[henryiii]

This is the output of uvx flake8-lazy --apply src/**.py. See post: https://iscinumpy.dev/post/flake8-lazy

Opening this as draft for awareness and to get eyes on the changes; it's
possible I've forgotten some or added too many. Probably should wait to put
this in until 3.15b1. It's easy to rerun if there are conflicts in the future.

I don't really know how to test this, other than running a python process and checking to see what's in sys.modules (which I did locally), or checking -Ximporttime, etc. We don't have a CLI (yet?)) to check import times on.

[notatallshaw]

FYI, I'm adding a mechanism in pip where we will prevent imports from running after a certain point, before we vendor libraries that use lazy imports we will also need to add a mechanism to resolve all lazy imports and import them: pypa/pip#13829

This is not a blocker for packaging, we vendor packaging and can make sure one doesn't land without the other. But I do believe my approach is best practices for all Python package installers that use the same Python environment they are installing into, so once this lands on pip side I do think we should communicate this to known Python packaging installers that use packaging, before packaging adopts lazy imports.

[henryiii]

It seems like a "resolve all lazy imports" function would be really useful - package installers could resolve them all right before installing packages, and still get the benefits of lazy imports for things like --help.

The vendoring tool could learn to remove lazy and __lazy_modules__. Or a tool could strip them from all code (ruff does have a check to disallow them, I think).

[notatallshaw]

FYI, I'm now +1 on lazy modules, pip is ready.

[henryiii]

Two updates (besides the bump in flake8-lazy): this is now maintained by a pre-commit hook, and I'm using the set style, which is faster if there are more than 1-2 entries (checked by __contains__).

[henryiii]

I’ll rebase. Are we fine with the two changes (set and pre-commit hook)?