Mock源码阅读

写单元测试的时候，经常要借助Mock来完成某些替换，鉴于 Python 的动态性，Python中 写单元测试还是很方便的。但是需要吐槽的是，也是因为Python的动态性，Python的功能 必须要由高覆盖率的测试来保证。

Mock源码以及注释

https://github.com/jiajunhuang/cpython/blob/annotation/Lib/unittest/mock.py

感受

看完mock的代码感觉有两点：

• 标准库的代码并不是那么“标准” --- 并不怎么遵守pep8

• 标准库的代码测试覆盖率超高

看完mock的代码以后给CPython提交了两个patch，一个是优化了一下写法，一个是修复了 一个bug：

• http://bugs.python.org/issue28919

• http://bugs.python.org/issue28961

造个小轮子

简单实现一个mock如下：

class Mock:
    return_value = NotImplemented

    def __new__(cls, *args, **kwargs):
        new = type(cls.__name__, (cls, ), {'__doc__': cls.__doc__})
        return object.__new__(new)

    def __init__(self, return_value):
        self.return_value = return_value

    def __call__(self):
        return self.return_value if self.return_value else None


class PatchObject:
    def __init__(self, instance, attr_name, value):
        self.attr_origin = getattr(instance, attr_name, None)
        self.attr_name = attr_name
        self.instance = instance
        self.value = value

    def __enter__(self):
        """perform the patch"""
        mocked = Mock(self.value)
        setattr(self.instance, self.attr_name, mocked)
        return mocked

    def __exit__(self, exc_type, exc_val, exc_rb):
        """unpatch"""
        if self.attr_origin:
            setattr(self.instance, self.attr_name, self.attr_origin)
        else:
            delattr(self.instance, self.attr_name)


if __name__ == "__main__":
    # test mock
    class Foo:
        def foo(self):
            pass

    foo = Foo()
    foo.foo = Mock(1)

    bar = Foo()
    bar.foo = Mock(2)

    print(foo.foo(), bar.foo())

    # test patch object
    baz = Foo()
    with PatchObject(baz, "baz", 3) as mocked_baz:
        print(mocked_baz())
    print(dir(baz))

运行结果：

(1, 2)
3
['__doc__', '__module__', 'foo']