Skip to contents

A low-level reader for 'Zarr' arrays (https://zarr-specs.readthedocs.io/), versions 2 and 3 including sharded version 3, native or virtual, from local disk, HTTP(S), Amazon 'S3' and Google Cloud Storage. Virtual stores read the bytes of existing 'NetCDF', 'HDF5' and 'GeoTIFF' files in place, from 'Kerchunk' 'JSON' and 'Parquet' references, 'VirtualiZarr' manifests and 'byterefs' tables ('SQLite', or the flat 'Parquet' export read by row group). Every read is a plan of byte references (one row per chunk) fetched through one 'curl' connection pool, then decoded in R. Arrays can also be opened as lazy R arrays (via 'altarr') with 'CF' unpacking, and 'Icechunk' repositories read (via 'tabularberg'). 'Arrow' and 'GDAL' (via 'gdalraster') are optional, for 'S3' and Google Cloud Storage listing and credentials and for some codecs.

Author

Maintainer: Michael Sumner mdsumner@gmail.com (ORCID)

Authors:

Other contributors:

  • Hugo Gruson (ORCID) [contributor]