N5 support for the zarrs ecosystem.
- Read-only
- No partial chunk reading
- "default" chunk mode (i.e. not varlen or object)
- Compression support:
- N5 core
- gzip
- bzip2
- lz4
- xz
- N5 extensions
- zstd https://github.com/JaneliaSciComp/n5-zstandard
- blosc https://github.com/saalfeldlab/n5-blosc
- jpeg is implemented in java but not well documented
- PRs welcome but I'm unlikely to prioritise this unless a Zarr JPEG codec were stabilised
- N5 core
- The handling of edge chunks is quite inefficient
- Zarr groups must have metadata documents, but N5 groups may not. This may lead to unexpected behaviour when discovering hierarchy structure. This library allows inferring a group with empty attributes when a metadata document is missing.
- N5 hierarchies have a root (metadata contains the
"n5"key), but Zarr hierarchies do not; in practice this doesn't matter much
- constantinpape/z5: C++/Python reader over a common subset of Zarr and N5
- aschampion/rust-n5: rust implementation of N5
- zarr-developers/zarr-python: Zarr v2's N5 support
- zarr-developers/n5py: python plugins for reading N5 through Zarr v3
- saalfeldlab/n5-zarr: Java implementation reading zarr data through the N5 interface