In climate science, a very common task involves concatenating NetCDF files with identical variables, dimensions, and coordinates along a brand new "ensemble member" or "record" dimension. With the NetCDF Operators, this is accomplished using ncecat.

MCVE Code Sample

Currently, it seems the correct way to do this in xarray is with xarray.combine_nested as follows:

python import xarray as xr files = ['member1.nc', 'member2.nc', ...] ds = xr.open_mfdataset( files, combine='nested', concat_dim='record', )

Problem Description

While this works, there does not seem to be any mention of this use case in the combine_nested or open_mfdataset docs... and using combine='nested' to concatenate along a brand new dimension feels quite unintuitive to me.

It would be nice to have examples in combine_nested and/or open_mfdataset with this special usage or mention the possibility of creating brand new dimensions with concat_dim. For example:

python In [1]: import xarray as xr ...: datasets = [ ...: xr.Dataset({'temp': (('x', 'y'), np.random.rand(10, 20))}) ...: for i in range(3) ...: ] ...: xr.combine_nested(datasets, concat_dim='record') Out[1]: <xarray.Dataset> Dimensions: (record: 3, x: 10, y: 20) Dimensions without coordinates: record, x, y Data variables: temp (record, x, y) float64 0.32 0.4897 0.2659 ... 0.3485 0.0251 0.399

Output of xr.show_versions()

n/a

What happened: DataArray.interp strips the dimension coordinate attributes.

What you expected to happen: Preserved coordinate attributes.

Minimal Complete Verifiable Example:

Input:

```python import xarray as xr import numpy as np

data = xr.DataArray( np.random.rand(5), dims='x', coords={'x': ('x', np.arange(5), {'foo': 'bar'})} )

print(data.x.attrs) # initial attributes print(data.sel(x=2).x.attrs) # sel and isel preserve attributes print(data.interp(x=2.5).x.attrs) # interp does not preserve attributes with xr.set_options(keep_attrs=True): print(data.interp(x=2.5).x.attrs) # keep_attrs does nothing ```

Output:

python {'foo': 'bar'} {'foo': 'bar'} {} {}

Environment:

Xarray's HTML Dataset and DataArray reprs are evidently not compatible with "dark" jupyter notebook themes. They seem to work fine with the dark jupyter lab theme, and since jupyter lab is the way of the future perhaps this issue is obsolete, but thought I'd mention it.

The below example is from a jupyter notebook with the "onedork" dark theme from jupyter-themes. It results in black text against a dark background for the section headers (Coordinates, Dimensions, etc.) and DataArray data tables, and a light background for the coordinate and Dataset data tables.

```python

Dataset repr

import numpy as np import xarray as xr ds = xr.Dataset( { 'temp': (('x', 'y'), np.random.rand(10, 20), {'long_name': 'temperature', 'units': 'degrees_Celsius'}), 'x': ('x', np.arange(10)), 'y': ('y', np.arange(20)), }, attrs={'description': 'example dataset'} ) ds ```

```python

DataArray repr

ds.temp ```

Note that, by contrast, the text repr is dark theme friendly:

```python

Text repr

xr.set_options(display_style='text') ds ```

Versions