Skip to content

zarr_filenames

anomaly_match.prediction.zarr_filenames

Derive per-image filenames for a Zarr store.

Shared by :mod:subprocess_scripts.prediction_process_zarr (to tag its store_results writes) and :mod:anomaly_match.pipeline.session (to detect when an entire Zarr file is already scored and skip the subprocess spawn entirely). Kept in one place so the two callers agree byte-for-byte on what name a given index resolves to.

derive_filenames(zarr_path)

Return the (filename ↔ zarr index) mapping used by prediction.

Looks for a parquet sidecar — pointed at either by the zarr's metadata_file attribute or by <name>_metadata.parquet next to the store — and pulls per-row filenames from original_filename / filename / source_id (in that order of preference). Falls back to {prefix}__image_{i:06d} when no metadata is found or the row count doesn't match.

Parameters:

Name Type Description Default
zarr_path str | Path

Path to the zarr store (file or directory).

 required

Returns:

Type Description
list[str]

List of filename strings, length equal to the zarr's images
list[str]

array length. Empty list if the images array is missing or
list[str]

unreadable.
Source code in anomaly_match/prediction/zarr_filenames.py 
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
def derive_filenames(zarr_path: str | Path) -> list[str]:
    """Return the ``(filename ↔ zarr index)`` mapping used by prediction.

    Looks for a parquet sidecar — pointed at either by the zarr's
    ``metadata_file`` attribute or by ``<name>_metadata.parquet`` next
    to the store — and pulls per-row filenames from
    ``original_filename`` / ``filename`` / ``source_id`` (in that order
    of preference).  Falls back to ``{prefix}__image_{i:06d}`` when no
    metadata is found or the row count doesn't match.

    Args:
        zarr_path: Path to the zarr store (file or directory).

    Returns:
        List of filename strings, length equal to the zarr's ``images``
        array length.  Empty list if the ``images`` array is missing or
        unreadable.
    """
    zarr_path = Path(zarr_path)
    try:
        root = zarr.open_group(str(zarr_path), mode="r")
    except Exception as exc:
        logger.warning("Cannot open Zarr store {} for filename derivation: {}", zarr_path, exc)
        return []

    if "images" not in root:
        logger.warning("Zarr store {} has no 'images' array", zarr_path)
        return []

    num_images = root["images"].shape[0]
    prefix = _zarr_prefix(zarr_path)
    metadata_file = _locate_metadata_file(zarr_path, dict(root.attrs))

    if metadata_file is None:
        return _generated_names(prefix, num_images)

    try:
        metadata_df = pd.read_parquet(metadata_file)
    except Exception as exc:
        logger.warning("Failed to read zarr metadata {}: {}", metadata_file, exc)
        return _generated_names(prefix, num_images)

    for column in _FILENAME_COLUMN_PRIORITY:
        if column in metadata_df.columns:
            filenames = [str(v) for v in metadata_df[column].tolist()]
            break
    else:
        logger.warning("No filename column in {}", metadata_file)
        return _generated_names(prefix, num_images)

    if len(filenames) != num_images:
        logger.warning(
            "Filename count ({}) does not match image count ({}) in {}",
            len(filenames),
            num_images,
            zarr_path,
        )
        return _generated_names(prefix, num_images)

    return filenames