issues

Expanding the dimensions of a variable with a dimension having the same name as a coordinate variable will also change the attributes (including encoding attributes) of this coordinate variable!

The following examples prints the attrs and encoding of coordinate variable time before/after expanding the dimensions of the variable sea_surface_temperature

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

grid = xr.Dataset({ 'time': (('time',), [np.datetime64('2018-01-01')]), 'lat': (('lat'), np.arange(-80, 80, 1)), 'lon': (('lon',), np.arange(-180, 180, 1)), 'sea_surface_temperature': (('lat', 'lon',), np.ones(shape=(160, 360))) })

grid['time'].encoding['units'] = 'seconds since 1981-01-01 00:00:00' grid['time'].attrs['comment'] = 'some comment'

print(grid['time'].encoding, grid['time'].attrs) grid['sea_surface_temperature'] = grid['sea_surface_temperature'].expand_dims('time', axis=0) print(grid['time'].encoding, grid['time'].attrs)

``` The current output is:

{'units': 'seconds since 1981-01-01 00:00:00'} {'comment': 'some comment'} {} {}

Expected Output

I would not expect any change to other coordinates or variables than the expanded one (here sea_surface_temperature)

{'units': 'seconds since 1981-01-01 00:00:00'} {'comment': 'some comment'} {'units': 'seconds since 1981-01-01 00:00:00'} {'comment': 'some comment'}

Output of xr.show_versions()

Using open_mfdataset on a series of netcdf files having variable attributes with type list will fail with the following exception, when these attributes have different values from one file to another:

ValueError: The truth value of an array with more than one element is ambiguous. Use a.any() or a.all()

An example of such variable is provided below:

double sea_ice_fraction(time) ; sea_ice_fraction:least_significant_digit = 2LL ; sea_ice_fraction:_FillValue = 1.e+20 ; sea_ice_fraction:long_name = "sea ice fraction" ; sea_ice_fraction:standard_name = "sea_ice_fraction" ; sea_ice_fraction:authority = "CF 1.7" ; sea_ice_fraction:units = "1" ; sea_ice_fraction:coverage_content_type = "auxiliaryInformation" ; sea_ice_fraction:coordinates = "time lon lat" ; sea_ice_fraction:source = "CCI Sea Ice" ; sea_ice_fraction:institution = "ESA" ; string sea_ice_fraction:source_files = "ice_conc_nh_ease2-250_cdr-v2p0_199912011200.nc", "ice_conc_sh_ease2-250_cdr-v2p0_199912011200.nc" ;

The exception will occur when the source_files attribute have a different values in the file time series I am trying to concatenate. I had to use the preprocess argument to remove first this attribute to avoid this exception.

This is caused by the equivalent method in xarray/core/utils.py that does not account for this case:

python def equivalent(first, second): """Compare two objects for equivalence (identity or equality), using array_equiv if either object is an ndarray """ # TODO: refactor to avoid circular import from . import duck_array_ops if isinstance(first, np.ndarray) or isinstance(second, np.ndarray): return duck_array_ops.array_equiv(first, second) else: return ((first is second) or (first == second) or (pd.isnull(first) and pd.isnull(second)))