failed to set key 'endStep' in to_grib

I read the temperature from GFS and tried to write the data to a new GRIB file .

from eccodes import *
import numpy as np
import xarray as xr 
from cfgrib.xarray_to_grib import to_grib

filename = "gfs.2022010500f24"
ds = xr.open_dataset(filename, engine='cfgrib',
                     backend_kwargs={'filter_by_keys': {'typeOfLevel': 'isobaricInhPa', 
                                                        'name'       : 'Temperature'}})

to_grib(ds, 'out1.grib', grib_keys={'edition': 2})

Running the script directly will result in an error

KeyError: 1641340800000000000

following #272 , it works with a warning :

failed to set key 'endStep' to numpy.timedelta64(24,'ns')

and the time information in the grib file is incorrect

1:0:d=2022010500:TMP Temperature [K]:1000 mb:anl:

which should be

542:117403445:d=2022010500:TMP Temperature [K]:1000 mb:24 hour fcst:

Environment: Python 3.8 from conda-forge on centos7

pandas 1.3.4 cfgrib 0.9.9.1 xarray 0.20.1

[jiaobf]

Also, make sure you’re using cfgrib from the master branch - this fix was important but is not yet in a release: 3cc01e3

It works!

[iainrussell]

Also, make sure you’re using cfgrib from the master branch - this fix was important but is not yet in a release: https://github.com/ecmwf/cfgrib/commit/3cc01e395cd66a376a279c0c603cbedcaeda7f50