xylaaaaa commented on issue #69884: URL: https://github.com/apache/doris/issues/69884#issuecomment-6654170183
Pushed the generated baseline for the Delta Parquet deletion-vector regression to [PR #69904](https://github.com/apache/doris/pull/69904) in [`7e31fe75db`](https://github.com/apache/doris/pull/69904/commits/7e31fe75dbe9614087015cea4c1fabd7ca8a590a). The two existing baselines passed, and generation and ordinary verification matched across all 22 result blocks. The generated file is the only change in this commit. I also ran five read-only checks through each of the filesystem and HMS catalogs on the retained Spark fixture. Both catalogs produced the same results. | Check | Result | | --- | --- | | Non-adjacent pages in one row group, `id IN (5,200,201,204)` | Only 201 and 204 returned. The profile showed two selected ranges and one row group read. | | Predicate matching only deleted IDs 199 and 200 | Zero rows. `COUNT(*), MIN(id), MAX(id)` returned `0, NULL, NULL`. | | `id IN (101,102,9005,9006)`, 32 KiB split size | Four planned splits. Only 102 and 9006 returned. The two scans of the catalogs each skipped 9,800 rows by page index. | | All live IDs | All 9,990 IDs matched the expected ordered sequence after excluding the ten known deletions. | The existing positive range profiles select one contiguous range each. The `abs(id)` controls showed zero row groups filtered by min/max, zero rows filtered by page, and zero page-index reads in both official runs. The additional checks above are manual and do not yet guard future builds. Priorities for permanent coverage are the disjoint/empty and split-boundary checks, then a fixture with row-specific payloads so a value selected from the wrong row is observable. No incorrect results were observed. -- 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] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
