Add support for nump types in sel()

[sandorkertesz]

Is your feature request related to a problem? Please describe.

At the moment sel() does not accept numpy types. So while this works:

ds.sel(param="t")

this call gives empty result:

ds.sel(param=np._str("t"))

Describe the solution you'd like

Allow using numpy types in sel.

Describe alternatives you've considered

No response

Additional context

No response

Organisation

ECMWF

[alioacar]

Hi, I believe the issue lies within the Selection class in index.py. The item and shape methods can be utilized to determine whether the variable is a np scalar or an array.

if hasattr(v, 'item') and callable(v.item):
    if not hasattr(v, 'shape') or not v.shape:
        v = v.item()
    else:
        v = v.tolist()

If it is a np array, we check each element to support scenarios where the array may contain mixed types. (Not sure if this is the best way)

v = [item.item() if hasattr(item, 'item') and
      callable(item.item) and (not hasattr(item, 'shape')
      or not item.shape) else item for item in v]

This logic accommodates all the following combinations

np.str_('t')
't'
['t','2t']
np.array((np.str('2t'),'t'))

Additionally, we need to pass the assertion for np arrays in normalize_selection from select.py