dr_catalog_openlineage()

Sends a START event using the recorded start time, followed by COMPLETE or FAIL, after execution has finished. This is buffered historical lineage, not live progress monitoring. Both events keep the same deterministic UUID across dataraft.core::dr_retry_catalogs() calls. Delivery is at least once; a retry can repeat an already accepted START event. Failed and blocked runs have no output datasets. No data rows or executable transformation definitions are sent. Output schema fields are included when available.

Usage

dr_catalog_openlineage(
  endpoint,
  namespace = "dataraft",
  request = NULL,
  datasets = list()
)

Arguments

endpoint: Full HTTP(S) OpenLineage ingestion URL, for example https://lineage.example/api/v1/lineage. Supply authentication through request, never embedded credentials or URL query parameters.
namespace: Job namespace. Dataset namespaces identify physical sources.
request: Optional httr2 request or function accepting a request and returning a configured request. A zero-argument request factory is also accepted. Factories can obtain fresh credentials at delivery time. The adapter always restores its configured destination URL.
datasets: Named list keyed by logical asset or source ID. Each value contains namespace and name following OpenLineage naming conventions, for example list(namespace = "postgres://warehouse:5432", name = "insurance.public.contracts"). Descriptor dataset identities take precedence. Unknown locations use an explicitly logical DataRaft namespace and cannot merge with physical datasets from other tools. Supply only public identifiers without credentials or signed URLs.

Value

A catalog adapter for dataraft.core::dr_add_catalog().

Details

External business catalogs consume descriptive metadata. They are distinct from dataraft.lake::dr_registry(), which records the local lake publication lifecycle.

Examples

catalog <- dr_catalog_openlineage("https://lineage.example/api/v1/lineage")
dataraft.core::dr_inspect(catalog)