26.5. unittest.mock — 仿造对象库 ¶

26.5.2.1. 调用 ¶

Mock objects are callable. The call will return the value set as the return_value attribute. The default return value is a new Mock object; it is created the first time the return value is accessed (either explicitly or by calling the Mock) - but it is stored and the same one returned each time.

Calls made to the object will be recorded in the attributes like call_args and call_args_list .

若 side_effect is set then it will be called after the call has been recorded, so if side_effect raises an exception the call is still recorded.

The simplest way to make a mock raise an exception when called is to make side_effect an exception class or instance:

>>> m = MagicMock(side_effect=IndexError)
>>> m(1, 2, 3)
Traceback (most recent call last):
  ...
IndexError
>>> m.mock_calls
[call(1, 2, 3)]
>>> m.side_effect = KeyError('Bang!')
>>> m('two', 'three', 'four')
Traceback (most recent call last):
  ...
KeyError: 'Bang!'
>>> m.mock_calls
[call(1, 2, 3), call('two', 'three', 'four')]
									

若 side_effect is a function then whatever that function returns is what calls to the mock return. The side_effect function is called with the same arguments as the mock. This allows you to vary the return value of the call dynamically, based on the input:

>>> def side_effect(value):
...     return value + 1
...
>>> m = MagicMock(side_effect=side_effect)
>>> m(1)
2
>>> m(2)
3
>>> m.mock_calls
[call(1), call(2)]
									

If you want the mock to still return the default return value (a new mock), or any set return value, then there are two ways of doing this. Either return mock.return_value from inside side_effect , or return DEFAULT :

>>> m = MagicMock()
>>> def side_effect(*args, **kwargs):
...     return m.return_value
...
>>> m.side_effect = side_effect
>>> m.return_value = 3
>>> m()
3
>>> def side_effect(*args, **kwargs):
...     return DEFAULT
...
>>> m.side_effect = side_effect
>>> m()
3
									

To remove a side_effect , and return to the default behaviour, set the side_effect to None :

>>> m = MagicMock(return_value=6)
>>> def side_effect(*args, **kwargs):
...     return 3
...
>>> m.side_effect = side_effect
>>> m()
3
>>> m.side_effect = None
>>> m()
6
									

side_effect can also be any iterable object. Repeated calls to the mock will return values from the iterable (until the iterable is exhausted and a StopIteration is raised):

>>> m = MagicMock(side_effect=[1, 2, 3])
>>> m()
1
>>> m()
2
>>> m()
3
>>> m()
Traceback (most recent call last):
  ...
StopIteration
									

If any members of the iterable are exceptions they will be raised instead of returned:

>>> iterable = (33, ValueError, 66)
>>> m = MagicMock(side_effect=iterable)
>>> m()
33
>>> m()
Traceback (most recent call last):
 ...
ValueError
>>> m()
66
									

26.5.2.2. Deleting Attributes ¶

Mock objects create attributes on demand. This allows them to pretend to be objects of any type.

You may want a mock object to return False 到 hasattr() call, or raise an AttributeError when an attribute is fetched. You can do this by providing an object as a spec for a mock, but that isn’t always convenient.

You “block” attributes by deleting them. Once deleted, accessing an attribute will raise an AttributeError .

>>> mock = MagicMock()
>>> hasattr(mock, 'm')
True
>>> del mock.m
>>> hasattr(mock, 'm')
False
>>> del mock.f
>>> mock.f
Traceback (most recent call last):
    ...
AttributeError: f
									

26.5.2.3. Mock names and the name attribute ¶

Since “name” is an argument to the Mock constructor, if you want your mock object to have a “name” attribute you can’t just pass it in at creation time. There are two alternatives. One option is to use configure_mock() :

>>> mock = MagicMock()
>>> mock.configure_mock(name='my_name')
>>> mock.name
'my_name'
									

A simpler option is to simply set the “name” attribute after mock creation:

>>> mock = MagicMock()
>>> mock.name = "foo"
									

26.5.2.4. Attaching Mocks as Attributes ¶

When you attach a mock as an attribute of another mock (or as the return value) it becomes a “child” of that mock. Calls to the child are recorded in the method_calls and mock_calls attributes of the parent. This is useful for configuring child mocks and then attaching them to the parent, or for attaching mocks to a parent that records all calls to the children and allows you to make assertions about the order of calls between mocks:

>>> parent = MagicMock()
>>> child1 = MagicMock(return_value=None)
>>> child2 = MagicMock(return_value=None)
>>> parent.child1 = child1
>>> parent.child2 = child2
>>> child1(1)
>>> child2(2)
>>> parent.mock_calls
[call.child1(1), call.child2(2)]
									

The exception to this is if the mock has a name. This allows you to prevent the “parenting” if for some reason you don’t want it to happen.

>>> mock = MagicMock()
>>> not_a_child = MagicMock(name='not-a-child')
>>> mock.attribute = not_a_child
>>> mock.attribute()
<MagicMock name='not-a-child()' id='...'>
>>> mock.mock_calls
[]
									

Mocks created for you by patch() are automatically given names. To attach mocks that have names to a parent you use the attach_mock() 方法：

>>> thing1 = object()
>>> thing2 = object()
>>> parent = MagicMock()
>>> with patch('__main__.thing1', return_value=None) as child1:
...     with patch('__main__.thing2', return_value=None) as child2:
...         parent.attach_mock(child1, 'child1')
...         parent.attach_mock(child2, 'child2')
...         child1('one')
...         child2('two')
...
>>> parent.mock_calls
[call.child1('one'), call.child2('two')]
									

26.5.3.1. patch ¶

unittest.mock. patch ( target , new=DEFAULT , spec=None , create=False , spec_set=None , autospec=None , new_callable=None , **kwargs ) ¶

patch() acts as a function decorator, class decorator or a context manager. Inside the body of the function or with statement, the target is patched with a new object. When the function/with statement exits the patch is undone.

若 new is omitted, then the target is replaced with a MagicMock 。若 patch() is used as a decorator and new is omitted, the created mock is passed in as an extra argument to the decorated function. If patch() is used as a context manager the created mock is returned by the context manager.

target should be a string in the form 'package.module.ClassName' 。 target is imported and the specified object replaced with the new object, so the target must be importable from the environment you are calling patch() from. The target is imported when the decorated function is executed, not at decoration time.

spec and spec_set keyword arguments are passed to the MagicMock if patch is creating one for you.

In addition you can pass spec=True or spec_set=True , which causes patch to pass in the object being mocked as the spec/spec_set object.

new_callable allows you to specify a different class, or callable object, that will be called to create the new object. By default MagicMock 被使用。

A more powerful form of spec is autospec . If you set autospec=True then the mock will be created with a spec from the object being replaced. All attributes of the mock will also have the spec of the corresponding attribute of the object being replaced. Methods and functions being mocked will have their arguments checked and will raise a TypeError if they are called with the wrong signature. For mocks replacing a class, their return value (the ‘instance’) will have the same spec as the class. See the create_autospec() 函数和 Autospeccing .

Instead of autospec=True you can pass autospec=some_object to use an arbitrary object as the spec instead of the one being replaced.

默认情况下 patch() will fail to replace attributes that don’t exist. If you pass in create=True , and the attribute doesn’t exist, patch will create the attribute for you when the patched function is called, and delete it again afterwards. This is useful for writing tests against attributes that your production code creates at runtime. It is off by default because it can be dangerous. With it switched on you can write passing tests against APIs that don’t actually exist!

注意

3.5 版改变： If you are patching builtins in a module then you don’t need to pass create=True , it will be added by default.

Patch can be used as a TestCase class decorator. It works by decorating each test method in the class. This reduces the boilerplate code when your test methods share a common patchings set. patch() finds tests by looking for method names that start with patch.TEST_PREFIX . By default this is 'test' , which matches the way unittest finds tests. You can specify an alternative prefix by setting patch.TEST_PREFIX .

Patch can be used as a context manager, with the with statement. Here the patching applies to the indented block after the with statement. If you use “as” then the patched object will be bound to the name after the “as”; very useful if patch() is creating a mock object for you.

patch() takes arbitrary keyword arguments. These will be passed to the Mock (或 new_callable ) on construction.

patch.dict(...) , patch.multiple(...) and patch.object(...) are available for alternate use-cases.

patch() as function decorator, creating the mock for you and passing it into the decorated function:

>>> @patch('__main__.SomeClass')
... def function(normal_argument, mock_class):
...     print(mock_class is SomeClass)
...
>>> function(None)
True
									

Patching a class replaces the class with a MagicMock instance . If the class is instantiated in the code under test then it will be the return_value of the mock that will be used.

If the class is instantiated multiple times you could use side_effect to return a new mock each time. Alternatively you can set the return_value to be anything you want.

To configure return values on methods of instances on the patched class you must do this on the return_value 。例如：

>>> class Class:
...     def method(self):
...         pass
...
>>> with patch('__main__.Class') as MockClass:
...     instance = MockClass.return_value
...     instance.method.return_value = 'foo'
...     assert Class() is instance
...     assert Class().method() == 'foo'
...
									

若使用 spec or spec_set and patch() is replacing a class , then the return value of the created mock will have the same spec.

>>> Original = Class
>>> patcher = patch('__main__.Class', spec=True)
>>> MockClass = patcher.start()
>>> instance = MockClass()
>>> assert isinstance(instance, Original)
>>> patcher.stop()
									

new_callable argument is useful where you want to use an alternative class to the default MagicMock for the created mock. For example, if you wanted a NonCallableMock to be used:

>>> thing = object()
>>> with patch('__main__.thing', new_callable=NonCallableMock) as mock_thing:
...     assert thing is mock_thing
...     thing()
...
Traceback (most recent call last):
  ...
TypeError: 'NonCallableMock' object is not callable
									

Another use case might be to replace an object with an io.StringIO instance:

>>> from io import StringIO
>>> def foo():
...     print('Something')
...
>>> @patch('sys.stdout', new_callable=StringIO)
... def test(mock_stdout):
...     foo()
...     assert mock_stdout.getvalue() == 'Something\n'
...
>>> test()
									

当 patch() is creating a mock for you, it is common that the first thing you need to do is to configure the mock. Some of that configuration can be done in the call to patch. Any arbitrary keywords you pass into the call will be used to set attributes on the created mock:

>>> patcher = patch('__main__.thing', first='one', second='two')
>>> mock_thing = patcher.start()
>>> mock_thing.first
'one'
>>> mock_thing.second
'two'
									

As well as attributes on the created mock attributes, like the return_value and side_effect , of child mocks can also be configured. These aren’t syntactically valid to pass in directly as keyword arguments, but a dictionary with these as keys can still be expanded into a patch() call using ** :

>>> config = {'method.return_value': 3, 'other.side_effect': KeyError}
>>> patcher = patch('__main__.thing', **config)
>>> mock_thing = patcher.start()
>>> mock_thing.method()
3
>>> mock_thing.other()
Traceback (most recent call last):
  ...
KeyError
									

26.5.3.2. patch.object ¶

patch. object ( target , 属性 , new=DEFAULT , spec=None , create=False , spec_set=None , autospec=None , new_callable=None , **kwargs ) ¶

patch the named member ( 属性 ) on an object ( target ) with a mock object.

patch.object() can be used as a decorator, class decorator or a context manager. Arguments new , spec , create , spec_set , autospec and new_callable have the same meaning as for patch() 。像 patch() , patch.object() takes arbitrary keyword arguments for configuring the mock object it creates.

When used as a class decorator patch.object() honours patch.TEST_PREFIX for choosing which methods to wrap.

You can either call patch.object() with three arguments or two arguments. The three argument form takes the object to be patched, the attribute name and the object to replace the attribute with.

When calling with the two argument form you omit the replacement object, and a mock is created for you and passed in as an extra argument to the decorated function:

>>> @patch.object(SomeClass, 'class_method')
... def test(mock_method):
...     SomeClass.class_method(3)
...     mock_method.assert_called_with(3)
...
>>> test()
									

spec , create and the other arguments to patch.object() have the same meaning as they do for patch() .

26.5.3.3. patch.dict ¶

patch. dict ( in_dict , values=() , clear=False , **kwargs ) ¶

Patch a dictionary, or dictionary like object, and restore the dictionary to its original state after the test.

in_dict can be a dictionary or a mapping like container. If it is a mapping then it must at least support getting, setting and deleting items plus iterating over keys.

in_dict can also be a string specifying the name of the dictionary, which will then be fetched by importing it.

values can be a dictionary of values to set in the dictionary. values can also be an iterable of (key, value) pairs.

若 clear is true then the dictionary will be cleared before the new values are set.

patch.dict() can also be called with arbitrary keyword arguments to set values in the dictionary.

patch.dict() can be used as a context manager, decorator or class decorator. When used as a class decorator patch.dict() honours patch.TEST_PREFIX for choosing which methods to wrap.

patch.dict() can be used to add members to a dictionary, or simply let a test change a dictionary, and ensure the dictionary is restored when the test ends.

>>> foo = {}
>>> with patch.dict(foo, {'newkey': 'newvalue'}):
...     assert foo == {'newkey': 'newvalue'}
...
>>> assert foo == {}
									

>>> import os
>>> with patch.dict('os.environ', {'newkey': 'newvalue'}):
...     print(os.environ['newkey'])
...
newvalue
>>> assert 'newkey' not in os.environ
									

Keywords can be used in the patch.dict() call to set values in the dictionary:

>>> mymodule = MagicMock()
>>> mymodule.function.return_value = 'fish'
>>> with patch.dict('sys.modules', mymodule=mymodule):
...     import mymodule
...     mymodule.function('some', 'args')
...
'fish'
									

patch.dict() can be used with dictionary like objects that aren’t actually dictionaries. At the very minimum they must support item getting, setting, deleting and either iteration or membership test. This corresponds to the magic methods __getitem__() , __setitem__() , __delitem__() and either __iter__() or __contains__() .

>>> class Container:
...     def __init__(self):
...         self.values = {}
...     def __getitem__(self, name):
...         return self.values[name]
...     def __setitem__(self, name, value):
...         self.values[name] = value
...     def __delitem__(self, name):
...         del self.values[name]
...     def __iter__(self):
...         return iter(self.values)
...
>>> thing = Container()
>>> thing['one'] = 1
>>> with patch.dict(thing, one=2, two=3):
...     assert thing['one'] == 2
...     assert thing['two'] == 3
...
>>> assert thing['one'] == 1
>>> assert list(thing) == ['one']
									

26.5.3.4. patch.multiple ¶

patch. multiple ( target , spec=None , create=False , spec_set=None , autospec=None , new_callable=None , **kwargs ) ¶

Perform multiple patches in a single call. It takes the object to be patched (either as an object or a string to fetch the object by importing) and keyword arguments for the patches:

with patch.multiple(settings, FIRST_PATCH='one', SECOND_PATCH='two'):
    ...
											

使用 DEFAULT as the value if you want patch.multiple() to create mocks for you. In this case the created mocks are passed into a decorated function by keyword, and a dictionary is returned when patch.multiple() is used as a context manager.

patch.multiple() can be used as a decorator, class decorator or a context manager. The arguments spec , spec_set , create , autospec and new_callable have the same meaning as for patch() . These arguments will be applied to all patches done by patch.multiple() .

When used as a class decorator patch.multiple() honours patch.TEST_PREFIX for choosing which methods to wrap.

If you want patch.multiple() to create mocks for you, then you can use DEFAULT as the value. If you use patch.multiple() as a decorator then the created mocks are passed into the decorated function by keyword.

>>> thing = object()
>>> other = object()
									

>>> @patch.multiple('__main__', thing=DEFAULT, other=DEFAULT)
... def test_function(thing, other):
...     assert isinstance(thing, MagicMock)
...     assert isinstance(other, MagicMock)
...
>>> test_function()
									

patch.multiple() can be nested with other patch decorators, but put arguments passed by keyword after any of the standard arguments created by patch() :

>>> @patch('sys.exit')
... @patch.multiple('__main__', thing=DEFAULT, other=DEFAULT)
... def test_function(mock_exit, other, thing):
...     assert 'other' in repr(other)
...     assert 'thing' in repr(thing)
...     assert 'exit' in repr(mock_exit)
...
>>> test_function()
									

若 patch.multiple() is used as a context manager, the value returned by the context manger is a dictionary where created mocks are keyed by name:

>>> with patch.multiple('__main__', thing=DEFAULT, other=DEFAULT) as values:
...     assert 'other' in repr(values['other'])
...     assert 'thing' in repr(values['thing'])
...     assert values['thing'] is thing
...     assert values['other'] is other
...
									

26.5.3.5. patch methods: start and stop ¶

All the patchers have start() and stop() methods. These make it simpler to do patching in setUp methods or where you want to do multiple patches without nesting decorators or with statements.

To use them call patch() , patch.object() or patch.dict() as normal and keep a reference to the returned patcher object. You can then call start() to put the patch in place and stop() to undo it.

If you are using patch() to create a mock for you then it will be returned by the call to patcher.start .

>>> patcher = patch('package.module.ClassName')
>>> from package import module
>>> original = module.ClassName
>>> new_mock = patcher.start()
>>> assert module.ClassName is not original
>>> assert module.ClassName is new_mock
>>> patcher.stop()
>>> assert module.ClassName is original
>>> assert module.ClassName is not new_mock
									

A typical use case for this might be for doing multiple patches in the setUp method of a TestCase :

>>> class MyTest(TestCase):
...     def setUp(self):
...         self.patcher1 = patch('package.module.Class1')
...         self.patcher2 = patch('package.module.Class2')
...         self.MockClass1 = self.patcher1.start()
...         self.MockClass2 = self.patcher2.start()
...
...     def tearDown(self):
...         self.patcher1.stop()
...         self.patcher2.stop()
...
...     def test_something(self):
...         assert package.module.Class1 is self.MockClass1
...         assert package.module.Class2 is self.MockClass2
...
>>> MyTest('test_something').run()
									

>>> class MyTest(TestCase):
...     def setUp(self):
...         patcher = patch('package.module.Class')
...         self.MockClass = patcher.start()
...         self.addCleanup(patcher.stop)
...
...     def test_something(self):
...         assert package.module.Class is self.MockClass
...
										

It is also possible to stop all patches which have been started by using patch.stopall() .

patch. stopall ( ) ¶

Stop all active patches. Only stops patches started with start .

26.5.3.6. patch builtins ¶

You can patch any builtins within a module. The following example patches builtin ord() :

>>> @patch('__main__.ord')
... def test(mock_ord):
...     mock_ord.return_value = 101
...     print(ord('c'))
...
>>> test()
101
									

26.5.3.7. TEST_PREFIX ¶

All of the patchers can be used as class decorators. When used in this way they wrap every test method on the class. The patchers recognise methods that start with 'test' as being test methods. This is the same way that the unittest.TestLoader finds test methods by default.

It is possible that you want to use a different prefix for your tests. You can inform the patchers of the different prefix by setting patch.TEST_PREFIX :

>>> patch.TEST_PREFIX = 'foo'
>>> value = 3
>>>
>>> @patch('__main__.value', 'not three')
... class Thing:
...     def foo_one(self):
...         print(value)
...     def foo_two(self):
...         print(value)
...
>>>
>>> Thing().foo_one()
not three
>>> Thing().foo_two()
not three
>>> value
3
									

26.5.3.8. Nesting Patch Decorators ¶

If you want to perform multiple patches then you can simply stack up the decorators.

You can stack up multiple patch decorators using this pattern:

>>> @patch.object(SomeClass, 'class_method')
... @patch.object(SomeClass, 'static_method')
... def test(mock1, mock2):
...     assert SomeClass.static_method is mock1
...     assert SomeClass.class_method is mock2
...     SomeClass.static_method('foo')
...     SomeClass.class_method('bar')
...     return mock1, mock2
...
>>> mock1, mock2 = test()
>>> mock1.assert_called_once_with('foo')
>>> mock2.assert_called_once_with('bar')
									

Note that the decorators are applied from the bottom upwards. This is the standard way that Python applies decorators. The order of the created mocks passed into your test function matches this order.

26.5.3.9. Where to patch ¶

patch() works by (temporarily) changing the object that a name points to with another one. There can be many names pointing to any individual object, so for patching to work you must ensure that you patch the name used by the system under test.

The basic principle is that you patch where an object is looked up , which is not necessarily the same place as where it is defined. A couple of examples will help to clarify this.

Imagine we have a project that we want to test with the following structure:

a.py
    -> Defines SomeClass
b.py
    -> from a import SomeClass
    -> some_function instantiates SomeClass
									

Now we want to test some_function but we want to mock out SomeClass 使用 patch() . The problem is that when we import module b, which we will have to do then it imports SomeClass from module a. If we use patch() to mock out a.SomeClass then it will have no effect on our test; module b already has a reference to the real SomeClass and it looks like our patching had no effect.

The key is to patch out SomeClass where it is used (or where it is looked up ). In this case some_function will actually look up SomeClass in module b, where we have imported it. The patching should look like:

@patch('b.SomeClass')
									

However, consider the alternative scenario where instead of from a import SomeClass module b does import a and some_function 使用 a.SomeClass . Both of these import forms are common. In this case the class we want to patch is being looked up in the module and so we have to patch a.SomeClass instead:

@patch('a.SomeClass')
									

26.5.3.10. Patching Descriptors and Proxy Objects ¶

Both patch and patch.object correctly patch and restore descriptors: class methods, static methods and properties. You should patch these on the class rather than an instance. They also work with some objects that proxy attribute access, like the django settings object .

26.5.4.1. Mocking Magic Methods ¶

Mock supports mocking the Python protocol methods, also known as “magic methods”. This allows mock objects to replace containers or other objects that implement Python protocols.

Because magic methods are looked up differently from normal methods [2] , this support has been specially implemented. This means that only specific magic methods are supported. The supported list includes almost all of them. If there are any missing that you need please let us know.

You mock magic methods by setting the method you are interested in to a function or a mock instance. If you are using a function then it must take self as the first argument [3] .

>>> def __str__(self):
...     return 'fooble'
...
>>> mock = Mock()
>>> mock.__str__ = __str__
>>> str(mock)
'fooble'
									

>>> mock = Mock()
>>> mock.__str__ = Mock()
>>> mock.__str__.return_value = 'fooble'
>>> str(mock)
'fooble'
									

>>> mock = Mock()
>>> mock.__iter__ = Mock(return_value=iter([]))
>>> list(mock)
[]
									

One use case for this is for mocking objects used as context managers in a with 语句：

>>> mock = Mock()
>>> mock.__enter__ = Mock(return_value='foo')
>>> mock.__exit__ = Mock(return_value=False)
>>> with mock as m:
...     assert m == 'foo'
...
>>> mock.__enter__.assert_called_with()
>>> mock.__exit__.assert_called_with(None, None, None)
									

Calls to magic methods do not appear in method_calls , but they are recorded in mock_calls .

The full list of supported magic methods is:

• __hash__ , __sizeof__ , __repr__ and __str__
• __dir__ , __format__ and __subclasses__
• __floor__ , __trunc__ and __ceil__
• Comparisons: __lt__ , __gt__ , __le__ , __ge__ , __eq__ and __ne__
• Container methods: __getitem__ , __setitem__ , __delitem__ , __contains__ , __len__ , __iter__ , __reversed__ and __missing__
• Context manager: __enter__ and __exit__
• Unary numeric methods: __neg__ , __pos__ and __invert__
• The numeric methods (including right hand and in-place variants): __add__ , __sub__ , __mul__ , __matmul__ , __div__ , __truediv__ , __floordiv__ , __mod__ , __divmod__ , __lshift__ , __rshift__ , __and__ , __xor__ , __or__ ，和 __pow__
• Numeric conversion methods: __complex__ , __int__ , __float__ and __index__
• Descriptor methods: __get__ , __set__ and __delete__
• Pickling: __reduce__ , __reduce_ex__ , __getinitargs__ , __getnewargs__ , __getstate__ and __setstate__

The following methods exist but are not supported as they are either in use by mock, can’t be set dynamically, or can cause problems:

• __getattr__ , __setattr__ , __init__ and __new__
• __prepare__ , __instancecheck__ , __subclasscheck__ , __del__

26.5.4.2. Magic Mock ¶

There are two MagicMock variants: MagicMock and NonCallableMagicMock .

class unittest.mock. MagicMock ( *args , **kw ) ¶

MagicMock 是子类对于 Mock with default implementations of most of the magic methods. You can use MagicMock without having to configure the magic methods yourself.

The constructor parameters have the same meaning as for Mock .

若使用 spec or spec_set arguments then only magic methods that exist in the spec will be created.

class unittest.mock. NonCallableMagicMock ( *args , **kw ) ¶

A non-callable version of MagicMock .

The constructor parameters have the same meaning as for MagicMock , with the exception of return_value and side_effect which have no meaning on a non-callable mock.

The magic methods are setup with MagicMock objects, so you can configure them and use them in the usual way:

>>> mock = MagicMock()
>>> mock[3] = 'fish'
>>> mock.__setitem__.assert_called_with(3, 'fish')
>>> mock.__getitem__.return_value = 'result'
>>> mock[2]
'result'
									

By default many of the protocol methods are required to return objects of a specific type. These methods are preconfigured with a default return value, so that they can be used without you having to do anything if you aren’t interested in the return value. You can still set the return value manually if you want to change the default.

Methods and their defaults:

• __lt__ : NotImplemented
• __gt__ : NotImplemented
• __le__ : NotImplemented
• __ge__ : NotImplemented
• __int__ : 1
• __contains__ : False
• __len__ : 0
• __iter__ : iter([])
• __exit__ : False
• __complex__ : 1j
• __float__ : 1.0
• __bool__ : True
• __index__ : 1
• __hash__ : default hash for the mock
• __str__ : default str for the mock
• __sizeof__ : default sizeof for the mock

例如：

>>> mock = MagicMock()
>>> int(mock)
1
>>> len(mock)
0
>>> list(mock)
[]
>>> object() in mock
False
									

The two equality methods, __eq__() and __ne__() , are special. They do the default equality comparison on identity, using the side_effect attribute, unless you change their return value to return something else:

>>> MagicMock() == 3
False
>>> MagicMock() != 3
True
>>> mock = MagicMock()
>>> mock.__eq__.return_value = True
>>> mock == 3
True
									

The return value of MagicMock.__iter__() can be any iterable object and isn’t required to be an iterator:

>>> mock = MagicMock()
>>> mock.__iter__.return_value = ['a', 'b', 'c']
>>> list(mock)
['a', 'b', 'c']
>>> list(mock)
['a', 'b', 'c']
									

If the return value is an iterator, then iterating over it once will consume it and subsequent iterations will result in an empty list:

>>> mock.__iter__.return_value = iter(['a', 'b', 'c'])
>>> list(mock)
['a', 'b', 'c']
>>> list(mock)
[]
									

MagicMock has all of the supported magic methods configured except for some of the obscure and obsolete ones. You can still set these up if you want.

Magic methods that are supported but not setup by default in MagicMock 是：

• __subclasses__
• __dir__
• __format__
• __get__ , __set__ and __delete__
• __reversed__ and __missing__
• __reduce__ , __reduce_ex__ , __getinitargs__ , __getnewargs__ , __getstate__ and __setstate__
• __getformat__ and __setformat__

26.5.5.1. sentinel ¶

unittest.mock. sentinel ¶

sentinel object provides a convenient way of providing unique objects for your tests.

Attributes are created on demand when you access them by name. Accessing the same attribute will always return the same object. The objects returned have a sensible repr so that test failure messages are readable.

sentinel attributes don’t preserve their identity when they are copied or pickled .

Sometimes when testing you need to test that a specific object is passed as an argument to another method, or returned. It can be common to create named sentinel objects to test this. sentinel provides a convenient way of creating and testing the identity of objects like this.

In this example we monkey patch 方法 to return sentinel.some_object :

>>> real = ProductionClass()
>>> real.method = Mock(name="method")
>>> real.method.return_value = sentinel.some_object
>>> result = real.method()
>>> assert result is sentinel.some_object
>>> sentinel.some_object
sentinel.some_object
									

26.5.5.2. DEFAULT ¶

unittest.mock. DEFAULT ¶

DEFAULT object is a pre-created sentinel (actually sentinel.DEFAULT ). It can be used by side_effect functions to indicate that the normal return value should be used.

26.5.5.3. 调用 ¶

unittest.mock. call ( *args , **kwargs ) ¶

call() is a helper object for making simpler assertions, for comparing with call_args , call_args_list , mock_calls and method_calls . call() can also be used with assert_has_calls() .

>>> m = MagicMock(return_value=None)
>>> m(1, 2, a='foo', b='bar')
>>> m()
>>> m.call_args_list == [call(1, 2, a='foo', b='bar'), call()]
True
											

调用。 call_list ( ) ¶

For a call object that represents multiple calls, call_list() returns a list of all the intermediate calls as well as the final call.

call_list is particularly useful for making assertions on “chained calls”. A chained call is multiple calls on a single line of code. This results in multiple entries in mock_calls on a mock. Manually constructing the sequence of calls can be tedious.

call_list() can construct the sequence of calls from the same chained call:

>>> m = MagicMock()
>>> m(1).method(arg='foo').other('bar')(2.0)
<MagicMock name='mock().method().other()()' id='...'>
>>> kall = call(1).method(arg='foo').other('bar')(2.0)
>>> kall.call_list()
[call(1),
 call().method(arg='foo'),
 call().method().other('bar'),
 call().method().other()(2.0)]
>>> m.mock_calls == kall.call_list()
True
									

A call object is either a tuple of (positional args, keyword args) or (name, positional args, keyword args) depending on how it was constructed. When you construct them yourself this isn’t particularly interesting, but the call objects that are in the Mock.call_args , Mock.call_args_list and Mock.mock_calls attributes can be introspected to get at the individual arguments they contain.

call objects in Mock.call_args and Mock.call_args_list are two-tuples of (positional args, keyword args) whereas the call objects in Mock.mock_calls , along with ones you construct yourself, are three-tuples of (name, positional args, keyword args).

You can use their “tupleness” to pull out the individual arguments for more complex introspection and assertions. The positional arguments are a tuple (an empty tuple if there are no positional arguments) and the keyword arguments are a dictionary:

>>> m = MagicMock(return_value=None)
>>> m(1, 2, 3, arg='one', arg2='two')
>>> kall = m.call_args
>>> args, kwargs = kall
>>> args
(1, 2, 3)
>>> kwargs
{'arg2': 'two', 'arg': 'one'}
>>> args is kall[0]
True
>>> kwargs is kall[1]
True
									

>>> m = MagicMock()
>>> m.foo(4, 5, 6, arg='two', arg2='three')
<MagicMock name='mock.foo()' id='...'>
>>> kall = m.mock_calls[0]
>>> name, args, kwargs = kall
>>> name
'foo'
>>> args
(4, 5, 6)
>>> kwargs
{'arg2': 'three', 'arg': 'two'}
>>> name is m.mock_calls[0][0]
True
									

26.5.5.4. create_autospec ¶

unittest.mock. create_autospec ( spec , spec_set=False , instance=False , **kwargs ) ¶

Create a mock object using another object as a spec. Attributes on the mock will use the corresponding attribute on the spec object as their spec.

Functions or methods being mocked will have their arguments checked to ensure that they are called with the correct signature.

若 spec_set is True then attempting to set attributes that don’t exist on the spec object will raise an AttributeError .

If a class is used as a spec then the return value of the mock (the instance of the class) will have the same spec. You can use a class as the spec for an instance object by passing instance=True . The returned mock will only be callable if instances of the mock are callable.

create_autospec() also takes arbitrary keyword arguments that are passed to the constructor of the created mock.

见 Autospeccing for examples of how to use auto-speccing with create_autospec() 和 autospec 自变量为 patch() .

26.5.5.5. ANY ¶

unittest.mock. ANY ¶

Sometimes you may need to make assertions about some of the arguments in a call to mock, but either not care about some of the arguments or want to pull them individually out of call_args and make more complex assertions on them.

To ignore certain arguments you can pass in objects that compare equal to everything . Calls to assert_called_with() and assert_called_once_with() will then succeed no matter what was passed in.

>>> mock = Mock(return_value=None)
>>> mock('foo', bar=object())
>>> mock.assert_called_once_with('foo', bar=ANY)
									

ANY can also be used in comparisons with call lists like mock_calls :

>>> m = MagicMock(return_value=None)
>>> m(1)
>>> m(1, 2)
>>> m(object())
>>> m.mock_calls == [call(1), call(1, 2), ANY]
True
									

26.5.5.6. FILTER_DIR ¶

unittest.mock. FILTER_DIR ¶

FILTER_DIR is a module level variable that controls the way mock objects respond to dir() (only for Python 2.6 or more recent). The default is True , which uses the filtering described below, to only show useful members. If you dislike this filtering, or need to switch it off for diagnostic purposes, then set mock.FILTER_DIR = False .

With filtering on, dir(some_mock) shows only useful attributes and will include any dynamically created attributes that wouldn’t normally be shown. If the mock was created with a spec (或 autospec of course) then all the attributes from the original are shown, even if they haven’t been accessed yet:

>>> dir(Mock())
['assert_any_call',
 'assert_called_once_with',
 'assert_called_with',
 'assert_has_calls',
 'attach_mock',
 ...
>>> from urllib import request
>>> dir(Mock(spec=request))
['AbstractBasicAuthHandler',
 'AbstractDigestAuthHandler',
 'AbstractHTTPHandler',
 'BaseHandler',
 ...
									

Many of the not-very-useful (private to Mock rather than the thing being mocked) underscore and double underscore prefixed attributes have been filtered from the result of calling dir() 在 Mock . If you dislike this behaviour you can switch it off by setting the module level switch FILTER_DIR :

>>> from unittest import mock
>>> mock.FILTER_DIR = False
>>> dir(mock.Mock())
['_NonCallableMock__get_return_value',
 '_NonCallableMock__get_side_effect',
 '_NonCallableMock__return_value_doc',
 '_NonCallableMock__set_return_value',
 '_NonCallableMock__set_side_effect',
 '__call__',
 '__class__',
 ...
									

Alternatively you can just use vars(my_mock) (instance members) and dir(type(my_mock)) (type members) to bypass the filtering irrespective of mock.FILTER_DIR .

26.5.5.7. mock_open ¶

unittest.mock. mock_open ( mock=None , read_data=None ) ¶

A helper function to create a mock to replace the use of open() . It works for open() called directly or used as a context manager.

mock argument is the mock object to configure. If None (the default) then a MagicMock will be created for you, with the API limited to methods or attributes available on standard file handles.

read_data is a string for the read() , readline() ，和 readlines() methods of the file handle to return. Calls to those methods will take data from read_data until it is depleted. The mock of these methods is pretty simplistic: every time the mock is called, the read_data is rewound to the start. If you need more control over the data that you are feeding to the tested code you will need to customize this mock for yourself. When that is insufficient, one of the in-memory filesystem packages on PyPI can offer a realistic filesystem for testing.

3.4 版改变： 添加 readline() and readlines() support. The mock of read() changed to consume read_data rather than returning it on each call.

3.5 版改变： read_data is now reset on each call to the mock .

使用 open() as a context manager is a great way to ensure your file handles are closed properly and is becoming common:

with open('/some/path', 'w') as f:
    f.write('something')
									

The issue is that even if you mock out the call to open() it is the returned object that is used as a context manager (and has __enter__() and __exit__() called).

Mocking context managers with a MagicMock is common enough and fiddly enough that a helper function is useful.

>>> m = mock_open()
>>> with patch('__main__.open', m):
...     with open('foo', 'w') as h:
...         h.write('some stuff')
...
>>> m.mock_calls
[call('foo', 'w'),
 call().__enter__(),
 call().write('some stuff'),
 call().__exit__(None, None, None)]
>>> m.assert_called_once_with('foo', 'w')
>>> handle = m()
>>> handle.write.assert_called_once_with('some stuff')
									

And for reading files:

>>> with patch('__main__.open', mock_open(read_data='bibble')) as m:
...     with open('foo') as h:
...         result = h.read()
...
>>> m.assert_called_once_with('foo')
>>> assert result == 'bibble'
									

26.5.5.8. Autospeccing ¶

Autospeccing is based on the existing spec feature of mock. It limits the api of mocks to the api of an original object (the spec), but it is recursive (implemented lazily) so that attributes of mocks only have the same api as the attributes of the spec. In addition mocked functions / methods have the same call signature as the original so they raise a TypeError if they are called incorrectly.

Before I explain how auto-speccing works, here’s why it is needed.

Mock is a very powerful and flexible object, but it suffers from two flaws when used to mock out objects from a system under test. One of these flaws is specific to the Mock api and the other is a more general problem with using mock objects.

First the problem specific to Mock . Mock has two assert methods that are extremely handy: assert_called_with() and assert_called_once_with() .

>>> mock = Mock(name='Thing', return_value=None)
>>> mock(1, 2, 3)
>>> mock.assert_called_once_with(1, 2, 3)
>>> mock(1, 2, 3)
>>> mock.assert_called_once_with(1, 2, 3)
Traceback (most recent call last):
 ...
AssertionError: Expected 'mock' to be called once. Called 2 times.
									

Because mocks auto-create attributes on demand, and allow you to call them with arbitrary arguments, if you misspell one of these assert methods then your assertion is gone:

>>> mock = Mock(name='Thing', return_value=None)
>>> mock(1, 2, 3)
>>> mock.assret_called_once_with(4, 5, 6)
									

Your tests can pass silently and incorrectly because of the typo.

The second issue is more general to mocking. If you refactor some of your code, rename members and so on, any tests for code that is still using the old api but uses mocks instead of the real objects will still pass. This means your tests can all pass even though your code is broken.

Note that this is another reason why you need integration tests as well as unit tests. Testing everything in isolation is all fine and dandy, but if you don’t test how your units are “wired together” there is still lots of room for bugs that tests might have caught.

mock already provides a feature to help with this, called speccing. If you use a class or instance as the spec for a mock then you can only access attributes on the mock that exist on the real class:

>>> from urllib import request
>>> mock = Mock(spec=request.Request)
>>> mock.assret_called_with
Traceback (most recent call last):
 ...
AttributeError: Mock object has no attribute 'assret_called_with'
									

The spec only applies to the mock itself, so we still have the same issue with any methods on the mock:

>>> mock.has_data()
<mock.Mock object at 0x...>
>>> mock.has_data.assret_called_with()
									

Auto-speccing solves this problem. You can either pass autospec=True to patch() / patch.object() or use the create_autospec() function to create a mock with a spec. If you use the autospec=True 自变量为 patch() then the object that is being replaced will be used as the spec object. Because the speccing is done “lazily” (the spec is created as attributes on the mock are accessed) you can use it with very complex or deeply nested objects (like modules that import modules that import modules) without a big performance hit.

Here’s an example of it in use:

>>> from urllib import request
>>> patcher = patch('__main__.request', autospec=True)
>>> mock_request = patcher.start()
>>> request is mock_request
True
>>> mock_request.Request
<MagicMock name='request.Request' spec='Request' id='...'>
									

You can see that request.Request has a spec. request.Request takes two arguments in the constructor (one of which is self ). Here’s what happens if we try to call it incorrectly:

>>> req = request.Request()
Traceback (most recent call last):
 ...
TypeError: <lambda>() takes at least 2 arguments (1 given)
									

The spec also applies to instantiated classes (i.e. the return value of specced mocks):

>>> req = request.Request('foo')
>>> req
<NonCallableMagicMock name='request.Request()' spec='Request' id='...'>
									

Request objects are not callable, so the return value of instantiating our mocked out request.Request is a non-callable mock. With the spec in place any typos in our asserts will raise the correct error:

>>> req.add_header('spam', 'eggs')
<MagicMock name='request.Request().add_header()' id='...'>
>>> req.add_header.assret_called_with
Traceback (most recent call last):
 ...
AttributeError: Mock object has no attribute 'assret_called_with'
>>> req.add_header.assert_called_with('spam', 'eggs')
									

In many cases you will just be able to add autospec=True to your existing patch() calls and then be protected against bugs due to typos and api changes.

As well as using autospec through patch() there is a create_autospec() for creating autospecced mocks directly:

>>> from urllib import request
>>> mock_request = create_autospec(request)
>>> mock_request.Request('foo', 'bar')
<NonCallableMagicMock name='mock.Request()' spec='Request' id='...'>
									

This isn’t without caveats and limitations however, which is why it is not the default behaviour. In order to know what attributes are available on the spec object, autospec has to introspect (access attributes) the spec. As you traverse attributes on the mock a corresponding traversal of the original object is happening under the hood. If any of your specced objects have properties or descriptors that can trigger code execution then you may not be able to use autospec. On the other hand it is much better to design your objects so that introspection is safe [4] .

A more serious problem is that it is common for instance attributes to be created in the __init__() method and not to exist on the class at all. autospec can’t know about any dynamically created attributes and restricts the api to visible attributes.

>>> class Something:
...   def __init__(self):
...     self.a = 33
...
>>> with patch('__main__.Something', autospec=True):
...   thing = Something()
...   thing.a
...
Traceback (most recent call last):
  ...
AttributeError: Mock object has no attribute 'a'
									

There are a few different ways of resolving this problem. The easiest, but not necessarily the least annoying, way is to simply set the required attributes on the mock after creation. Just because autospec doesn’t allow you to fetch attributes that don’t exist on the spec it doesn’t prevent you setting them:

>>> with patch('__main__.Something', autospec=True):
...   thing = Something()
...   thing.a = 33
...
									

There is a more aggressive version of both spec and autospec that does prevent you setting non-existent attributes. This is useful if you want to ensure your code only sets valid attributes too, but obviously it prevents this particular scenario:

>>> with patch('__main__.Something', autospec=True, spec_set=True):
...   thing = Something()
...   thing.a = 33
...
Traceback (most recent call last):
 ...
AttributeError: Mock object has no attribute 'a'
									

Probably the best way of solving the problem is to add class attributes as default values for instance members initialised in __init__() . Note that if you are only setting default attributes in __init__() then providing them via class attributes (shared between instances of course) is faster too. e.g.

class Something:
    a = 33
									

This brings up another issue. It is relatively common to provide a default value of None for members that will later be an object of a different type. None would be useless as a spec because it wouldn’t let you access any attributes or methods on it. As None is never going to be useful as a spec, and probably indicates a member that will normally of some other type, autospec doesn’t use a spec for members that are set to None . These will just be ordinary mocks (well - MagicMocks):

>>> class Something:
...     member = None
...
>>> mock = create_autospec(Something)
>>> mock.member.foo.bar.baz()
<MagicMock name='mock.member.foo.bar.baz()' id='...'>
									

If modifying your production classes to add defaults isn’t to your liking then there are more options. One of these is simply to use an instance as the spec rather than the class. The other is to create a subclass of the production class and add the defaults to the subclass without affecting the production class. Both of these require you to use an alternative object as the spec. Thankfully patch() supports this - you can simply pass the alternative object as the autospec argument:

>>> class Something:
...   def __init__(self):
...     self.a = 33
...
>>> class SomethingForTest(Something):
...   a = 33
...
>>> p = patch('__main__.Something', autospec=SomethingForTest)
>>> mock = p.start()
>>> mock.a
<NonCallableMagicMock name='Something.a' spec='int' id='...'>