For comprehensive examples, see the unit tests included in the full source michael@0: distribution.
michael@0:Here are some more examples for some slightly more advanced scenarios than in michael@0: the getting started guide.
michael@0:Mocking chained calls is actually straightforward with mock once you michael@0: understand the return_value attribute. When a mock is called for michael@0: the first time, or you fetch its return_value before it has been called, a michael@0: new Mock is created.
michael@0:This means that you can see how the object returned from a call to a mocked michael@0: object has been used by interrogating the return_value mock:
michael@0:>>> mock = Mock()
michael@0: >>> mock().foo(a=2, b=3)
michael@0: <Mock name='mock().foo()' id='...'>
michael@0: >>> mock.return_value.foo.assert_called_with(a=2, b=3)
michael@0: From here it is a simple step to configure and then make assertions about michael@0: chained calls. Of course another alternative is writing your code in a more michael@0: testable way in the first place...
michael@0:So, suppose we have some code that looks a little bit like this:
michael@0:>>> class Something(object):
michael@0: ... def __init__(self):
michael@0: ... self.backend = BackendProvider()
michael@0: ... def method(self):
michael@0: ... response = self.backend.get_endpoint('foobar').create_call('spam', 'eggs').start_call()
michael@0: ... # more code
michael@0: Assuming that BackendProvider is already well tested, how do we test michael@0: method()? Specifically, we want to test that the code section # more michael@0: code uses the response object in the correct way.
michael@0:As this chain of calls is made from an instance attribute we can monkey patch michael@0: the backend attribute on a Something instance. In this particular case michael@0: we are only interested in the return value from the final call to michael@0: start_call so we don’t have much configuration to do. Let’s assume the michael@0: object it returns is ‘file-like’, so we’ll ensure that our response object michael@0: uses the builtin file as its spec.
michael@0:To do this we create a mock instance as our mock backend and create a mock michael@0: response object for it. To set the response as the return value for that final michael@0: start_call we could do this:
michael@0:michael@0:michael@0:mock_backend.get_endpoint.return_value.create_call.return_value.start_call.return_value = mock_response.
We can do that in a slightly nicer way using the configure_mock() michael@0: method to directly set the return value for us:
michael@0:>>> something = Something()
michael@0: >>> mock_response = Mock(spec=file)
michael@0: >>> mock_backend = Mock()
michael@0: >>> config = {'get_endpoint.return_value.create_call.return_value.start_call.return_value': mock_response}
michael@0: >>> mock_backend.configure_mock(**config)
michael@0: With these we monkey patch the “mock backend” in place and can make the real michael@0: call:
michael@0:>>> something.backend = mock_backend
michael@0: >>> something.method()
michael@0: Using mock_calls we can check the chained call with a single michael@0: assert. A chained call is several calls in one line of code, so there will be michael@0: several entries in mock_calls. We can use call.call_list() to create michael@0: this list of calls for us:
michael@0:>>> chained = call.get_endpoint('foobar').create_call('spam', 'eggs').start_call()
michael@0: >>> call_list = chained.call_list()
michael@0: >>> assert mock_backend.mock_calls == call_list
michael@0: In some tests I wanted to mock out a call to datetime.date.today() to return michael@0: a known date, but I didn’t want to prevent the code under test from michael@0: creating new date objects. Unfortunately datetime.date is written in C, and michael@0: so I couldn’t just monkey-patch out the static date.today method.
michael@0:I found a simple way of doing this that involved effectively wrapping the date michael@0: class with a mock, but passing through calls to the constructor to the real michael@0: class (and returning real instances).
michael@0:The patch decorator is used here to michael@0: mock out the date class in the module under test. The side_effect michael@0: attribute on the mock date class is then set to a lambda function that returns michael@0: a real date. When the mock date class is called a real date will be michael@0: constructed and returned by side_effect.
michael@0:>>> from datetime import date
michael@0: >>> with patch('mymodule.date') as mock_date:
michael@0: ... mock_date.today.return_value = date(2010, 10, 8)
michael@0: ... mock_date.side_effect = lambda *args, **kw: date(*args, **kw)
michael@0: ...
michael@0: ... assert mymodule.date.today() == date(2010, 10, 8)
michael@0: ... assert mymodule.date(2009, 6, 8) == date(2009, 6, 8)
michael@0: ...
michael@0: Note that we don’t patch datetime.date globally, we patch date in the michael@0: module that uses it. See where to patch.
michael@0:When date.today() is called a known date is returned, but calls to the michael@0: date(...) constructor still return normal dates. Without this you can find michael@0: yourself having to calculate an expected result using exactly the same michael@0: algorithm as the code under test, which is a classic testing anti-pattern.
michael@0:Calls to the date constructor are recorded in the mock_date attributes michael@0: (call_count and friends) which may also be useful for your tests.
michael@0:An alternative way of dealing with mocking dates, or other builtin classes, michael@0: is discussed in this blog entry.
michael@0:A Python generator is a function or method that uses the yield statement to michael@0: return a series of values when iterated over [1].
michael@0:A generator method / function is called to return the generator object. It is michael@0: the generator object that is then iterated over. The protocol method for michael@0: iteration is __iter__, so we can michael@0: mock this using a MagicMock.
michael@0:Here’s an example class with an “iter” method implemented as a generator:
michael@0:>>> class Foo(object):
michael@0: ... def iter(self):
michael@0: ... for i in [1, 2, 3]:
michael@0: ... yield i
michael@0: ...
michael@0: >>> foo = Foo()
michael@0: >>> list(foo.iter())
michael@0: [1, 2, 3]
michael@0: How would we mock this class, and in particular its “iter” method?
michael@0:To configure the values returned from the iteration (implicit in the call to michael@0: list), we need to configure the object returned by the call to foo.iter().
michael@0:>>> mock_foo = MagicMock()
michael@0: >>> mock_foo.iter.return_value = iter([1, 2, 3])
michael@0: >>> list(mock_foo.iter())
michael@0: [1, 2, 3]
michael@0: | [1] | There are also generator expressions and more advanced uses of generators, but we aren’t michael@0: concerned about them here. A very good introduction to generators and how michael@0: powerful they are is: Generator Tricks for Systems Programmers. |
If you want several patches in place for multiple test methods the obvious way michael@0: is to apply the patch decorators to every method. This can feel like unnecessary michael@0: repetition. For Python 2.6 or more recent you can use patch (in all its michael@0: various forms) as a class decorator. This applies the patches to all test michael@0: methods on the class. A test method is identified by methods whose names start michael@0: with test:
michael@0:>>> @patch('mymodule.SomeClass')
michael@0: ... class MyTest(TestCase):
michael@0: ...
michael@0: ... def test_one(self, MockSomeClass):
michael@0: ... self.assertTrue(mymodule.SomeClass is MockSomeClass)
michael@0: ...
michael@0: ... def test_two(self, MockSomeClass):
michael@0: ... self.assertTrue(mymodule.SomeClass is MockSomeClass)
michael@0: ...
michael@0: ... def not_a_test(self):
michael@0: ... return 'something'
michael@0: ...
michael@0: >>> MyTest('test_one').test_one()
michael@0: >>> MyTest('test_two').test_two()
michael@0: >>> MyTest('test_two').not_a_test()
michael@0: 'something'
michael@0: An alternative way of managing patches is to use the patch methods: start and stop. michael@0: These allow you to move the patching into your setUp and tearDown methods.
michael@0:>>> class MyTest(TestCase):
michael@0: ... def setUp(self):
michael@0: ... self.patcher = patch('mymodule.foo')
michael@0: ... self.mock_foo = self.patcher.start()
michael@0: ...
michael@0: ... def test_foo(self):
michael@0: ... self.assertTrue(mymodule.foo is self.mock_foo)
michael@0: ...
michael@0: ... def tearDown(self):
michael@0: ... self.patcher.stop()
michael@0: ...
michael@0: >>> MyTest('test_foo').run()
michael@0: If you use this technique you must ensure that the patching is “undone” by michael@0: calling stop. This can be fiddlier than you might think, because if an michael@0: exception is raised in the setUp then tearDown is not called. unittest2 cleanup functions make this simpler:
michael@0:>>> class MyTest(TestCase):
michael@0: ... def setUp(self):
michael@0: ... patcher = patch('mymodule.foo')
michael@0: ... self.addCleanup(patcher.stop)
michael@0: ... self.mock_foo = patcher.start()
michael@0: ...
michael@0: ... def test_foo(self):
michael@0: ... self.assertTrue(mymodule.foo is self.mock_foo)
michael@0: ...
michael@0: >>> MyTest('test_foo').run()
michael@0: Whilst writing tests today I needed to patch an unbound method (patching the michael@0: method on the class rather than on the instance). I needed self to be passed michael@0: in as the first argument because I want to make asserts about which objects michael@0: were calling this particular method. The issue is that you can’t patch with a michael@0: mock for this, because if you replace an unbound method with a mock it doesn’t michael@0: become a bound method when fetched from the instance, and so it doesn’t get michael@0: self passed in. The workaround is to patch the unbound method with a real michael@0: function instead. The patch() decorator makes it so simple to michael@0: patch out methods with a mock that having to create a real function becomes a michael@0: nuisance.
michael@0:If you pass autospec=True to patch then it does the patching with a michael@0: real function object. This function object has the same signature as the one michael@0: it is replacing, but delegates to a mock under the hood. You still get your michael@0: mock auto-created in exactly the same way as before. What it means though, is michael@0: that if you use it to patch out an unbound method on a class the mocked michael@0: function will be turned into a bound method if it is fetched from an instance. michael@0: It will have self passed in as the first argument, which is exactly what I michael@0: wanted:
michael@0:>>> class Foo(object):
michael@0: ... def foo(self):
michael@0: ... pass
michael@0: ...
michael@0: >>> with patch.object(Foo, 'foo', autospec=True) as mock_foo:
michael@0: ... mock_foo.return_value = 'foo'
michael@0: ... foo = Foo()
michael@0: ... foo.foo()
michael@0: ...
michael@0: 'foo'
michael@0: >>> mock_foo.assert_called_once_with(foo)
michael@0: If we don’t use autospec=True then the unbound method is patched out michael@0: with a Mock instance instead, and isn’t called with self.
michael@0:mock has a nice API for making assertions about how your mock objects are used.
michael@0:>>> mock = Mock()
michael@0: >>> mock.foo_bar.return_value = None
michael@0: >>> mock.foo_bar('baz', spam='eggs')
michael@0: >>> mock.foo_bar.assert_called_with('baz', spam='eggs')
michael@0: If your mock is only being called once you can use the michael@0: assert_called_once_with() method that also asserts that the michael@0: call_count is one.
michael@0:>>> mock.foo_bar.assert_called_once_with('baz', spam='eggs')
michael@0: >>> mock.foo_bar()
michael@0: >>> mock.foo_bar.assert_called_once_with('baz', spam='eggs')
michael@0: Traceback (most recent call last):
michael@0: ...
michael@0: AssertionError: Expected to be called once. Called 2 times.
michael@0: Both assert_called_with and assert_called_once_with make assertions about michael@0: the most recent call. If your mock is going to be called several times, and michael@0: you want to make assertions about all those calls you can use michael@0: call_args_list:
michael@0:>>> mock = Mock(return_value=None)
michael@0: >>> mock(1, 2, 3)
michael@0: >>> mock(4, 5, 6)
michael@0: >>> mock()
michael@0: >>> mock.call_args_list
michael@0: [call(1, 2, 3), call(4, 5, 6), call()]
michael@0: The call helper makes it easy to make assertions about these calls. You michael@0: can build up a list of expected calls and compare it to call_args_list. This michael@0: looks remarkably similar to the repr of the call_args_list:
michael@0:>>> expected = [call(1, 2, 3), call(4, 5, 6), call()]
michael@0: >>> mock.call_args_list == expected
michael@0: True
michael@0: Another situation is rare, but can bite you, is when your mock is called with michael@0: mutable arguments. call_args and call_args_list store references to the michael@0: arguments. If the arguments are mutated by the code under test then you can no michael@0: longer make assertions about what the values were when the mock was called.
michael@0:Here’s some example code that shows the problem. Imagine the following functions michael@0: defined in ‘mymodule’:
michael@0:def frob(val):
michael@0: pass
michael@0:
michael@0: def grob(val):
michael@0: "First frob and then clear val"
michael@0: frob(val)
michael@0: val.clear()
michael@0: When we try to test that grob calls frob with the correct argument look michael@0: what happens:
michael@0:>>> with patch('mymodule.frob') as mock_frob:
michael@0: ... val = set([6])
michael@0: ... mymodule.grob(val)
michael@0: ...
michael@0: >>> val
michael@0: set([])
michael@0: >>> mock_frob.assert_called_with(set([6]))
michael@0: Traceback (most recent call last):
michael@0: ...
michael@0: AssertionError: Expected: ((set([6]),), {})
michael@0: Called with: ((set([]),), {})
michael@0: One possibility would be for mock to copy the arguments you pass in. This michael@0: could then cause problems if you do assertions that rely on object identity michael@0: for equality.
michael@0:Here’s one solution that uses the side_effect michael@0: functionality. If you provide a side_effect function for a mock then michael@0: side_effect will be called with the same args as the mock. This gives us an michael@0: opportunity to copy the arguments and store them for later assertions. In this michael@0: example I’m using another mock to store the arguments so that I can use the michael@0: mock methods for doing the assertion. Again a helper function sets this up for michael@0: me.
michael@0:>>> from copy import deepcopy
michael@0: >>> from mock import Mock, patch, DEFAULT
michael@0: >>> def copy_call_args(mock):
michael@0: ... new_mock = Mock()
michael@0: ... def side_effect(*args, **kwargs):
michael@0: ... args = deepcopy(args)
michael@0: ... kwargs = deepcopy(kwargs)
michael@0: ... new_mock(*args, **kwargs)
michael@0: ... return DEFAULT
michael@0: ... mock.side_effect = side_effect
michael@0: ... return new_mock
michael@0: ...
michael@0: >>> with patch('mymodule.frob') as mock_frob:
michael@0: ... new_mock = copy_call_args(mock_frob)
michael@0: ... val = set([6])
michael@0: ... mymodule.grob(val)
michael@0: ...
michael@0: >>> new_mock.assert_called_with(set([6]))
michael@0: >>> new_mock.call_args
michael@0: call(set([6]))
michael@0: copy_call_args is called with the mock that will be called. It returns a new michael@0: mock that we do the assertion on. The side_effect function makes a copy of michael@0: the args and calls our new_mock with the copy.
michael@0:Note
michael@0:If your mock is only going to be used once there is an easier way of michael@0: checking arguments at the point they are called. You can simply do the michael@0: checking inside a side_effect function.
michael@0:>>> def side_effect(arg):
michael@0: ... assert arg == set([6])
michael@0: ...
michael@0: >>> mock = Mock(side_effect=side_effect)
michael@0: >>> mock(set([6]))
michael@0: >>> mock(set())
michael@0: Traceback (most recent call last):
michael@0: ...
michael@0: AssertionError
michael@0: An alternative approach is to create a subclass of Mock or MagicMock that michael@0: copies (using copy.deepcopy) the arguments. michael@0: Here’s an example implementation:
michael@0:>>> from copy import deepcopy
michael@0: >>> class CopyingMock(MagicMock):
michael@0: ... def __call__(self, *args, **kwargs):
michael@0: ... args = deepcopy(args)
michael@0: ... kwargs = deepcopy(kwargs)
michael@0: ... return super(CopyingMock, self).__call__(*args, **kwargs)
michael@0: ...
michael@0: >>> c = CopyingMock(return_value=None)
michael@0: >>> arg = set()
michael@0: >>> c(arg)
michael@0: >>> arg.add(1)
michael@0: >>> c.assert_called_with(set())
michael@0: >>> c.assert_called_with(arg)
michael@0: Traceback (most recent call last):
michael@0: ...
michael@0: AssertionError: Expected call: mock(set([1]))
michael@0: Actual call: mock(set([]))
michael@0: >>> c.foo
michael@0: <CopyingMock name='mock.foo' id='...'>
michael@0: When you subclass Mock or MagicMock all dynamically created attributes, michael@0: and the return_value will use your subclass automatically. That means all michael@0: children of a CopyingMock will also have the type CopyingMock.
michael@0:You can use PropertyMock to mimic the behaviour of properties. This michael@0: includes raising exceptions when an attribute is accessed.
michael@0:Here’s an example raising a ValueError when the ‘foo’ attribute is accessed:
michael@0:>>> m = MagicMock()
michael@0: >>> p = PropertyMock(side_effect=ValueError)
michael@0: >>> type(m).foo = p
michael@0: >>> m.foo
michael@0: Traceback (most recent call last):
michael@0: ....
michael@0: ValueError
michael@0: Because every mock object has its own type, a new subclass of whichever mock michael@0: class you’re using, all mock objects are isolated from each other. You can michael@0: safely attach properties (or other descriptors or whatever you want in fact) michael@0: to type(mock) without affecting other mock objects.
michael@0:Note
michael@0:In mock 1.0 the handling of iterable side_effect was changed. Any michael@0: exceptions in the iterable will be raised instead of returned.
michael@0:Handling code that needs to behave differently on subsequent calls during the michael@0: test can be tricky. For example you may have a function that needs to raise michael@0: an exception the first time it is called but returns a response on the second michael@0: call (testing retry behaviour).
michael@0:One approach is to use a side_effect function that replaces itself. The michael@0: first time it is called the side_effect sets a new side_effect that will michael@0: be used for the second call. It then raises an exception:
michael@0:>>> def side_effect(*args):
michael@0: ... def second_call(*args):
michael@0: ... return 'response'
michael@0: ... mock.side_effect = second_call
michael@0: ... raise Exception('boom')
michael@0: ...
michael@0: >>> mock = Mock(side_effect=side_effect)
michael@0: >>> mock('first')
michael@0: Traceback (most recent call last):
michael@0: ...
michael@0: Exception: boom
michael@0: >>> mock('second')
michael@0: 'response'
michael@0: >>> mock.assert_called_with('second')
michael@0: Another perfectly valid way would be to pop return values from a list. If the michael@0: return value is an exception, raise it instead of returning it:
michael@0:>>> returns = [Exception('boom'), 'response']
michael@0: >>> def side_effect(*args):
michael@0: ... result = returns.pop(0)
michael@0: ... if isinstance(result, Exception):
michael@0: ... raise result
michael@0: ... return result
michael@0: ...
michael@0: >>> mock = Mock(side_effect=side_effect)
michael@0: >>> mock('first')
michael@0: Traceback (most recent call last):
michael@0: ...
michael@0: Exception: boom
michael@0: >>> mock('second')
michael@0: 'response'
michael@0: >>> mock.assert_called_with('second')
michael@0: Which approach you prefer is a matter of taste. The first approach is actually michael@0: a line shorter but maybe the second approach is more readable.
michael@0:Using patch as a context manager is nice, but if you do multiple patches you michael@0: can end up with nested with statements indenting further and further to the michael@0: right:
michael@0:>>> class MyTest(TestCase):
michael@0: ...
michael@0: ... def test_foo(self):
michael@0: ... with patch('mymodule.Foo') as mock_foo:
michael@0: ... with patch('mymodule.Bar') as mock_bar:
michael@0: ... with patch('mymodule.Spam') as mock_spam:
michael@0: ... assert mymodule.Foo is mock_foo
michael@0: ... assert mymodule.Bar is mock_bar
michael@0: ... assert mymodule.Spam is mock_spam
michael@0: ...
michael@0: >>> original = mymodule.Foo
michael@0: >>> MyTest('test_foo').test_foo()
michael@0: >>> assert mymodule.Foo is original
michael@0: With unittest2 cleanup functions and the patch methods: start and stop we can michael@0: achieve the same effect without the nested indentation. A simple helper michael@0: method, create_patch, puts the patch in place and returns the created mock michael@0: for us:
michael@0:>>> class MyTest(TestCase):
michael@0: ...
michael@0: ... def create_patch(self, name):
michael@0: ... patcher = patch(name)
michael@0: ... thing = patcher.start()
michael@0: ... self.addCleanup(patcher.stop)
michael@0: ... return thing
michael@0: ...
michael@0: ... def test_foo(self):
michael@0: ... mock_foo = self.create_patch('mymodule.Foo')
michael@0: ... mock_bar = self.create_patch('mymodule.Bar')
michael@0: ... mock_spam = self.create_patch('mymodule.Spam')
michael@0: ...
michael@0: ... assert mymodule.Foo is mock_foo
michael@0: ... assert mymodule.Bar is mock_bar
michael@0: ... assert mymodule.Spam is mock_spam
michael@0: ...
michael@0: >>> original = mymodule.Foo
michael@0: >>> MyTest('test_foo').run()
michael@0: >>> assert mymodule.Foo is original
michael@0: You may want to mock a dictionary, or other container object, recording all michael@0: access to it whilst having it still behave like a dictionary.
michael@0:We can do this with MagicMock, which will behave like a dictionary, michael@0: and using side_effect to delegate dictionary access to a real michael@0: underlying dictionary that is under our control.
michael@0:When the __getitem__ and __setitem__ methods of our MagicMock are called michael@0: (normal dictionary access) then side_effect is called with the key (and in michael@0: the case of __setitem__ the value too). We can also control what is returned.
michael@0:After the MagicMock has been used we can use attributes like michael@0: call_args_list to assert about how the dictionary was used:
michael@0:>>> my_dict = {'a': 1, 'b': 2, 'c': 3}
michael@0: >>> def getitem(name):
michael@0: ... return my_dict[name]
michael@0: ...
michael@0: >>> def setitem(name, val):
michael@0: ... my_dict[name] = val
michael@0: ...
michael@0: >>> mock = MagicMock()
michael@0: >>> mock.__getitem__.side_effect = getitem
michael@0: >>> mock.__setitem__.side_effect = setitem
michael@0: Note
michael@0:An alternative to using MagicMock is to use Mock and only provide michael@0: the magic methods you specifically want:
michael@0:>>> mock = Mock()
michael@0: >>> mock.__setitem__ = Mock(side_effect=getitem)
michael@0: >>> mock.__getitem__ = Mock(side_effect=setitem)
michael@0: A third option is to use MagicMock but passing in dict as the spec michael@0: (or spec_set) argument so that the MagicMock created only has michael@0: dictionary magic methods available:
michael@0:>>> mock = MagicMock(spec_set=dict)
michael@0: >>> mock.__getitem__.side_effect = getitem
michael@0: >>> mock.__setitem__.side_effect = setitem
michael@0: With these side effect functions in place, the mock will behave like a normal michael@0: dictionary but recording the access. It even raises a KeyError if you try michael@0: to access a key that doesn’t exist.
michael@0:>>> mock['a']
michael@0: 1
michael@0: >>> mock['c']
michael@0: 3
michael@0: >>> mock['d']
michael@0: Traceback (most recent call last):
michael@0: ...
michael@0: KeyError: 'd'
michael@0: >>> mock['b'] = 'fish'
michael@0: >>> mock['d'] = 'eggs'
michael@0: >>> mock['b']
michael@0: 'fish'
michael@0: >>> mock['d']
michael@0: 'eggs'
michael@0: After it has been used you can make assertions about the access using the normal michael@0: mock methods and attributes:
michael@0:>>> mock.__getitem__.call_args_list
michael@0: [call('a'), call('c'), call('d'), call('b'), call('d')]
michael@0: >>> mock.__setitem__.call_args_list
michael@0: [call('b', 'fish'), call('d', 'eggs')]
michael@0: >>> my_dict
michael@0: {'a': 1, 'c': 3, 'b': 'fish', 'd': 'eggs'}
michael@0: There are various reasons why you might want to subclass Mock. One reason michael@0: might be to add helper methods. Here’s a silly example:
michael@0:>>> class MyMock(MagicMock):
michael@0: ... def has_been_called(self):
michael@0: ... return self.called
michael@0: ...
michael@0: >>> mymock = MyMock(return_value=None)
michael@0: >>> mymock
michael@0: <MyMock id='...'>
michael@0: >>> mymock.has_been_called()
michael@0: False
michael@0: >>> mymock()
michael@0: >>> mymock.has_been_called()
michael@0: True
michael@0: The standard behaviour for Mock instances is that attributes and the return michael@0: value mocks are of the same type as the mock they are accessed on. This ensures michael@0: that Mock attributes are Mocks and MagicMock attributes are MagicMocks michael@0: [2]. So if you’re subclassing to add helper methods then they’ll also be michael@0: available on the attributes and return value mock of instances of your michael@0: subclass.
michael@0:>>> mymock.foo
michael@0: <MyMock name='mock.foo' id='...'>
michael@0: >>> mymock.foo.has_been_called()
michael@0: False
michael@0: >>> mymock.foo()
michael@0: <MyMock name='mock.foo()' id='...'>
michael@0: >>> mymock.foo.has_been_called()
michael@0: True
michael@0: Sometimes this is inconvenient. For example, one user is subclassing mock to michael@0: created a Twisted adaptor. michael@0: Having this applied to attributes too actually causes errors.
michael@0:Mock (in all its flavours) uses a method called _get_child_mock to create michael@0: these “sub-mocks” for attributes and return values. You can prevent your michael@0: subclass being used for attributes by overriding this method. The signature is michael@0: that it takes arbitrary keyword arguments (**kwargs) which are then passed michael@0: onto the mock constructor:
michael@0:>>> class Subclass(MagicMock):
michael@0: ... def _get_child_mock(self, **kwargs):
michael@0: ... return MagicMock(**kwargs)
michael@0: ...
michael@0: >>> mymock = Subclass()
michael@0: >>> mymock.foo
michael@0: <MagicMock name='mock.foo' id='...'>
michael@0: >>> assert isinstance(mymock, Subclass)
michael@0: >>> assert not isinstance(mymock.foo, Subclass)
michael@0: >>> assert not isinstance(mymock(), Subclass)
michael@0: | [2] | An exception to this rule are the non-callable mocks. Attributes use the michael@0: callable variant because otherwise non-callable mocks couldn’t have callable michael@0: methods. |
One situation where mocking can be hard is where you have a local import inside michael@0: a function. These are harder to mock because they aren’t using an object from michael@0: the module namespace that we can patch out.
michael@0:Generally local imports are to be avoided. They are sometimes done to prevent michael@0: circular dependencies, for which there is usually a much better way to solve michael@0: the problem (refactor the code) or to prevent “up front costs” by delaying the michael@0: import. This can also be solved in better ways than an unconditional local michael@0: import (store the module as a class or module attribute and only do the import michael@0: on first use).
michael@0:That aside there is a way to use mock to affect the results of an import. michael@0: Importing fetches an object from the sys.modules dictionary. Note that it michael@0: fetches an object, which need not be a module. Importing a module for the michael@0: first time results in a module object being put in sys.modules, so usually michael@0: when you import something you get a module back. This need not be the case michael@0: however.
michael@0:This means you can use patch.dict() to temporarily put a mock in place michael@0: in sys.modules. Any imports whilst this patch is active will fetch the mock. michael@0: When the patch is complete (the decorated function exits, the with statement michael@0: body is complete or patcher.stop() is called) then whatever was there michael@0: previously will be restored safely.
michael@0:Here’s an example that mocks out the ‘fooble’ module.
michael@0:>>> mock = Mock()
michael@0: >>> with patch.dict('sys.modules', {'fooble': mock}):
michael@0: ... import fooble
michael@0: ... fooble.blob()
michael@0: ...
michael@0: <Mock name='mock.blob()' id='...'>
michael@0: >>> assert 'fooble' not in sys.modules
michael@0: >>> mock.blob.assert_called_once_with()
michael@0: As you can see the import fooble succeeds, but on exit there is no ‘fooble’ michael@0: left in sys.modules.
michael@0:This also works for the from module import name form:
michael@0:>>> mock = Mock()
michael@0: >>> with patch.dict('sys.modules', {'fooble': mock}):
michael@0: ... from fooble import blob
michael@0: ... blob.blip()
michael@0: ...
michael@0: <Mock name='mock.blob.blip()' id='...'>
michael@0: >>> mock.blob.blip.assert_called_once_with()
michael@0: With slightly more work you can also mock package imports:
michael@0:>>> mock = Mock()
michael@0: >>> modules = {'package': mock, 'package.module': mock.module}
michael@0: >>> with patch.dict('sys.modules', modules):
michael@0: ... from package.module import fooble
michael@0: ... fooble()
michael@0: ...
michael@0: <Mock name='mock.module.fooble()' id='...'>
michael@0: >>> mock.module.fooble.assert_called_once_with()
michael@0: The Mock class allows you to track the order of method calls on michael@0: your mock objects through the method_calls attribute. This michael@0: doesn’t allow you to track the order of calls between separate mock objects, michael@0: however we can use mock_calls to achieve the same effect.
michael@0:Because mocks track calls to child mocks in mock_calls, and accessing an michael@0: arbitrary attribute of a mock creates a child mock, we can create our separate michael@0: mocks from a parent one. Calls to those child mock will then all be recorded, michael@0: in order, in the mock_calls of the parent:
michael@0:>>> manager = Mock()
michael@0: >>> mock_foo = manager.foo
michael@0: >>> mock_bar = manager.bar
michael@0:
michael@0: >>> mock_foo.something()
michael@0: <Mock name='mock.foo.something()' id='...'>
michael@0: >>> mock_bar.other.thing()
michael@0: <Mock name='mock.bar.other.thing()' id='...'>
michael@0:
michael@0: >>> manager.mock_calls
michael@0: [call.foo.something(), call.bar.other.thing()]
michael@0: We can then assert about the calls, including the order, by comparing with michael@0: the mock_calls attribute on the manager mock:
michael@0:>>> expected_calls = [call.foo.something(), call.bar.other.thing()]
michael@0: >>> manager.mock_calls == expected_calls
michael@0: True
michael@0: If patch is creating, and putting in place, your mocks then you can attach michael@0: them to a manager mock using the attach_mock() method. After michael@0: attaching calls will be recorded in mock_calls of the manager.
michael@0:>>> manager = MagicMock()
michael@0: >>> with patch('mymodule.Class1') as MockClass1:
michael@0: ... with patch('mymodule.Class2') as MockClass2:
michael@0: ... manager.attach_mock(MockClass1, 'MockClass1')
michael@0: ... manager.attach_mock(MockClass2, 'MockClass2')
michael@0: ... MockClass1().foo()
michael@0: ... MockClass2().bar()
michael@0: ...
michael@0: <MagicMock name='mock.MockClass1().foo()' id='...'>
michael@0: <MagicMock name='mock.MockClass2().bar()' id='...'>
michael@0: >>> manager.mock_calls
michael@0: [call.MockClass1(),
michael@0: call.MockClass1().foo(),
michael@0: call.MockClass2(),
michael@0: call.MockClass2().bar()]
michael@0: If many calls have been made, but you’re only interested in a particular michael@0: sequence of them then an alternative is to use the michael@0: assert_has_calls() method. This takes a list of calls (constructed michael@0: with the call object). If that sequence of calls are in michael@0: mock_calls then the assert succeeds.
michael@0:>>> m = MagicMock()
michael@0: >>> m().foo().bar().baz()
michael@0: <MagicMock name='mock().foo().bar().baz()' id='...'>
michael@0: >>> m.one().two().three()
michael@0: <MagicMock name='mock.one().two().three()' id='...'>
michael@0: >>> calls = call.one().two().three().call_list()
michael@0: >>> m.assert_has_calls(calls)
michael@0: Even though the chained call m.one().two().three() aren’t the only calls that michael@0: have been made to the mock, the assert still succeeds.
michael@0:Sometimes a mock may have several calls made to it, and you are only interested michael@0: in asserting about some of those calls. You may not even care about the michael@0: order. In this case you can pass any_order=True to assert_has_calls:
michael@0:>>> m = MagicMock()
michael@0: >>> m(1), m.two(2, 3), m.seven(7), m.fifty('50')
michael@0: (...)
michael@0: >>> calls = [call.fifty('50'), call(1), call.seven(7)]
michael@0: >>> m.assert_has_calls(calls, any_order=True)
michael@0: Using the same basic concept as ANY we can implement matchers to do more michael@0: complex assertions on objects used as arguments to mocks.
michael@0:Suppose we expect some object to be passed to a mock that by default michael@0: compares equal based on object identity (which is the Python default for user michael@0: defined classes). To use assert_called_with() we would need to pass michael@0: in the exact same object. If we are only interested in some of the attributes michael@0: of this object then we can create a matcher that will check these attributes michael@0: for us.
michael@0:You can see in this example how a ‘standard’ call to assert_called_with isn’t michael@0: sufficient:
michael@0:>>> class Foo(object):
michael@0: ... def __init__(self, a, b):
michael@0: ... self.a, self.b = a, b
michael@0: ...
michael@0: >>> mock = Mock(return_value=None)
michael@0: >>> mock(Foo(1, 2))
michael@0: >>> mock.assert_called_with(Foo(1, 2))
michael@0: Traceback (most recent call last):
michael@0: ...
michael@0: AssertionError: Expected: call(<__main__.Foo object at 0x...>)
michael@0: Actual call: call(<__main__.Foo object at 0x...>)
michael@0: A comparison function for our Foo class might look something like this:
michael@0:>>> def compare(self, other):
michael@0: ... if not type(self) == type(other):
michael@0: ... return False
michael@0: ... if self.a != other.a:
michael@0: ... return False
michael@0: ... if self.b != other.b:
michael@0: ... return False
michael@0: ... return True
michael@0: ...
michael@0: And a matcher object that can use comparison functions like this for its michael@0: equality operation would look something like this:
michael@0:>>> class Matcher(object):
michael@0: ... def __init__(self, compare, some_obj):
michael@0: ... self.compare = compare
michael@0: ... self.some_obj = some_obj
michael@0: ... def __eq__(self, other):
michael@0: ... return self.compare(self.some_obj, other)
michael@0: ...
michael@0: Putting all this together:
michael@0:>>> match_foo = Matcher(compare, Foo(1, 2))
michael@0: >>> mock.assert_called_with(match_foo)
michael@0: The Matcher is instantiated with our compare function and the Foo object michael@0: we want to compare against. In assert_called_with the Matcher equality michael@0: method will be called, which compares the object the mock was called with michael@0: against the one we created our matcher with. If they match then michael@0: assert_called_with passes, and if they don’t an AssertionError is raised:
michael@0:>>> match_wrong = Matcher(compare, Foo(3, 4))
michael@0: >>> mock.assert_called_with(match_wrong)
michael@0: Traceback (most recent call last):
michael@0: ...
michael@0: AssertionError: Expected: ((<Matcher object at 0x...>,), {})
michael@0: Called with: ((<Foo object at 0x...>,), {})
michael@0: With a bit of tweaking you could have the comparison function raise the michael@0: AssertionError directly and provide a more useful failure message.
michael@0:As of version 1.5, the Python testing library PyHamcrest provides similar functionality, michael@0: that may be useful here, in the form of its equality matcher michael@0: (hamcrest.library.integration.match_equality).
michael@0:This recipe, for easier configuration of mock objects, is now part of Mock. michael@0: See the configure_mock() method.
michael@0:This example is now built in to mock. See ANY.
michael@0:This example is now built in to mock. See PropertyMock.
michael@0:This example is now built in to mock. See mock_open().
michael@0:This example is now built in to mock. See Deleting Attributes.
michael@0: