issue_comments

If we changed isel() to only modify data variables, then we would be in trouble with something like isel(x=slice(3)). The coordinate system would be inconsistent if we slice only the data variables but not the coordinates.

There's a bit of a conflict here between two desirable properties:

1. Methods on a Dataset only modify data variables, leaving coordinates unchanged
2. Methods on a Dataset keep the entire coordinate system for the Dataset consistent, including coordinates

What I think would be more useful:

<xarray.DataArray (x: 5)> array([0., 0., 0., 0., 0.]) Coordinates: y int64 42 e (y) int64 10 <---- Note the (y) Dimensions without coordinates: x

Thanks for clarifying!

One problem with this -- at least for now -- is that xarray currently doesn't allow coordinates on DataArray objects to have dimensions that don't appear on the DataArray itself.

It might also be surprising that this would make squeeze('y') inconsistent with isel(y=0)

Hi @chrisroat, thanks for the clear bug report!

It indeed be nice if squeeze followed by expand_dims preserved the original inputs, but I don't think that is possible in general -- the squeeze operation removes information.

For example, this array does currently satisfy your desired property, but wouldn't if we made the change you request: python arr1 = xr.DataArray(np.zeros((1,5)), dims=['y', 'x'], coords={'e': 10}) arr2 = arr1.squeeze('y').expand_dims('y') xr.testing.assert_identical(arr1, arr2) # passes

I suspect our best option for achieving this behavior would be to add another optional argument to expand_dims, e.g., perhaps - expand_dims(..., expand_coords=False): don't expand coordinates (default behavior) - expand_dims(..., expand_coords=True): expand all coordinates - expand_dims(..., expand_coords=['e']): only expand the coordinate 'e'