Re: [boost] Re: Python 2.2a1 & Boost.Python

From: David Abrahams (david.abrahams_at_[hidden])
Date: 2001-07-19 13:33:50

----- Original Message -----
From: <rwgk_at_[hidden]>
To: <boost_at_[hidden]>
Sent: Wednesday, July 18, 2001 9:44 PM
Subject: [boost] Re: Python 2.2a1 & Boost.Python

> --- In boost_at_y..., "David Abrahams" <david.abrahams_at_r...> wrote:
> > I think it is somewhat important. It means Python 2.2 is breaking
> source
> > compatibility on some existing code and limiting the flexibility of
> > Boost.Python. It would be better if everything could be made to
> continue to
> > work.
>
> Apparently it is Guido's goal not to break existing code in 2.2
> (http://www.python.org/2.2/descrintro.html):
>
> > While the ultimate goal of PEP 253 is to do away with
> > ExtensionClass, I believe that ExtensionClass should still
> > work in 2.2, breaking it in 2.3.
>
> What is the best way of communicating our findings to Guido?

I think a copy of your previous message would be fine.

> How would you describe the problem in two sentences?

How about in one sentence?
With versions earlier than 2.2, we were able to derive extension classes
from a mixture of built-in classes and extension classes, /even if the
built-in class was the first listed base/, but that no longer works:
  <code evidence>

> > On the other hand, perhaps it would be best if under Python 2.2,
> > Boost.Python classes were just subtypes of built-in classes.
> > That seems to
> > be the whole point of that PEP.
>
> Guido writes:
> > However, you can create mix-in classes by inheriting from "object".
> > This a new built-in, naming the featureless base type of all
> > built-in types under the new system.
>
> I am guessing that this is what we will have to use under
> Python 2.3.

I think it might be possible to extend Class rather than object. That would
certainly cut down on the size of the code needed for extension class
support in Boost.Python.

-Dave