Mock is a way to replace parts of your system in order to test your unit independently of the other one. Mocks can be see like an object that allow any method, attribute to be called/accessed. The behavior can - of course - be defined if you want a specific response to a given method. Note that for a specific element (for example a method) the same mock is always returned. Last but not least mock record every call which allow you to check that expected calls are made.

The goal of a unit test is to test if execution result is as expected, it should not test the implementation details. In other word implementation may change, unit test should still pass.

Here the test is really low level to cover a wide range of mock usage with a minimal code base, but depending of your testing strategy, your test can “zoom” less in the interaction details.

This test case will try to show you a quick overview of how mock can be in order to test a class. I have tried to cover the most common case but there is still many things. See python mock documentation for more details.

In this example test case I assume that you understand how unit test are written in python: test case classes, make assertion and how to run it (nose, pytest, …). If it is not the case, you may find more informations on official python documentation.

Materials & method

This blog post will be written as a commented python test case to help you understand clearly interactions between code and tests.

Code was written for python 2.7 and file organization is the follow:

|-- pyqueue
|   |--
|   |--
|-- tests
|   `--

Dependencies (can be installed using pip in a virtualenv of course):


To be test code

Here is a simple queue wrapper ( ease the use of Amazon SQS.

from boto.sqs import connect_to_region
from boto.sqs.message import Message

class Queue(object):

    def __init__(self, name):
        self._queue = self._get_queue(name)

    def _get_queue(self, name):
        sqs_connection = connect_to_region('eu-west-1')
        return sqs_connection.get_queue(name)

    def is_empty(self):
        return self._queue.count() == 0

    def push(self, *messages):
        for message in messages:
            envelope = Message()

    def pop(self):
        if self.is_empty:
            return None
        message =
        return message.get_body()

Let’s write some tests!

import unittest
from boto.sqs.message import Message
from mock import patch, Mock, PropertyMock

from pyqueue.queue import Queue

class QueueTestCase(unittest.TestCase):

    @patch.object(Queue, '_get_queue')
    # This is the same thing
    # @patch('pyqueue.queue.Queue._get_queue')
    def test_queue_initialization(self, get_queue_mock):
        Can use either path or patch.object as Queue object is

         * To check that a method called only once:
         * To check the last call: `assert_called_with`
         * To check that a particular call is done among other:

        queue = Queue('foo')
        assert queue._queue == get_queue_mock.return_value

    # Mock the imported module
    def test_get_queue(self, connect_to_region_mock):
        When mocking object, should be done were it will be used.
        Here connect_to_region comes from boto but it is imported and
        used in pyqueue.queue

        Here connect_to_region returns a connection object from which
        we call the get_queue method. That's why we need the
        connect_to_region_mock to return the sqs_connection_mock.

        Two way to know if a method (i.e. a mock) have been called:
         * my_mock.called: returns a boolean regardless the number
           of call
         * my_mock.call_count: returns the actual number of call

        sqs_connection_mock = Mock()
        sqs_connection_mock.get_queue.return_value = 'bar'

        connect_to_region_mock.return_value = sqs_connection_mock

        queue = Queue('foo')
        assert connect_to_region_mock.called
        assert queue._queue == 'bar'

    @patch.object(Queue, '_get_queue')
    def test_is_empty_should_return_false(self, get_queue_mock):
        If you understand the previous examples this test is

        We create a mocked queue object that will respond to count
        with our value.
        queue_mock = Mock()
        queue_mock.count.return_value = 10

        get_queue_mock.return_value = queue_mock

        queue = Queue('foo')
        assert queue.is_empty is False

    @patch('pyqueue.queue.Message', spec=Message)
    # Notice the argument order
    @patch.object(Queue, '_get_queue')
    def test_push_multiple_messages(self,
        Notice the decoration and parameter order: the first
        parameter is the closest to the function name (thing how
        decorator are called).

        Same as previously we start by mocking the queue

        Then we create a Message container (envelope) that match the
        specification of a real Message object with spec=Message.
        It means that if the object signature change or if we have a
        typo in the code and the test it will raise (Mock object has
        no attribute ...).

        Finally we check that every message is well written in the
        queue_mock = Mock()
        get_queue_mock.return_value = queue_mock

        envelope_mock = Mock(spec=Message)
        message_mock.return_value = envelope_mock

        queue = Queue('foo')
        queue.push('foo', 'bar')

        assert queue_mock.write.call_count == 2

    @patch.object(Queue, '_get_queue')
    # by default new_callable=Mock
    @patch.object(Queue, 'is_empty', new_callable=PropertyMock)
    def test_pop_empty_queue_should_return_none(self,
        The new thing here is the new_callable=PropertyMock, it tells
        mock that the mocked element is a property and should be
        called as it.
        If not set mock will tell you that it was not called
        (Expected to be called once).

        To sum up, the my_mock.return_value is returned when the
        object is called think parenthesis () but also if
        new_callable is set to PropertyMock. An attribute of a mock
        can simply be set with = 'bar'.
        queue_mock = Mock()
        get_queue_mock.return_value = queue_mock
        is_empty_mock.return_value = True

        queue = Queue('foo')
        assert queue.pop() is None
        assert is False


This example show you a quick overview of mock’s most common features, but there is many more!

To really master mock it takes some time and practice. Just give it a try on your next project, you will gain some skills, your project will be more robust and your user more happy with fewer bugs!

To finish a good news for python 3.3+ users, mock have been added to standard library, you can directly import it from unitest.mock