When I read code from others, I found someone defines __reduce__ for the class accepting arguments. For instance, here. If you are not aware, __reduce__ magic method is for pickling a Python object.
def __reduce__(self, args=(), kwargs=None):
kwargs = {} if not kwargs else kwargs
return (unpickle_backend, (self.__class__, args, kwargs))
So I quote the document:
The interface is currently defined as follows. The
__reduce__()method takes no argument and shall return either a string or preferably a tuple (the returned object is often referred to as the “reduce value”).
有人帮我理解吗？
The pickle machinery will only ever call
__reduce__without arguments. Celery has chosen to overload their__reduce__implementations to support arguments; these arguments are used by Celery internally, and are not provided by the pickle machinery. Since the extra arguments are optional, these__reduce__methods can still handle no-argument calls, and they are still compatible with the__reduce__protocol.For example,
celery.backends.rpc.RPCBackend.__reduce__looks like this:It passes a non-empty
kwargsargument tosuper().__reduce__. The delegated-to__reduce__implementation can handle that, but it's not a call the pickle machinery would ever make directly.