That test is passing on my end, I don't know what the problem is 🤔

Looks like the test just timed out. Let’s see if a re-spin helps…

Edit: apparently not.

• Nowvert=True won't emit a deprecation warning, regardless of how it's set. It also won't overrideorientation.

• vert=False will emit a deprecation warning and overrideorientation.

• Settingmpl.rcParams['boxplot.vertical'] to False will trigger an rcparam deprecation warning.

Thanks@saranti, this is looking great!

It took me a while to get my head around why we are not warning whenvert=True is explicitly passed. I guess it's because it's too complicated to do forbxp as we can't tell whether it came from the rcParam. However would it be worth adding an extra warning inboxplotspecifically for whenvert=True? Apologies if I've missed some discussion on this somewhere.

Good catch. Since we have the tri-state True/False/None, we can detect whenever True/False was passed explicitly. If we push rcParams resolution down tobxp() we can even detect that also forbxp().

If we push rcParams resolution down to bxp() we can even detect that also for bxp().

That would mean the rcParam affects cases where the user callsbxp directly, which it didn’t before. However I think it’s a little odd that it didn’t, so we could consider that a bugfix 🤔

Could we also set the rcParam in the default matplotlibrc toNone and then add it to this dictionary (highlighted by@QuLogic above)?

If I have understood, that would mean we get the warning if the rcParam is explicitly set to eitherTrue orFalse by the user.

If we push rcParams resolution down to bxp() we can even detect that also for bxp().

That would mean the rcParam affects cases where the user callsbxp directly, which it didn’t before. However I think it’s a little odd that it didn’t, so we could consider that a bugfix 🤔

Indeed. But it would only affectbxp() calls, wherevert is not passedand the rcParam is modified. I'd be willing to accept that change because being able to warn on all explicitbxp(..., vert=) is IMHO more important. And as you say, it can be considered a bugfix. And the rcParam gets deprecated, so the user has to adapt their code anyway.

Re rcParam None: I have not been able / spent enough time to fully understand the mechanics (see#28074 (comment)) but the solution there to warn on rcParam readout seems good enough. We only don't warn if you configure the rcParam explicitly to True (but why would one do that when it's the default behavior anyway).

I didn't see any reason why thevert = mpl.rcParams['boxplot.vertical'] needed to be inboxplot so I moved it tobxp. Now it also gives a warning tobxp calls where it previously didn't.
Explicitly settingvert=True will now also emit a deprecation warning.