Small cleanup in models.base

[Ivan Illarionov]

fixed # TODO: Checking for the presence of '_meta' is hackish.

[Ivan Illarionov]

Since each base can only be a Model subclass it always has the same metaclass and therefore always has a _meta attribute except the Model class itself.

[Chris Beaven]

Push to a design decision, since this sounds about right but such a core change needs a bit of thinking.

[Ivan Illarionov]

I can also suggest the renaming of the first argument of ModelBase.__new__ from cls to metacls for clarity, because it really is a ModelBase metaclass. It can be a source of confusion in future development and for people trying to understand what's happening since it's not clear from the code that cls here is NOT a newly created Model or Model-based class.

Oh, and ModelBase can be renamed to something like _ModelMetaclass because it is not a BASE class for Model it is a CLASS of Model class. And with first underscore if it is not intended for importing or customization.

[Ivan Illarionov]

And, if we are talking about the design decision, I would also note that classmethods of Model (add_to_class and _prepare) can be safely moved to its metaclass as normal (not class) methods. It has the advantage of separating the logic of Model classes and Model instances.

[Chris Beaven]

Perhaps you could try to get a discussion going in the django-dev group.

[Ivan Illarionov]

fixes the previous patch

[Ivan Illarionov]

I mistakenly thought that only Model subclasses can be in bases. New patch fixes this.

[Ivan Illarionov]

small optimization of previous patch

[Ivan Illarionov]

even more optimisations

[eratothene]

Link to the discussion:
​http://groups.google.com/group/django-developers/browse_thread/thread/0073b756f35b2030?hl=en

[Ivan Illarionov]

filter(lambda b...) changed to list comprehension

[Malcolm Tredinnick]

(In [7098]) Fixed #6214 -- Added some small optimisations to model class initialisation.
Thanks, Ivan Illarionov.