Erwin XML

The Erwin adapter accepts XML exports from current and prior versions of Erwin Data Modeler. Logical entities, physical tables, attributes, columns, domains, primary/foreign keys, indexes, and subject areas are all preserved with their relationships intact.

The logical/physical split is preserved in the MetaModel rather than being collapsed. A logical entity `Customer` and its physical realization `dbo.tbl_customer_v2` are linked by a `realizes` edge in the MetaModel, so a Hierarchical Schema Inspection view can render either layer without re-deriving the relationship.

Erwin domains — reusable type definitions like `MoneyAmount` or `EmailAddress` — become first-class MetaModel domain objects. Re-applying a domain in a downstream mapping inherits its constraints, descriptions, and classification labels automatically, instead of requiring re-annotation per use.

Each ingest writes a `SchemaIngested` entry with the importer module identifier (`ErwinXmlImporter@x.y.z`), the SHA-256 of the XML bytes, and the assigned schema ULID. The Erwin model GUID is captured as a provenance annotation so a reviewer can correlate Magellan entries with the source Erwin model file.

When an Erwin model is re-exported and re-ingested after edits in the Erwin tool, the adapter computes the diff against the prior `SchemaVersionPublished` entry and writes `SchemaEvolutionEvaluated`. Additive changes auto-classify as minor; renames and type-narrowing classify as major; both require an Approver before reaching Published.

Erwin is often the system-of-record for enterprise data architecture. By keeping the Erwin model as the source artifact and the Magellan MetaModel as the governed projection, teams avoid the trap of two parallel models drifting out of sync — the Magellan side is always derivable from a known Erwin export ULID.