely numeric, so treat it specially:
            push @bit, '', $x;
        } else {
            # Consume the string.
            while(length $x) {
                push @bit, ($x =~ s/^(\D+)//s) ? lc($1) : '';
                push @bit, ($x =~ s/^(\d+)//s) ?    $1  :  0;
            }
        }

        \@bit;
    }
    @_;
}

This is a slightly more portable version rather than maintaining the Sort::Naturally dependency.

See also: Natural Sort Part Deux

This syntax still doesn't make much sense to me but here is how you sort a perl hash by value. This returns a list of all the keys of the hash sorted in the order you want. To reverse the sort simply change $a and $b locations with each other.

my @sort = sort{ $unique{$a} <=> $unique{$b} } keys %unique;