fix multiIndex min max issue #2923

[arabidopsis]

• Fixes Calling max() on a stacked dimension raises exception #2923
• Test added

I ran pytest and it didn't seem to produce any problems.

[pep8speaks]

Hello @arabidopsis! Thanks for updating this PR. We checked the lines you've touched for PEP 8 issues, and found:

There are currently no PEP 8 issues detected in this Pull Request. Cheers! 🍻

Comment last updated at 2019-06-27 16:06:21 UTC

[arabidopsis]

Oops... I usually use black -- which I turned off since it would really have stuffed the code :)

[shoyer]

Awesome, thank you @arabidopsis!

If you don't mind, it would be great if you could add a brief note about the bug fix to whats-new.rst. (if not, don't worry, I'll merge this in a few days either way)

You can ignore the test failure on Python 3.5. Something else seems to be going wrong there.

[arabidopsis]

No problem. Thanks for xarray!

Just a question. I put a guard in to test just for tuples (because that's what was breaking for me) but it seems to me that _nan_minmax_object should just return utils.to_0d_object_array(data)
unconditionally at that point no? It's a max or a min right.

Wasn't sure.

[shoyer]

Just a question. I put a guard in to test just for tuples (because that's what was breaking for me) but it seems to me that _nan_minmax_object should just return utils.to_0d_object_array(data)
unconditionally at that point no? It's a max or a min right.

Wasn't sure.

Yes, it does seem like a good idea to just return utils.to_0d_object_array() unilaterally in this case.

[dcherian]

Thanks @arabidopsis. Welcome to xarray!