python - 为什么我不能在 python 中像这样使用'__getattr__'？

Question

代码是这样的，我想用来__getattr__让State类有用，可以在上面挂载一些属性，以便我们以后可以使用它，例如：

state = State()
state.name1 = 1
print(state.name1)

完整的代码是这样的：

class State:
    """
    提供给用户使用的 state；

    为了实现断点重训，用户应当保证其保存的信息都是可序列化的；

    # TODO：可能需要提供保存该state的函数，但是用户可以自己在 callback 里实现，现在我们先不管；
    """

    def __init__(self):
        self._value = dict()

    def __setattr__(self, key, value):
        self._value[key] = value

    def __getattr__(self, item):
        if item in self._value:
            return self._value[item]
        else:
            raise AttributeError(f"{self.__name__} has no attribute {item}.")

    def state_dict(self):
        return self._value

    def load_state_dict(self, value: dict):
        if not isinstance(value, dict):
            raise ValueError("If you want to reload a state dict for reasons like breakpoint retraining, this parameter"
                             "value should be a dict type.")
        self._value = value

回溯是这样的：

Traceback (most recent call last):
File "state.py", line 165, in <module>
a = State()
File "state.py", line 52, in __init__
self._value = dict()
File "state.py", line 55, in __setattr__
self._value[key] = value
File "state.py", line 58, in __getattr__
if item in self._value:
File "state.py", line 58, in __getattr__
if item in self._value:
File "state.py", line 58, in __getattr__
if item in self._value:
[Previous line repeated 993 more times]
RecursionError: maximum recursion depth exceeded

Answer 1

为了避免 getattr 和 setattr 函数之间的递归，你必须调用父 setattr：

def __setattr__(self, key, value):
    super().__setattr__(key, value)
    self._value[key] = value