dict subclass and pickle bug (?)

[James Stroud]

Hello All,

I subclassed dict and overrode __setitem__. When instances are
unpickled, the __setstate__ is not called before the keys are assigned
via __setitem__ in the unpickling protocol.

I googled a bit and found that this a bug filed in 2003:

http://bugs.python.org/issue826897

It is still "open" with "normal" priority.

Am I missing something? Is there a workaround for this bug that makes
fixing it pointless or has it just fallen through the cracks for the
last 5 years?

Here is an example:

class DictPlus(dict):
def __init__(self, *args, **kwargs):
self.extra_thing = ExtraThingClass()
dict.__init__(self, *args, **kwargs)
def __setitem__(self, k, v):
do_something_with(self.extra_thing, k, v)
dict.__setitem__(self, k, v)

Upon unpickling, the error would be:

AttributeError: 'DictPlus' object has no attribute 'extra_thing'


I'm still using python 2.5.1.

James

--
James Stroud
UCLA-DOE Institute for Genomics and Proteomics
Box 951570
Los Angeles, CA 90095

http://www.jamesstroud.com

 

[James Stroud]

Hello All,

I subclassed dict and overrode __setitem__. When instances are
unpickled, the __setstate__ is not called before the keys are assigned
via __setitem__ in the unpickling protocol.

I googled a bit and found that this a bug filed in 2003:

http://bugs.python.org/issue826897

It is still "open" with "normal" priority.

Here is the ugly "fix" I'm basically going to have to live with, it seems:

class DictPlus(dict):
def __init__(self, *args, **kwargs):
self.extra_thing = ExtraThingClass()
dict.__init__(self, *args, **kwargs)
def __setitem__(self, k, v):
try:
do_something_with(self.extra_thing, k, v)
except AttributeError:
self.extra_thing = ExtraThingClass()
do_something_with(self.extra_thing, k, v)
dict.__setitem__(self, k, v)
def __setstate__(self, adict):
pass


This violates this:

Beautiful is better than ugly.


I can't imagine this bug has survived but I also can't imagine any
better way to do this without specifying a different protocol, which
would probably break other pickling I'm doing. I don't feel like finding
out right now.

Maybe repeal pep 307 ;o)


James


--
James Stroud
UCLA-DOE Institute for Genomics and Proteomics
Box 951570
Los Angeles, CA 90095

http://www.jamesstroud.com

 

[Terry Reedy]

It is an 'issue' reporting a possibly unexpected side-effect of protocol
working as designed and documented. Possibly a design flaw, but not a
bug in the narrow sense (in spite of the url of the issue tracker).

Here is the ugly "fix" I'm basically going to have to live with, it seems:

class DictPlus(dict):
def __init__(self, *args, **kwargs):
self.extra_thing = ExtraThingClass()
dict.__init__(self, *args, **kwargs)
def __setitem__(self, k, v):
try:
do_something_with(self.extra_thing, k, v)
except AttributeError:
self.extra_thing = ExtraThingClass()
do_something_with(self.extra_thing, k, v)
dict.__setitem__(self, k, v)
def __setstate__(self, adict):
pass

I took the liberty of adding this to the issue.

I can't imagine this bug has survived

because there is no bug to fix. I have suggesting closing.

Terry Jan Reedy

 

[James Stroud]

because there is no bug to fix. I have suggesting closing.

May I suggest to add something to this effect within the issue itself so
others won't spend time trying to figure out why the "bug" is still
open? If this is a more general feature of issues, then perhaps it would
be helpful to provide a footnote at the bottom of all issue pages via
the page template that explains why they are not bugs and suggests a
general course of action for the programmer.

James

 

[James Stroud]

May I suggest to add something to this effect within the issue itself so
others won't spend time trying to figure out why the "bug" is still
open?

Sorry, you did that.

James