Inconsistent behavior when using xr.concat() with pd.Index vs xr.DataArray as dim parameter for a new dimension

[raj-magesh]

summary

Using an xr.DataArray as the dim parameter to xr.concat does not name the new dimension with the name of the given DataArray (a generic dim_0 is used instead), though the coordinate information is captured correctly. This unexpected behavior does not happen when using a pd.Index as the dim parameter.

minimal reproducible example

import numpy as np
import pandas as pd
import xarray as xr

xarr1 = xr.DataArray(np.ones((2, 3)), dims=('existing_dim_0', 'existing_dim_1'))
xarr2 = xr.DataArray(np.ones((2, 3)), dims=('existing_dim_0', 'existing_dim_1'))

new_coord = ['xarr1', 'xarr2']

concatenating using pd.Index as dim parameter

xr.concat(
    [xarr1, xarr2],
    dim=pd.Index(new_coord, name='new_dim'),
)

Output: as expected, the new dimension has the name new_dim and the specified coordinate information

<xarray.DataArray (new_dim: 2, existing_dim_0: 2, existing_dim_1: 3)>
array([[[1., 1., 1.],
        [1., 1., 1.]],

       [[1., 1., 1.],
        [1., 1., 1.]]])
Coordinates:
  * new_dim  (new_dim) object 'xarr1' 'xarr2'
Dimensions without coordinates: existing_dim_0, existing_dim_1

concatenating using xr.DataArray as dim parameter

xr.concat(
    [xarr1, xarr2],
    dim=xr.DataArray(new_coord, name='new_dim'),
)

Output: inconsistent with the pd.Index method, the new dimension does NOT have the specified name new_dim, though it has the coordinate information

<xarray.DataArray (dim_0: 2, existing_dim_0: 2, existing_dim_1: 3)>
array([[[1., 1., 1.],
        [1., 1., 1.]],

       [[1., 1., 1.],
        [1., 1., 1.]]])
Coordinates:
    new_dim  (dim_0) <U5 'xarr1' 'xarr2'
Dimensions without coordinates: dim_0, existing_dim_0, existing_dim_1

Environment:

Output of xr.show_versions()

INSTALLED VERSIONS

commit: None
python: 3.9.5 | packaged by conda-forge | (default, Jun 19 2021, 00:32:32)
[GCC 9.3.0]
python-bits: 64
OS: Linux
OS-release: 3.10.0-1160.31.1.el7.x86_64
machine: x86_64
processor: x86_64
byteorder: little
LC_ALL: None
LANG: en_US.UTF-8
LOCALE: ('en_US', 'UTF-8')
libhdf5: 1.12.0
libnetcdf: 4.7.4

xarray: 0.20.1
pandas: 1.3.1
numpy: 1.21.1
scipy: 1.7.0
netCDF4: 1.5.7
pydap: None
h5netcdf: None
h5py: 3.3.0
Nio: None
zarr: None
cftime: 1.5.0
nc_time_axis: None
PseudoNetCDF: None
rasterio: None
cfgrib: None
iris: None
bottleneck: None
dask: 2021.09.0
distributed: None
matplotlib: 3.4.2
cartopy: None
seaborn: 0.11.1
numbagg: None
fsspec: 2021.08.1
cupy: None
pint: None
sparse: None
setuptools: 49.6.0.post20210108
pip: 21.3.1
conda: None
pytest: None
IPython: 7.24.1
sphinx: None

[max-sixty]

Thanks @raj-magesh, you're correct. We would definitely take a pull request to fix this.

[dcherian]

This looks like a duplicate of #1646

The design question is that we have two names from which to pick one: name of dataarray, and name of dimension in the DataArray.

[max-sixty]

Thanks @dcherian . Closing as duplicate