Checks failing locally but not on GitHub

[EvaJanouskova]

@tamuri, I'm now facing another issue with the Checks. Currently, for the branch, where I'm working on PR #1609, the Checks are failing locally, but not on GitHub. I suspect this is because we still haven't fixed the version of ruff in the tox setup, as I mentioned might be needed in #1687. GitHub is now using a newer version of ruff for the tox Checks:
https://github.com/astral-sh/ruff/releases/tag/0.13.0

Locally, it's failing due to import formatting in src/scripts/healthsystem/analysis_comparison_of_impact_in_different_scenarios.py.

Could we please fix the ruff version in the tox setup so that the outcomes are consistent between local and GitHub runs? Otherwise, I end up having to run tests multiple times on GitHub for bugs I could have resolved locally.

Also, when I look at the script where the Checks are failing locally, I see an import like from src.scripts., whereas most scripts use from scripts.. I'm not sure which is correct, or if both are acceptable?

[tamuri]

This is frustrating - the checks don't seem very stable, compared to other tooling. I suggest two changes:

1. As per your recommendation, fix version of ruff in dev.txt and tox.ini to be the same. And
2. Exclude src/scripts from ruff checks. I don't think failing linters in this directory should prevent merging. ruff --exclude src/scripts will do that.

[EvaJanouskova]

2. Exclude src/scripts from ruff checks. I don't think failing linters in this directory should prevent merging. ruff --exclude src/scripts will do that.

Will there be any disadvantages to do this?

[tamuri]

They'll be some but because of how often ruff updates it rules, we end up having to make changes more often in this directory than src/tlo. src/scripts is intended to be a place for exploratory analysis and half-finished scripts. If there's an unused import here (or other lint-style issues), I just don't think it matters.

[EvaJanouskova]

You might be right.

[EvaJanouskova]

Could there then be some testing purely for personal use, to check whether the script is correctly formatted at the time you're working on it, which we could run locally?