Command line¶

Installing fastgrib adds a fastgrib command. Without installing, use python -m fastgrib instead.

fastgrib FILE|URL                                  # the inventory, as wgrib2 -s
fastgrib FILE -match ':TMP:' -not ':0\.' -v        # with units, long names, valid times
fastgrib FILE -select name=TMP level=500 -stats    # min, max, mean, undefined points
fastgrib FILE -grid                                # the grids the fields lie on
fastgrib FILE -match ':HGT:[0-9]+ mb:' -o z.nc     # NetCDF (needs xarray)
fastgrib FILE -bin out.bin                         # float32, west-east and south-north
fastgrib FILE -csv out.csv                         # lon,lat,value for each point
fastgrib -warm                                     # compile every decoder now

Option	Meaning
`-match REGEX`	Fields whose inventory line matches, as `wgrib2 -match`
`-not REGEX`	Leave out fields that match
`-select key=value ...`	Pick by keyword, as `GribFile.select`
`-v`	Longer inventory lines: units, long names, valid times
`-stats`	Minimum, maximum, mean and number of undefined points
`-grid`	Describe the grids
`-o FILE.nc`	Write NetCDF through xarray
`-bin FILE`	Raw float32, undefined points 9.999e20, as `wgrib2 -order we:sn -no_header -bin`
`-csv FILE`	One `lon,lat,value` line per point of a field (pick one with `-match`)
`-gpu`	Decode on the GPU
`-time`	Report how long it took
`-warm`	Compile all decoders, for an install script or image build
`-version`	Print the version

Listing, matching and selecting never decode anything, so numba isn't even imported. An inventory of GFS 0.25° takes 0.27 s from start to finish, most of it starting Python and numpy.