[Pythoncard-users] Numerical Python (numpy) and PythonCard

[Kevin A. <al...@se...>]

I was playing with PyGame earlier today to see how drawing times compared to
my code. PyGame is definitely faster, but I don't know what portions of my
wxPython code can be optimized. It will be a while before I know what we can
and can't do with PyGame, but right now it looks like we'll be able to
leverage PyGame for drawing and sound.

Anyway, while trying things out in PyGame, the issue of using a surface
array and numPy came up. Basically, the window bitmap gets stored as a numPy
array which is where all the calculations and plotting take place and then
you convert the numPy array back to a bitmap for the screen. Chris Barker
sent me some code a few weeks ago that also uses NumPy to speed up some
scaling transformations. I don't know whether numPy or PIL is faster for
rotation and other effects.

Anyway, numPy is new to me, just like PIL, so I'm not sure what conversions
I need to support to make it convenient to use. Is it enough to simply
support getArray and setArray type methods for Bitmap and a drawArray for
BitmapCanvas? I would appreciate some feedback from anyone using numPy on
what would be the most helpful.

There is an example conversion of a numPy array to wxImage/wxBitmap at:
http://wxpython.org/cgi-bin/wiki/WorkingWithImages

and there is some array to PIL and PPM code as part of the numPy demos. So,
conversions are easy to add to PythonCard, I just need to know what kinds,
so hopefully someone already doing visualizations with numPy can provide
some guidance.

I will probably move the Bitmap class out of widget.py and into its own
module, which I'm thinking about simply calling image.py. That module will
also contain any generic image conversion and utility functions outside the
Bitmap class. If you are interested in working on that module let me know.

I would like to provide convenient ways to display data from PIL and numPy
and thus make PythonCard a useful tool for users of those packages without
requiring them to learn much new stuff in order to get work done.

ka