Volunteering for chores: Are maintainers open to changes?

[Dronakurl]

Hi,

thanks for the great package. When using it, I found some things that bother me, an I have a few ideas for chores.

• remove the root logger that messes up the log of the application that I am using
• improve pyproject.toml
• use uv as a package manager to manage dependencies
• use pre-commit-hooks to ensure formatting and linting
• use ruff instead of black and isort
• make sure, the type annotations are consistent
• more helpful comments

Would you guys be open to one or several PR with those things?
I am asking first, because maybe you have strong opinions about this stuff.
I want to keep the package as lightweight as it is, just improving the usability.

For my application I wanted to get my own DetectionModel and want to remove the rocks, I stumbled upon.

Thanks for you opinion on that and no hurt feelings if the answer is just "no, thanks" :-)

[fcakyon]

@Dronakurl I'm in love with the idea! Please open dedicated discussions before each PR to confirm the changes. Feel free to tag me directly in your PR's and discussions 👍🏻

• Instead of removing the logging, how about letting users set env var named SAHI_LOGLEVEL and use it in sahi logger?
• Migration from pip+setup.py to uv+pyproject.toml sounds perfect!
• I'm not sure about pre-commit since it might make contribution a bit harder for those that are not familiar with but you can use it in local.
• Migration to ruff is also good 💯
• We are always open for more type annotations + code simplification + comments & documentation 🚀

[Dronakurl]

Cool! Is it okay, when I work with draft PRs? This way, we can discuss directly on the proposed changes.
Also: Which python versions do you want to support? I suggest >3.10 is a good choice. Currently it is set to 3.8, but I think that's really ancient and this way, we cannot use some features especially in type annotations.

[fcakyon]

@Dronakurl Yes, we can discuss in draft PRs!

I understand your concerns about Python 3.8 being ancient, but we try to keep the legacy support in the ultralytics package as well since many people have to stick to ancient Python versions due to Jetson Nano and Nvidia Jetpack compatibility.

[Dronakurl]

Sure, I understand.
Formatting and proper type checking can really find some weird issues. Look at this gem here:

sahi/sahi/models/yolov5sparse.py

Line 104 in c9469f5

"giraffe" "backpack",

Python will interpret this as "giraffe backpack", not "'giraffe', 'backpack'" like intended. Maybe there is someone out there wondering why he can't detect backpacks. :-)

It might take a while to produce a consistent PR.

[Dronakurl]

I am currently working my way through CI.yml.
Did you realize that operating-system: [ubuntu-latest, windows-latest, macos-latest] does not actually run anything on windows or macos? You would have to include runs-on in the matrix strategy (see docs.github). Or did I miss something?

[fcakyon]

@Dronakurl Perfect catch! This is indeed a bug that needs to be fixed. Maybe Github changed the OS usage logic at some point, and we missed it.