Key error in to_netcdf

[jsadler2]

MCVE Code Sample

import xarray as xr

p = range(10)
ds = xr.Dataset({'precip': (['time'], p)}, coords={'time': range(10)})
ds.attrs['Conventions'] = ['COARDS', 'GrADS']
ds.to_netcdf('a.nc')

Expected Output

Saves data to a.nc without error

Problem Description

In version 0.12.1 this works without error. In 0.12.2 it throws an error:

  File "xarray_error.py", line 6, in <module>
    ds.to_netcdf('a.nc')
  File "C:\Users\jsadler\AppData\Local\Continuum\anaconda3\envs\nwm\lib\site-packages\xarray\core\dataset.py", line 1365, in to_netcdf
    compute=compute)
  File "C:\Users\jsadler\AppData\Local\Continuum\anaconda3\envs\nwm\lib\site-packages\xarray\backends\api.py", line 900, in to_netcdf
    store.close()
  File "C:\Users\jsadler\AppData\Local\Continuum\anaconda3\envs\nwm\lib\site-packages\xarray\backends\scipy_.py", line 226, in close
    self._manager.close()
  File "C:\Users\jsadler\AppData\Local\Continuum\anaconda3\envs\nwm\lib\site-packages\xarray\backends\file_manager.py", line 188, in close
    file.close()
  File "C:\Users\jsadler\AppData\Local\Continuum\anaconda3\envs\nwm\lib\site-packages\scipy\io\netcdf.py", line 299, in close
    self.flush()
  File "C:\Users\jsadler\AppData\Local\Continuum\anaconda3\envs\nwm\lib\site-packages\scipy\io\netcdf.py", line 409, in flush
    self._write()
  File "C:\Users\jsadler\AppData\Local\Continuum\anaconda3\envs\nwm\lib\site-packages\scipy\io\netcdf.py", line 420, in _write
    self._write_gatt_array()
  File "C:\Users\jsadler\AppData\Local\Continuum\anaconda3\envs\nwm\lib\site-packages\scipy\io\netcdf.py", line 442, in _write_gatt_array
    self._write_att_array(self._attributes)
  File "C:\Users\jsadler\AppData\Local\Continuum\anaconda3\envs\nwm\lib\site-packages\scipy\io\netcdf.py", line 450, in _write_att_array
    self._write_att_values(values)
  File "C:\Users\jsadler\AppData\Local\Continuum\anaconda3\envs\nwm\lib\site-packages\scipy\io\netcdf.py", line 562, in _write_att_values
    nc_type = REVERSE[values.dtype.char, values.dtype.itemsize]
KeyError: ('U', 24)

If instead the attribute is saved as as string instead of a list, it works:

import xarray as xr

p = range(10)
ds = xr.Dataset({'precip': (['time'], p)}, coords={'time': range(10)})
ds.attrs['Conventions'] = "['COARDS', 'GrADS']"
ds.to_netcdf('a.nc')

Output of xr.show_versions()

INSTALLED VERSIONS ------------------ commit: None python: 3.7.4 (default, Aug 9 2019, 18:34:13) [MSC v.1915 64 bit (AMD64)] python-bits: 64 OS: Windows OS-release: 10 machine: AMD64 processor: Intel64 Family 6 Model 158 Stepping 10, GenuineIntel byteorder: little LC_ALL: None LANG: en_US.UTF-8 LOCALE: None.None libhdf5: None libnetcdf: None

xarray: 0.12.2
pandas: 0.25.1
numpy: 1.17.1
scipy: 1.3.1
netCDF4: None
pydap: installed
h5netcdf: None
h5py: None
Nio: None
zarr: 2.3.2
cftime: None
nc_time_axis: None
PseudonetCDF: None
rasterio: None
cfgrib: None
iris: None
bottleneck: None
dask: 2.3.0
distributed: 2.5.1
matplotlib: 3.1.1
cartopy: None
seaborn: None
numbagg: None
setuptools: 41.2.0
pip: 19.2.3
conda: None
pytest: 5.1.2
IPython: 7.8.0
sphinx: None

[jhamman]

I amble to reproduce this using the scipy backend using xarray=0.13. I don't recall all of the specifics but I know netCDF3/scipy had a limited support for some attribute types so this may be to blame here. If you can use the netCDF4 library, this doesn't seem to be an issue.

[jsadler2]

Thanks for the workaround, @jhamman. After I installing the netcdf4 library it runs fine.

[dnowacki-usgs]

The NC_STRING type enables arrays of strings to be serialized as attributes in nc files. NC_STRING is not in netcdf3, only netcdf4, so @jhamman is right. xarray has supported NC_STRING in conjunction with netcdf4 since 0.10.4. Not sure how it wasn't throwing an error for @jsadler2 on 0.12.1 sans netcdf4 🤷‍♂.

[kmuehlbauer]

Is this still relevant? It looks like it works with netcdf4-backend and should not work with scipy backend, which is the case with latest xarray, netCDF4 and scipy.

[jhamman]

Agreed. I think we can close this now.