description.yml

extension:
  name: cozip
  description: Cloud-Optimized ZIP reader for DuckDB
  version: 1.2.0
  language: C++
  build: cmake
  license: MIT
  maintainers:
    - csaybar
    - ryali93
repo:
  github: asterisk-labs/cozip_reader
  ref: 0000000000000000000000000000000000000000
docs:
  hello_world: |
    INSTALL cozip FROM community;
    LOAD cozip;

    SELECT *
    FROM read_cozip('https://huggingface.co/datasets/Major-TOM/Core-VIIRS-Nighttime-Light/resolve/main/2024/MAJORTOM-VIIRS-NTL_2024_median_000.zip')
    LIMIT 10;
  extended_description: |
    cozip replaces the ZIP Central Directory scan with a Parquet metadata
    file located through a fixed 51-byte header at byte 0 of the archive.
    read_cozip(path) reads that Parquet directly through a virtual
    cozip-subfile filesystem, so range requests flow lazily through the
    underlying transport. Works on local files and remote URLs (HTTPS, S3,
    GCS, Azure, HuggingFace), on native and WebAssembly. Every row gets
    an extra cozip:gdal_vsi column with a /vsisubfile/ path that opens
    the referenced inner file in GDAL or rasterio without re-downloading
    the archive.