Add an __all__ variable to constants.c

[MyreMylar]

An experiment to see if this will fix the LGTM errors.

[lgtm-com]

This pull request fixes 1 alert when merging 6087114 into 6df1c80 - view on LGTM.com

fixed alerts:

• 1 for Unused import

[lgtm-com]

This pull request fixes 1 alert when merging 7242bb6 into 6df1c80 - view on LGTM.com

fixed alerts:

• 1 for Unused import

[illume]

What was the result of your testing? Does this help any tool?

• LGTM
• pylint
• python help(pygame), help(pygame.locals)
• jedi
• pycharm

I expect that some analyze the python code statically, so that the __all__ would need to also be static, not dynamic?

[MyreMylar]

What was the result of your testing? Does this help any tool?

• LGTM
• pylint
• python help(pygame), help(pygame.locals)
• jedi
• pycharm

I expect that some analyze the python code statically, so that the __all__ would need to also be static, not dynamic?

From my testing it seems that pylint can pick up the __all__ variable the way I have defined it here. Most of the alerts and errors we got previously have been covered by the .pyi files in pylint, jedi & PyCharm.

using 2.0.0.dev8 if you do:

pylint --extension-pkg-whitelist=pygame --allow-wildcard-with-all y test.py

on a file called test.py containing:

"""
Doc
"""
from pygame.constants import *
print(K_END)

you will get:

test.py:4:0: W0401: Wildcard import pygame.constants (wildcard-import)

But with this pull request compiled you won't. allow-wildcard-with-all is disabled by default though so I guess the more reliable way to fix pylint would be for us to never use wildcard imports in pygame and instead import each constant individually into pygame's __init__.py and locals.py. The downside with that is that it is tedious and basically duplicates the long list of constants in two more places in the code (at least)

LGTM is a mystery to me.

[MyreMylar]

On LGTM,

I don't know how it is checking for the __all__ variable for its alerts, or if the incremental check we get here on pull requests works the same as a full scan (which seem to happen on a slower schedule in my experience). It's possible the version of __all__ I generate in the c file here is not what the LGTM checker is expecting, but without knowing how it checks it's hard to know. I could try emailing the LGTM team at GitHub?

I suspect, as with pylint, the full-proof way to eliminate these alerts is to directly import each constant individually.

[MyreMylar]

I'll guess I'll try emailing semmle (creators of LGTM) about this just to see what they say, unless anyone has a better idea?

[MyreMylar]

I have now tried sending an email, will see what response I get.

[illume]

👍 Thanks