gh-136380: Fix import behavior for concurrent.futures.InterpreterPoolExecutor

[aisk]

It's hard to write the test, but on my local machine, when remove the _interpreters module or raise a ModuleNotFoundError in the try branch, the behavior is correct compare to a normal non-exist field:

>>> from concurrent.futures import InterpreterPoolExecutor Traceback (most recent call last): File "<python-input-0>", line 1, in <module> from concurrent.futures import InterpreterPoolExecutor ImportError: cannot import name 'InterpreterPoolExecutor' from 'concurrent.futures' (C:\Users\xxxxx\Source\cpython\Lib\concurrent\futures\__init__.py) >>> from concurrent import futures >>> futures.InterpreterPoolExecutor Traceback (most recent call last): File "<python-input-2>", line 1, in <module> futures.InterpreterPoolExecutor File "C:\Users\xxxxx\Source\cpython\Lib\concurrent\futures\__init__.py", line 69, in __getattr__ raise AttributeError(f"module {__name__!r} has no attribute {name!r}") AttributeError: module 'concurrent.futures' has no attribute 'InterpreterPoolExecutor' 

>>> from concurrent.futures import NotExist Traceback (most recent call last): File "<python-input-0>", line 1, in <module> from concurrent.futures import NotExist ImportError: cannot import name 'NotExist' from 'concurrent.futures' (C:\Users\xxxxx\Source\cpython\Lib\concurrent\futures\__init__.py) >>> from concurrent import futures >>> futures.NotExist Traceback (most recent call last): File "<python-input-2>", line 1, in <module> futures.NotExist File "C:\Users\xxxxx\Source\cpython\Lib\concurrent\futures\__init__.py", line 69, in __getattr__ raise AttributeError(f"module {__name__!r} has no attribute {name!r}") AttributeError: module 'concurrent.futures' has no attribute 'NotExist' 

We must import _interpreters in __init__.py to determine whether to append the InterpreterPoolExecutor to __all__. This may have a negative impact on importing performance.

• Issue: Inconsistent import behavior when concurrent.futures.InterpreterPoolExecutor not exist #136380

[serhiy-storchaka]

The downside of this is that _interpreters is now always imported.

Could you add tests? It is not easy, the tests should remove concurrent and all its submodules recursively from sys.module, and restore sys.module at the end. There are helpers for this somewhere, they can be private.

[ZeroIntensity]

cc @ericsnowcurrently

[aisk]

The downside of this is that _interpreters is now always imported.

Could you add tests? It is not easy, the tests should remove concurrent and all its submodules recursively from sys.module, and restore sys.module at the end. There are helpers for this somewhere, they can be private.

Found a way to test it by running the test within a subprocess.

[serhiy-storchaka]

LGTM. 👍

[ZeroIntensity]

I didn't mean to hit the update button, sorry about that.

[sobolevn]

Thank you!

[miss-islington-app]

Thanks @aisk for the PR, and @sobolevn for merging it 🌮🎉.. I'm working now to backport this PR to: 3.14.
🐍🍒⛏🤖

[bedevere-app]

GH-136420 is a backport of this pull request to the 3.14 branch.