Checking call order across multiple mocks

Question:

I have three functions that I’m trying to test the call order of.

Let’s say that in module module.py I have the following

# module.py    

def a(*args):
    # do the first thing

def b(*args):
    # do a second thing

def c(*args):
    # do a third thing


def main_routine():
    a_args = ('a')
    b_args = ('b')
    c_args = ('c')

    a(*a_args)
    b(*b_args)
    c(*c_args)

I want to check that b is called after a, and before c. So getting a mock for each of a, b and c is easy:

# tests.py

@mock.patch('module.a')
@mock.patch('module.b')
@mock.patch('module.c')
def test_main_routine(c_mock, b_mock, a_mock):
    # test all the things here

Checking that each of the individial mocks are called is easy, too. How do I check the order of the calls relative to one another?

call_args_list won’t work as it’s maintained separately for each mock.

I’ve tried using a side effect to log each of the calls:

calls = []
def register_call(*args):
    calls.append(mock.call(*args))
    return mock.DEFAULT

a_mock.side_effect = register_call
b_mock.side_effect = register_call
c_mock.side_effect = register_call

But this only gives me the args that the mocks were called with, but not the actual mock that the call was made against. I can add a bit more logic:

# tests.py
from functools import partial

def register_call(*args, **kwargs):
    calls.append(kwargs.pop('caller', None), mock.call(*args, **kwargs))
    return mock.DEFAULT

a_mock.side_effect = partial(register_call, caller='a')
b_mock.side_effect = partial(register_call, caller='b')
c_mock.side_effect = partial(register_call, caller='c')

And that seems to get the job done… Is there a better way though? It feels like there should already be something in the API that can do this that I’m missing.

Asked By: Shaun O'Keefe

||
Source

Answers:

Define a Mock manager and attach mocks to it via attach_mock(). Then check for the mock_calls:

@patch('module.a')
@patch('module.b')
@patch('module.c')
def test_main_routine(c, b, a):
    manager = Mock()
    manager.attach_mock(a, 'a')
    manager.attach_mock(b, 'b')
    manager.attach_mock(c, 'c')

    module.main_routine()

    expected_calls = [call.a('a'), call.b('b'), call.c('c')]
    assert manager.mock_calls == expected_calls

Just to test that it works, change the order of function calls in the main_routine() function add see that it throws AssertionError.

See more examples at Tracking order of calls and less verbose call assertions (link is dead; substitute: https://docs.python.org/3/library/unittest.mock.html#attaching-mocks-as-attributes)

Hope that helps.

Answered By: alecxe

I needed this answer today, but the example code in the question is really hard to read because the call args are the same as the names of the mocks on the manager and in the scope of the test. Here’s the official documentation on this concept, and below is a clearer example for non-robots. All the modules I’m patching are made-up for the sake of the example:

@patch('module.file_reader')
@patch('module.json_parser')
@patch('module.calculator')
def test_main_routine(mock_calculator, mock_json_parser, mock_file_reader):
    manager = Mock()

    # First argument is the mock to attach to the manager.
    # Second is the name for the field on the manager that holds the mock.
    manager.attach_mock(mock_file_reader, 'the_mock_file_reader')
    manager.attach_mock(mock_json_parser, 'the_mock_json_parser')
    manager.attach_mock(mock_calculator, 'the_mock_calculator')
    
    module.main_routine()

    expected_calls = [
        call.the_mock_file_reader('some file'),
        call.the_mock_json_parser('some json'),
        call.the_mock_calculator(1, 2)
    ]
    assert manager.mock_calls == expected_calls

Note that you have to use attach_mock in this case because your mocks were created by patch. Mocks with names, including those created by patch, must be attached via attach_mock for this code to work. You don’t have to use attach_mock if you make your own Mock objects without names:

def test_main_routine(mock_calculator, mock_json_parser, mock_file_reader):
    manager = Mock()

    mock_file_reader = Mock()
    mock_json_parser = Mock()
    mock_calculator = Mock()

    manager.the_mock_file_reader = mock_file_reader
    manager.the_mock_json_parser = mock_json_parser
    manager.the_mock_calculator = mock_calculator
    
    module.main_routine()

    expected_calls = [
        call.the_mock_file_reader('some file'),
        call.the_mock_json_parser('some json'),
        call.the_mock_calculator(1, 2)
    ]
    assert manager.mock_calls == expected_calls

If you want a clear assertion failed message when the order or expected calls are missing, use the following assert line instead.

self.assertListEqual(manager.mock_calls, [
    call.the_mock_file_reader('some file'),
    call.the_mock_json_parser('some json'),
    call.the_mock_calculator(1, 2)
])
Answered By: lortimer

A cleaner solution would be to wrap your functions into a class, then mock the class in the test. This will eliminate the need to do any patching (always a plus).

# module.py

class Wrapper:
    def a(self, *args):
        pass

    def b(self, *args):
        pass

    def c(self, *args):
        pass

    def main_routine(self):
        a_args = ('arg for a',)
        b_args = ('arg for b',)
        c_args = ('arg for c',)

        self.a(*a_args)
        self.b(*b_args)
        self.c(*c_args)

In the test file, you create a mock wrapper class, then insert the mock wrapper in as the argument self when calling Wrapper.main_method (notice that this does not instantiate the class).

# module_test.py

from unittest.mock import MagicMock, call

from module import Wrapper


def test_main_routine():
    mock_wrapper = MagicMock()
    Wrapper.main_routine(mock_wrapper)
    expected_calls = [call.a('arg for a'),
                      call.b('arg for b'),
                      call.c('arg for c')]
    mock_wrapper.assert_has_calls(expected_calls)

Benefits:

  • No patching needed
  • In the test, you only need to type the name of the method being called once (instead of 2-3 times)
  • Uses assert_has_calls instead of comparing the mock_calls attribute to a list of calls.
  • Can be made into a general check_for_calls function (see below)
# module_better_test.py

from unittest.mock import MagicMock, call

from module import Wrapper


def test_main_routine():
    expected_calls = [call.a('arg for a'),
                      call.b('arg for b'),
                      call.c('arg for c')]
    check_for_calls('main_routine', expected_calls)


def check_for_calls(method, expected_calls):
    mock_wrapper = MagicMock()
    getattr(Wrapper, method)(mock_wrapper)
    mock_wrapper.assert_has_calls(expected_calls)
Answered By: Chris Collett