GRIB

The gribberish library (a Rust-backed GRIB reader) provides a GribberishParser for VirtualiZarr. Both GRIB1 and GRIB2 are supported. Each GRIB message becomes a single chunk, and reading a chunk back decodes the referenced message bytes through gribberish's registered zarr codec.

Packing templates

gribberish decodes GRIB2 grid-point fields (including complex/second-order packing) and GRIB1 simple packing. GRIB1 messages using complex or spherical-harmonic packing are not yet supported and raise an error on read. See the gribberish repository for the current coverage.

Install it alongside VirtualiZarr with:

pip install "virtualizarr[grib]"

Then point virtualizarr.open_virtual_dataset (or virtualizarr.open_virtual_datatree, for files whose messages don't fit a single hypercube) at a GRIB file:

from obspec_utils.registry import ObjectStoreRegistry
from obstore.store import LocalStore

from gribberish.virtualizarr import GribberishParser
from virtualizarr import open_virtual_dataset

registry = ObjectStoreRegistry({"file://": LocalStore()})
parser = GribberishParser()
vds = open_virtual_dataset(
    url="file:///path/to/file.grib2", parser=parser, registry=registry
)

See the gribberish documentation for the full parser API and its filtering options (only_variables, drop_variables, etc.).