Provenance

See Export to d3rlpy for writing a sidecar during export.

class ehr2rl.provenance.DatasetProvenance(bigquery_job_ids, query_hash, itemid_map_version, feature_preset, extraction_timestamp, mimic_version)

Metadata needed to reproduce a BigQuery-backed dataset extraction.

Parameters:

  • bigquery_job_ids (tuple[str, ...]) – IDs of the BigQuery jobs that ran. Cached queries have none.

  • query_hash (str) – Cache keys of every query in the load, joined with ":".

  • itemid_map_version (str) – Version of the itemid map used.

  • feature_preset (str) – Name of the feature preset used.

  • extraction_timestamp (str) – UTC time of the extraction, in ISO 8601 format.

  • mimic_version (str) – MIMIC-IV version, such as "3.1".

ehr2rl.provenance.write_provenance(path, provenance)

Write provenance metadata to a JSON file.

Parameters:

  • path (str | Path) – File to write. An existing file is replaced.

  • provenance (DatasetProvenance) – Metadata to write.

Return type:

None

ehr2rl.provenance.read_provenance(path)

Read provenance metadata from a JSON file.

Parameters:

path (str | Path) – File written by write_provenance.

Returns:

The metadata.

Return type:

DatasetProvenance

Raises:

ValueError – If a required field is missing.