hutiefang76 commented on PR #12498: URL: https://github.com/apache/seatunnel/pull/12498#issuecomment-5875719878
Synced this branch with current dev and removed the XA documentation overlap with #12505. The bulk implementation is unchanged; JDK 8/17 each pass the 51 focused tests, and MarkdownTest passes 4/4. I also checked the staging cost through the actual SinkFactory/writer with identical generated-SQL settings in both modes, toggling only `ducklake_bulk_write`: | Setup | Rows / batch size | Ordinary / bulk time | Ordinary / bulk active files | | --- | --- | --- | --- | | SQLite metadata, local data | 2,000 / 512 | 0.858s / 0.246s | 2,000 / 4 | | PostgreSQL 17 metadata, MinIO | 500 / 128 | 3.673s / 0.292s | 500 / 4 | Each row included DECIMAL(18,4), a microsecond TIMESTAMPTZ and a 1KiB text value or null. A new connection read back and compared every value. Both modes added four snapshots in each listed case, so the file reduction does not change commit semantics. These are one-run smoke measurements with a fresh JVM/lake per case, not a steady-state benchmark. Whole-process RSS includes native DuckDB and readback; I am not claiming a general memory reduction or a fixed speedup. The opt-in/insert-only delivery limits remain unchanged. -- 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]
