The zarr library is a Python implementation of the Zarr storage format. zarr delivers compressed, chunked, N-dimensional arrays that work well for parallel computing and object storage. See the documentation for more information.

• Create N-dimensional arrays with NumPy-compatible dtypes.
• Chunk arrays along any dimension.
• Encode chunks using a variety of useful encodings (e.g., compression).
• Store arrays in memory, on disk, inside a zip file, on S3, etc...
• Read an array concurrently from multiple threads or processes.
• Write to an array concurrently from multiple threads or processes.
• Organize arrays into hierarchies via groups.

Zarr can be installed from PyPI using pip:

pip install zarr

or via conda:

conda install -c conda-forge zarr

For more details, including how to install from source, see the installation documentation.

In addition to the primary zarr implementation, this repository contains other packages that provide specialized functionality with minimal dependencies:

• zarr-metadata: Tools for Zarr metadata. Install with pip install zarr-metadata.
• zarr-indexing: Tools for lazily indexing chunked arrays. Install with pip install zarr-indexing.
• zarr-http-server: An HTTP server implementation targeting Zarr data. Install with pip install zarr-http-server.