在新样式类中，实例 __ getattr __ 方法不再拦截内置操作所进行的调用。使用内置操作时，搜索从类级别而不是实例级别开始。不过，对方法名称的显式调用将起作用：

In new style classes, the instance __getattr__ method no longer intercepts calls made by built-in operations. When the built-in operation is used, the search begins at the class level instead of at the instance level. Nevertheless, an explicit call to the method name will work:

>>> class Li(list): pass
>>> m, n= Li([1]), Li([2])
>>> def r(*args, **kwargs):
        raise Exception('hop')

>>> setattr(m, '__iadd__', r)
>>> m += n           # no exception raised

>>> m.__iadd__(n)    # explicitly calling the method searches the instance first
Traceback (most recent call last):
  File "<pyshell#76>", line 1, in <module>
    m.__iadd__(n)
  File "<pyshell#73>", line 1, in r
    def r(*args, **kwargs): raise Exception('hop')
Exception: hop

实现第二个目标的一种方法是使用元类，尤其是如果您只想创建属性一次（根据您的评论）：

One way to accomplish your second goal is to use a metaclass, especially if you only want to create the attributes once (as per your comments):

>>> class Meta(type):
    def __new__(meta, name, bases, attrdict):
        print 'adding class attributes!'
        for s in {'__iadd__','append'}:
            attrdict[s] = r
        return super(Meta, meta).__new__(meta, name, bases, attrdict)
    def __init__(cls, name, bases, attrdict):
        super(Meta, cls).__init__(name, bases, attrdict)


>>> class Li(list):
    __metaclass__ = Meta


adding class attributes!          # attributes added here
>>> t = Li()                      # but not on instance creation
>>> Li.__dict__
dict_proxy({'__module__': '__main__', '__metaclass__': <class '__main__.Meta'>, '__iadd__': <function r at 0x02AAA270>, '__dict__': <attribute '__dict__' of 'Li' objects>, '__weakref__': <attribute '__weakref__' of 'Li' objects>, '__doc__': None, 'append': <function r at 0x02AAA270>})