interpolate_na drops DataArray attributes

[RobWatersMet]

When performing interpolate_na on a DataArray the attributes are removed. Unless I am doing something incorrect.

MCVE Code Sample

print(dhi_da)
dhi_ds = dhi_ds.interpolate_na(dim='time')
print(dhi_da)

Output first print statement

<xarray.DataArray 'dhi' (south_north: 90, west_east: 90, time: 105408)>
dask.array<rechunk-merge, shape=(90, 90, 105408), dtype=float64, chunksize=(90, 90, 105408), chunktype=numpy.ndarray>
Coordinates:
    lat      (south_north, west_east) float32 dask.array<chunksize=(90, 90), meta=np.ndarray>
    lon      (south_north, west_east) float32 dask.array<chunksize=(90, 90), meta=np.ndarray>
  * time     (time) datetime64[ns] 2016-12-31 ... 2019-01-01T23:50:00.000000064
Dimensions without coordinates: south_north, west_east
Attributes:
    long_name:        Diffuse Horizontal Irradiance
    standard_name:    dhi
    units:            W/m2
    dhi_calc_method:  pvlib.irradiance.erbs

Output secondprint statement

<xarray.DataArray 'dhi' (south_north: 90, west_east: 90, time: 105408)>
dask.array<vectorize_func_interpolate_na, shape=(90, 90, 105408), dtype=float64, chunksize=(90, 90, 105408), chunktype=numpy.ndarray>
Coordinates:
    lat      (south_north, west_east) float32 dask.array<chunksize=(90, 90), meta=np.ndarray>
    lon      (south_north, west_east) float32 dask.array<chunksize=(90, 90), meta=np.ndarray>
  * time     (time) datetime64[ns] 2016-12-31 ... 2019-01-01T23:50:00.000000064
Dimensions without coordinates: south_north, west_east

Problem Description

I do not believe attributes should ever be dropped like this as they typically contain persistent information.

Versions

Output of `xr.show_versions()`

INSTALLED VERSIONS

commit: None
python: 3.7.6 | packaged by conda-forge | (default, Mar 23 2020, 23:03:20)
[GCC 7.3.0]
python-bits: 64
OS: Linux
OS-release: 3.10.0-862.14.4.el7.x86_64
machine: x86_64
processor:
byteorder: little
LC_ALL: C.UTF-8
LANG: C.UTF-8
LOCALE: en_US.UTF-8
libhdf5: 1.10.5
libnetcdf: 4.7.4

xarray: 0.15.1
pandas: 1.0.3
numpy: 1.18.1
scipy: 1.4.1
netCDF4: 1.5.3
pydap: None
h5netcdf: None
h5py: None
Nio: None
zarr: None
cftime: 1.1.1.2
nc_time_axis: None
PseudoNetCDF: None
rasterio: None
cfgrib: None
iris: None
bottleneck: None
dask: 2.14.0
distributed: 2.14.0
matplotlib: None
cartopy: None
seaborn: None
numbagg: None
setuptools: 46.1.3.post20200325
pip: 20.0.2
conda: 4.8.3
pytest: None
IPython: None
sphinx: None

[TomNicholas]

Hi Rob,

Thanks for flagging this. It was a bug, which I've now fixed in a PR.

My fix adds a keep_attrs argument to interpolate_na in order to be consistent with the rest of the API. That means to preserve the attrs you'll either have to pass keep_attrs=True explicitly, or set the global option with xr.set_options(keep_attrs=True). I would recommend doing the latter, and then please raise another issue if you spot any other methods which don't preserve attrs.

I do not believe attributes should ever be dropped like this as they typically contain persistent information.

I basically agree, but we're discussing the details of whether or not to change to this behaviour as the default over in #3891 .

[RobWatersMet]

Hi Tom,

thanks for the information and timely response.

Cheers,

Rob