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


Algorithm::MarkovChain - Object oriented Markov chain generator


  use Algorithm::MarkovChain;

  my $chain = Algorithm::MarkovChain::->new();

  # learn about things from @symbols
  $chain->seed(symbols => \@symbols,
               longest => 6);

  # attempt to tell me something about the sky
  my @newness = $chain->spew(length   => 20,
                             complete => [ qw( the sky is ) ]);


Algorithm::MarkovChain is an implementation of the Markov Chain algorithm within an object container.

It is implemented as a base class, Algorithm::MarkovChain::Base, with storage implementations of a hash (Algorithm::MarkovChain), and an fairly memory efficent implementation using glib (Algorithm::MarkovChain::GHash). DBI and MLDBM-friendly versions are planned.

Deriving alternate representations is intended to be straightforward.


Algorithm::MarkovChain::->new() or $obj->new()

Creates a new instance of the Algorithm::MarkovChain class.

Takes one optional parameter: recover_symbols

recover_symbols has meaning if your symbols differ from their true values when stringifyed. With this option enabled steps are taken to ensure that the original values for symbols are returned by the spew method.


Seeds the markov chains from an example symbol stream.

Takes two parameters, one required symbols, one optional longest

symbols presents the symbols to seed from

longest sets an upper limit on the longest chain to construct. (defaults to 4)


Uses the constructed chains to produce symbol streams

Takes four optional parameters complete, length, longest_subchain, force_length, stop_at_terminal and strict_start

complete provides a starting point for the generation of output. Note: the algorithm will discard elements of this list if it does not find a starting chain that matches it, this is infinite-loop avoidance.

length specifies the minimum number of symbols desired (default is 30)

stop_at_terminal directs the spew to stop chaining at the first terminal point reached

force_length ensures you get exactly length symbols returned (note this overrides the behaviour of stop_at_terminal)

strict_start makes the spew operation always take a known start state rather than selecting a sequence at random

$obj->increment_seen($sequence, $symbol)

Increments the seeness of a symbol following a sequence.


Recompute the probabilities for a branch of the tree. Called towards the end of the seed operation for 'dirty' sequences.


Returns possible next symbols and probablities as a hash.



I need to explain Markov Chains, and flesh out the examples some more.

Fix bugs/respond to feature requests

Just email me <> and I'll hit it with hammers...


Hopefully not, though if they probably arise from my not understanding Markov chaining as well as I thought I did when coding commenced.

That or they're jst stupid mistakes :)


Richard Clamp <>



2 POD Errors

The following errors were encountered while parsing the POD:

Around line 180:

You forgot a '=back' before '=head2'

Around line 184:

=back without =over