Someone using PyGtk

Jeff Shannon jeff at ccvcorp.com
Wed Apr 3 13:30:30 EST 2002


In article <Xns91E4ADCD75B59RASXnewsDFE1 at 130.133.1.4>, 
starx at pacbell.net says...
> Jeff Shannon || Tue 02 Apr 2002 01:55:39p:
> 
> > Hm, wxPython seemed to perform adequately for me, on a (Win95) 
> > P200 with only 64MB RAM.  
> 
> I use WX too, and it seems that you pretty much pay a fixed price for 
> importing the modlue, no matter how much you need or how big your program 
> is. I'd think most other frameworks would have the same problem, an easy 
> way to tell would be something like:
> 
> print "Loading wxPython...",
> from wxPython.wx import *
> print "done"
> 
> ... I bet most of your execuation time for a hello world program would be 
> between those prints.

Indeed, testing with this simple script: 

from time import time
start = time()
from wxPython.wx import *
print "import took %f seconds" % (time() - start)

seems to indicate a 2 1/2 - 4 1/2 second import time on my (as 
noted, extremely slow) machine.  (Faster times happen when 
repeated in quick succession, slower times after some time has 
passed; I suspect caching and virtual memory paging to be the 
significant variables.) This, added to the time to start the 
Python interpreter, means that there is a significant startup 
time for even a very simple wxPython program.  But that startup 
time increases fairly slowly with the complexity of the program.  
And keep in mind that starting Notepad.exe, an optimized C 
program that consists almost entirely of an OS-builtin control, 
takes almost a second on this same machine, and any complex 
application (such as Excel or Photoshop) can take 45 seconds to 
a minute to load.  Also keep in mind that slow startup time is 
not necessarily correlated with poor responsiveness once the 
application is running.  Therefore, I would say that, unless 
there are *serious* timing constraints, complaints about wxPython 
being slow are on questionable ground.  I'd consider algorithm 
optimization to be a more fruitful area to work on, than worrying 
about the toolkit.

(I'm sure that Philip is in agreement on this; I'm clarifying my 
point for the O.P.'s benefit here.)

-- 

Jeff Shannon
Technician/Programmer
Credit International



More information about the Python-list mailing list