String coordinates in netcdf datasets

[corentincarton]

earthkit-data/earthkit/data/readers/netcdf.py

Line 36 in fe88303

def as_level(self, level):

I'm trying to read a netcdf dataset with string coordinates. The netcdf field is a list of points, each point has an ID (for instance G0001) and a time series, so two dimensions: station and time.
I get the following error when trying to read with earthkit:

File /etc/ecmwf/nfs/dh1_home_a/macw/git/test/hat/venv/hat/lib/python3.10/site-packages/earthkit/data/readers/netcdf.py:37, in as_level(self, level)
     36 def as_level(self, level):
---> 37     n = float(level)
     38     if int(n) == n:
     39         return int(n)

ValueError: could not convert string to float: 'G0001'

Do we really need to force a casting to float for the not-recognised dimensions?

[sandorkertesz]

@corentincarton, thank you for reporting this issue. The netcdf parser should not crash like that so it has to be fixed. Please can you make the data file available?

As for the data, I have to admit I am not sure if we need to treat it as a field/fieldset.

[corentincarton]

Hi @sandorkertesz, sorry I should have added a small code to reproduce the error. Here it is:

import xarray as xr
ds = xr.DataArray([[1,2,3], [4,5,6]], coords={'ids': ["Index1", "Index2" ], 'x':[1,2,3]})
ds.to_netcdf("test.nc")
earthkit.data.from_source("file", "test.nc")

Not sure about the field or fieldset... something we need to discuss indeed.

[sandorkertesz]

Hi @corentincarton, the string coordinate issue can be fixed. However, at the moment when we create a field/fieldlist out of netcdf we assume it has a pair of horizontal coordinates (lat/lon or x/y). Without this only a NetCDFReader is created not a FieldList.
I will fix the string coordinate problem within this issue and open another one to discuss the netcdf field/fieldlist problem.

[corentincarton]

Thanks @sandorkertesz !