pytest-when

Inspired by mokito-scala, pytest-when provides a fixture for pytest to simplify the mocking of python objects:

Purpose

More readable than the full Given...When...Then pattern, pytest-when is meant for developers who want to test for behaviour, without any extra overhead. Enable the mock only for a specific argument's values to make code more readable.

Benefits

In this example, when you specify the first two arguments and any third argument, the attribute will be mocked,

(
    when(some_object, "attribute")
    .called_with(1, 2, when.markers.any)
    .then_return("attribute mocked")
    # or
    # .then_call(some_callable)
    # or
    # .then_raise(SomeException())
)

Note that the .called_with method arguments are compared with the real callable signature. This gives additional protection against changing the real callable interface.

With when fixture, the following expression:

when(example_module, "some_normal_function").called_with(
    "a",
    1,
    kwarg1="b",
    kwarg2=2,
).then_return("Mocked")

is roughly equal to:

mock_only_for_calls = (
    call("a", 1, kwarg1="b", kwarg2=2),
    "Mocked",
)

def matches(
        call: Call,
        mocked_calls_registry: tuple[tuple[call, ReturnType], ...],
) -> bool:
    ...

def create_call(*args, **kwargs) -> Call:
    ...

def side_effect_callback(*args, **kwargs):
    if matches(create_call(*args, **kwargs), mock_only_for_calls):
        return mock_only_for_calls[1]
    return unittest.mock.DEFAULT


mocker.patch.object(
    example_module,
    "some_normal_function",
    autospec=True,
    side_effect=side_effect_callback,
)

where logic of matches and create_call is not trivial

Installation

Install the package into your development environment, from pypi, using pip, for example:

pip install pytest-when

Implementation

Onced installed, the when fixture will be available just like the rest of the pytest plugins. See the following example of how to use it:

# class which we're going to mock in the test
class Klass1:
    def some_method(
            self,
            arg1: str,
            arg2: int,
            *,
            kwarg1: str,
            kwarg2: str,
    ) -> str:
        return "some_method not mocked"


def test_should_properly_patch_calls(when):
    when(Klass1, "some_method").called_with(
        "a",
        when.markers.any,
        kwarg1="b",
        kwarg2=when.markers.any,
    ).then_return("some method mocked")

    assert (
            Klass1().some_method(
                "a",
                1,
                kwarg1="b",
                kwarg2="c",
            )
            == "some method mocked"
    )
    assert (
            Klass1().some_method(
                "not mocked param",
                1,
                kwarg1="b",
                kwarg2="c",
            )
            == "some method not mocked"
    )


# if you need to patch a function
def test_patch_a_function(when):
    when(example_module, "some_normal_function").called_with(
        "a",
        when.markers.any,
        kwarg1="b",
        kwarg2=when.markers.any,
    ).then_return("some_normal_function mocked")

    assert (
            example_module.some_normal_function(
                "a",
                1,
                kwarg1="b",
                kwarg2="c",
            )
            == "some_normal_function mocked"
    )
    assert (
            example_module.some_normal_function(
                "not mocked param",
                1,
                kwarg1="b",
                kwarg2="c",
            )
            == "some_normal_function not mocked"
    )

It is possible to use when with class methods and standalone functions (in this case cls parameter will become a python module).

You can patch the same object multiple times using different called_with parameters in a single test.

You can also patch multiple targets (cls, method)

See more examples at: test_integration

Setup for local developement

The project can be extended by cloning the repo and installing the PDM build tool So, the development environment requires:

1. pdm https://pdm.fming.dev/latest/#installation
2. python3.8 or greater

pdm install

To run tests and linters use:

make test
make lint