python的描述符(descriptor)、装饰器（property）造成的一个无限递归问题分享

分享一下刚遇到的一个小问题，我有一段类似于这样的python代码：

# coding: utf-8
class A(object):

    @property

    def _value(self):

#        raise AttributeError("test")

        return {"v": "This is a test."}
    def __getattr__(self, key):

        print "__getattr__:", key

        return self._value[key]
if __name__ == '__main__':

    a = A()

    print a.v

运行后可以得到正确的结果

__getattr__: v

This is a test.

但是注意，如果把

#        raise AttributeError("test")

这行的注释去掉的话，即在_value方法里面抛出AttributeError异常，事情就会变得有些奇怪。程序运行的时候并不会抛出异常，而是会进入一个无限递归：

File "attr_test.py", line 12, in __getattr__

    return self._value[key]

  File "attr_test.py", line 12, in __getattr__

    return self._value[key]

RuntimeError: maximum recursion depth exceeded while calling a Python object

通过多方查找后发现是property装饰器的问题，property实际上是一个descriptor。在python doc中可以发现这样的文字：

object.__get__(self, instance, owner)
Called to get the attribute of the owner class (class attribute access) or of an instance of that class (instance attribute access). owner is always the owner class, while instance is the instance that the attribute was accessed through, or None when the attribute is accessed through the owner. This method should return the (computed) attribute value or raise an AttributeError exception.

这样当用户访问._value时，抛出了AttributeError从而调用了__getattr__方法去尝试获取。这样程序就变成了无限递归。

这个问题看上去不复杂，但是当你的_value方法是比较隐晦的抛出AttributeError的话，调试起来就会比较困难了。