Re: [pygame] BUG: required modules not raising error when failing to import was Re: [pygame] Surfarray.array_alpha() bugfix, py2exe/py2app segfault

[Lenard Lindstrom <len-l@xxxxxxxxx>]

Brian Fisher wrote:
> On Thu, Feb 5, 2009 at 12:21 PM, Zack Schilling 
> <zack.schilling@xxxxxxxxx <mailto:zack.schilling@xxxxxxxxx>> wrote:
>
>     I don't think manually including bufferproxy when it might not be
>     needed is a good solution, nor do I think getting py2exe to search
>     for compiled C modules calling other compiled C modules is likely
>     to happen.
>
> I agree that dependencies should not be added - but "import pygame" 
> already tries to import pygame.surface and pygame.mixer, and those 2 
> modules depend on bufferproxy (and always try to import it)
>
> so given that pygame has always included surface and mixer and those 2 
> always import bufferproxy (if available) then pygame is already always 
> importing bufferproxy - whether it is "needed" or not.
>
> So added an in python import of the module really doesn't change 
> anything other than making py2exe and py2app work out of the box.
>
>  
> .. not that things couldn't be better though. My point is just that 
> the hack fix here doesn't make anything worse.
Why don't we just have a pygame module called everything. It has top 
level import statements of all pygame modules and major dependencies, 
such as NumPy and OpenGL, but is not itself imported by __init__.py. 
That way a game packaged with py2exe/py2app can just import 
pygame.everything and be sure no modules are overlooked. The everything 
module could even be copied into a game and pruned.

Included in the pygame distribution directory is an experimental 
program, check_test.py, that tracks Pygame modules that are imported and 
accessed by a running unit test. It can be generalized to run a game. So 
one would see what a game needs, and what can be omitted. And it could 
generate a game specific pygame everything module.

--
Lenard Lindstrom
<len-l@xxxxxxxxx>