[Dev] Python startup times

[Alec Flett]

Phillip J. Eby wrote:
> At 09:45 AM 7/27/2005 -0700, Alec Flett wrote:
>> I think unfortunately this isn't really testing the zipping of python 
>> files - just executing "pass" does not load any of the libraries from 
>> the .zip
>
> Actually, it does.  Try this experiment for yourself:
>
sorry, my bad - I meant (and should have written) "does not load a 
significant number of libraries"

perhaps we could do a test to see what percentage of the load time comes 
from python loading, and what comes from the rest of chandler.

If you imagined starting your stopwatch when ./release/RunChandler was 
run, then hitting on the first line of Chandler.py, and then hitting it 
again when the chandler window is done rendering. Now those would be 
some interesting numbers to see with/without some zipping!

Maybe a wx person (or john, for a CPIA answer) can tell us how we know 
when the main view has finished rendering?

Alec

>     python -c "import sys; print sys.modules.keys()"
>
> If you go through the list of module names printed, you'll find that 
> quite a few of them (e.g. warnings, exceptions, site, types, some 
> parts of re, os.path and others) are coming from .py files that are 
> imported by Python as part of its normal startup operations.  So yes, 
> "pass" does load quite a few libraries from the .zip, and that's what 
> accounts for the 350ms difference on the Mac start time, and the 40ms 
> or so difference on the faster machine.
>