Adds an IcechunkParser for reading existing icechunk repositories as virtual datasets without copying data, chunk-aligned indexing on ManifestArray (so xarray.Dataset.isel works end-to-end on virtual datasets), and limited sub-chunk slicing for uncompressed arrays.
What's Changed
- docs: add guidance on fill value + scale/offset to custom parsers by @maxrjones in #974
- Chunk-aligned indexing on ManifestArray by @TomNicholas in #994
- ci: decouple upstream dev-version testing from the pixi workspace by @TomNicholas in #997
- Sub-chunk slicing for uncompressed ManifestArrays (part of #86) by @TomNicholas in #996
- fix: also require kerchunk for test_read_netcdf3 by @TomNicholas in #998
- feat(parsers): add IcechunkParser for converting icechunk sessions to ManifestStores by @TomNicholas in #991
- docs(parsers): align all parser docstrings with IcechunkParser by @TomNicholas in #999
- Release notes for v2.6.2 by @TomNicholas in #1000
Full Changelog: v2.6.1...v2.6.2
Contributors
maxrjones and TomNicholas