Metadata Workflow Example

This notebook demonstrates the stable, user-facing API without requiring a live catalog. It covers three common tasks:

• build a canonical ScanResult

• summarize it into a user-facing DatasetSummary

• join it to profile data in Arrow

import pyarrow as pa

from iceberg_bioimage import (
    ImageAsset,
    ScanResult,
    join_profiles_with_scan_result,
    summarize_scan_result,
)
from iceberg_bioimage.publishing.chunk_index import scan_result_to_chunk_rows
from iceberg_bioimage.publishing.image_assets import scan_result_to_rows

scan_result = ScanResult(
    source_uri="data/example.ome.zarr",
    format_family="zarr",
    image_assets=[
        ImageAsset(
            uri="data/example.ome.zarr",
            array_path="0",
            shape=[1, 1, 256, 256],
            dtype="uint16",
            chunk_shape=[1, 1, 128, 128],
            metadata={
                "axes": "czyx",
                "channel_count": 1,
                "storage_variant": "zarr-v2",
            },
            image_id="example:0",
        )
    ],
)

scan_result.to_dict()

summary = summarize_scan_result(scan_result)
summary.to_dict()

profiles = pa.table(
    {
        "dataset_id": ["example"],
        "image_id": ["example:0"],
        "cell_count": [42],
    }
)

joined = join_profiles_with_scan_result(scan_result, profiles, include_chunks=True)
joined.to_pydict()

{
    "image_assets": scan_result_to_rows(scan_result),
    "chunk_index_count": len(scan_result_to_chunk_rows(scan_result)),
}

Optional OME-Arrow path

If you install the optional ome-arrow extra, the package also exposes a small bridge for Arrow-native image payload workflows:

from iceberg_bioimage import create_ome_arrow, scan_ome_arrow

oa = create_ome_arrow("image.ome.tiff")
lazy_oa = scan_ome_arrow("image.ome.parquet")

That keeps metadata registration and Arrow-native image handling adjacent, without pulling OME-Arrow into the required dependency set.