Test::BDD::Cucumber::Manual::Architecture - Structural Overview


version 0.83


This short document exists to give you an idea how the different components of this distribution fit together. Components fall into three categories: one for dealing with definition of the step functions (Perl code), the other for dealing with feature files. And then there's the third group, with Test::BDD::Cucumber::Executor which links the two groups together; building on the steps to execute the features.

In the first group are Test::BDD::Cucumber::StepFile and Test::BDD::Cucumber::Loader. The second group is bigger and is comprised of

  • Test::BDD::Cucumber::Parser

  • Test::BDD::Cucumber::Model::*

The third group holds - next to Test::BDD::Cucumber::Executor:

  • Test::BDD::Cucumber::Harness::*

  • Test::BDD::Cucumber::StepContext

  • Test::BDD::Cucumber::Errors

  • Test::BDD::Cucumber::Extension

Please note that the Test::BDD::Cucumber::Harness should not be confused with TAP::Harness or Test2::Harness; this harness is TBC's own and optionally forwards events to the Test::Builder::Test harness.


The core of a Cucumber-based test suite are the feature files and the step definitions files. By convention, these are saved under /features/ and /features/step_definitions/ respectively.

The feature files are encapsulated by the classes in Test::BDD::Cucumber::Model.

                  one to one
      |                               |
      +-------------------+           |
      | has many          | has a     | has many
      V                   |           V
 TBCM::Scenario           +----->TBCM::Line
      |                            ^  ^
      +----------------------------+  |
      | has many                      |
      V                               |


We build up a Test::BDD::Cucumber::Executor object, in to which we load the step definitions. We then pass this in a Test::BDD::Cucumber::Model::Feature object, along with a Test::BDD::Cucumber::Harness object, which controls interaction with the outside world.


While running step functions, Test::BDD::Cucumber::Executor reroutes the flow of test events (calls to ok, fail, etc) to itself. Based on the collected data, the step itself is reported as a success or failure to the test driver.

Confusing about this situation is that both the channel to report through to the actual test driver is an instance of Test::Builder as well as the method used to route the stream of test events to itself uses a Test::Builder instance.


Extensions allow hooking into the execution of the steps, with pre- and post hooks for steps, scenarios, features and the entire execution. Extensions can provide additional step directories from which steps will be made available. The feature and scenario stashes are passed to the extension hooks allowing for a means of communication between the hooks and the steps.

Additionally, as of 0.60, it's possible to define meta data for a step upon step definition like this example:

  Given qr/a step with meta data/ => { meta => 'data' }, sub { };

This allows step function authors to communicate with extensions, because extensions receive this meta data in their pre_step callback:

  sub pre_step {
    my ($stepdef, $step_context) = @_;
    my ($verb, $metadata, $stepfn) = @$stepdef;

    die 'All steps should have meta -> data'
      unless $metadata->{meta} eq 'data';

Extensions - when loaded by the pherkin test executor - receive their configuration from the pherkin.yaml configuration file, which works similar to the configuration of extensions in Behat.

Note: when using extensions in combination with the TAP::Parser::SourceHandler::Feature plugin for prove, there is no guarantee that the pre_execute and post_execute hooks execute exactly once or even execute at all. This is a current limitation to be lifted in a future release.


Peter Sergeant


  Copyright 2019-2021, Erik Huelsmann
  Copyright 2011-2019, Peter Sergeant; Licensed under the same terms as Perl