open_dataset crashes for coordinate grids with a single node

Hello cfgrib gurus,

Experienced python user here, but complete novice to cfgrib and the GRIB format itself. I am trying to look at an ERA5 GRIB file in which the coordinate grid contains a unique latitude/longitude entry (i.e. looking at the ERA5 data for a single point on Earth).

Calling xr.open_dataset() gets me the following error:

~/anaconda3/lib/python3.7/site-packages/cfgrib/dataset.py in build_geography_coordinates(index, encode_cf, errors, log)
    366             dimensions=('latitude',), data=latitudes, attributes=COORD_ATTRS['latitude'].copy()
    367         )
--> 368         if latitudes[0] > latitudes[-1]:
    369             geo_coord_vars['latitude'].attributes['stored_direction'] = 'decreasing'
    370         geo_coord_vars['longitude'] = Variable(

IndexError: too many indices for array

I went in cfgrib/dataset.py to see if I might bypass this using an easy-but-far-from-perfect fix, doing:

if latitudes.ndim > 0: # <= this just by-passes the check
    if latitudes[0] > latitudes[-1]:
        geo_coord_vars['latitude'].attributes['stored_direction'] = 'decreasing'

However, this only displaces the problem further down the line, as I am then faced with the following error:

[...]
~/anaconda3/lib/python3.7/site-packages/xarray/backends/cfgrib_.py in open_store_variable(self, name, var)
     53         encoding["original_shape"] = var.data.shape
     54 
---> 55         return Variable(var.dimensions, data, var.attributes, encoding)
     56 
     57     def get_variables(self):

~/anaconda3/lib/python3.7/site-packages/xarray/core/variable.py in __init__(self, dims, data, attrs, encoding, fastpath)
    302         """
    303         self._data = as_compatible_data(data, fastpath=fastpath)
--> 304         self._dims = self._parse_dimensions(dims)
    305         self._attrs = None
    306         self._encoding = None

~/anaconda3/lib/python3.7/site-packages/xarray/core/variable.py in _parse_dimensions(self, dims)
    498             raise ValueError(
    499                 "dimensions %s must have the same length as the "
--> 500                 "number of data dimensions, ndim=%s" % (dims, self.ndim)
    501             )
    502         return dims

ValueError: dimensions ('latitude',) must have the same length as the number of data dimensions, ndim=0

To me, it looks like the “real” issue is that for coordinate grids with single nodes, the latitudes and longitudes values are stored in 0-dimensional arrays ?

Running Mac OS 10.13.6 with Python 3.7.3, cfgrib 0.9.8.1, and xarray 0.15.1 and pyeccodes (in place of the ecCodes library). For completeness:

$ python -m cfgrib selfcheck
Found: ecCodes v2.17.0.
Your system is ready.

Here is a demo ERA5 GRIB file that allows to reproduce the problem. ERA5_single-grid-point_test.grib.zip

[fpavogt]

@alexamici: great ! Thanks a lot for the fix @iainrussell !

[iainrussell]

Looks like this is what I fixed for #199.