You will need to fix failing tests. Basically, it should be docstring test in most cases.
We also need an additional non-regression test: I think that we should be a PR curve and check that the last point is equivalent to a decision rule that always predicts the positive class.

@glemaitre I added 3 commits:

• fixing existing unit tests
• fixing existing doc tests
• adding regression test (I check that it is failing if I revert the change)

Please add an entry to the change log at doc/whats_new/v1.1.rst. Like the other entries there, please reference this pull request with :pr: and credit yourself (and other contributors if applicable) with :user:.

It seems the doc is failing, but I cannot understand why:

Sphinx Warnings in affected files
doc/whats_new/v1.1.rst:105: WARNING: Unexpected indentation.

nothing wrong at line 105.

For sure this is not linked with your change. We can ignore it.

I merge main into your branch. The problem was solved in main by the following PR: #23246

It went very smoothly, I'm happy I did this first contribution, thank you @glemaitre.
I will most probably continue to contribute by adding uncertainty band on precision-recall curves soon.

@stephanecollot I started a POC on the subject: #21211

The idea is to use cross-validation to get uncertainty bounds. I will probably find time to carry on some work in the next release. However, we will need to breakdown the PR into smaller PR to facilitate the review process:

1. Add the return_indices parameter to cross_validate
2. Then add a new from_cv_results class method for each display