An interactive Jupyter notebook that teaches the Zarr data model from first principles, then introduces Icechunk for version-controlled Zarr.
- Zarr as a protocol — stores, groups, arrays, chunks, and how they map to key-value backends
- Backend independence — the same API over local files, in-memory stores, or cloud storage
- Icechunk — git-like versioning (commits, time travel) without changing the Zarr API
- On-disk internals — how Icechunk organizes refs, snapshots, manifests, chunks, and transaction logs
A companion StoreExplorer widget (built with anywidget) displays three live panels alongside the notebook: the logical Zarr data model, the store keys, and the actual files on disk.
Requires Python 3.12+.
uv syncuv run jupyter lab zarr-intro.ipynbRun cells top to bottom. The StoreExplorer opens in a side panel and auto-refreshes after each cell.
MIT