Ctypes experiments start well (Characterising the problems...)


Today began work in earnest on creating a PyOpenGL implementation with ctypes. This is basically a matter of figuring out how to make the API look like the old SWIG API without introducing too much overhead, either in run-time processing, or development effort.

So far it's going well. There are only about 5 or 6 primitives required to build the interface for the functions I'm wrapping. The idea is to work my way through the base library, characterising the (extra) processing required for each function. I'm then building Python versions of the code required (with only very minor C extensions where something hasn't yet been exposed to Python). As I go, I'm generalising the code to handle more and more of the cases. The idea being that by the time I finish going through the library I'll solve most of the use-cases and then I'll go back and worry about how to make it all efficient.

Shiva and Shademan also came by for a few hours, and I did some shopping, so I haven't done a full day on it, but there's enough work done to make me think I've earned a bit of a break :) .

Comments

Comments are closed.

Pingbacks

Pingbacks are closed.