You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I have been working on the zarr python support in kerchunk in #516, but i need to cycle off of it for a bit. So I am going to list out what is left to be done here and how it can be split off as best I can in hopes that someone else can help me to move this along in the meantime:
Heres where the functionality is at in that PR.
Generating references:
HDF
netcdf3
grib2
zarr (not tested, might work)
tiff (not tested, might work)
others (not sure)
Reading with xarray
Currently creating a store works, however there are caveats:
zarr 3 RemoteStore issues: zarr python 3 requires that an fsspec filesystem used for a RemoteStore is an AsyncFilesystem. ReferenceFilesystem supports this, however this also means that the remote filesystem within the ReferenceFilesystem must be async. This means that all usage where data files are on a non async filesystem will not work. This blocks most of the tests in kerchunk as it stands today because LocalFilesystem is not async
Codecs
Codecs (filters, compressors) are treated differently in zarr 3. zarr 2 stores read with zarr 3 will still use numcodecs codecs, however if users want to ever use these codecs with another store (say to read a grib virtual dataset with Icechunk) the codecs from this package need to conform to zarr's Codec abc
Codecs support zarr 3 Codec API
I think that this is mostly it, and im happy to help anyone who is interested in helping drive this forward
Zarr-Python 3.0 is getting close to a full release. This issue tracks the integration of the 3.0 release with Kerchunk.
Here's a running list of issues we expect we need to solve either here or upstream:
Eventually, we may also want to:
xref: #504
The text was updated successfully, but these errors were encountered: