hutiefang76 commented on PR #12592: URL: https://github.com/apache/seatunnel/pull/12592#issuecomment-5986154599
@DanielLeens > try `rs.getObject(idx, LocalDateTime.class)` first and only fall back to the plain `getTimestamp` Addressed in eb693c5. With DuckDB JDBC 1.3.1.0, I reproduced the Los Angeles DST gap (2024-03-10 02:30:00.123456 becoming 03:30:00.123456) and the Gregorian cutover (1582-10-10 becoming 1582-10-20). Typed reads preserve those standard TIMESTAMP values. The new real SourceFactory regression covers gap/overlap, cutover and NULL through table_path and query in UTC, Shanghai and Los Angeles; it failed before the fix and passes now. The driver rejects typed reads for TIMESTAMP_S/MS/NS. Their plain fallback remains for compatibility, with unsupported capability cached per result-set column to avoid per-row exceptions. The EN/ZH docs and PR description now state the aliases' remaining DST/cutover limitation and recommend an SQL CAST to VARCHAR where database text must be retained. All three aliases are also listed in the mapping table. > please re-run the failed job (or sync with latest dev, which is 15 commits ahead) Synced with current dev (eb2eb1d) in 743a440, resolving documentation/test conflicts while retaining the merged BIT/ENUM changes. The exact final tree passed 121 focused tests on JDK 17, the affected nine-module verify/Spotless checks and all four MarkdownTest checks. Engine E2E tests were not run. The new upstream Build check is queued; these local results do not establish a remote CI pass. -- 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]
