a d中的事件链

Question

假设我有一个客户deleter属性和实例本身的类：

class Hello:
    def __init__(self):
        self._is_open = None
    @property
    def is_open(self):
        return self._is_open
    @is_open.deleter
    def is_open(self):
        print("Using custom deleter!")
        del self._is_open
    def __delattr__(self, attr):
        print ('Deleting attr %s' % attr)
        super().__delattr__(attr)

并称之为：

>>> hello = Hello()
>>> del hello.is_open
Deleting attr is_open
Using custom deleter!
Deleting attr _is_open

看起来它首先调用is_open上的__delattr__，然后调用@is_open.deleter，然后调用_is_open上的__delattr__。为什么删除程序的事件链是这样工作的？你知道吗

Answer 1

Python properties是descriptors。它们通过descriptor protocol实现。你知道吗

datamodel钩子^{}优先于描述符协议。因此，如果您定义了一个定制的__delattr__方法，那么将优先调用属性deleter。你知道吗

事实上，它是__delattr__的默认实现，如果需要，它将去调用描述符，您可以通过注释掉以super开头的行来验证这一点（您应该看到属性deleter现在根本不会被调用）。你知道吗

有了这个推理，你就可以理解这样的一系列事件：

Deleting attr is_open
# the del statement `del hello.is_open` is directly invoking Hello.__delattr__,
# passing in attr="is_open" as argument

# Now, the implementation of `Hello.__delattr__` calls
# `super().__delattr__(attr)`, passing along the argument attr="is_open", which
# then invokes a descriptor for that attribute (i.e. the function
# `Hello.is_open.fdel` is about to be called)

Using custom deleter!
# This comes from within the property (`Hello.is_open`) deleter.

Deleting attr _is_open
# The implementation of the `is_open` deleter actually uses another del
# statement, i.e. `del self._is_open`. This invokes again `Hello.__delattr__`,
# passing attr="_is_open" as an argument. However, this time there is no
# descriptor with the name `_is_open` present so an attribute gets deleted from
# the instance namespace instead. Note that the attribute `self._is_open` was
# there in `self.__dict__` already because it gets created during the __init__
# method when `hello = Hello()` is executed.

需要注意的是，__delattr__第一次和第二次收到不同的参数："is_open"第一次然后"_is_open"第二次。你知道吗