Assigning an indexed coordinate to a data variable still uses an IndexVariable

[shoyer]

What is your issue?

Consider the following example:

In [1]: import xarray

In [2]: ds = xarray.Dataset(coords={'x': [1, 2, 3]})

In [3]: ds['y'] = ds['x']

In [4]: ds
Out[4]:
<xarray.Dataset> Size: 48B
Dimensions:  (x: 3)
Coordinates:
  * x        (x) int64 24B 1 2 3
Data variables:
    y        (x) int64 24B 1 2 3

In [5]: ds['y'].variable
Out[5]:
<xarray.IndexVariable 'x' (x: 3)> Size: 24B
array([1, 2, 3])

In [6]: ds['x'].variable is ds['y'].variable
Out[6]: False

The assigned variable 'y' still uses an IndexVariable under the hood, which means it has immutable data. This is not very useful.

Instead, assigning a DataArray that contains a wrapped IndexVariable to a Dataset should create a new varaible of the base Variable type.

[mjwillson]

This is a problem when assigning to non-index coordinates too. It can break some xarray code which expects to be able to assign to the .data of a non-index coordinate variable. See #10829.