gribjump/eccodes as a virtual zarr reader?

[TomNicholas]

Is your feature request related to a problem? Please describe.

I built VirtualiZarr to provide zarr-compliant access to archival file formats such as netCDF without copying the data. but we are still in need of a really good and flexible GRIB reader. (see zarr-developers/VirtualiZarr#312)

Describe the solution you'd like

Some organisation that has a lot of GRIB data expertise build and maintain a virtual zarr reader for GRIB files. This could be very simple, and I suspect that your C++ library can already parse all the information needed.

Describe alternatives you've considered

Everyone continues to make their own stack instead of recognising that they all have the same problem and can solve it with common tools...

Additional context

A virtual zarr reader would allow storing references to GRIB files in Icechunk, which allows for much more efficient access using Zarr. See this blog post on doing the same to access netCDF files via zarr.

Organisation

Earthmover, but also the Pangeo community.

[ChrisspyB]

Hi Tom, your suggestion sounds somewhat similar to the prototype being developed in https://github.com/ecmwf/zfdb, where they are building a zarr interface which uses gribjump and eccodes to access GRIB files archived in FDB (https://github.com/ecmwf/fdb). However, I suspect it is too FDB-specific to match your requirements.

@Ozaq @tbkr I wonder if this issue interests you? Is it within the scope of what you are currently working on?