BUG: xarray data access

[jklenzing]

There is a possible bug with how we access xarray data found in #468. While that specific issue was solved by using xarray notation to bypass the index acces via pysat, the error analysis is documented here to help with future flake8 compliance pulls and potential bugs in other instruments. The issue arises when trying to set 2D data via indices, ie,

inst = pysat.Instrument(platform='icon', name='mighti', tag='vector_wind_green', clean_level='none')
inst.load(2020, 2)
idx, idy, = np.where(inst['Wind_Quality'] != 1)
inst[idx, idy, 'Zonal_Wind'] = np.nan

This can be traced to the except path here:

pysat/pysat/_instrument.py

Lines 565 to 569 in ec81344

	try:
	self.data[key[-1]].loc[indict] = in_data
	except:
	indict['time'] = self.index[indict['time']]
	self.data[key[-1]].loc[indict] = in_data

The error triggered at 568 is

IndexError: arrays used as indices must be of integer or boolean type

It's possible that the real error is masked by the bare except statement. If I remove the try/except statement to force the main line, I get the following error at 566

KeyError: "not all values found in index 'Altitude'"

Solution needed: sort out what error the except statement is supposed to catch, then debug from there.

[jklenzing]

Tracked down the source of the error. When passing >1D data, the except statement winds up trying to pass mixed indices (in the case above, a datetime index and integers). This needs to be updated using the latest xarray standards.

https://xarray.pydata.org/en/stable/indexing.html#

[jklenzing]

NOTE: this issue affects setitem and getitem in the Instrument class

[rstoneback]

Worth reviewing the whole setitem and getitem codes while working on this one.

[jklenzing]

Closing with merge of #647