[Zope3-dev] Interface package broken for Python 2.2 metaclasses

[Phillip J. Eby]

I'm not sure where to send/post this.  I'm working on porting TransWarp to 
use Python 2.2 metaclasses instead of ExtensionClass, and I have found a 
problem with the Interface package (which TransWarp makes some pretty heavy 
use of).

The problem is that the routines in the Interface.iclass module check 
whether a type is in 'ClassTypes' in order to know if an object is a class 
(and thus whether '__implements__' or '__class_implements__' should be 
used).  If you have a class based on a Python metaclass, the Interface 
package will think the class implements what its instances implement, which 
leads to the collapse of the type system as we know it.  :)

Unfortunately, despite some hacking around with it myself, I can't seem to 
wrap my brain around a decent solution.  I tried registering the metaclass 
with 'Interface.assertTypeImplements()', but that didn't help because it 
checks the  '__implements__' attribute *before* checking the type registry.

(It sort of seems to me that 'assertTypeImplements' may be wrongly named 
anyhow, since it actually asserts that *instances* of the type implement 
something.  But that's entirely beside my point at present...)

Would it perhaps make sense to change ClassTypes to a dictionary, and 
supply an 'Interface.declareMetaclass()' call that would add a metaclass to 
ClassTypes?  This seems messy, but so far I haven't been able to figure out 
any better way to intelligently detect that a Python metaclass is a 
class-like thing.

I've made a hack equivalent to 'declareMetaClass()' in my own copy of the 
Interface package for 2.2, and it solves my problem for now, but what 
should be done to the official package?  Anyone have any better ideas?