++ed by:

10 PAUSE users
4 non-PAUSE users.

Curtis "Ovid" Poe


Test::Class::Moose - Serious testing for serious Perl


version 0.54


    package TestsFor::DateTime;
    use Test::Class::Moose;
    use DateTime;

    # methods that begin with test_ are test methods.
    sub test_constructor {
        my $test = shift;
        $test->test_report->plan(3);    # strictly optional

        can_ok 'DateTime', 'new';
        my %args = (
            year  => 1967,
            month => 6,
            day   => 20,
        isa_ok my $date = DateTime->new(%args), 'DateTime';
        is $date->year, $args{year}, '... and the year should be correct';



See Test::Class::Moose hompage for a summary.

`Test::Class::Moose` is a powerful testing framework for Perl. Out of the box you get:

  • Reporting

  • Extensibility

  • Tagging tests

  • Parallel testing

  • Test inheritance

  • Write your tests using Moose

  • All the testing functions and behavior from Test::Most

  • Event handlers for startup, setup, teardown, and shutdown of test classes

Better docs will come later. You should already know how to use Moose and Test::Class.


Inheriting from Test::Class::Moose

Just use Test::Class::Moose. That's all. You'll get all Test::Most test functions, too, along with strict and warnings. You can use all Moose behavior, too.

Declare a test method

All method names that begin with test_ are test methods. Methods that do not are not test methods.

 sub test_this_is_a_method {
     my $test = shift;

     ok 1, 'whee!';

 sub this_is_not_a_test_method {
    my $test = shift;
    # but you can, of course, call it like normal

You may specify Test and Tests method attributes, just like in Test::Class and the method will automatically be a test method, even if does not start with test_:

    sub this_is_a_test : Test {
        pass 'we have a single test';

    sub another_test_method : Tests { # like "no_plan"
        # a bunch of tests

    sub yet_another_test_method : Tests(7) { # sets plan to 7 tests

Note: Prior to version 0.51, this feature only worked if you had the optional Sub::Attribute installed.


No plans needed. The test suite declares a plan of the number of test classes.

Each test class is a subtest declaring a plan of the number of test methods.

Each test method relies on an implicit done_testing call.

If you prefer, you can declare a plan in a test method:

    sub test_something {
        my $test = shift;

Or with a Tests attribute:

    sub test_something : Tests(3) {
        my $test = shift;

You may call plan() multiple times for a given test method. Each call to plan() will add that number of tests to the plan. For example, with a method modifier:

    before 'test_something' => sub {
        my $test = shift;

        # more tests

Please note that if you call plan, the plan will still show up at the end of the subtest run, but you'll get the desired failure if the number of tests run does not match the plan.

Inheriting from another Test::Class::Moose class

List it as the extends in the import list.

 package TestsFor::Some::Class::Subclass;
 use Test::Class::Moose extends => 'TestsFor::Some::Class';

 sub test_me {
     my $test  = shift;
     my $class = $test->test_class;
     ok 1, "I overrode my parent! ($class)";

 before 'test_this_baby' => sub {
     my $test  = shift;
     my $class = $test->test_class;
     pass "This should run before my parent method ($class)";

 sub this_should_not_run {
     my $test = shift;
     fail "We should never see this test";

 sub test_this_should_be_run {
     for ( 1 .. 5 ) {
         pass "This is test number $_ in this method";



Do not run tests in test control methods. This will cause the test control method to fail (this is a feature, not a bug). If a test control method fails, the class/method will fail and testing for that class should stop.

Every test control method will be passed two arguments. The first is the $test invocant. The second is an object implementing Test::Class::Moose::Role::Reporting. You may find that the notes hashref is a handy way of recording information you later wish to use if you call $test_suite->test_report.

These are:

  • test_startup

     sub test_startup {
        my $test = shift;
        # more startup

    Runs at the start of each test class. If you need to know the name of the class you're running this in (though usually you shouldn't), use $test->test_class, or you can do this:

        sub test_startup {
            my $test                 = shift;
            my $report               = $test->test_report;
            my $class                = $report->current_class->name;
            my $upcoming_test_method = $report->current_method->name;

    The $test->test_report object is a Test::Class::Moose::Report::Class object.

  • test_setup

     sub test_setup {
        my $test = shift;
        # more setup

    Runs at the start of each test method. If you must know the name of the test you're about to run, you can do this:

     sub test_setup {
        my $test = shift;
        my $test_method = $test->test_report->current_method->name;
        # do something with it
  • test_teardown

     sub test_teardown {
        my $test = shift;
        # more teardown

    Runs at the end of each test method.

  • test_shutdown

     sub test_shutdown {
         my $test = shift;
         # more teardown

    Runs at the end of each test class.

To override a test control method, just remember that this is OO:

 sub test_setup {
     my $test = shift;
     $test->next::method; # optional to call parent test_setup
     # more setup code here


We strongly recommend using Test::Class::Moose::Load as the driver for your test suite. Simply point it at the directory or directories containing your test classes:

 use Test::Class::Moose::Load 't/lib';

By running Test::Class::Moose with a single driver script like this, all classes are loaded once and this can be a significant performance boost. This does mean a global state will be shared, so keep this in mind.

You can also pass arguments to Test::Class::Moose's contructor.

 my $test_suite = My::Base::Class->new({
     show_timing => 1,
     randomize   => 0,
     statistics  => 1,
 # do something

The attributes passed in the constructor are not directly available from the Test::Class::Moose instance. They're available in Test::Class::Moose::Config and to avoid namespace pollution, we do not delegate the attributes directly as a result. If you need them at runtime, you'll need to access the test_configuration attribute:

 my $builder = $test_suite->test_configuration->builder;

Note that you can call Test::Class::Moose->new instead of My::Base::Class->new, but we recommend that you instantiate an instance of your base class instead of Test::Class::Moose. There are times when you may apply a role to your base class and modify it, but running it in the context of Test::Class::Moose will not always pick up those modifications.

In other words, create an instance of your base class, not Test::Class::Moose.

Contructor Attributes

  • show_timing

    Boolean. Will display verbose information on the amount of time it takes each test class/test method to run. Defaults to false, but see use_environment.

  • statistics

    Boolean. Will display number of classes, test methods and tests run. Defaults to false, but see use_environment.

  • use_environment

    If this is true, then the default value for show_timing and statistics will be true if the HARNESS_IS_VERBOSE environment variable is true. This is set when running prove -v ..., for example.

  • randomize

    Boolean. Will run test methods in a random order.

  • builder

    Defaults to Test::Builder->new. You can supply your own builder if you want, but it must conform to the Test::Builder interface. We make no guarantees about which part of the interface it needs.

  • test_classes

    Takes a class name or an array reference of class names. If it is present, only these test classes will be run. This is very useful if you wish to run an individual class as a test:

            test_classes => $ENV{TEST_CLASS}, # ignored if undef

    You can also achieve this effect by writing a subclass and overriding the test_classes method, but this makes it trivial to do this:

        TEST_CLASS=TestsFor::Our::Company::Invoice prove -lv t/test_classes.t


            test_classes => \@ARGV, # ignored if empty

    That lets you use the arisdottle to provide arguments to your test driver script:

        prove -lv t/test_classes.t :: TestsFor::Our::Company::Invoice TestsFor::Something::Else
  • include

    Regex. If present, only test methods whose name matches include will be included. However, they must still start with test_.

    For example:

     my $test_suite = Test::Class::Moose->new({
         include => qr/customer/,

    The above constructor will let you match test methods named test_customer and test_customer_account, but will not suddenly match a method named default_customer.

    By enforcing the leading test_ behavior, we don't surprise developers who are trying to figure out why default_customer is being run as a test. This means an include such as /^customer.*/ will never run any tests.

  • exclude

    Regex. If present, only test methods whose names don't match exclude will be included. However, they must still start with test_. See include.

  • include_tags

    Array ref of strings matching method tags (a single string is also ok). If present, only test methods whose tags match include_tags or whose tags don't match exclude_tags will be included. However, they must still start with test_.

    For example:

     my $test_suite = Test::Class::Moose->new({
         include_tags => [qw/api database/],

    The above constructor will only run tests tagged with api or database.

  • exclude_tags

    The same as include_tags, but will exclude the tests rather than include them. For example, if your network is down:

     my $test_suite = Test::Class::Moose->new({
         exclude_tags => [ 'network' ],
     # or
     my $test_suite = Test::Class::Moose->new({
         exclude_tags => 'network',

Skipping Classes and Methods

If you wish to skip a class, set the reason in the test_startup method.

    sub test_startup {
        my $test = shift;
        $test->test_skip("I don't want to run this class");

If you wish to skip an individual method, do so in the test_setup method.

    sub test_setup {
        my $test = shift;
        my $test_method = $test->test_report->current_method;
        if ( 'test_time_travel' eq $test_method->name ) {
            $test->test_skip("Time travel not yet available");

The "Tests" and "Test" Attributes

If you're comfortable with Test::Class, know test methods methods are declared in Test::Class with Test (for a method with a single test) or Tests, for a method with multiple tests. This also works for Test::Class::Moose. Test methods declared this way do not need to start with test_.

    sub something_we_want_to_check : Test {
        # this method may have only one test

    sub something_else_to_check : Tests {
        # this method may have multiple tests

    sub another_test_method : Tests(3) {
        # this method must have exactly 3 tests

If a test method overrides a parent test method and calls it, their plans will be added together:

    package TestsFor::Parent;

    use Test::Class::Moose;

    sub some_test : Tests(3) {
        # three tests

And later:

    package TestsFor::Child;

    use Test::Class::Moose extends => 'TestsFor::Parent';

    sub some_test : Tests(2) {
        my $test = shift;
        # 2 tests here

In the above example, TestsFor::Parent::some_test will run three tests, but TestsFor::Child::some_test will run five tests (two tests, plus the three from the parent).

Note that if a plan is explicitly declared, any modifiers or overriding methods calling the original method will also have to assert the number of tests to ensure the plan is correct. The above TestsFor::Parent and TestsFor::Child code would fail if the child's some_test method attribute was Tests without the number of tests asserted.

Do not use Test or Tests with test control methods becase you don't run tests in those.

Tagging Methods

Sometimes you want to be able to assign metadata to help you better manage your test suite. You can do this with tags:

    sub test_save_poll_data : Tags(api network) {

Tags are strictly optional and you can provide one or more tags for each test method with a space separated list of tags. You can use this to filter your tests suite, if desired. For example, if your network goes down and all tests which rely on a network are tagged with network, you can skip those tests with this:

    My::Base::Class->new( exclude_tags => 'network' )->runtests;

Or maybe you want to run all api and database tests, but skip those marked deprecated:

        include_tags => [qw/api database/],
        exclude_tags => 'deprecated',

You can also inspect tags withing your test classes:

    sub test_setup {
        my $test          = shift;
        my $method_to_run = $test->test_report->current_method;
        if ( $method_to_run->has_tag('db') ) {

Tagging support relies on Sub::Attribute. If this module is not available, include_tags and exclude_tags will be ignored, but a warning will be issued if those are seen. Prior to version 0.51, Sub::Attribute was optional. Now it's mandatory, so those features should always work.


If you want to run the tests in parallel, see the experimental Test::Class::Moose::Role::Parallel role. Read the documentation carefully as it can take a while to understand. You only need to use the role and (optionally) provide a schedule() method. Any tests tagged with noparallel will be run sequentially after the parallel tests (unless you provide your own schedule, in which case you can do anything you want).


... but probably shouldn't.

As a general rule, methods beginning with /^test_/ are reserved for Test::Class::Moose. This makes it easier to remember what you can and cannot override. However, any test with Test or Tests are test methods regardless of their names.


 my $test_configuration = $test->test_configuration;

Returns the Test::Class::Moose::Config object.


 my $report = $test->test_report;

Returns the Test::Class::Moose::Report object. Useful if you want to do your own reporting and not rely on the default output provided with the statistics boolean option.

You can also call it in test classes (most useful in the test_setup() method):

    sub test_setup {
        my $test = shift;
        my $report= $test->test_report;
        my $class = $test->current_class;
        my $method = $test->current_method; # the test method we're about to run
        if ( $method->name =~ /customer/ ) {
        # or better still
        if ( $method->has_tag('customer') ) {


 my $class = $test->test_class;

Returns the name for this test class. Useful if you rebless an object (such as applying a role at runtime) and don't want to lose the original class name.


You may override this in a subclass. Currently returns a sorted list of all loaded classes that inherit directly or indirectly through Test::Class::Moose


You may override this in a subclass. Currently returns all methods in a test class that start with test_ (except for the test control methods).

Please note that the behavior for include and exclude is also contained in this method. If you override it, you will need to account for those yourself.


If you really, really want to change how this module works, you can override the runtests method. We don't recommend it.

Returns the Test::Class::Moose instance.


Sadly, we have an import method. This is used to automatically provide you with all of the Test::Most behavior.


We use nested tests (subtests) at each level:

    # Executing tests for TestsFor::Basic::Subclass
        # TestsFor::Basic::Subclass->test_me()
            ok 1 - I overrode my parent! (TestsFor::Basic::Subclass)
        ok 1 - test_me
        # TestsFor::Basic::Subclass->test_this_baby()
            ok 1 - This should run before my parent method (TestsFor::Basic::Subclass)
            ok 2 - whee! (TestsFor::Basic::Subclass)
        ok 2 - test_this_baby
        # TestsFor::Basic::Subclass->test_this_should_be_run()
            ok 1 - This is test number 1 in this method
            ok 2 - This is test number 2 in this method
            ok 3 - This is test number 3 in this method
            ok 4 - This is test number 4 in this method
            ok 5 - This is test number 5 in this method
        ok 3 - test_this_should_be_run
    ok 1 - TestsFor::Basic::Subclass
    # Executing tests for TestsFor::Basic
        # TestsFor::Basic->test_me()
            ok 1 - test_me() ran (TestsFor::Basic)
            ok 2 - this is another test (TestsFor::Basic)
        ok 1 - test_me
        # TestsFor::Basic->test_this_baby()
            ok 1 - whee! (TestsFor::Basic)
        ok 2 - test_this_baby
    ok 2 - TestsFor::Basic
    # Test classes:    2
    # Test methods:    5
    # Total tests run: 11
    All tests successful.
    Files=1, Tests=2,  2 wallclock secs ( 0.03 usr  0.00 sys +  0.27 cusr  0.01 csys =  0.31 CPU)
    Result: PASS


See Test::Class::Moose::Report for more detailed information on reporting.

Reporting features are subject to change.

Sometimes you want more information about your test classes, it's time to do some reporting. Maybe you even want some tests for your reporting. If you do that, run the test suite in a subtest (because the plans will otherwise be wrong).

    #!/usr/bin/env perl
    use lib 'lib';
    use Test::Most;
    use Test::Class::Moose::Load qw(t/lib);
    my $test_suite = My::Base::Class->new;

    subtest 'run the test suite' => sub {
    my $report = $test_suite->test_report;

    foreach my $class ( $report->all_test_classes ) {
        my $class_name = $class->name;
        ok !$class->is_skipped, "$class_name was not skipped";

        subtest "$class_name methods" => sub {
            foreach my $method ( $class->all_test_methods ) {
                my $method_name = $method->name;
                ok !$method->is_skipped, "$method_name was not skipped";
                cmp_ok $method->num_tests, '>', 0,
                  '... and some tests should have been run';
                diag "Run time for $method_name: ".$method->time->duration;
        my $time   = $class->time;
        diag "Run time for $class_name: ".$class->time->duration;

        my $real   = $time->real;
        my $user   = $time->user;
        my $system = $time->system;
        # do with these as you will
    diag "Number of test classes: " . $report->num_test_classes;
    diag "Number of test methods: " . $report->num_test_methods;
    diag "Number of tests:        " . $report->num_tests;


If you just want to output reporting information, you do not need to run the test suite in a subtest:

    my $test_suite = My::Base::Class->new->runtests;
    my $report     = $test_suite->test_report;

Or even shorter:

    my $report = My::Base::Class->new->runtests->test_report;


If you would like Test::Class::Moose to take care of loading your classes for you, see Test::Class::Moose::Role::AutoUse in this distribution.


  • test_reporting

    As of version .40, the long deprecated method test_reporting has now been removed.

  • $report argument to methods deprecated

    Prior to version .40, you used to have a second argument to all test methods and test control methods:

        sub test_something {
            my ( $test, $report ) = @_;

    This was annoying. It was doubly annoying in test control methods in case you forgot it:

        sub test_setup {
            my ( $test, $report ) = @_;
            $test->next::method; # oops, needed $report

    That second argument is still passed, but it's deprecated. It's now recommended that you call the $test->test_report method to get that. Instead of this:

        sub test_froblinator {
            my ( $test, $report ) = @_;

    You write this:

        sub test_froblinator {
            my $test = shift;


  • Callbacks for tags (for example, 'critical' tags could bailout)

  • New test phases - start and end suite, not just start and end class/method


Please report any bugs or feature requests to bug-test-class-moose at rt.cpan.org, or through the web interface at http://rt.cpan.org/NoAuth/ReportBug.html?Queue=Test-Class-Moose. I will be notified, and then you'll automatically be notified of progress on your bug as I make changes.


You can find documentation for this module with the perldoc command.

    perldoc Test::Class::Moose

You can also look for information at:



  • Dave Rolsky <autarch@urth.org>

  • Doug Bell <doug.bell@baml.com>

  • Gregory Oschwald <goschwald@maxmind.com>

  • Jonathan C. Otsuka <djgoku@gmail.com>

  • Neil Bowers <neil@bowers.com>

  • Olaf Alders <olaf@wundersolutions.com>

  • Ovid <curtis\_ovid\_poe@yahoo.com>

  • Paul Boyd <pboyd@dev3l.net>

  • Petrea Corneliu Stefan <stefan@garage-coding.com>

  • Stuckdownawell <stuckdownawell@gmail.com>

  • Tom Beresford <tom.beresford@bskyb.com>

  • Tom Heady <tom@punch.net>

  • Udo Oji <Velti@signor.com>


Curtis "Ovid" Poe <ovid@cpan.org>


This software is copyright (c) 2014 by Curtis "Ovid" Poe.

This is free software; you can redistribute it and/or modify it under the same terms as the Perl 5 programming language system itself.