http://qs321.pair.com?node_id=476580

Limbic~Region has asked for the wisdom of the Perl Monks concerning the following question:

All,
Intrepid asked in the CB for a piece of code that would satisfy the following requirements:

These have been generalized a bit as hashref keys were involved, but I provided the following straightforward solution:

my @min_keys = sort { length($a) <=> length($b) || $a cmp $b } keys %$ +hashref; my $min = length $min_keys[0]; for ( 1 .. $#min_keys ) { if ( length $min_keys[$_] != $min ) { splice @min_keys, $_; last; } } # maxstr & random print join "\t", $min_keys[-1], $min_keys[ rand @min_keys ];
There was a follow on discussion about how this could be done efficiently. While some were interested in a general solution that may be incorporated into List::Util, I felt a variation of the watermark algorithm was all that was needed to satisfy all the requirements in a single pass:
my @list = qw(zero one two three four five six seven eight nine ten el +even twelve); my ($maxstr, $random) = find_shortest( @list ); print join "\t", $maxstr, $random; sub find_shortest { my ($pos, $len, $maxstr, @min) = (0, undef, '', 0, undef); my %dispatch = ( -1 => sub { ($len, $pos) = (length($_), 0); ($min[$pos], $maxstr) = ($_, $_); }, 0 => sub { $min[++$pos] = $_; $maxstr = $_ if $_ gt $maxstr; }, 1 => sub { return }, ); for ( @_ ) { if ( ! defined $len ) { $len = length(); ($min[$pos], $maxstr) = ($_, $_); next; } $dispatch{ length($_) <=> $len }->(); } return ($maxstr, (@min[0..$pos])[rand $pos]); }

There are still ways to improve on it, but I thought it was neat and only required a minimal amount of extra effort. How could you improve (keeping the original 3 requirements in mind)?
Cheers - L~R