Mocks and Monkeypatching in Python

Post originally published on http://krzysztofzuraw.com/. Republished with author’s permission.

Introduction

In this post I will look into the essential part of testing — mocks.

First of all, what I want to accomplish here is to give you basic examples of how to mock data using two tools — mock and pytest monkeypatch.

Why bother mocking?

Some of the parts of our application may have dependencies for other libraries or objects. To isolate the behaviour of our parts, we need to substitute external dependencies. Here comes the mocking. We mock an external API to check certain behaviours, such as proper return values, that we previously defined.

Mocking function

Let’s say we have a module called function.py:

def square(value): return value ** 2 def cube(value): return value ** 3 def main(value): return square(value) + cube(value)

Then let’s see how these functions are mocked using the mock library:

 try: import mock except ImportError: from unittest import mock import unittest from function import square, main class TestNotMockedFunction(unittest.TestCase): @mock.patch('__main__.square', return_value=1) def test_function(self, mocked_square): # because you need to patch in exact place where function that has to be mocked is called self.assertEquals(square(5), 1) @mock.patch('function.square') @mock.patch('function.cube') def test_main_function(self, mocked_square, mocked_cube): # underling function are mocks so calling main(5) will return mock mocked_square.return_value = 1 mocked_cube.return_value = 0 self.assertEquals(main(5), 1) mocked_square.assert_called_once_with(5) mocked_cube.assert_called_once_with(5) if __name__ == '__main__': unittest.main()

What is happening here? Lines 1-4 are for making this code compatible between Python 2 and 3. In Python 3, mock is part of the standard library, whereas in Python 2 you need to install it by pip install mock.

In line 13, I patched the square function. You have to remember to patch it in the same place you use it. For instance, I’m calling square(5) in the test itself so I need to patch it in __main__. This is the case if I’m running this by using python tests/test_function.py. If I’m using pytest for that, I need to patch it as test_function.square.

In lines 18-19, I patch the square and cube functions in their module because they are used in the main function. The last two asserts come from the mock library, and are there to make sure that mock was called with proper values.

The same can be accomplished using mokeypatching for py.test:

from function import square, main def test_function(monkeypatch): monkeypatch.setattr(“test_function_pytest.square”, lambda x: 1) assert square(5) == 1 def test_main_function(monkeypatch): monkeypatch.setattr(‘function.square’, lambda x: 1) monkeypatch.setattr(‘function.cube’, lambda x: 0) assert main(5) == 1

As you can see, I’m using monkeypatch.setattr for setting up a return value for given functions. I still need to monkeypatch it in proper places — test_function_pytest and function.

Mocking classes

I have a module called square:

import math class Square(object): def __init__(radius): self.radius = radius def calculate_area(self): return math.sqrt(self.radius) * math.pi

and mocks using standard lib:

try: import mock except ImportError: from unittest import mock import unittest from square import Square class TestClass(unittest.TestCase): @mock.patch('__main__.Square') # depends in witch from is run def test_mocking_instance(self, mocked_instance): mocked_instance = mocked_instance.return_value mocked_instance.calculate_area.return_value = 1 sq = Square(100) self.assertEquals(sq.calculate_area(), 1) def test_mocking_classes(self): sq = Square sq.calculate_area = mock.MagicMock(return_value=1) self.assertEquals(sq.calculate_area(), 1) @mock.patch.object(Square, 'calculate_area') def test_mocking_class_methods(self, mocked_method): mocked_method.return_value = 20 self.assertEquals(Square.calculate_area(), 20) if __name__ == ‘__main__’: unittest.main()

At line 13, I patch the class Square. Lines 15 and 16 present a mocking instance. mocked_instance is a mock object which returns another mock by default, and to these mock.calculate_area I add return_value 1. In line 23, I’m using MagicMock, which is a normal mock class, except in that it also retrieves magic methods from the given object. Lastly, I use patch.object to mock the method in the Square class.

The same using pytest:

try: from mock import MagicMock except ImportError: from unittest.mock import MagicMock from square import Square def test_mocking_class_methods(monkeypatch): monkeypatch.setattr('test_class_pytest.Square.calculate_area', lambda: 1) assert Square.calculate_area() == 1 def test_mocking_classes(monkeypatch): monkeypatch.setattr('test_class_pytest.Square', MagicMock(Square)) sq = Square sq.calculate_area.return_value = 1 assert sq.calculate_area() == 1

The issue here is with test_mocking_class_methods, which works well in Python 3, but not in Python 2.

All examples can be found in this repo.

If you have any questions and comments, feel free to leave them in the section below.

References:

• What is Mocking?
• Mocking With Kung Fu Panda