父级访问在Python中的子级中定义的类变量？

Question

I am learning OOP in python, and I am surprised to find in another post that a parent class can access a class variable defined in child, like the example below. I can understand that child can access a class variable defined in parent bc of inheritance (or is it not inheritance but something else?). What makes this possible and does it imply a pool of class variables that a subclass can adds to (and its superclasses can access) when it inherits from a superclass?

class Parent(object):
    def __init__(self):
        for attr in self.ATTRS: // accessing class variable defined in child?
            setattr(self, attr, 0)

class Child(Parent):
    #class variable
    ATTRS = ['attr1', 'attr2', 'attr3']

    def __init__(self):
        super(Child, self).__init__()

Answer 1

The example is a bit twisted. The definition of the parent class is not complete, it references non-existing class variables, which only becomes apparent when instantiating the class or sub-classing it.

If you subclass as follows:

class Child2(Parent):
    def __init__(self):
        super(Child2, self).__init__()

Then instantiating the child class won't work:

Traceback (most recent call last):
  File "./cl.py", line 25, in <module>
    y=Child2()
  File "./cl.py", line 20, in __init__
    super(Child2, self).__init__()
   File "./cl.py", line 6, in __init__
    for attr in self.ATTRS:
AttributeError: 'Child2' object has no attribute 'ATTRS'

You would have to fix your parent as follows to make it work:

class Parent(object):
    ATTRS = None
    def __init__(self):
        if self.ATTRS:
            for attr in self.ATTRS:
                setattr(self, attr, 0)

The reason it works in your original example is that Python is an interpreted language, so the code is only evaluated at run-time.

Answer 2

父不访问类变量，因为self在__init__(self)是一个实例Child 。