[MNT] homogenize CI workflows with GC.OS repositories

[tingiskhan]

Homogenizes the main workflows of pyportfolioopt with that of GC.OS repositories - template used is sktime @fkiraly

[tingiskhan]

@fkiraly Would it make sense to have a dedicated devops repo associated with sktime where re-usable workflow components are stored?

[fkiraly]

@fkiraly Would it make sense to have a dedicated devops repo associated with sktime where re-usable workflow components are stored?

Yes, that sounds like an excellent idea. I think @tschm has set this up for their repos.

However, I think it is not a priority currently when compared to getting ppfopt back on track.

[fkiraly]

I would go for uv pip install rather than uv sync. See long discussion on the topic in sktime.

[fkiraly]

why move it here? I think we should just leave it in non-uv format

[tingiskhan]

This is actually the pip standard as of 25.1: https://packaging.python.org/en/latest/specifications/dependency-groups/

[tingiskhan]

The biggest benefit is that you can install just the dependencies inside that group, without having to install the full suite of packages - something I've longed for!

[fkiraly]

indeed, you are right - introduced in 2025! I must have missed this.

Yes, `notebook-test´ should then also be here.

We should also think about how and whether to move this in the other GC.OS packages. sktime will move to 1.0 soon, that is an opportunity to move all the dev/testing related depsets to dependency-groups.

[fkiraly]

I reverted, since it would be a breaking change for environment setup - pip install pyportfolioopt[dev] no longer works

[fkiraly]

Thanks!

Looks mostly good, but in some places it notably diverges from the "close to pip setup" to more uv based idioms. I think these may be problematic and shuld be reverted:

• use of uv sync instead of uv pip install, this can cause problems with incompatible depsets
• use of uv depset definition in pyproject.toml which no longer conforms with PEP

[fkiraly]

I see, thanks!

• opinion on uv sync remains
• also, we should not rename release.yml, since it is registered as a trusted publisher with pypi now. Not uniform, but less of a hassle...

[fkiraly]

the yml now has conflicts with main, I think due to a dependabot update

[fkiraly]

accidentally added some tabs here, I think?

[fkiraly]

LGTM now - I think there is are a few stray whitespaces though, if you want to remove them. Does the code quality workflow not check the actions ymls?

[tingiskhan]

I'll take a look at the white spaces and fix the errors in the workflow

[fkiraly]

Wanted to merge but noticed that there are substantial changes in notebook 2 - is this an accident that should be reverted?

[tingiskhan]

Wanted to merge but noticed that there are substantial changes in notebook 2 - is this an accident that should be reverted?

The notebook fails stochastically due to yfinance not always being able to fetch the prices for the optimization universe for the latest date - it fails between runs both locally and in CI. We could either (i) do as I do and simply drop the partia NaN rows, (ii) forward fill the prices, or (iii) select the previous day (no guarantee for non-missing observations though).

But just to make sure, you are referring to the NaN-handling, right? Shouldn’t be any other differences (other than adding plotly to pip install)

[fkiraly]

I would suggest to revert changes in the notebook, since otherwise this PR would conflict with the more programmatic fix in #721

[tingiskhan]

I would suggest to revert changes in the notebook, since otherwise this PR would conflict with the more programmatic fix in #721

Ah, nice, had missed that - fixing!

[fkiraly]

Thanks!