Skip to content

zarr-tree v0.3.0

Choose a tag to compare

@ronfinn ronfinn released this 29 Aug 12:05
· 19 commits to master since this release

zarr-tree v0.3.0

This release expands zarr-tree from a local Zarr metadata inspector into a
remote and SpatialData-aware inspection tool.

Remote stores

  • inspect Zarr stores directly on S3
  • inspect HTTP/HTTPS stores using WebDAV-compatible listing
  • traverse static HTTP stores using consolidated metadata
  • support Zarr V2 .zmetadata
  • support Zarr V3 inline consolidated_metadata

SpatialData

  • inspect Parquet footer metadata for Points and Shapes
  • report rows, columns, file counts and schemas without reading records
  • summarize AnnData-backed Tables
  • report observations, variables, X representation, obs/var columns and
    SpatialData region linkage
  • distinguish unavailable Points payload metadata with parquet files: ?

Efficiency

  • arrays remain leaves; chunk keys are not enumerated
  • remote Parquet metadata uses bounded footer reads
  • expression matrices are not read
  • consolidated metadata can traverse an entire HTTP hierarchy with a single
    metadata document

Validation

Validated against real public data including:

  • IDR OME-Zarr 0.3, 0.4 and 0.5
  • OME-Zarr HCS
  • Janelia COSEM
  • Pangeo consolidated Zarr
  • SpatialData Xenium
  • SpatialData Visium
  • SpatialData MERFISH
  • SpatialData MIBI-TOF

The final public-data validation found no correctness bugs.

Scope

zarr-tree remains a read-only metadata inspection tool.

No chunk/pixel reads, Parquet row reads or expression-matrix reads are
performed.