Release Release v1.0.0 · datagems-eosc/ap-explanation

v1.0.0 — 2026-08-26

provsql 1.12.0 required — Existing database volumes upgrade automatically on restart via ALTER EXTENSION provsql UPDATE in the entrypoint; annotating a table afterwards (add_semiring) migrates its mapping to the new format
AP labels now follow the MoMa vocabulary — Relational_Database → RelationalDatabase, CSV_Set → CsvSet, and the contain edge → containedIn. APs using the old labels are rejected. The Provenance_SQL_Operator / Provenance_Annotate_Dataset_Operator operators are unchanged — MoMa has no provenance node types yet.
APs are validated against the MoMa graph schema — node and edge ids must be valid UUIDs, every node must carry a properties object whose keys match ^[a-z@][a-zA-Z0-9_]*$, and edge labels are restricted to MoMa's closed set. Any string was previously accepted.

Stable row references: Instead of using CTIDs, every annotated table now carries a provsql_ref column instead of relying on physical row location. CtidMapping is removed; KeyMapping replaces it.
AP models come from moma-domain — the hand-maintained PgJson copy is gone; AnalyticalPattern now builds on moma_management.domain, so the graph vocabulary stays owned by moma-management. Only moma's StructureStep runs, not its full SchemaStep & StructureStep & MappingStep chain, since SchemaStep would reject the provenance operators.

Provsql >=1.11.0 add_provenance() and remove_provenance() are now idempotent, unnecessary workaround were removed as a consequence.

Full Changelog: v0.14.0...v1.0.0