Meta-class inheritance problem (Python2.0 bug?)

[Lorien Dunn]

Gordon McMillan wrote:

> 
> That's the same code as Python 1.5. Did this work before?

        I don't know. I don't have python 1.5 anymore, and its a new extension I'm 
writing. David Abrahams (the major author of the boost::python lib) said 
that isinstance() and issubclass() have been fixed to deal with metaclasses 
in Python 2.0, but it seems this check of self hasn't.
> 
> From your previous post, it does not appear to be using the Don Beaudry
> hook, but then again, I don't know what type(MyBoostClass) is.

        A boost class is a class that uses the boost::python library to reflect a 
C++ class in python. It does use the Don Beaudry hook. IMHO Boost seems to 
be the only reasonable way to wrap C++ (and I've tried SWIG, Siloon, SIP, 
CXX, and Boost), in that it supports exceptions, overidable virtual 
functions, derivation from C++ classes without shadow classes, automatic 
type conversion, function overloading, etc. It is also MUCH easier to use 
than Zope. However it puts a lot of stress on the compiler in terms of 
adherance to the ansi standard. And a lot of stress on the system. The 
wrapping code of my extension takes about  10 minutes 160 megs of ram to 
compile (and its not that large). 
        Boost as a whole is a C++ library that seems to be almost a draft for the 
next ansi standard. If you use C++ it will make your life more pleasant. 
Have a look at http://www.boost.org.

Thanks,
Lorien Dunn