ump45nose opened a new pull request, #3502: URL: https://github.com/apache/iceberg-rust/pull/3502
## Which issue does this PR close? - Closes #3494. ## What changes are included in this PR? `RowGroupMetricsEvaluator::visit_inequality` only checked the literal for NaN. When a row group's min statistic was NaN (parquet-mr 1.9 and earlier record a NaN that starts a row group as both min and max), `f < 5.0` and `f <= 5.0` pruned the row group even though it could hold smaller values. This makes `visit_inequality` return might-match when the bound it compares against is NaN, the same way `eq` and `in` in this evaluator already treat a NaN bound as unreliable. ## Are these changes tested? Yes. Four new tests in `row_group_metrics_evaluator.rs` evaluate `<`, `<=`, `>` and `>=` against a row group whose min and max are both NaN. The `<` and `<=` tests fail on `main` (`219571a`) and pass with the fix; the `>`/`>=` ones already pass (a NaN max compares greater than the literal) and are there to pin the behaviour for both bounds. `cargo test -p iceberg --lib expr::visitors` and `cargo clippy -p iceberg --all-targets -- -D warnings` pass locally. ## AI Disclosure I used an AI assistant to help write the fix and tests, then reviewed the change and ran the tests myself. -- 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]
