Something strange with 'object'

[Dan P]

Boris Boutillier wrote:

> hi all,
> 
> I made a python program and had a bug that turn out to be a problem with
> 'object' heritage, here is a tiny script that reproduce the bug :
> Buggous version :
> 
>>class a(object):
>>        def __add__(self,other):
>>                print 'a.add'
>>class b(a):
>>        def __radd__(self,other):
>>                print 'b.radd'
>>x = b()
>>y = b()
>>x.__class__ = a
>>x+y
> b.radd
> 
> This means that a.add hasn't been called for x.
> 
> If you remove the object heritage, the buggous behaviour disappear :
>>class a:
>>       def __add__(self,other):
>>                print 'a.add'
>>class b(a):
>>        def __radd__(self,other):
>>                print 'b.radd'
>>x = b()
>>y = b()
>>x.__class__ = a
>>x+y
> a.add
> 
> which is what you want.
> 
> Is this a real bug ? or did I miss something about use of 'object' ? I
> tried to find documentation in python reference guide but found nothing
> helpfull.
> Someone's got an idea ?

My guess is that it's the __mro__. You changed the instance's class, but 
it's method resolution order (stored in the __mro__ attribute) still has 
class b as its first element.

Tell me what you find out,
Dan

 
> Boris Boutillier