Several issues came up in xarray >= 2022.6 with integration. Particularly:
Casting energy to a Pandas data frame and dropping duplicates no longer works well to generate an integrator list, so I use np.unique instead.
More importantly, when the integrator inspects the xarray's indexes object, xarray >= 2022.6 patches in a fake index label for each multiindex level, which had to be undone for our auto-stacking to work.
Several issues came up in xarray >= 2022.6 with integration. Particularly:
Casting energy to a Pandas data frame and dropping duplicates no longer works well to generate an integrator list, so I use np.unique instead.
More importantly, when the integrator inspects the xarray's
indexes
object, xarray >= 2022.6 patches in a fake index label for each multiindex level, which had to be undone for our auto-stacking to work.