The London Perl and Raku Workshop takes place on 26th Oct 2024. If your company depends on Perl, please consider sponsoring and/or attending.


Regexp::Genex - get the strings a regex will match, with a regex


 # first try:
 $ perl -MRegexp::Genex=:all -le 'print for strings(qr/a(b|c)d{2,3}e*/)'

 $ perl -x `pmpath Regexp::Genex`
#!/usr/bin/perl -l

 use Regexp::Genex qw(:all);

 $regex = shift || "a(b|c)d{2,4}?";

 print "Trying: $regex";
 print for strings($regex);
 # abdd
 # abddd
 # abdddd
 # acdd
 # acddd
 # acdddd

 print "\nThe regex code for that was:\nqr/";
 print strings_rx($regex);
 print "/x\n";

 my $generator = generator($regex);
 print "Taking first two using generator";
 print $generator->() for 1..2;

 my $big_rx = 'b*?c*?d*?';   # * becomes {0,20}

 my $big = generator($big_rx, ($max_length = 100) );

 print "Taking string 100 of $big_rx";
 print $big->(100); # (caveats below)
 # ccccdddddddddddddddd   NOT 'd'x100 as you may expect



This is alpha code that relies on experimental features of perl (regex (?{ }) and friends) and avoiding optimizations in the regex engine. New optimizations could break this module.

The interface is also quite likely to change.


This module uses the regex engine to generate the strings that a given regex would match.

Some ideas for uses:

  Test and debug your regex.
  Generate test data.
  Generate combinations.
  Generate data according to a lexical pattern (urls, etc)
  Edit the regex code to do your things (eg. add assertions)
  Generate strings, reverse & alternate for pseudo-variable look behind


Nothing by default, everything with the :all tag.

@list = strings( $regex, [ $max_length = 10 ] )

Produce a list of strings that would match the regex.

$regex_string = strings_rx( $regex )

Returns the regex string used to implement the above. You'll need to use re 'eval' for this and maybe no warnings 'regexp'

$generator = generator( $regex, [ $max_length = 10 ] )

Return a closure to access the strings one at a time.

Calling $generator->() will return the next string (starting from 0). Calling $generator->($n) will reset the iterator to string $n and return it.

$regex_string = generator_rx( $regex )

Returns the regex string used to implement the above. You'll need to use re 'eval' for this and maybe no warnings 'regexp'

Gx Package

Small package which is not installed by default, nor officially approved as a namespace. It's not part of the public interface, don't use it in modules. is just a short cut to import Regexp::Genex qw(:all) mainly useful from the command line:

 perl -MGx -le 'print for strings(qr/a(b|c){2,4}/);'


Many regex elements such as anchors (^ $ \A \G), look ahead, look-behind, code elements and conditionals are not implemented. Some may be in the future. I'm considering making a pattern not wrapped in ^ $ generate leading and trailing junk. Look-ahead inparticular, is unlikely to ever get implemented. Perhaps for finite languages.

Regex elements which could match a number of things such as . [class] \w \s \D currently select a few items from the set of possibilities and the randomly select one at runtime. So . may become ("~","`","\307","9","\266")[rand 5]. The rand call is only repeat if the element is backtracked over. Try these a few times:

 perl -MRegexp::Genex=:all -e 'print strings_rx(qr/\d\w/);'
 perl -MRegexp::Genex=:all -le 'print for strings(qr/\d\w/);'
 perl -MRegexp::Genex=:all -le 'print for strings(qr/\d{1,2}\t\w{1,2}/);'

If you pick apart the generated expression you'll note that the quantifier * translates to {0,20} (+ to {1,20}). This can be set (but don't tell ayone it was me that told you) with $Regexp::Genex::MAX_QUANTIFIER. 32767 is what perl uses. MAX_QUANTIFIER keeps string generation to smaller sizes.

The generator actually has to replay the match up to where it was in order to get the next one. Pretty inefficient but I can't suspend/yield from within the regex. Best way forward might be to fork and use pipes for lazy generation.

The /ismx mode handling is probably not all it could be, 'x' isn't very relevant, 'm' relates to unimplemented anchors, 'i' will mess with the case of you text items and 's' mean dot might produce newlines.


 perl -MRegexp::Genex=:all -e 'print strings_rx(qr/aBc/i);'
 perl -MRegexp::Genex=:all -le 'print for strings(qr/aBc/i);'

Currently, a small patch is required to YAPE::Regex to get this module to work correctly, see the end of this file. Hopefully it will be fixed soon (vers currently 3.01)


  keep funky state in %_
  work out a good max_length
  dynamically select chars in classes
  unimplemented: anchors, lookbehind, code

  testing code
  could upload with patch
  note modifiers in effect in comment


Brad Bowman, <>


YAPE::Regex String::Random