On 04/18/10 12:49, Tim Diels wrote:
Hi
I was thinking of writing a GUI toolkit from scratch using a basic '2D
library'. I have already come across the Widget Construction Kit.
My main question is: Could I build a GUI toolkit of reasonable
performance with the Widget Construction Kit, would it still feel more
or less lightweight? By reasonable I mean that the user wouldn't think
of the interface as being slow or unresponsive.
I've also thought of using pyglet to build widgets with, but this would
seem to be overkill. As a side question: by using opengl, the work would
be delegated to the GPU rather than the CPU; is this always a good
thing, or does it have downsides as well (performance, power usage, ...)?
Are there any other libraries that may be of interest to me?
Thanks in advance
It probably depends on how low level you want to go, I have pondered
about the possibility myself to have an all python(ic) gui toolkit,
capable of writing a (x11) windowing manager itself with.
But I decided that using tkinter and just live with its rough corners is
more bang for the buck for me than to reimplement tkinter badly.
However as a thought exercise I did spend some energy on it and I had
the following ideas.
- Need to have direct access to at least x11, cocoa and win32gui;
or even lower than that (if possible/reasonable).
- Only need to abstract enough so I can display a borderless window full
screen or on any position/size.
- Need to provide a wrapper for the input devices too, e.g.: keyboard,
mouse, joystick, touchscreen, etc.
- Optionally graphical acceleration (OpenGL, DirectX, SDL?)
- It would be good that fonts, windows, decoration and icons are all SVG
so that all these items can scale independently.
I also had some more questions:
- How about providing an interface for video playback?
- How about audio?
- How about printing?
- How about multiple displays?
- How about odd sized displays (round, triangle, etc)?
- How to handle 'legacy' gui applications?
- Need to remain completely BSD licensed so that it is possible that it
might some day be incorporated in the standard distribution.
So I gave up on it as it seem to me much to much work for not enough
ROI, but I still would welcome anyone giving it a shot :-)
--
mph
--
http://mail.python.org/mailman/listinfo/python-list
