That would require travelling the regex node tree. It doesn't seem very possible, even with reading the results of re.pm output... I've thought about making a text-to-regex translator, but I stop everytime I think about it, since that would require parsing Perl in Perl, which is not a very nice idea. With simple regexes like yours, though, you can hack it together yourself. I honestly have no idea why I wrote this. It is far too much for such a simple problem.

package MatchRegex;

use strict;
use vars qw( @ISA @EXPORT_OK @digits @alpha );

require Exporter;
@ISA = 'Exporter';
@EXPORT_OK = qw( digits alpha UCalpha LCalpha );

@digits = (0..9);
@alpha = ('a'..'z', 'A'..'Z');

sub digits {
  ref $_[0] and local @digits = @{shift()};  # could use my()
  my $len;
  if (!@_) { $len = 1 }
  elsif (@_ == 1) { $len = $_[0] }
  else { $len = $_[0] + rand($_[1] + 1 - $_[0]) }
  return map $digits[rand @digits], 1 .. $len;
}

sub LCalpha { local @alpha = @alpha[0..25]; α }

sub UCalpha { local @alpha = @alpha[26..51]; α }

sub alpha {
  ref $_[0] and local @alpha = @{shift()};  # could use my()
  my $len;
  if (!@_) { $len = 1 }
  elsif (@_ == 1) { $len = $_[0] }
  else { $len = $_[0] + rand($_[1] + 1 - $_[0]) }
  return map $alpha[rand @alpha], 1 .. $len;
}
[download]

And then:

use MatchRegex qw( digits UCalpha );

$filename = join "", (
  digits(2),     # \d{2}
  UCalpha(2,5),  # [A-Z]{2,5}
  '_',           # _
  digits(2),     # \d{2}
  digits(2),     # \d{2}
  digits(4),     # \d{4}
);
[download]

This is a glorified "make a random string as per my request" module. It can be expanded. But it can't interpret your regex for you. That's left up to you. Maybe MJD or Ilya or someone else could help hack at the regex engine to make something like this more likely.

$_="goto+F.print+chop;\n=yhpaj";F1:eval