Claude Code Zarr Skill for Access-Aligned Array Chunks
Choose chunk shapes from expected reads. Use small local arrays to compare how many chunks a representative slice touches before storing larger data.
中文Install in Claude Code
Use Claude Code with the pinned skill and the project's Python, Zarr, and NumPy versions. Use temporary local storage.
- Get the complete `cli-tool/components/skills/scientific/zarr-python` directory from the pinned GitHub checkout.
- Copy it under the target project's `.claude/skills/`. Confirm `.claude/skills/zarr-python/SKILL.md`, open Claude Code at the project root, and call `/zarr-python`.
Example task
Compare chunks (1,64,64) and (24,16,16) for a fictional 24 by 64 by 64 hourly grid when most reads request one full hourly plane.What to check
The first shape touches one chunk for an hourly plane. The second touches sixteen spatial chunks. Measure your actual workload before selecting a layout.
Original SKILL.md on GitHub · Updated
Frequently asked question
Why not simply choose the largest chunk?
Chunk shape controls how many chunks and unrelated values a read touches. Start with typical slices and check storage and I/O costs.