Wow, is that ugly. I frequently have a difficult time with nested parens in regexes, so I wrote a quick test script to use this regex to figure out where $1, $2 etc would go:

#!/usr/bin/perl -w

$string="foobar";
if ($string =~ /(f(o(o(b(a(r)?)?)?)?)?)$/) {
  print "$1\n";
  print "$2\n";
  print "$3\n";
}
[download]

There is nothing really earth-shattering in the results:

foobar
oobar
obar
[download]

So essentially, this will do very much what I suppose John wants to do, as the most complete match goes to $1.

Very nice, Sifmole, but I bet this tends to be a pretty slow regex.

Scott