Another way to spell __metaclass__ ?

[Michael Hudson]

Thomas Heller <theller at python.net> writes:

> Mike Dean <klaatu at evertek.net> writes:
> 
> > On 13 Dec 2002 19:34:04 +0100 Thomas Heller <theller at python.net> wrote:
> > > Didn't Michael Hudson some time ago have a patch which allowed
> > > to write something like this?

Yep.

> > > class IUnknown(bases) [interface]:
> > >    ...
> > 
> > > or
> > 
> > > def IUnknown() [a, b, c]:
> > >    ...
> > 
> > I think I remember seeing something like that a long time ago (well,
> > long time relative to my Python activity).  Was it for metaclasses, or
> > was it for some kind of interface mechanism?

It apparently resembles something from C#, so it could be that.

> IIRC, the list on the right was a list of callables which would have
> been invoked before or after the object on the left (class or
> function) is or was created.  So it could be used to make class- or
> static methods, or to hook the class creation.

I only implemented it for `def' statements.  It's slightly harder to
see how it would work in `class' statements.  I guess

class Foo(base) [whatever]:
    a = b

could be made equivalent to

Foo = whatever('Foo', (base,), {'a': b})

(so morally `whatever' defaults to `type'), but it's then hard to see
how that would chain (the righthand most function would seem to have
to have a different signature to the others).

> I don't know how long your Python activity is, but I think it was this
> year.  Maybe a search on the python-dev archives will find it.

I don't think I posted the patch anywhere, and I don't think I still
have it.  OTOH, it wasn't very hard -- I'm pretty sure I could
recreate it in a couple of hours.

I have a cunning plan to write a PEP targeted at Python 2.4 for this
fairly soon after 2.3 gets out.  It's probably too late for 2.3 and it
rather defeats the "no new syntax" goal...

Cheers,
M.

-- 
  Hey, if I thought I was wrong, I'd change my mind.  :)
                                    -- Grant Edwards, comp.lang.python