Attribute Lookup: 当访问实例的属性时，发生了什么

原文地址, 拒绝转载: https://www.jianshu.com/p/b2691cf186d4

Attribute Lookup

假设 Cls 是类，instance 是类 Cls 的一个实例，当调用 instance.attr 时，到底发生了什么呢？下面就来一一探讨属性访问的调用流程

1 descriptor

什么是 desciptor, 官方文档给出的回答是

A descriptor is what we call any object that defines __get__(), __set__(), or __delete__().

即包含了任意 __get__ 或者 __set__ 或者 __delete__ 函数的方法的 object 都是 descriptor

2 data descriptor 与 non-data descriptors

If an object defines __set__() or __delete__(), it is considered a data descriptor. Descriptors that only define __get__() are called non-data descriptors (they are typically used for methods but other uses are possible).

如果一个 descriptor 只定义了 __get__ 方法，那么就是 non-data descriptor

如果一个 object 定义了 __set__ 或者 __delete__ 方法，那么就是 data descriptor

class DataDescriptor:
    """
    包含了 __set__ 方法，所以这个类的实例是 data-descritptor
    """
    def __init__(self, init_value):
        self.value = init_value

    def __get__(self, instance, typ):
        return 'DataDescriptor __get__' + str(typ)

    def __set__(self, instance, value):
        print ('DataDescriptor __set__')
        self.value = value

class NonDataDescriptor:
    """
    只定义了 __get__ 方法，所以这个类的实例是 non-data descriptor
    """
    def __init__(self, init_value):
        self.value = init_value

    def __get__(self, instance, typ):
        return'NonDataDescriptor __get__' + str(typ)

3 当调用 instance.attr 时，发生了什么

假设 cls 是类，instance 是类 cls 的一个实例，当调用 instance.attr 时，调用流程如下

如果在cls 或者其基类中的 __dict__ 找到了 attr，并且 attr 是 data descriptor 则调用其 __get__方法，即 __dict__['attr'].__get(instance, cls)

class Base(object):
    dd_base = DataDescriptor(0)
    ndd_base = NonDataDescriptor(0)

class Derive(Base):
    dd_derive = DataDescriptor(0)
    ndd_derive = NonDataDescriptor(0)
    ndd_derive2 = NonDataDescriptor(1)
    not_descriptor_in_class = "Derive not descriptor in class"
    
    def __getattr__(self, key):
        return '__getattr__ with key %s in Derive' % key

print(Base.__dict__)
"""
{
'__module__': '__main__', 
'dd_base': <__main__.DataDescriptor object at 0x7fc5c5b68a58>, 
'ndd_base': <__main__.NonDataDescriptor object at 0x7fc5c5b68a90>, 
'__dict__': <attribute '__dict__' of 'Base' objects>, 
'__weakref__': <attribute '__weakref__' of 'Base' objects>, 
'__doc__': None}
"""
print(Derive.__dict__)
"""
{'__module__': '__main__', 
'dd_derive': <__main__.DataDescriptor object at 0x7f9e74ac79b0>, 
'ndd_derive': <__main__.NonDataDescriptor object at 0x7f9e74ac79e8>, 'same_name_attr': 'attr in class', 
'__doc__': None}
"""
b = Base()
# 打印: DataDescriptor __get__<class '__main__.Base'>
print(b.dd_base)
d = Derive()
# 打印: DataDescriptor __get__<class '__main__.Derive'>
print(d.dd_base)
# 打印: DataDescriptor __get__<class '__main__.Derive'>
print(d.dd_derive)

# 即使我们更改了 instance 的 __dict__ 属性，访问时仍然从 data descriptor 中读取
# 不会从 instance.__dict__ 中读取
b.__dict__['dd_base'] = 'changed in dict dd base'
# 打印: DataDescriptor __get__<class '__main__.Base'>
print(b.dd_base)
d.__dict__['dd_derive'] = 'changed in dict dd derive'
# 打印: DataDescriptor __get__<class '__main__.Derive'>
print(d.dd_derive)

如果 attr 出现在 instance.__dict__ 中，则返回 instance.__dict__['attr']。否则，执行下面的流程

# 更改了 instance 的 __dict__
# 如果访问的不是 data descriptor, 则直接中 instance.__dict__ 中读取 attr
b.__dict__['ndd_base'] = 'changed in dict ndd base'
# 打印: changed in dict ndd base
print(b.ndd_base)
d.__dict__['ndd_derive'] = 'changed in dict ndd derive'
# 打印: changed in dict ndd derive
print(d.ndd_derive)

如果 attr 出现在类或者基类的 __dict__ 中

如果是 non-data descriptor, 则调用 __get__方法

# 打印: NonDataDescriptor __get__<class '__main__.Derive'>
print(d.ndd_derive2)

如果不是 descriptor , 则返回 __dict__['attr']

# 打印: Derive not descriptor in class
print(d.not_descriptor_in_class)

如果仍未找到，如果类或者其基类有 __getattr__ 方法，则调用 __getattr__ 方法
```
# 打印: __getattr__ with key no_exist_key in Derive
print(d.no_exist_key)
```

否则抛出 AttributeError

try:
    b.no_exists_key
except Exception as e:
    # 打印: True
    print(isinstance(e, AttributeError))

4 完整测试代码

class DataDescriptor:
    """
    包含了 __set__ 方法，所以这个类的实例是 data-descritptor
    """

    def __init__(self, init_value):
        self.value = init_value

    def __get__(self, instance, typ):
        return 'DataDescriptor __get__' + str(typ)

    def __set__(self, instance, value):
        print('DataDescriptor __set__')
        self.value = value


class NonDataDescriptor:
    """
    只定义了 __get__ 方法，所以这个类的实例是 non-data descriptor
    """

    def __init__(self, init_value):
        self.value = init_value

    def __get__(self, instance, typ):
        return 'NonDataDescriptor __get__' + str(typ)


class Base(object):
    dd_base = DataDescriptor(0)
    ndd_base = NonDataDescriptor(0)


class Derive(Base):
    dd_derive = DataDescriptor(0)
    ndd_derive = NonDataDescriptor(0)
    ndd_derive2 = NonDataDescriptor(1)
    not_descriptor_in_class = "Derive not descriptor in class"

    def __getattr__(self, key):
        return '__getattr__ with key %s in Derive' % key


if __name__ == '__main__':
    b = Base()
    # 打印: DataDescriptor __get__<class '__main__.Base'>
    print(b.dd_base)
    d = Derive()
    # 打印: DataDescriptor __get__<class '__main__.Derive'>
    print(d.dd_base)
    # 打印: DataDescriptor __get__<class '__main__.Derive'>
    print(d.dd_derive)

    # 即使我们更改了 instance 的 __dict__ 属性，访问时仍然从 data descriptor 中读取
    # 不会从 instance.__dict__ 中读取
    b.__dict__['dd_base'] = 'changed in dict dd base'
    # 打印: DataDescriptor __get__<class '__main__.Base'>
    print(b.dd_base)
    d.__dict__['dd_derive'] = 'changed in dict dd derive'
    # 打印: DataDescriptor __get__<class '__main__.Derive'>
    print(d.dd_derive)

    # 更改了 instance 的 __dict__
    # 如果访问的不是 data descriptor, 则直接中 instance.__dict__ 中读取 attr
    b.__dict__['ndd_base'] = 'changed in dict ndd base'
    # 打印: changed in dict ndd base
    print(b.ndd_base)
    d.__dict__['ndd_derive'] = 'changed in dict ndd derive'
    # 打印: changed in dict ndd derive
    print(d.ndd_derive)

    # 打印: NonDataDescriptor __get__<class '__main__.Derive'>
    print(d.ndd_derive2)
    # 打印: Derive not descriptor in class
    print(d.not_descriptor_in_class)

    # 打印: __getattr__ with key no_exist_key
    print(d.no_exist_key)

    try:
        b.no_exists_key
    except Exception as e:
        # 打印: True
        print(isinstance(e, AttributeError))

5 参考链接

https://blog.peterlamut.com/2018/11/04/python-attribute-lookup-explained-in-detail/
https://docs.python.org/3/howto/descriptor.html
https://www.cnblogs.com/xybaby/p/6270551.html

6 转载一下调用流程图片

图片是从参考链接的第一个博客中复制的，把调用的流程描述的很清晰，值得一看