mocking.txt

Mocking out objects and methods
===============================

.. py:currentmodule:: testfixtures

Mocking is the process of replacing objects used in your code with ones
that make testing easier, but only while the tests are running.

This may mean replacing resources or dependencies, such as database
connections or file paths, with ones that are isolated for testing.
It may also mean replacing chunks of complex functionality
that aren't the subject of the test with mock objects that allow you
to check that the mocked out functionality is being used as expected.

What to mock with
-----------------

Python has a standard mock implementation in the form of :mod:`unittest.mock`
which is also available as a `rolling backport`__ so that the latest features
and bugfixes can be used in any version of Python.

__ https://mock.readthedocs.io

For convenience, TestFixtures provides a facade over both of these in the form
of :mod:`testfixtures.mock`. The contents are identical and preference is given
to the rolling backport if it is present. The facade also contains any bugfixes
that are critical to the operation of functionality provided by TestFixtures.

TestFixtures also provides specialised mocks for dealing with
:doc:`dates and times <datetime>` and :doc:`subprocesses <popen>`.

How to mock
-----------

TestFixtures provides :class:`Replace`, :class:`Replacer` and the :func:`replace`
decorator to mock out objects. These work in a similar way to
:func:`unittest.mock.patch`, and have been around longer. They still provide a little
more flexibility than :func:`~unittest.mock.patch`, so you whichever feels best in
your codebase.

Methods of replacement
----------------------

There are three different methods of mocking out
functionality that can be used to replace functions, classes
or even individual methods on a class. Consider the following module:

.. topic:: testfixtures.tests.sample1
   :class: module

   .. literalinclude:: ../testfixtures/tests/sample1.py
      :pyobject: X

.. do the import quietly

  >>> from testfixtures.tests.sample1 import X

We want to mock out the ``y`` method of the ``X`` class, with,
for example, the following function:

.. code-block:: python

  def mock_y(self):
       return 'mock y'

The context manager
~~~~~~~~~~~~~~~~~~~

For replacement of a single thing, it's easiest to use the
:class:`~testfixtures.Replace` context manager:

.. code-block:: python

  from testfixtures import Replace

  def test_function():
      with Replace('testfixtures.tests.sample1.X.y', mock_y):
          print(X().y())

For the duration of the ``with`` block, the replacement is used:

>>> test_function()
mock y

For multiple replacements to do, or where the you need access to the replacement
within the code block under test, the :class:`~testfixtures.Replacer` context
manager can be used instead:


.. code-block:: python

  from testfixtures.mock import Mock
  from testfixtures import Replacer

  def test_function():
      with Replacer() as replace:
          mock_y = replace('testfixtures.tests.sample1.X.y', Mock())
          mock_y.return_value = 'mock y'
          print(X().y())

For the duration of the ``with`` block, the replacement is used:

>>> test_function()
mock y

The decorator
~~~~~~~~~~~~~

If you are working in a traditional :mod:`unittest` environment and
want to replace different things in different test functions, you may
find the decorator suits your needs better:

.. code-block:: python

  from testfixtures import replace

  @replace('testfixtures.tests.sample1.X.y', mock_y)
  def test_function():
      print(X().y())

When using the decorator, the replacement is used for the duration of
the decorated callable's execution:

>>> test_function()
mock y

If you need to manipulate or inspect the object that's used as a
replacement, you can add an extra parameter to your function. The
decorator will see this and pass the replacement in it's place:

.. code-block:: python

  from testfixtures.mock import Mock, call
  from testfixtures import compare, replace

  @replace('testfixtures.tests.sample1.X.y', Mock())
  def test_function(mocked_y):
      mocked_y.return_value = 'mock y'
      print(X().y())
      compare(mocked_y.mock_calls, expected=[call()])

The above still results in the same output:

>>> test_function()
mock y

.. note::
    This method is not compatible with pytest's fixture discovery stuff.
    Instead, put a fixture such as the following in your `conftest.py`:

    .. code-block:: python

      from testfixtures import Replace
      import pytest

      @pytest.fixture()
      def mocked_y():
          m = Mock()
          with Replace('testfixtures.tests.sample1.X.y', m):
              yield m

Manual usage
~~~~~~~~~~~~

If you want to replace something for the duration of a doctest or you
want to replace something for every test in a
:class:`~unittest.TestCase`, then you can use the
:class:`~testfixtures.Replacer` manually.

The instantiation and replacement are done in the ``setUp`` function
of the :class:`~unittest.TestCase` or passed to the
:class:`~doctest.DocTestSuite` constructor:

>>> from testfixtures import Replacer
>>> replace = Replacer()
>>> replace('testfixtures.tests.sample1.X.y', mock_y)
<...>

The replacement then stays in place until removed:

>>> X().y()
'mock y'

Then, in the ``tearDown`` function
of the :class:`~unittest.TestCase` or passed to the
:class:`~doctest.DocTestSuite` constructor, the replacement is
removed:

>>> replace.restore()
>>> X().y()
'original y'

The :meth:`~testfixtures.Replacer.restore` method can also be added as an
:meth:`~unittest.TestCase.addCleanup` if that is easier or more compact in your test
suite.

Replacing more than one thing
-----------------------------

Both the :class:`~testfixtures.Replacer` and the
:func:`~testfixtures.replace` decorator can be used to replace more
than one thing at a time. For the former, this is fairly obvious:

.. code-block:: python

  def test_function():
      with Replacer() as replace:
          y = replace('testfixtures.tests.sample1.X.y', Mock())
          y.return_value = 'mock y'
          aMethod = replace('testfixtures.tests.sample1.X.aMethod', Mock())
          aMethod.return_value = 'mock method'
          x = X()
          print(x.y(), x.aMethod())

.. the result:

   >>> test_function()
   mock y mock method

For the decorator, it's less obvious but still pretty easy:

.. code-block:: python

  from testfixtures import replace

  @replace('testfixtures.tests.sample1.X.y', Mock())
  @replace('testfixtures.tests.sample1.X.aMethod', Mock())
  def test_function(aMethod, y):
      print(aMethod, y)
      aMethod().return_value = 'mock method'
      y().return_value = 'mock y'
      x = X()
      print(aMethod, y)
      print(x.y(), x.aMethod())

You'll notice that you can still get access to the replacements, even
though there are several of them.

Replacing things that may not be there
--------------------------------------

The following code shows a situation where ``hpy`` may or may not be
present depending on whether the ``guppy`` package is installed or
not.


.. topic:: testfixtures.tests.sample2
   :class: module

   .. literalinclude:: ../testfixtures/tests/sample2.py
      :lines: 10-19

To test the behaviour of the code that uses ``hpy`` in both of
these cases, regardless of whether or not the ``guppy`` package is
actually installed, we need to be able to mock out both ``hpy`` and the
``guppy`` global. This is done by doing non-strict replacement, as
shown in the following :class:`~unittest.TestCase`:

.. imports

  >>> import unittest,sys

.. code-block:: python

    from testfixtures.tests.sample2 import dump
    from testfixtures import replace
    from testfixtures.mock import Mock, call

    class Tests(unittest.TestCase):

        @replace('testfixtures.tests.sample2.guppy', True)
        @replace('testfixtures.tests.sample2.hpy', Mock(), strict=False)
        def test_method(self, hpy):

            dump('somepath')

            compare([
                     call(),
                     call().heap(),
                     call().heap().stat.dump('somepath')
                   ], hpy.mock_calls)

        @replace('testfixtures.tests.sample2.guppy', False)
        @replace('testfixtures.tests.sample2.hpy', Mock(), strict=False)
        def test_method_no_heapy(self,hpy):

            dump('somepath')

            compare(hpy.mock_calls,[])

.. the result:

   >>> from testfixtures.compat import StringIO
   >>> suite = unittest.TestLoader().loadTestsFromTestCase(Tests)
   >>> unittest.TextTestRunner(verbosity=0,stream=StringIO()).run(suite)
   <unittest...TextTestResult run=2 errors=0 failures=0>


The :meth:`~testfixtures.Replacer.replace` method and calling
a :class:`Replacer` also supports non-strict replacement using the same
keyword parameter.

Replacing items in dictionaries and lists
-----------------------------------------

:class:`~testfixtures.Replace`, :class:`~testfixtures.Replacer` and the
:func:`~testfixtures.replace` decorator can be used to replace items
in dictionaries and lists.

For example, suppose you have a data structure like the following:

.. topic:: testfixtures.tests.sample1
   :class: module

   .. literalinclude:: ../testfixtures/tests/sample1.py
      :lines: 67-70

You can mock out the value associated with ``key`` and the second
element in the ``complex_key`` list as follows:

.. code-block:: python

  from pprint import pprint
  from testfixtures import Replacer
  from testfixtures.tests.sample1 import someDict

  def test_function():
      with Replacer() as replace:
          replace('testfixtures.tests.sample1.someDict.key', 'foo')
          replace('testfixtures.tests.sample1.someDict.complex_key.1', 42)
          pprint(someDict)

While the replacement is in effect, the new items are in place:

>>> test_function()
{'complex_key': [1, 42, 3], 'key': 'foo'}

When it is no longer in effect, the originals are returned:

>>> pprint(someDict)
{'complex_key': [1, 2, 3], 'key': 'value'}

.. _removing_attr_and_item:

Removing attributes and dictionary items
----------------------------------------

:class:`~testfixtures.Replace`, :class:`~testfixtures.Replacer` and the
:func:`~testfixtures.replace` decorator can be used to remove
attributes from objects and remove items from dictionaries.

For example, suppose you have a data structure like the following:

.. topic:: testfixtures.tests.sample1
   :class: module

   .. literalinclude:: ../testfixtures/tests/sample1.py
      :lines: 67-70

If you want to remove the ``key`` for the duration of a test, you can
do so as follows:

.. code-block:: python

  from testfixtures import Replace, not_there
  from testfixtures.tests.sample1 import someDict

  def test_function():
      with Replace('testfixtures.tests.sample1.someDict.key', not_there):
          pprint(someDict)

While the replacement is in effect, ``key`` is gone:

>>> test_function()
{'complex_key': [1, 2, 3]}

When it is no longer in effect, ``key`` is returned:

>>> pprint(someDict)
{'complex_key': [1, 2, 3], 'key': 'value'}

If you want the whole ``someDict`` dictionary to be removed for the
duration of a test, you would do so as follows:

.. code-block:: python

  from testfixtures import Replace, not_there
  from testfixtures.tests import sample1

  def test_function():
      with Replace('testfixtures.tests.sample1.someDict', not_there):
          print(hasattr(sample1, 'someDict'))

While the replacement is in effect, ``key`` is gone:

>>> test_function()
False

When it is no longer in effect, ``key`` is returned:

>>> pprint(sample1.someDict)
{'complex_key': [1, 2, 3], 'key': 'value'}


Gotchas
-------

- Make sure you replace the object where it's used and not where it's
  defined. For example, with the following code from the
  ``testfixtures.tests.sample1`` package:

  .. literalinclude:: ../testfixtures/tests/sample1.py
     :lines: 30-34

  You might be tempted to mock things as follows:

  >>> replace = Replacer()
  >>> replace('time.time', Mock())
  <...>

  But this won't work:

  >>> from testfixtures.tests.sample1 import str_time
  >>> type(float(str_time()))
  <... 'float'>

  You need to replace :func:`~time.time` where it's used, not where
  it's defined:

  >>> replace('testfixtures.tests.sample1.time', Mock())
  <...>
  >>> str_time()
  "<...Mock...>"

  .. cleanup

   >>> replace.restore()

  A corollary of this is that you need to replace *all* occurrences of
  an original to safely be able to test. This can be tricky when an
  original is imported into many modules that may be used by a
  particular test.

- You can't replace whole top level modules, and nor should you want
  to! The reason being that everything up to the last dot in the
  replacement target specifies where the replacement will take place,
  and the part after the last dot is used as the name of the thing to
  be replaced:

  >>> Replacer().replace('sys', Mock())
  Traceback (most recent call last):
   ...
  ValueError: target must contain at least one dot!