26.6. unittest.mock --- 入門

バージョン 3.3 で追加.

26.6.1. Mock を使う

26.6.1.1. Mock のパッチ用メソッド

一般的な Mock の使い方の中には次のものがあります:

• メソッドにパッチを当てる
• オブジェクトに対するメソッド呼び出しを記録する

システムの他の部分からメソッドが正しい引数で呼び出されたかどうかを確認するために、そのオブジェクトのメソッドを置き換えることができます:

>>> real = SomeClass()
>>> real.method = MagicMock(name='method')
>>> real.method(3, 4, 5, key='value')
<MagicMock name='method()' id='...'>

モック(上の例では real.method)が利用された場合、どう使われたかを assert できるようにする属性やメソッドがモックにあります。

注釈

この例のような場合、たいてい Mock と MagicMock は互換です。 MagicMock の方が強力なので、デフォルトではこちらを使うといいでしょう。

モックが呼び出されると、その called 属性が True に設定されます。そして assert_called_with() や assert_called_once_with() メソッドを使ってそのメソッドが正しい引数で呼び出されたかどうかをチェックできます。

次の例では ProductionClass().method が something メソッドを呼び出したことをテストしています:

>>> class ProductionClass:
...     def method(self):
...         self.something(1, 2, 3)
...     def something(self, a, b, c):
...         pass
...
>>> real = ProductionClass()
>>> real.something = MagicMock()
>>> real.method()
>>> real.something.assert_called_once_with(1, 2, 3)

26.6.1.2. オブジェクトのメソッド呼び出しに対するモック

上の例ではオブジェクトのメソッドに対して直接パッチを当てて、それが正しく呼び出されていたかどうかをテストしていました。もう一つのよくあるユースケースが、モックをメソッド (またはテスト対象のシステムのどこか) に渡して、それが正しく利用されたかどうかをチェックする方法です。

次の例で、ProductionClass は closer メソッドを持っています。このメソッドは渡されたオブジェクトの close メソッドを呼び出します。

>>> class ProductionClass:
...     def closer(self, something):
...         something.close()
...

ですからこれををテストするには、close メソッドを持ったオブジェクトを渡して、それが正しく呼び出されたかどうかをテストしなければなりません。

>>> real = ProductionClass()
>>> mock = Mock()
>>> real.closer(mock)
>>> mock.close.assert_called_with()

モックに close メソッドを持たせるために何か準備する必要はありません。 close メソッドにアクセスすると自動的にそれが作られます。なので、もし close が呼び出されなかったとしてもテスト時に生成されるのですが、 assert_called_with() が failure 例外を発生させます。

26.6.1.3. クラスをモックする

他のよくあるユースケースが、テスト対象のコードによってインスタンス化されているクラスをモックに置き換えることです。クラスに patch すると、そのクラスがモックに置き換えられます。インスタンスは クラスを呼び出した時に 作られます。なので、モックの戻り値を使うことで、「モックのインスタンス」にアクセスできます。

次の例では、 some_function という関数が Foo をインスタンス化し、その method を呼び出しています。 patch() を呼び出すと Foo クラスをモックに置き換えます。 Foo のインスタンスはモックを呼び出して作られるので、モックの return_value を変更することでカスタマイズできます。

>>> def some_function():
...     instance = module.Foo()
...     return instance.method()
...
>>> with patch('module.Foo') as mock:
...     instance = mock.return_value
...     instance.method.return_value = 'the result'
...     result = some_function()
...     assert result == 'the result'

26.6.1.4. モックに名前をつける

モックに名前をつけると便利なことがあります。その名前はモックを repr したときに表示されるので、モックがテスト失敗のメッセージ内に現れた時に便利です。また、モックの名前はそのモックの属性やメソッドにも伝播します:

>>> mock = MagicMock(name='foo')
>>> mock
<MagicMock name='foo' id='...'>
>>> mock.method
<MagicMock name='foo.method' id='...'>

26.6.1.5. 全ての呼び出しのトラッキング

メソッドの複数回の呼び出しをトラックしたいことがあります。 mock_calls 属性は、そのモックの子属性やさらにその子孫に対する呼び出しすべてを記録しています。

>>> mock = MagicMock()
>>> mock.method()
<MagicMock name='mock.method()' id='...'>
>>> mock.attribute.method(10, x=53)
<MagicMock name='mock.attribute.method()' id='...'>
>>> mock.mock_calls
[call.method(), call.attribute.method(10, x=53)]

mock_calls に対して assert すると、予期していないメソッド呼び出しがあったときにその assert が失敗します。これはあるメソッド呼び出しが期待通りに実行されたかどうかだけでなく、その呼び出し順序や期待した以外の呼び出しが起こらなかったことまでテストできるので便利です:

mock_calls と比較するリストを作るために call オブジェクトを利用できます:

>>> expected = [call.method(), call.attribute.method(10, x=53)]
>>> mock.mock_calls == expected
True

26.6.1.6. 戻り値や属性を設定する

モックオブジェクトに戻り値を設定するのはとっても簡単です:

>>> mock = Mock()
>>> mock.return_value = 3
>>> mock()
3

もちろん同じことがモックのメソッドに対しても行えます:

>>> mock = Mock()
>>> mock.method.return_value = 3
>>> mock.method()
3

コンストラクターで戻り値を設定することもできます:

>>> mock = Mock(return_value=3)
>>> mock()
3

モックに属性を設定したかったら、普通に設定するだけです:

>>> mock = Mock()
>>> mock.x = 3
>>> mock.x
3

mock.connection.cursor().execute("SELECT 1") のような複雑なケースでモックを使いたい場合もあります。この呼出があるリストを返すようにしたい場合、このネストした呼び出しを構成しなければなりません。

call を使って "chained call" 内の呼び出しを構成して、 assert で使うことができます:

>>> mock = Mock()
>>> cursor = mock.connection.cursor.return_value
>>> cursor.execute.return_value = ['foo']
>>> mock.connection.cursor().execute("SELECT 1")
['foo']
>>> expected = call.connection.cursor().execute("SELECT 1").call_list()
>>> mock.mock_calls
[call.connection.cursor(), call.connection.cursor().execute('SELECT 1')]
>>> mock.mock_calls == expected
True

call object を chained call を表す list にするために .call_list() を使います。

26.6.1.7. モックから例外を発生させる

side_effect という便利な属性があります。この属性に例外クラスやそのインスタンスを設定すると、モックが呼ばれた時にその例外を発生させます。

>>> mock = Mock(side_effect=Exception('Boom!'))
>>> mock()
Traceback (most recent call last):
  ...
Exception: Boom!

26.6.1.8. side_effect の関数と iterable

side_effect には関数や iterable を設定することもできます。side_effect に iterable を設定するユースケースは、そのモックが複数回呼び出され、そのたびに違う値を返したい場合です。side_effect に iterable を設定すると、そのモックに対するすべての呼び出しは iterable の次の値を返します:

>>> mock = MagicMock(side_effect=[4, 5, 6])
>>> mock()
4
>>> mock()
5
>>> mock()
6

より高度なユースケースとして、mock が呼び出された時の引数によって戻り値を変化させたい場合は、side_effect に関数を設定することができます。その関数は mock と同じ引数で呼び出されます。その関数の戻り値がそのモック呼び出しの戻り値になります:

>>> vals = {(1, 2): 1, (2, 3): 2}
>>> def side_effect(*args):
...     return vals[args]
...
>>> mock = MagicMock(side_effect=side_effect)
>>> mock(1, 2)
1
>>> mock(2, 3)
2

26.6.1.9. 既存のオブジェクトから Mock を作る

mock を使いすぎることの問題の一つは、テストが実際のコードではなく mock の実装をテストするようになってしまうことです。some_method というメソッドを実装したクラスがあるとします。他のクラスをテストするときに、some_method を提供する mock を使います。最初のクラスをリファクタリングして some_method がなくなった時、コードは壊れているのにテストは通る状態になってしまいます

Mock は spec というキーワード引数で mock の定義となるオブジェクトを指定できます。定義オブジェクトに存在しないメソッドや属性にアクセスすると AttributeError を発生させます。定義となるクラスの実装を変更した場合、テストの中でそのクラスをインスタンス化させなくても、テストを失敗させる事ができます。

>>> mock = Mock(spec=SomeClass)
>>> mock.old_method()
Traceback (most recent call last):
   ...
AttributeError: object has no attribute 'old_method'

Using a specification also enables a smarter matching of calls made to the mock, regardless of whether some parameters were passed as positional or named arguments:

>>> def f(a, b, c): pass
...
>>> mock = Mock(spec=f)
>>> mock(1, 2, 3)
<Mock name='mock()' id='140161580456576'>
>>> mock.assert_called_with(a=1, b=2, c=3)

If you want this smarter matching to also work with method calls on the mock, you can use auto-speccing.

任意の属性の参照だけでなく代入も禁止するより強い定義を利用したい場合は、spec の代わりに spec_set を使います。

26.6.2. patch デコレータ

注釈

patch() では探索される名前空間内のオブジェクトにパッチをあてることが重要です。通常は単純ですが、クイックガイドには where-to-patch を読んでください。

テストの中でクラス属性やモジュール属性、例えば組み込み関数や、テスト対象モジュールにあるインスタンス化されるクラスに対してパッチしたいことがあります。モジュールやクラスは実際はグローバルなので、パッチするときは必ずテスト後にパッチを解除しないと、そのパッチが永続化されて他のテストに影響を与え、解析しにくい問題になります。

mock はこのために3つの便利なデコレータを提供しています: patch(), patch.object(), patch.dict() です。patch はパッチ対象を指定する package.module.Class.attribute の形式の文字列を引数に取ります。オプションでその属性 (やクラスなど) を置き換えるオブジェクトを渡すことができます。 'patch.object' はオブジェクトとパッチしたい属性名、それにオプションで置き換えるオブジェクトを受け取ります。

patch.object:

>>> original = SomeClass.attribute
>>> @patch.object(SomeClass, 'attribute', sentinel.attribute)
... def test():
...     assert SomeClass.attribute == sentinel.attribute
...
>>> test()
>>> assert SomeClass.attribute == original

>>> @patch('package.module.attribute', sentinel.attribute)
... def test():
...     from package.module import attribute
...     assert attribute is sentinel.attribute
...
>>> test()

モジュール (builtins を含む) をパッチしようとする場合、 patch.object() の代わりに patch() を使用してください:

>>> mock = MagicMock(return_value=sentinel.file_handle)
>>> with patch('builtins.open', mock):
...     handle = open('filename', 'r')
...
>>> mock.assert_called_with('filename', 'r')
>>> assert handle == sentinel.file_handle, "incorrect file handle returned"

モジュール名は必要に応じて package.module のようにドットを含むことができます:

>>> @patch('package.module.ClassName.attribute', sentinel.attribute)
... def test():
...     from package.module import ClassName
...     assert ClassName.attribute == sentinel.attribute
...
>>> test()

テストメソッド自体をデコレートするのは良いパターンです:

>>> class MyTest(unittest.TestCase):
...     @patch.object(SomeClass, 'attribute', sentinel.attribute)
...     def test_something(self):
...         self.assertEqual(SomeClass.attribute, sentinel.attribute)
...
>>> original = SomeClass.attribute
>>> MyTest('test_something').test_something()
>>> assert SomeClass.attribute == original

Mock を使ってパッチしたい場合は、 patch() を1引数で (または patch.object() を2引数で) 使うことができます。mock が自動で生成され、テスト関数/メソッドに渡されます:

>>> class MyTest(unittest.TestCase):
...     @patch.object(SomeClass, 'static_method')
...     def test_something(self, mock_method):
...         SomeClass.static_method()
...         mock_method.assert_called_with()
...
>>> MyTest('test_something').test_something()

次のパターンのように patch デコレータを重ねることができます:

>>> class MyTest(unittest.TestCase):
...     @patch('package.module.ClassName1')
...     @patch('package.module.ClassName2')
...     def test_something(self, MockClass2, MockClass1):
...         self.assertIs(package.module.ClassName1, MockClass1)
...         self.assertIs(package.module.ClassName2, MockClass2)
...
>>> MyTest('test_something').test_something()

patch デコレータをネストした際、モックは (デコレータを適用する python の通常の) 順に適用されます。つまり引数は下から上の順になり、よって上記の例では test_module.ClassName2 が先になります。

また、 patch.dict() を使うと、スコープ内だけで辞書に値を設定し、テスト終了時には元の状態に復元されます:

>>> foo = {'key': 'value'}
>>> original = foo.copy()
>>> with patch.dict(foo, {'newkey': 'newvalue'}, clear=True):
...     assert foo == {'newkey': 'newvalue'}
...
>>> assert foo == original

patch, patch.object, patch.dict は全てコンテキストマネージャーとしても利用できます。

patch() に mock を生成させる場合、その参照を with 文の as を使って受け取れます:

>>> class ProductionClass:
...     def method(self):
...         pass
...
>>> with patch.object(ProductionClass, 'method') as mock_method:
...     mock_method.return_value = None
...     real = ProductionClass()
...     real.method(1, 2, 3)
...
>>> mock_method.assert_called_with(1, 2, 3)

他の方法として、patch, patch.object, patch.dict はクラスデコレータとしても利用できます。その場合、"test" で始まる全てのメソッドにデコレータを適用するのと同じになります。

26.6.3. さらなる例

より高度なシナリオを想定した例をあげていきます。

26.6.3.1. chained call をモックする

chained call を mock するのは、一度 return_value 属性を理解してしまえば簡単です。mock が最初に呼ばれた時や、呼び出される前に return_value を参照した場合、新しい Mock が生成されます。

つまり、戻り値のオブジェクトがどう利用されたかは、return_value mock を調べれば分かります:

>>> mock = Mock()
>>> mock().foo(a=2, b=3)
<Mock name='mock().foo()' id='...'>
>>> mock.return_value.foo.assert_called_with(a=2, b=3)

これをもとに、mock を構成して chained call に対する assert を行うのは簡単です。もちろん、元のコードを最初からもっとテストしやすく書くという選択肢もありますが...

では、例として次のようなコードがあるとします:

>>> class Something:
...     def __init__(self):
...         self.backend = BackendProvider()
...     def method(self):
...         response = self.backend.get_endpoint('foobar').create_call('spam', 'eggs').start_call()
...         # more code

BackendProvider はすでに十分テストされているとします。method() をどうテストしましょうか？特に、response オブジェクトを使う # more code の部分のコードをテストしたいとします。

この chained call はインスタンス属性を起点にしているので、 Something インスタンスの backend 属性に対してモンキーパッチすることができます。今回の場合だと、最後の start_call の呼び出しが返す値にだけ興味があるので、あまり多くの構成は必要ありません。このメソッドが返すのが 'file-like' オブジェクトだとしましょう。そうすると、 response オブジェクトは組み込みの open() を spec として利用できます。

これをするために、backend のモックとしてモックインスタンスを作成し、それに対するモックの response オブジェクトを作成します。最終的な start_call の返り値として response をセットすると、このようにすることができます:

mock_backend.get_endpoint.return_value.create_call.return_value.start_call.return_value = mock_response

これより少し良いやり方として、返り値を直接セットする configure_mock() メソッドを使用して次のようにすることができます:

>>> something = Something()
>>> mock_response = Mock(spec=open)
>>> mock_backend = Mock()
>>> config = {'get_endpoint.return_value.create_call.return_value.start_call.return_value': mock_response}
>>> mock_backend.configure_mock(**config)

これらによって「モックの backend」をその場で monkey patch して、実際の呼び出しを行うことができます:

>>> something.backend = mock_backend
>>> something.method()

mock_calls を使用すると、チェーンされた呼び出しを単一のアサーションでチェックすることができます。チェーンされた呼び出しは、1行のコードの中で行われる複数の呼び出しです。したがって、 mock_calls には複数のエントリーがあるでしょう。 call.call_list() を使用することで、この呼び出しのリストを作成することができます:

>>> chained = call.get_endpoint('foobar').create_call('spam', 'eggs').start_call()
>>> call_list = chained.call_list()
>>> assert mock_backend.mock_calls == call_list

26.6.3.2. 部分的なモック

テストによっては、 datetime.date.today() の呼び出しを既知の date を返すようにモックしたいと思うことがありました。しかし、テスト対象のコードが新しい date オブジェクトを生成するのを妨げたくはありません。不運にも、 datetime.date は C で書かれています。したがって、静的な date.today() メソッドを単にモンキーパッチすることはできませんでした。

私はこれを行う単純な方法を見つけました。それは、date クラスをモックで事実上ラップして、しかしコンストラクタの呼び出しを実際のクラスへと素通りさせる (そして、実際のインスタンスを返す) というものです。

ここで、テスト対象のモジュール中の date クラスをモックするために patch decorator が使用されています。そして、モックの date クラスの side_effect 属性は、本物の date を返すラムダ関数にセットされます。モックの date クラスが呼ばれる時、 side_effect によって本物の date が構築されて返されます。

>>> from datetime import date
>>> with patch('mymodule.date') as mock_date:
...     mock_date.today.return_value = date(2010, 10, 8)
...     mock_date.side_effect = lambda *args, **kw: date(*args, **kw)
...
...     assert mymodule.date.today() == date(2010, 10, 8)
...     assert mymodule.date(2009, 6, 8) == date(2009, 6, 8)
...

Note that we don't patch datetime.date globally, we patch date in the module that uses it. See where to patch.

When date.today() is called a known date is returned, but calls to the date(...) constructor still return normal dates. Without this you can find yourself having to calculate an expected result using exactly the same algorithm as the code under test, which is a classic testing anti-pattern.

Calls to the date constructor are recorded in the mock_date attributes (call_count and friends) which may also be useful for your tests.

An alternative way of dealing with mocking dates, or other builtin classes, is discussed in this blog entry.

26.6.3.3. ジェネレータ method をモックする

A Python generator is a function or method that uses the yield statement to return a series of values when iterated over [1].

A generator method / function is called to return the generator object. It is the generator object that is then iterated over. The protocol method for iteration is __iter__(), so we can mock this using a MagicMock.

Here's an example class with an "iter" method implemented as a generator:

>>> class Foo:
...     def iter(self):
...         for i in [1, 2, 3]:
...             yield i
...
>>> foo = Foo()
>>> list(foo.iter())
[1, 2, 3]

How would we mock this class, and in particular its "iter" method?

To configure the values returned from the iteration (implicit in the call to list), we need to configure the object returned by the call to foo.iter().

>>> mock_foo = MagicMock()
>>> mock_foo.iter.return_value = iter([1, 2, 3])
>>> list(mock_foo.iter())
[1, 2, 3]

[1]	There are also generator expressions and more advanced uses of generators, but we aren't concerned about them here. A very good introduction to generators and how powerful they are is: Generator Tricks for Systems Programmers.

26.6.3.4. 同じパッチを全てのメソッドに適用する

If you want several patches in place for multiple test methods the obvious way is to apply the patch decorators to every method. This can feel like unnecessary repetition. For Python 2.6 or more recent you can use patch() (in all its various forms) as a class decorator. This applies the patches to all test methods on the class. A test method is identified by methods whose names start with test:

>>> @patch('mymodule.SomeClass')
... class MyTest(TestCase):
...
...     def test_one(self, MockSomeClass):
...         self.assertIs(mymodule.SomeClass, MockSomeClass)
...
...     def test_two(self, MockSomeClass):
...         self.assertIs(mymodule.SomeClass, MockSomeClass)
...
...     def not_a_test(self):
...         return 'something'
...
>>> MyTest('test_one').test_one()
>>> MyTest('test_two').test_two()
>>> MyTest('test_two').not_a_test()
'something'

An alternative way of managing patches is to use the patch のメソッド: start と stop. These allow you to move the patching into your setUp and tearDown methods.

>>> class MyTest(TestCase):
...     def setUp(self):
...         self.patcher = patch('mymodule.foo')
...         self.mock_foo = self.patcher.start()
...
...     def test_foo(self):
...         self.assertIs(mymodule.foo, self.mock_foo)
...
...     def tearDown(self):
...         self.patcher.stop()
...
>>> MyTest('test_foo').run()

If you use this technique you must ensure that the patching is "undone" by calling stop. This can be fiddlier than you might think, because if an exception is raised in the setUp then tearDown is not called. unittest.TestCase.addCleanup() makes this easier:

>>> class MyTest(TestCase):
...     def setUp(self):
...         patcher = patch('mymodule.foo')
...         self.addCleanup(patcher.stop)
...         self.mock_foo = patcher.start()
...
...     def test_foo(self):
...         self.assertIs(mymodule.foo, self.mock_foo)
...
>>> MyTest('test_foo').run()

26.6.3.5. Mocking Unbound Methods

Whilst writing tests today I needed to patch an unbound method (patching the method on the class rather than on the instance). I needed self to be passed in as the first argument because I want to make asserts about which objects were calling this particular method. The issue is that you can't patch with a mock for this, because if you replace an unbound method with a mock it doesn't become a bound method when fetched from the instance, and so it doesn't get self passed in. The workaround is to patch the unbound method with a real function instead. The patch() decorator makes it so simple to patch out methods with a mock that having to create a real function becomes a nuisance.

If you pass autospec=True to patch then it does the patching with a real function object. This function object has the same signature as the one it is replacing, but delegates to a mock under the hood. You still get your mock auto-created in exactly the same way as before. What it means though, is that if you use it to patch out an unbound method on a class the mocked function will be turned into a bound method if it is fetched from an instance. It will have self passed in as the first argument, which is exactly what I wanted:

>>> class Foo:
...   def foo(self):
...     pass
...
>>> with patch.object(Foo, 'foo', autospec=True) as mock_foo:
...   mock_foo.return_value = 'foo'
...   foo = Foo()
...   foo.foo()
...
'foo'
>>> mock_foo.assert_called_once_with(foo)

If we don't use autospec=True then the unbound method is patched out with a Mock instance instead, and isn't called with self.

26.6.3.6. モックで複数回の呼び出しをチェックする

mock has a nice API for making assertions about how your mock objects are used.

>>> mock = Mock()
>>> mock.foo_bar.return_value = None
>>> mock.foo_bar('baz', spam='eggs')
>>> mock.foo_bar.assert_called_with('baz', spam='eggs')

If your mock is only being called once you can use the assert_called_once_with() method that also asserts that the call_count is one.

>>> mock.foo_bar.assert_called_once_with('baz', spam='eggs')
>>> mock.foo_bar()
>>> mock.foo_bar.assert_called_once_with('baz', spam='eggs')
Traceback (most recent call last):
    ...
AssertionError: Expected to be called once. Called 2 times.

Both assert_called_with and assert_called_once_with make assertions about the most recent call. If your mock is going to be called several times, and you want to make assertions about all those calls you can use call_args_list:

>>> mock = Mock(return_value=None)
>>> mock(1, 2, 3)
>>> mock(4, 5, 6)
>>> mock()
>>> mock.call_args_list
[call(1, 2, 3), call(4, 5, 6), call()]

The call helper makes it easy to make assertions about these calls. You can build up a list of expected calls and compare it to call_args_list. This looks remarkably similar to the repr of the call_args_list:

>>> expected = [call(1, 2, 3), call(4, 5, 6), call()]
>>> mock.call_args_list == expected
True

26.6.3.7. ミュータブルな引数に対処する

Another situation is rare, but can bite you, is when your mock is called with mutable arguments. call_args and call_args_list store references to the arguments. If the arguments are mutated by the code under test then you can no longer make assertions about what the values were when the mock was called.

Here's some example code that shows the problem. Imagine the following functions defined in 'mymodule':

def frob(val):
    pass

def grob(val):
    "First frob and then clear val"
    frob(val)
    val.clear()

When we try to test that grob calls frob with the correct argument look what happens:

>>> with patch('mymodule.frob') as mock_frob:
...     val = {6}
...     mymodule.grob(val)
...
>>> val
set()
>>> mock_frob.assert_called_with({6})
Traceback (most recent call last):
    ...
AssertionError: Expected: (({6},), {})
Called with: ((set(),), {})

One possibility would be for mock to copy the arguments you pass in. This could then cause problems if you do assertions that rely on object identity for equality.

Here's one solution that uses the side_effect functionality. If you provide a side_effect function for a mock then side_effect will be called with the same args as the mock. This gives us an opportunity to copy the arguments and store them for later assertions. In this example I'm using another mock to store the arguments so that I can use the mock methods for doing the assertion. Again a helper function sets this up for me.

>>> from copy import deepcopy
>>> from unittest.mock import Mock, patch, DEFAULT
>>> def copy_call_args(mock):
...     new_mock = Mock()
...     def side_effect(*args, **kwargs):
...         args = deepcopy(args)
...         kwargs = deepcopy(kwargs)
...         new_mock(*args, **kwargs)
...         return DEFAULT
...     mock.side_effect = side_effect
...     return new_mock
...
>>> with patch('mymodule.frob') as mock_frob:
...     new_mock = copy_call_args(mock_frob)
...     val = {6}
...     mymodule.grob(val)
...
>>> new_mock.assert_called_with({6})
>>> new_mock.call_args
call({6})

copy_call_args is called with the mock that will be called. It returns a new mock that we do the assertion on. The side_effect function makes a copy of the args and calls our new_mock with the copy.

注釈

If your mock is only going to be used once there is an easier way of checking arguments at the point they are called. You can simply do the checking inside a side_effect function.

>>> def side_effect(arg):
...     assert arg == {6}
...
>>> mock = Mock(side_effect=side_effect)
>>> mock({6})
>>> mock(set())
Traceback (most recent call last):
    ...
AssertionError

An alternative approach is to create a subclass of Mock or MagicMock that copies (using copy.deepcopy()) the arguments. Here's an example implementation:

>>> from copy import deepcopy
>>> class CopyingMock(MagicMock):
...     def __call__(self, *args, **kwargs):
...         args = deepcopy(args)
...         kwargs = deepcopy(kwargs)
...         return super(CopyingMock, self).__call__(*args, **kwargs)
...
>>> c = CopyingMock(return_value=None)
>>> arg = set()
>>> c(arg)
>>> arg.add(1)
>>> c.assert_called_with(set())
>>> c.assert_called_with(arg)
Traceback (most recent call last):
    ...
AssertionError: Expected call: mock({1})
Actual call: mock(set())
>>> c.foo
<CopyingMock name='mock.foo' id='...'>

When you subclass Mock or MagicMock all dynamically created attributes, and the return_value will use your subclass automatically. That means all children of a CopyingMock will also have the type CopyingMock.

26.6.3.8. patch をネストする

Using patch as a context manager is nice, but if you do multiple patches you can end up with nested with statements indenting further and further to the right:

>>> class MyTest(TestCase):
...
...     def test_foo(self):
...         with patch('mymodule.Foo') as mock_foo:
...             with patch('mymodule.Bar') as mock_bar:
...                 with patch('mymodule.Spam') as mock_spam:
...                     assert mymodule.Foo is mock_foo
...                     assert mymodule.Bar is mock_bar
...                     assert mymodule.Spam is mock_spam
...
>>> original = mymodule.Foo
>>> MyTest('test_foo').test_foo()
>>> assert mymodule.Foo is original

With unittest cleanup functions and the patch のメソッド: start と stop we can achieve the same effect without the nested indentation. A simple helper method, create_patch, puts the patch in place and returns the created mock for us:

>>> class MyTest(TestCase):
...
...     def create_patch(self, name):
...         patcher = patch(name)
...         thing = patcher.start()
...         self.addCleanup(patcher.stop)
...         return thing
...
...     def test_foo(self):
...         mock_foo = self.create_patch('mymodule.Foo')
...         mock_bar = self.create_patch('mymodule.Bar')
...         mock_spam = self.create_patch('mymodule.Spam')
...
...         assert mymodule.Foo is mock_foo
...         assert mymodule.Bar is mock_bar
...         assert mymodule.Spam is mock_spam
...
>>> original = mymodule.Foo
>>> MyTest('test_foo').run()
>>> assert mymodule.Foo is original

26.6.3.9. MagicMock で辞書をモックする

You may want to mock a dictionary, or other container object, recording all access to it whilst having it still behave like a dictionary.

We can do this with MagicMock, which will behave like a dictionary, and using side_effect to delegate dictionary access to a real underlying dictionary that is under our control.

When the __getitem__() and __setitem__() methods of our MagicMock are called (normal dictionary access) then side_effect is called with the key (and in the case of __setitem__ the value too). We can also control what is returned.

After the MagicMock has been used we can use attributes like call_args_list to assert about how the dictionary was used:

>>> my_dict = {'a': 1, 'b': 2, 'c': 3}
>>> def getitem(name):
...      return my_dict[name]
...
>>> def setitem(name, val):
...     my_dict[name] = val
...
>>> mock = MagicMock()
>>> mock.__getitem__.side_effect = getitem
>>> mock.__setitem__.side_effect = setitem

注釈

An alternative to using MagicMock is to use Mock and only provide the magic methods you specifically want:

>>> mock = Mock()
>>> mock.__getitem__ = Mock(side_effect=getitem)
>>> mock.__setitem__ = Mock(side_effect=setitem)

A third option is to use MagicMock but passing in dict as the spec (or spec_set) argument so that the MagicMock created only has dictionary magic methods available:

>>> mock = MagicMock(spec_set=dict)
>>> mock.__getitem__.side_effect = getitem
>>> mock.__setitem__.side_effect = setitem

With these side effect functions in place, the mock will behave like a normal dictionary but recording the access. It even raises a KeyError if you try to access a key that doesn't exist.

>>> mock['a']
1
>>> mock['c']
3
>>> mock['d']
Traceback (most recent call last):
    ...
KeyError: 'd'
>>> mock['b'] = 'fish'
>>> mock['d'] = 'eggs'
>>> mock['b']
'fish'
>>> mock['d']
'eggs'

After it has been used you can make assertions about the access using the normal mock methods and attributes:

>>> mock.__getitem__.call_args_list
[call('a'), call('c'), call('d'), call('b'), call('d')]
>>> mock.__setitem__.call_args_list
[call('b', 'fish'), call('d', 'eggs')]
>>> my_dict
{'a': 1, 'c': 3, 'b': 'fish', 'd': 'eggs'}

26.6.3.10. Mock のサブクラスと属性

There are various reasons why you might want to subclass Mock. One reason might be to add helper methods. Here's a silly example:

>>> class MyMock(MagicMock):
...     def has_been_called(self):
...         return self.called
...
>>> mymock = MyMock(return_value=None)
>>> mymock
<MyMock id='...'>
>>> mymock.has_been_called()
False
>>> mymock()
>>> mymock.has_been_called()
True

The standard behaviour for Mock instances is that attributes and the return value mocks are of the same type as the mock they are accessed on. This ensures that Mock attributes are Mocks and MagicMock attributes are MagicMocks [2]. So if you're subclassing to add helper methods then they'll also be available on the attributes and return value mock of instances of your subclass.

>>> mymock.foo
<MyMock name='mock.foo' id='...'>
>>> mymock.foo.has_been_called()
False
>>> mymock.foo()
<MyMock name='mock.foo()' id='...'>
>>> mymock.foo.has_been_called()
True

Sometimes this is inconvenient. For example, one user is subclassing mock to created a Twisted adaptor. Having this applied to attributes too actually causes errors.

Mock (in all its flavours) uses a method called _get_child_mock to create these "sub-mocks" for attributes and return values. You can prevent your subclass being used for attributes by overriding this method. The signature is that it takes arbitrary keyword arguments (**kwargs) which are then passed onto the mock constructor:

>>> class Subclass(MagicMock):
...     def _get_child_mock(self, **kwargs):
...         return MagicMock(**kwargs)
...
>>> mymock = Subclass()
>>> mymock.foo
<MagicMock name='mock.foo' id='...'>
>>> assert isinstance(mymock, Subclass)
>>> assert not isinstance(mymock.foo, Subclass)
>>> assert not isinstance(mymock(), Subclass)

[2]	An exception to this rule are the non-callable mocks. Attributes use the callable variant because otherwise non-callable mocks couldn't have callable methods.

26.6.3.11. patch.dict で import をモックする

One situation where mocking can be hard is where you have a local import inside a function. These are harder to mock because they aren't using an object from the module namespace that we can patch out.

Generally local imports are to be avoided. They are sometimes done to prevent circular dependencies, for which there is usually a much better way to solve the problem (refactor the code) or to prevent "up front costs" by delaying the import. This can also be solved in better ways than an unconditional local import (store the module as a class or module attribute and only do the import on first use).

That aside there is a way to use mock to affect the results of an import. Importing fetches an object from the sys.modules dictionary. Note that it fetches an object, which need not be a module. Importing a module for the first time results in a module object being put in sys.modules, so usually when you import something you get a module back. This need not be the case however.

This means you can use patch.dict() to temporarily put a mock in place in sys.modules. Any imports whilst this patch is active will fetch the mock. When the patch is complete (the decorated function exits, the with statement body is complete or patcher.stop() is called) then whatever was there previously will be restored safely.

Here's an example that mocks out the 'fooble' module.

>>> mock = Mock()
>>> with patch.dict('sys.modules', {'fooble': mock}):
...    import fooble
...    fooble.blob()
...
<Mock name='mock.blob()' id='...'>
>>> assert 'fooble' not in sys.modules
>>> mock.blob.assert_called_once_with()

As you can see the import fooble succeeds, but on exit there is no 'fooble' left in sys.modules.

This also works for the from module import name form:

>>> mock = Mock()
>>> with patch.dict('sys.modules', {'fooble': mock}):
...    from fooble import blob
...    blob.blip()
...
<Mock name='mock.blob.blip()' id='...'>
>>> mock.blob.blip.assert_called_once_with()

With slightly more work you can also mock package imports:

>>> mock = Mock()
>>> modules = {'package': mock, 'package.module': mock.module}
>>> with patch.dict('sys.modules', modules):
...    from package.module import fooble
...    fooble()
...
<Mock name='mock.module.fooble()' id='...'>
>>> mock.module.fooble.assert_called_once_with()

26.6.3.12. Tracking order of calls and less verbose call assertions

The Mock class allows you to track the order of method calls on your mock objects through the method_calls attribute. This doesn't allow you to track the order of calls between separate mock objects, however we can use mock_calls to achieve the same effect.

Because mocks track calls to child mocks in mock_calls, and accessing an arbitrary attribute of a mock creates a child mock, we can create our separate mocks from a parent one. Calls to those child mock will then all be recorded, in order, in the mock_calls of the parent:

>>> manager = Mock()
>>> mock_foo = manager.foo
>>> mock_bar = manager.bar

>>> mock_foo.something()
<Mock name='mock.foo.something()' id='...'>
>>> mock_bar.other.thing()
<Mock name='mock.bar.other.thing()' id='...'>

>>> manager.mock_calls
[call.foo.something(), call.bar.other.thing()]

We can then assert about the calls, including the order, by comparing with the mock_calls attribute on the manager mock:

>>> expected_calls = [call.foo.something(), call.bar.other.thing()]
>>> manager.mock_calls == expected_calls
True

If patch is creating, and putting in place, your mocks then you can attach them to a manager mock using the attach_mock() method. After attaching calls will be recorded in mock_calls of the manager.

>>> manager = MagicMock()
>>> with patch('mymodule.Class1') as MockClass1:
...     with patch('mymodule.Class2') as MockClass2:
...         manager.attach_mock(MockClass1, 'MockClass1')
...         manager.attach_mock(MockClass2, 'MockClass2')
...         MockClass1().foo()
...         MockClass2().bar()
...
<MagicMock name='mock.MockClass1().foo()' id='...'>
<MagicMock name='mock.MockClass2().bar()' id='...'>
>>> manager.mock_calls
[call.MockClass1(),
 call.MockClass1().foo(),
 call.MockClass2(),
 call.MockClass2().bar()]

If many calls have been made, but you're only interested in a particular sequence of them then an alternative is to use the assert_has_calls() method. This takes a list of calls (constructed with the call object). If that sequence of calls are in mock_calls then the assert succeeds.

>>> m = MagicMock()
>>> m().foo().bar().baz()
<MagicMock name='mock().foo().bar().baz()' id='...'>
>>> m.one().two().three()
<MagicMock name='mock.one().two().three()' id='...'>
>>> calls = call.one().two().three().call_list()
>>> m.assert_has_calls(calls)

Even though the chained call m.one().two().three() aren't the only calls that have been made to the mock, the assert still succeeds.

Sometimes a mock may have several calls made to it, and you are only interested in asserting about some of those calls. You may not even care about the order. In this case you can pass any_order=True to assert_has_calls:

>>> m = MagicMock()
>>> m(1), m.two(2, 3), m.seven(7), m.fifty('50')
(...)
>>> calls = [call.fifty('50'), call(1), call.seven(7)]
>>> m.assert_has_calls(calls, any_order=True)

26.6.3.13. More complex argument matching

Using the same basic concept as ANY we can implement matchers to do more complex assertions on objects used as arguments to mocks.

Suppose we expect some object to be passed to a mock that by default compares equal based on object identity (which is the Python default for user defined classes). To use assert_called_with() we would need to pass in the exact same object. If we are only interested in some of the attributes of this object then we can create a matcher that will check these attributes for us.

You can see in this example how a 'standard' call to assert_called_with isn't sufficient:

>>> class Foo:
...     def __init__(self, a, b):
...         self.a, self.b = a, b
...
>>> mock = Mock(return_value=None)
>>> mock(Foo(1, 2))
>>> mock.assert_called_with(Foo(1, 2))
Traceback (most recent call last):
    ...
AssertionError: Expected: call(<__main__.Foo object at 0x...>)
Actual call: call(<__main__.Foo object at 0x...>)

A comparison function for our Foo class might look something like this:

>>> def compare(self, other):
...     if not type(self) == type(other):
...         return False
...     if self.a != other.a:
...         return False
...     if self.b != other.b:
...         return False
...     return True
...

And a matcher object that can use comparison functions like this for its equality operation would look something like this:

>>> class Matcher:
...     def __init__(self, compare, some_obj):
...         self.compare = compare
...         self.some_obj = some_obj
...     def __eq__(self, other):
...         return self.compare(self.some_obj, other)
...

全てをつなぎ合わせて:

>>> match_foo = Matcher(compare, Foo(1, 2))
>>> mock.assert_called_with(match_foo)

The Matcher is instantiated with our compare function and the Foo object we want to compare against. In assert_called_with the Matcher equality method will be called, which compares the object the mock was called with against the one we created our matcher with. If they match then assert_called_with passes, and if they don't an AssertionError is raised:

>>> match_wrong = Matcher(compare, Foo(3, 4))
>>> mock.assert_called_with(match_wrong)
Traceback (most recent call last):
    ...
AssertionError: Expected: ((<Matcher object at 0x...>,), {})
Called with: ((<Foo object at 0x...>,), {})

With a bit of tweaking you could have the comparison function raise the AssertionError directly and provide a more useful failure message.

As of version 1.5, the Python testing library PyHamcrest provides similar functionality, that may be useful here, in the form of its equality matcher (hamcrest.library.integration.match_equality).