chunked, compressed, N-dimensional arrays for Python

 Zarr is a Python package providing an implementation of compressed,
 chunked, N-dimensional arrays, designed for use in parallel
 computing. Some highlights:
 .
   - Create N-dimensional arrays with any NumPy dtype.
   - Chunk arrays along any dimension.
   - Compress chunks using the fast Blosc meta-compressor or
     alternatively using zlib, BZ2 or LZMA.
   - Store arrays in memory, on disk, inside a Zip file, on S3, ...
   - 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.
   - Use filters to preprocess data and improve compression.