MNT Update main-ci lock files#29388
Conversation
|
Oh well, so it does look like the doc build segmentation fault is a real thing 😱 ... |
|
Oh well polars, what else 😅? pola-rs/polars#17380 Seems an issue with polars and numpy 2. Edit: this is a known polars issue pola-rs/polars#16998 which is waiting for rust-numpy to support numpy 2: PyO3/rust-numpy#429 |
…nto update-main-lock-files
bff4879 to
9d5e610
Compare
|
@adrinjalali maybe this is one you could look at, being on triage duties? I tried to sum up the main differences in the top post to facilitate review. The rest of it is lock-file diff, personally I quickly glance over it making sure the changes are mostly changed versions. |
adrinjalali
left a comment
There was a problem hiding this comment.
Otherwise LGTM. (And I think @glemaitre and I are switching triage duty since he's on vacation next week)
| # - bump minimum matplotlib supported versions to 3.9 at one point | ||
| # - complicate the example code to do the right thing depending on | ||
| # maplotlib version | ||
| "matplotlib": "<3.9", |
There was a problem hiding this comment.
what about fixing the code, so that our code actually runs warning free for people with an up to date env, and ignore those warnings in the min-version of the CI?
There was a problem hiding this comment.
And I think @glemaitre and I are switching triage duty since he's on vacation next week
Right, I likely missed the update 😉
what about fixing the code, so that our code actually runs warning free for people with an up to date env, and ignore those warnings in the min-version of the CI?
Fixing the code is the "complicate the example code" option (if matplotlib recent enough use the new way else use the old way) I tried to describe in the comment, maybe let's leave this for a separate PR?
For completeness, this is a DeprecationWarning in matplotlib 3.9:
matplotlib._api.deprecation.MatplotlibDeprecationWarning: The 'labels' parameter of boxplot() has been renamed 'tick_labels' since Matplotlib 3.9; support for the old name will be dropped in 3.11.
3 examples fail (DeprecationWarning turned into error):
- examples/inspection/plot_permutation_importance_multicollinear.py
- examples/ensemble/plot_gradient_boosting_regression.py failed
- examples/release_highlights/plot_release_highlights_0_22_0.py
See build log from one of the commit in this PR.
There was a problem hiding this comment.
yeah I saw the same thing in another library and there we just did the "complicated" thing by having something in fixes.py.
I don't mind having it for a separate PR, but we always have the danger of forgetting to unpin things for a while.
There was a problem hiding this comment.
I opened #29434 to not forget. In practice, one day or another I would have bumped into this TODO and be annoyed enough to do something about it 😉.
Take over and close #29276
Summary of the fixes:
matplotlib<3.9to avoid DeprecationWarning, see https://github.com/scikit-learn/scikit-learn/pull/29388/files#diff-ee7fc86c91edbd3dc5c55ac1653f23bb5d169fdcd3439da619d357bf4e42d6e4conftest.py: https://github.com/scikit-learn/scikit-learn/pull/29388/files#diff-3cdc9913d34276436e69cd506f11cdc77ef286069bd5465aaeeb73ac2bdee318pinarray-api-strict<2issue may be fixed by Quick fix for new version of array_api_strict #29387doc build: pinactually polars 1.1.0 released 7 July is compatible with numpy 2numpy<2to avoid polars segmentation fault may be fixed in conda-forge Release 1.0.0 conda-forge/polars-feedstock#240