model-writtenrun 820 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
- Bertrand Drouvot reported that DROP PROPERTY GRAPH could fail with an unsupported-object-class error while an event trigger was active, and proposed adding missing type-description and identity cases in #1.
- Michael Paquier pointed out that object-address functions also reach the missing cases and that event triggers are unsuitable for parallel regression groups, while Ashutosh Bapat reproduced the failure with a smaller test; Bertrand consequently moved trigger coverage into event_trigger.sql and added object-address tests in #4, following #2 and #3.
- Ashutosh withdrew his initial doubt about exposing these objects because ALTER PROPERTY GRAPH can manipulate labels and label properties independently, then proposed recursive dependency coverage in #5; Bertrand added that coverage in #6.
- Ashutosh supplied a revised patch in #7 that preserved missing label and property names in identities, replaced manual scans with get_catalog_object_by_oid(), reused a dependency view for tests, and exercised an edge and ALTER PROPERTY GRAPH.
- Bertrand observed locale-dependent ordering and added COLLATE "C", while also simplifying the existing description lookups in #8; after property-graph row types were removed elsewhere, #10 rebased the patch and dropped the unnecessary collation workaround.
- Alex Guo reviewed the patch positively in #11, and Michael split its unchanged description-lookup cleanup into 6827de95ee, then posted the remainder with the catalog-consistent name "property graph label property" in #13.
- Peter Eisentraut questioned whether exposing implementation-level catalogs to event triggers would constrain future catalog changes and suggested skipping them in #14; Michael argued for consistent dependency descriptions in #15, and Ashutosh explained their usefulness for tracking ALTER PROPERTY GRAPH changes across clusters in #16.
- Peter ultimately committed the remainder from #13 as 72498a8698, adding missing ObjectTypeMap entries and removing inappropriate translation markers, as reported in #18.
- The original bug was fixed, but Michael renewed his concern about ambiguous identities such as repeated names joined by "of" in #19, and Tom Lane argued that identifying the intermediate object types would improve readability in #20.
- Ashutosh distinguished machine-oriented identities from verbose object descriptions in #21, but Tom showed that existing identity formats already include object types where needed in #22, with Michael supporting that argument in #23.
- Ashutosh posted a separate readability patch in #24 that added "property graph", "element", and "label" to identity strings; Michael questioned the shorter label terminology in #25, then accepted Ashutosh's explanation in #26 after checking the standard in #27.
- Ashutosh marked the original open item resolved in #28, and Peter subsequently committed the readability patch unchanged apart from rebasing as 2a7e95b659, announcing completion in #30.
- The thread ends with Michael's acknowledgment in #31; the cleanup, original bug fix, and readability follow-up have all landed, with no attached patch left pending.