Python / C++ integration

Brandon Van Every vanevery at 3DProgrammer.com
Fri Feb 7 06:25:51 CET 2003


Ok, so Numarray addresses a lot of my 3D graphics concerns.

People have said "just drop down to C" when you want to do something
efficient.  That is not acceptable.  I don't drop down to C.  Friends don't
let friends do C.  If I'm going to drop down, first it's going to be to C++,
then to ASM.  There are lotsa valid reasons to keep an OO paradigm before
descending to ASM.  Python cannot be "the owner" of my OO paradigms in my
development.  It may own a layer of them, but I certainly won't let it own
all of them.

What are the pitfalls of trying to get classes to interoperate between
Python and C++?

One pitfall has already come up: Python consumes more storage for basic
types than does C++.  The objects will not be the same size.  This has
implications... some mitigable... but also opening a door to tedium.  I
might write some translation layers by hand, but I won't write a bunch.  If
it's about manual labor, then in practice, Python and C++ will be loosely
coupled.  Mostly off in their own worlds, really.

How about throwing exceptions?

How about allocation and deallocation of objects?

How about debuggers?

Any other gotchas?

--
Cheers,                         www.3DProgrammer.com
Brandon Van Every               Seattle, WA

20% of the world is real.
80% is gobbledygook we make up inside our own heads.





More information about the Python-list mailing list