DanielLeens commented on PR #12003: URL: https://github.com/apache/seatunnel/pull/12003#issuecomment-5473706182
Confirmed -- I independently compared the two diffs and @davidzollo's read is accurate. #12001 (merged into `dev` on 2026-08-30) fixes the exact same root cause as this PR, the same way: - Both stop re-parsing the raw `row.getTableId()` string back into a `TablePath` via the ambiguous `TablePath.of(String)` (which hardcodes `schemaFirst=false` and mis-resolves 2-part `schema.table` identifiers from Oracle/PostgreSQL/SQL Server/Redshift/DM/Xugu/Iris/Kingbase-style dialects). - Both switch the internal map to be keyed by the raw string identity instead: this PR uses `Map<String, TablePath>` keyed by `sourceTable.getFullName()`, #12001 uses `Map<String, String>` keyed by `sourceTable.toString()`. These are equivalent -- `TablePath.toString()` just delegates to `getFullName()` (`TablePath.java:130-131`), so both PRs converge on the identical key scheme, which also matches how `MultiTableSinkWriter` itself already routes rows by raw string ID. - Both ship an equivalent regression test for the same schema-first case (`PDC_SCHEMA.CUSTOMER` -> `target_db.CUSTOMER`), asserting the write count lands on the resolved table and that the `default.default.default` fallback bucket stays at 0. I don't see anything in this PR's diff that #12001 is missing -- no extra edge case, no additional test coverage, no different (more correct) behavior. Given #12001 is earlier and already merged, I'd recommend closing this one in favor of it to avoid duplicate/conflicting changes to the same map field. Thanks for tracking this bug down independently, @zhangshenghang -- the analysis and test here are solid, it's just a case of two contributors fixing the same issue in parallel. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
