Apache DataFusion SQL Query Engine
fix: fix panic when lo is greater than hi (#19099)
Trent Hauckc293854a3a27
diff --git a/datafusion/functions-aggregate-common/src/tdigest.rs b/datafusion/functions-aggregate-common/src/tdigest.rsindex 80f016f051..225c61b719 100644--- a/datafusion/functions-aggregate-common/src/tdigest.rs+++ b/datafusion/functions-aggregate-common/src/tdigest.rs@@ -193,7 +193,11 @@ impl TDigest {if lo.is_nan() || hi.is_nan() {return v;}- v.clamp(lo, hi)++ // Handle the case where floating point precision causes min > max.+ let (min, max) = if lo > hi { (hi, lo) } else { (lo, hi) };++ v.clamp(min, max)}// public for testing in other modules@@ -744,4 +748,22 @@ mod tests {assert_eq!(t.size(), 96);}++ #[test]+ fn test_identical_values_floating_point_precision() {+ // Regression test for https://github.com/apache/datafusion/issues/14855+ // When all values are the same, floating-point arithmetic during centroid+ // merging can cause slight precision differences between min and max,+ // which previously caused a panic in clamp().++ let t = TDigest::new(100);+ let values: Vec<_> = (0..215).map(|_| 15.699999988079073_f64).collect();++ let t = t.merge_unsorted_f64(values);++ // This should not panic+ let result = t.estimate_quantile(0.99);+ // The result should be approximately equal to the input value+ assert!((result - 15.699999988079073).abs() < 1e-10);+ }}