[Chicago] problem with __getstate__

[Kumar McMillan]

On Sun, May 18, 2008 at 2:15 PM, Massimo Di Pierro
<mdipierro at cs.depaul.edu> wrote:
> The fact is that I can use copy_reg for the class S and does not fix the
> problem. I can only fix the problem using copy_reg on the class A but only
> this simple toy example. I am still running into problems into more complex
> example although this should be straightforward. I guess the question is
> why does __getstate__ behaves differently depending on the super classes but
> the super.__getstate__ is not called?

attributes prefixed with double underscores have different scoping
rules in Python, they are essentially "private."  This *might* be the
source of confusion around __getstate__ although admittedly I haven't
spent time to completely understand the issue.

http://diveintopython.org/object_oriented_framework/private_functions.html

> Massimo
>
>
> On May 18, 2008, at 2:07 PM, Atul Varma wrote:
>
> Let me rephrase the first sentence of my last post: I don't think the issue
> is inheritance of just *any* class, but specifically inheritance of built-in
> types. :)
>
> - Atul
>
> On Sun, May 18, 2008 at 12:06 PM, Atul Varma <varmaa at gmail.com> wrote:
>>
>> I could be wrong, but I don't think the issue is inheritance--I think it's
>> actually because you're subclassing a built-in type, which requires doing
>> some special stuff with pickle (namely, registering with copy_reg as you've
>> done in your last email).
>>
>> If you use pickle instead of cPickle, you get a more informative
>> traceback; take a look at pickle.Pickler and notice how the dispatch table,
>> which appears to map types to serialization functions, maps the dictionary
>> type to Pickler.save_dict().  I think that has something to do with it.
>>
>> Hope that helps...  I have to run so I can't look into this more right
>> now, but let me know if you'd like me to.
>>
>> - Atul
>>
>> On Sun, May 18, 2008 at 11:58 AM, Massimo Di Pierro
>> <mdipierro at cs.depaul.edu> wrote:
>>>
>>> Odd... this works instead...
>>>
>>> from cPickle import *
>>> import copy_reg
>>>
>>> class S(dict):
>>>    def __getattr__(self,key): return self[key]
>>>    def __setattr__(self, key, value): self[key]=value
>>>
>>> class A(S):
>>>    def __init__(self): self.i=lambda u: u
>>>    #def __getstate__(self): return {}
>>>
>>> def unserialize_A(d):
>>>    return A()
>>>
>>> def serialize_A(o):
>>>    return unserialize_A, ({},)
>>>
>>> copy_reg.pickle(A, serialize_A)
>>>
>>> a=A()
>>> print a.i(3)
>>> x=dumps(a)
>>> b=loads(x)
>>> print b.i(4)
>>>
>>> I guess the rule is use copy_reg and not __getstate__ if you have
>>> inheritance.
>>>
>>>
>>> On May 18, 2008, at 1:44 PM, Massimo Di Pierro wrote:
>>>
>>>> does anybody know why the following code does not work?
>>>>
>>>>     from cPickle import *
>>>>
>>>>     class S(dict):
>>>>         def __setattr__(self, key, value): self[key]=value
>>>>     class A(S):
>>>>        def __init__(self): self.i=lambda u: u
>>>>        def __getstate__(self): return {}
>>>>
>>>>     a=A()
>>>>     x=dumps(a)
>>>>
>>>> I know that self.i is a lambda and thus not pickable but __getstate__
>>>> is called (I checked) and thus it should not matter. I guess I do not
>>>> undretansd how pickling under inheritance.
>>>>
>>>> Massimo
>>>> _______________________________________________
>>>> Chicago mailing list
>>>> Chicago at python.org
>>>> http://mail.python.org/mailman/listinfo/chicago
>>>
>>> _______________________________________________
>>> Chicago mailing list
>>> Chicago at python.org
>>> http://mail.python.org/mailman/listinfo/chicago
>>
>
> <ATT00001.txt>
>
> _______________________________________________
> Chicago mailing list
> Chicago at python.org
> http://mail.python.org/mailman/listinfo/chicago
>
>