Interesting list() un-optimization

[Roy Smith]

In article <513d18d6$0$6512$c3e8da3$5496439d at news.astraweb.com>,
 Steven D'Aprano <steve+comp.lang.python at pearwood.info> wrote:

> On Sun, 10 Mar 2013 18:34:58 -0400, Roy Smith wrote:
> 
> > Yeah, that's what I was afraid of.  The "obvious" solution of:
> > 
> > class QuerySet(mongoengine.queryset.QuerySet):
> >     def __init__(self, document, collection):
> >         super(QuerySet, self).__init__(document, collection) [...]
> >         del self.__len__
> > 
> > results in:
> > 
> > [rest of stack dump elided]
> >     del self.__len__
> > AttributeError: __len__
> > 
> > which I don't understand.
> 
> You don't define a per-instance custom QuerySet attribute called __len__, 
> so you can't delete it from the instance.
> 
> The existing __len__ attribute is attached to the mongoengine 
> queryset.QuerySet class, not the instance. You could monkeypatch the 
> parent class, but that will probably break something else:
> 
>     del mongoengine.queryset.QuerySet.__len__
> 
> 
> or you could try over-riding __len__ with a fake that pretends it doesn't 
> exist:
> 
> 
> def __len__(self):
>     raise AttributeError
> 
> 
> Try that and see it it works. (It may not.)

Yeah, that was one of the things I experimented with.  It seems to work, 
but like most of these other things, I suspect it's version specific.  
If list() does:

try:
    l = obj.__len__()
except AttributeErrror:
    l = 0

then we're good.  On the other hand, if it does:

if obj.getattr('__len__', None):
    # etc

then it fails.

At this point, I'm thinking the monkeypatch is the right solution.