NAME

Test::Shadow - override a class's methods in a scope, checking input/output

SYNOPSIS

Provides RSpec-like mocking with 'receive'/'and_return' functionality. However the interface is more explicit. This may be considered a feature.

    use Test::More;
    use Test::Shadow;

    use Foo;

    with_shadow Foo => inner_method => {
        in => [ 'list', 'of', 'parameters' ],
        out => 'barry',
        count => 3
    }, sub {
        my $foo = Foo->new;
        $foo->outer_method();
    };

EXPORTED FUNCTIONS

with_shadow

Exported by default

    with_shadow $class1 => $method1 => $args1, ..., $callback;

Each supplied class/method is overridden as per the specification in the supplied args. Finally, the callback is run with that specification.

The args passed are as follows:

in

A list of parameters to compare every call of the method against. This will be checked each time, until the first failure, if any. The parameters can be supplied as an arrayref:

    in => [ 'list', 'of', 'parameters' ]

or a hashref:

    in => { key => 'value', key2 => 'value2 },

and the comparison may be made using any of the extended routines in Test::Deep

    use Test::Deep;
    with_shadow Foo => inner_method => {
        in => { foo => any(1,2,3) },
        ...
out

Stub the return value. This can be

  • a simple (non-reference) scalar value

        ...
        out => 100,
  • a subroutine ref, which will be passed at every invocation the parameters ($orig, $self, @args).

Note that the subroutine args are the same as if you were creating a Moose or Class::Method::Modifiers around wrapper, but dynamically scoped to the test.

    out => sub { my ($orig, $self, @args) = @_; ... },

If you want to return a reference (including a subroutine reference) return this from the subroutine: We require wrapping in a subroutine ref for the same reason that Moose's default does: otherwise we would end up passing the same reference to each invocation, with possibly surprising results.

    out => sub { [] }, # return a new, empty arrayref on each invocation

Of course you can simply ignore the call args and invoke as a subroutine. See also the "iterate" function.

count

The number of times you expect the method to be called. This is checked at the end of the callback scope.

This may be an exact value:

    count => 4,

Or a hashref with one or both of min and max declared:

    count => { min => 5, max => 10 },

iterate

We provide a helper function to iterate over a number of scalar return values. This can be attached to out, and takes a list of values to be provided as the stubbed return value on each successive call.

    use Test::Shadow 'iterate';

    with_shadow ...
        out => iterate(1,2,3,4), # return 1 on first invocation, 2 on second, etc.
        ...

The values wrap if they run out: you may want to use a count argument to diagnose that this has happened.

As well as simple values, iterate handles method calls in exactly the same format as they are normally passed to out.

    with_shadow ...
        out => iterate(
            sub { my ($orig, $self, $arg) = @_; ... },
            ...

SEE ALSO

There are several other modules that deal with mocking objects. One of them may well serve your needs better. I was having RSpec envy, about the call expectation side of things (not about the "English-like" DSL, which I found both confusing, and slightly filthy) so Test::Shadow is designed to cover that use case with an API that is less magical and more Perlish (thanks to ribasushi, haarg, tobyink, vincent, ether on #perl-qa for pointing out that my first implementation with the lovely-but-frightening Scope::Upper may not have been the poster child for sanity I'd intended.)

  • Test::MockObject is the oldest CPAN library I'm aware of. It has a very different usage, where you create an object instance and stub methods on it, rather than mocking a class.

  • Test::MockModule does mock a class's methods, but hasn't been updated since 2005, and doesn't give the control over return value stubbing and call count tracing.

  • Mock::Quick looks like a more modern mocking implementation. Again, it looks like this works on an object instance.

  • Test::Spec looks like a good reimplementation of RSpec, which means that personally I dislike aspects of the API -- the monkey-patching and the confusing expects and returns keywords, but this may be a good choice. Note that the ::Mocks routines are "currently only usable from within tests built with the Test::Spec BDD* framework".

* my current (snarky) understanding is that "BDD" means something to do with using it and describe as synonyms for subtest.

AUTHOR and LICENSE

Copyright 2014 Hakim Cassimally <osfameron@cpan.org>

This module is released under the same terms as Perl.