-
Notifications
You must be signed in to change notification settings - Fork 413
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Reduce duplication in minimum dependencies #1842
Conversation
This was naming the wrong version for a release: "v1" instead of "v1.0".
conda-forge's packages are now automatically built and should stay that way.
Instead, use sphinx's literalinclude to directly pull the list of dependencies from `setup.cfg`.
Instead of having a separate file (that doesn't get updated by Dependabot), just read our minimum versions directly from setup.cfg and use this to generate a Minimum constraint file on CI. This also updates setup.cfg to reflect the current versions from Minimum.
A consideration during review: is it worth keeping a |
Yeah, I don't want the duplicated information. If you think it's worth it, I could extract the Python code that generates it into a script in |
Nah, the more I think about the point on testing minimums the less I think there's any serious use case for keeping that around on its own. |
Description Of Changes
As evidenced by #1839, we have a problem where we have duplication of information about the minimum supported versions of things. They currently exist in 3 places:
setup.cfg
ci/Minimum
-- so we can test with our minimum versionsThis PR changes things so that
setup.cfg
, which drives our PyPI package and gets (manually) synced to our conda-forge recipe, is the only source of this information. For CI, this deletes the static copy inci/Minimum
and instead CI parsessetup.cfg
and generates this file. For the install guide, this changes to use sphinx'sliteralinclude
directive to directly include the relevant portion ofsetup.cfg
. So that changes from:to
I don't love the new visual, but maybe at some point in the future we can address that with CSS. What I do love is that we only need to change one (sensibly canonical) spot to update our minimum supported versions of upstream dependencies, which is a good win for maintainability.
This PR also removes CI's hack for installing an older scipy for conda (and bumps us to scipy 1.6.3), which is no longer needed, as well as updates a few URLs that were redirecting.
Checklist