Miguel Jimenez

@MaceKuailv Is the issue with NaN's solvable by replacing the NaNs with a number (say -10000)? This is how some models deal with NaNs.

Will keep this open, as pining `xgcm` is only a quick fix.

The new version of `xarray` also contains some internal refactoring changes that changes some aspects of indexing. This new version of xarray was released last week, and so it didn't...

Makes sense, as there is some interpolation happening there.

Thanks for looking into this. It is re assuring that we don't need to make changes regarding `xgcm`. It wasn't clear to me whether that was true or not back...

I'll look into this later on Monday. Sent from my T-Mobile 4G LTE Device Get Outlook for Android ________________________________ From: ThomasHaine ***@***.***> Sent: Saturday, October 22, 2022 5:12:35 PM To:...

PR #270 unpins xarray and all tests pass

This is true! Nice catch @asiddi24 . The question now is if this issue is propagated to after transforming the dataset (after the cutout). Have you taken a look? This...

> This could only occur when there is a discontinuity of longitude in the cutout region. We can define a new x scale as "how much east is a point...

> I am testing it on ecco, there is a different issue coming up. Somehow HFacS and HFacW after the cutout become an array with both X and Xp1 in...