zhuxiangyi commented on PR #10144: URL: https://github.com/apache/paimon/pull/10144#issuecomment-5810876018
Thanks @JingsongLi for the careful review! All three points are addressed in 8c3cca696: 1. **Signed zero (P1):** comparisons, `IN` and `BETWEEN` on nested FLOAT/DOUBLE fields are no longer converted, so Flink evaluates them. `IS NULL` / `IS NOT NULL` are still pushed down. Covered by `testNestedFloatingPointComparisonsAreNotConverted` and, end to end, by `testNestedDoubleSignedZeroKeepsMatchingRows` / `testNestedFloatingPointComparisonsAreNotPushedToScan`. 2. **Field IDs on evolved schemas (P2):** nested predicates are now bound to the table's own `RowType` (new `PredicateConverter.convert(org.apache.paimon.types.RowType, ...)`, used by `FlinkTableSource`), so the transform keeps the real field IDs. `testNestedPredicateKeepsTableFieldIdsOnEvolvedSchema` reproduces the `changed identity` failure through `PredicateRemapper` without the fix. 3. **Real plan:** `testRealPlanPushesNestedPredicateToScan` plans a SQL query, takes the source from the plan, and checks that the nested predicate actually prunes the scan (3 of 6 rows read instead of 6). Separately, top-level DOUBLE columns have the same `-0.0` issue on master (`WHERE d = 0.0` can drop `-0.0` rows through parquet pruning). It isn't related to this PR, so I'll open a separate issue for it. -- 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]
