Update ahead of change of 'item in ds' behavior in xarray v0.11

[spencerahill]

Just updated to xarray v0.10 and started getting a lot of deprecation warnings when running aospy due to pydata/xarray#1645, c.f. pydata/xarray#1267. So I tried to find it everywhere this comes up and fix it.

@spencerkclark let me know if you see any reason why we shouldn't do this, or if I missed anything.

[spencerkclark]

@spencerahill thanks for making these updates. It's clear we will need to be careful about this update to xarray.

In these places it seems we were relying on __contains__ working differently for DataArrays and Datasets.

• For DataArrays we were checking if a name was a key in the dictionary of coordinates.
• For Datasets we were checking if a name was a key in either the dictionary of coordinates or data variables.

Do you know where exactly the warnings were emanating from? The warnings should only stem from when we use the __contains__ operator on a DataArray. If some of these methods are called using Datasets we probably want to check if the name is in the data variables dictionary as well (and therefore should use the original VAR_STR in ds logic).

My initial thought was that since _prep_time_data is called before _sel_var in data_loader.load_variable:

aospy/aospy/data_loader.py

Lines 278 to 280 in 389961a

	ds, min_year, max_year = _prep_time_data(ds)
	ds = set_grid_attrs_as_coords(ds)
	da = _sel_var(ds, var, self.upcast_float32)

ds must always be a Dataset, but it seems if this warning was triggered, it must sometimes be a DataArray. Did you encounter these warnings when doing calculations using aospy or when running the test suite?

[spencerahill]

Oy, this is an embarrassing one. I forgot that I was running on an old branch for one of my research projects. I just ran the example main script and the test suite on develop and no warnings are emitted.

Thanks @spencerkclark for responding, sorry for wasting your time!