Zarr.jl

Appearance

Changelog ​

Unreleased ​

v0.10.0 - 2026-04-24 ​

  • Enable sharding_indexed codec for Zarr v3 #241

    • outer chunks (shards) are now split into inner chunks with a byte-range index, enabling efficient partial reads of large arrays

    • Python zarr interoperability tests for sharded arrays using real fixtures generated via PythonCall/zarr-python

    • Function-based codec registration system with typed CodecEntry for extensible V3 codec parsing

    • sharding_indexed codec context propagation: inner codec parser now receives element-size context so codecs like blosc get the correct typesize

    • sharding_indexed decode double-shift bug for index_location=:start: shard index offsets are now treated as absolute byte offsets from the shard start in all cases

    • sharding_indexed decode fill value: zdecode! now accepts an optional fill_value argument so missing inner chunks are filled with the array fill value instead of zero(T)

    • pipeline_decode! for V2Pipeline now accepts the fill_value keyword argument (ignored, but required to match the unified call site in ZArray.readblock!)

    • Ragged inner chunks in ShardingCodec handled correctly

  • fixes UI str parsing #259

  • added zarr_format to ZGroup #258

  • Add support for S3Path and remove deprecated global_aws_config() #253

  • (docs) get started #246

  • update badges #245

  • setup vitepress #243

  • Registration system for chunk key encoding #242

  • test: Consolidate CondaPkg.add calls #240

  • test: Read Julia-generated v3 fixtures with Python zarr #239

  • Use CondaPkg.jl#206 for nightly, fix manifest_uuid_path #238

  • fix docs build #237

  • consolidate methods #235

  • Add AbstractChunkKeyEncoding #234

  • V3 codec pipeline, group creation, and Python interop #232

  • Add read support for big endian #230

  • Indent method/type signatures on docstrings #229

  • Promote header levels in tutorial.md #228

  • Continue on v3 PR #226

  • move AWSS3 into extension #224