Remove Xarray Private API and remove pinning

[pilotchute]

Describe the bug
Xarray 2023.03.0 breaks tests due to a call on line 1818 of xarray/core/common.py

To Reproduce
run tests

Expected behavior
no error

Screenshots
Failing test run
OS:
Both MacOS and Ubuntu

Environment
n/a

[anissa111]

This seems relevant: pydata/xarray#7645

[dcherian]

xr.core.common.contains_cftime_datetimes(d_arr)

This is private API, you should vendor it though the current version isn't the most friendly for vendoring. So perhaps reimplement it.

Also I recommend updating your upstream-dev-ci (e.g. corteva/rioxarray#654)

[cyschneck]

Part of this issue appears in 02-more_advanced_functionality and prevents graphs from being generated

# Minimal Example:
from geocat.comp import climatology_average
import xarray as xr
import geocat.datafiles as gdf

nc_file = gdf.get('netcdf_files/atm.20C.hourly6-1990-1995-TS.nc')
ds = xr.open_dataset(nc_file)
ds = ds.isel(member_id=0)  # select one model from the ensemble

temp = ds.TS
monthly_temp = climatology_average(temp, freq='month')

Throws: AttributeError: 'DataArray' object has no attribute '_data'

The issue seems to come from contains_cftime_datetimes no longer accepts DataArrays. So now it expects xarray.Variable

For now, while running locally by changing xr.core.common.contains_cftime_datetimes(d_arr) to xr.core.common.contains_cftime_datetimes(d_arr.variable) fixes the error and can now generate graphs in the geocat-tutorials

Until there is a public API this can be change in /home/user/miniconda3/envs/geocat-tutorials/lib/python3.10/site-packages/geocat/comp/climatologies.py on conda to run the tutorial

def _contains_datetime_like_objects(d_arr):
    """Check if a variable contains datetime like objects (either
    np.datetime64, or cftime.datetime)"""
    return np.issubdtype(
        d_arr.dtype,
        np.datetime64) or xr.core.common.contains_cftime_datetimes(d_arr.variable)

[cyschneck]

Currently, an open issue on geocat-comp. Can get around the error by setting xarray<=2023.02.0

A more general solution

data = encode_cf_variable(out_data).values.astype(numpy_dtype)
with
data = encode_cf_variable(out_data.variable).values.astype(numpy_dtype)