@qinhanmin2014, got it. Sorry for overlooking that. What version should the FutureWarning list for the change to take place?

Would it make sense to change the docstring to at least match the current default, since the two are still in conflict?

What version should the FutureWarning list for the change to take place?

0.23 (next release is 0.21)

Would it make sense to change the docstring to at least match the current default, since the two are still in conflict?

We usually add some notes in the docstring, e.g., see https://scikit-learn.org/dev/modules/generated/sklearn.metrics.auc.html.

Alright, so leave line 2492 unchanged, and just add a note that the default value stated there is incorrect and will be fixed in 0.23?

Alright, so leave line 2492 unchanged, and just add a note that the default value stated there is incorrect and will be fixed in 0.23?

Please refer to https://scikit-learn.org/dev/developers/contributing.html#change-the-default-value-of-a-parameter

Ah, of course. I feel quite ridiculous. Thanks for pointing me in the right direction again

Thanks for raising this! I agree it's an unnecessary surprise.

Thanks, @qinhanmin2014 and @jnothman! Looks like a few of the checks are failing because the FutureWarning just added is being raised.

So just to make sure I'm following the guidelines, I should just wrap the failing tests with @pytest.mark.filterwarnings(...), correct?

So just to make sure I'm following the guidelines, I should just wrap the failing tests with @pytest.mark.filterwarnings(...), correct?

Yes, please. Maybe add some comments, e.g., @pytest.mark.filterwarnings(...) # 0.23

The deprecation message seems strange, please try to follow the contributing guide. (e.g., The default value of XXX will change from XXX to XXX in XXX in order to.../because...)

@qinhanmin2014, regarding 08faa3c, is this the proper way to deal with deprecations in the docstring examples? Or should I use warnings.filterwarnings?

Thanks @HunterMcGushion!