scene_pipelines.ingestion.provenance

DuplicatePolicy

Bases: str, Enum

What to do when the tracked row's primary key is already stored.

DataJoint's skip_duplicates / replace flags only look at the primary key. This policy also compares the content_hash stored in the row-meta stamp, so callers can detect a re-registration with different content. A stamp of another HASH_VERSION is treated as missing: its hash is not comparable.

Attributes:

Name	Type	Description
`REJECT`		Raise `ValueError`.
`SKIP`		Leave the stored row and its stamp untouched.
`VERIFY`		Leave the stored row untouched when the stamp hash matches `payload`; raise `ValueError` when it differs or no stamp exists.
`OVERWRITE`		Overwrite the row and it's stamp. Warns when the hash changed. Omitted optional fields are reset to their default.

insert_tracked_row

insert_tracked_row(row_meta_table: type[RowMetaBase], row: DjRow, *, payload: dict[str, Any], deployment: DjRow[Deployment], if_exists: DuplicatePolicy, writer_version: str, parts: Mapping[type[Part], Sequence[DjRow]] | None = None) -> DjKey

Insert row into row_meta_table.tracked_table and stamp it.

A new primary key is inserted and stamped. An existing primary key is handled by if_exists. parts are written whenever the row is: inserted with a new row, replaced on OVERWRITE, and left untouched otherwise. The Deployment row is inserted when missing. All statements run atomically: inside the caller's open transaction when there is one, otherwise in a transaction opened here.

Parameters:

Name	Type	Description	Default
`row_meta_table`	`type[RowMetaBase]`	Row-meta table; its `tracked_table` receives `row`.	required
`row`	`DjRow`	Full insert dict for the tracked table.	required
`payload`	`dict[str, Any]`	Fields hashed into `content_hash` (caller-defined shape).	required
`deployment`	`DjRow[Deployment]`	Deployment row to stamp with.	required
`if_exists`	`DuplicatePolicy`	Policy when the tracked primary key already exists.	required
`writer_version`	`str`	Stamped `ingestion_version`: the version constant of the calling ingestion package, e.g. `SCENE_WRITER_VERSION`.	required
`parts`	`Mapping[type[Part], Sequence[DjRow]] \| None`	Optional part rows per part table, without the master key (it is added here), e.g. `{Session.Subject: [{"subject_id": ...}]}`. Include them in `payload` so the hash covers them.	`None`

Returns:

Type	Description
`DjKey`	Primary key of the tracked row.

Raises:

Type	Description
`ValueError`	When `if_exists` rejects the existing row.
`SchemaVersionError`	When a written database's schema version does not match the installed code (checked once per database and process).