http://qs321.pair.com?node_id=358966

@sorted = map { s/^\S+ // } sort map { my ( $x ) = ( /^_(\S+)/ ) ? ( $1 ) : ( /(\S+)$/ ); "$x $_" } @unsorted;
Explanation (going upwards from the bottom, which is how it works):
  1. The first map block (the one on the bottom) creates a lexically scoped variable $x, and assigns a single non-whitespace token (string) to it. The value of $x depends on whether $_ (the current string from @unprocessed that map is handling) begins with "_". If it does, $x gets the first token of the line (minus the "_"), otherwise, it gets the last token.
  2. Each iteration of the first map outputs the string "$x $_"
  3. The sort works in the normal way, working on the modified list of strings
  4. The second map block (at the top) reprocesses the sorted list of strings, by undoing the work of the first one, by removing the initial token from each item.

An adaptation of a code snippet from LadyAleena's scratchpad:

elsif ($type =~ /name/) { for ( $c, $d ) { my $x = ( /^_(.*)/ ) ? $1 : substr( $_, rindex( $_, ' ' )) +; $_ = "$x $_"; } return $c cmp $d; }