Pytest：模拟具有不同副作用的同一方法的多次调用

Question

我有一个如下所示的单元测试：

# utilities.py  
def get_side_effects():
    def side_effect_func3(self):
        # Need the "self" to do some stuff at run time.
        return {"final":"some3"} 

    def side_effect_func2(self):
        # Need the "self" to do some stuff at run time.
        return {"status":"some2"}
      
    def side_effect_func1(self):
        # Need the "self" to do some stuff at run time.
        return {"name":"some1"} 

    return side_effect_func1, side_effect_func2, side_effect_func2

#################

# test_a.py
def test_endtoend():
   
    s1, s2, s3 = utilities.get_side_effects()
    
    m1 = mock.MagicMock()
    m1.side_effect = s1

    m2 = mock.MagicMock()
    m2.side_effect = s2

    m3 = mock.MagicMock()
    m3.side_effect = s3
   
    with mock.patch("a.get_request", m3):
        with mock.patch("a.get_request", m2):
            with mock.patch("a.get_request", m1):
                foo = a() # Class to test
                result = foo.run() 
    
    

作为foo.run()代码运行的一部分，get_request被多次调用。我希望每次调用get_request方法时都有不同的 side_effect 函数，在本例中是side_effect_func1, side_effect_func2, side_effect_func3。但我注意到，只有m1模拟对象是活动的，即仅side_effect_func1被调用，而不是其他 2。我如何实现这一点？

我也尝试过下面的方法，但实际的 side_effect 函数不会被调用，它们总是返回function object，但实际上并不执行 side_effect 函数。

# utilities.py
def get_side_effects():
    def side_effect_func3(self):
        # Need the "self" to do some stuff at run time.
        return {"final":"some3"} 

    def side_effect_func2(self):
        # Need the "self" to do some stuff at run time.
        return {"status":"some2"}
      
    def side_effect_func1(self):
        # Need the "self" to do some stuff at run time.
        return {"name":"some1"} 

    all_get_side_effects = []
    all_get_side_effects.append(side_effect_func1)
    all_get_side_effects.append(side_effect_func2)
    all_get_side_effects.append(side_effect_func3)
     
    return all_get_side_effects

#########################
# test_a.py
def test_endtoend():

    all_side_effects = utilities.get_side_effects()

    m = mock.MagicMock()
    m.side_effect = all_side_effects

    with mock.patch("a.get_request", m):
       foo = a() # Class to test
       result = foo.run()

Answer 1

您的第一次尝试不起作用，因为每个模拟都只是替换了前一个模拟（外部两个模拟不执行任何操作）。

您的第二次尝试不起作用，因为副作用已超载，可以为可迭代对象提供不同的用途（文档）：

如果side_effect是一个可迭代对象，那么每次调用模拟都会返回可迭代对象的下一个值。

相反，您可以使用可调用类来实现副作用，即维护有关实际连续调用哪个底层函数的某种状态。

具有两个函数的基本示例：

>>> class SideEffect:
...     def __init__(self, *fns):
...         self.fs = iter(fns)
...     def __call__(self, *args, **kwargs):
...         f = next(self.fs)
...         return f(*args, **kwargs)
... 
>>> def sf1():
...     print("called sf1")
...     return 1
... 
>>> def sf2():
...     print("called sf2")
...     return 2
... 
>>> def foo():
...     print("called actual func")
...     return "f"
... 
>>> with mock.patch("__main__.foo", side_effect=SideEffect(sf1, sf2)) as m:
...     first = foo()
...     second = foo()
... 
called sf1
called sf2
>>> assert first == 1
>>> assert second == 2
>>> assert m.call_count == 2