Why does pickle.dumps call __getattr__?
pickle, python, serialization
Solution
According the documentation for cPickle: http://docs.python.org/library/pickle.html
object.__getstate__()
Classes can further influence how their instances are pickled; if the class defines the method `__getstate__()`, it is called and the return state is pickled as the contents for the instance, instead of the contents of the instance’s dictionary. If there is no `__getstate__()` method, the instance’s `__dict__` is pickled.
Note
At unpickling time, some methods like `__getattr__()`, `__getattribute__()`, or `__setattr__()` may be called upon the instance. In case those methods rely on some internal invariant being true, the type should implement either `__getinitargs__()` or `__getnewargs__()` to establish such an invariant; otherwise, neither `__new__()` nor `__init__()` will be called.
Since you are trying to assert that `hasattr(self, '_data')` is True, I believe that you need to use `__getinitargs__()` or `__getnewargs__()`. This is because when using pickle, a classes `__init__` method is not called.
Problem
``` import cPickle class Foo(object): def __init__(self): self._data = {'bar': 'baz'} def __getattr__(self, name): assert hasattr(self, '_data') return self._data[name] # I even had to define this just to stop KeyError: '__getstate__' def __getstate__(self): return self.__dict__ foo = Foo() bar = cPickle.dumps(foo) cPickle.loads(bar) ``` This raises an assertion error. I thought `pickle`/`cPickle` just turns `__dict__` into a string when dumping and then uses that string to set the `__dict__` of the new object directly when loading. Why would `dumps` need to call `bar.__getattr__`? How can I change `Foo` to avoid that?