This is an automated email from the ASF dual-hosted git repository.
zhengruifeng pushed a commit to branch master
in repository https://gitbox.apache.org/repos/asf/spark.git
The following commit(s) were added to refs/heads/master by this push:
new 15a091b94a4f [SPARK-58916][SQL] Handle infinite relativeError in
approxQuantile
15a091b94a4f is described below
commit 15a091b94a4fe733b95c60aa79e184de3966bcfd
Author: Ruifeng Zheng <[email protected]>
AuthorDate: Fri Aug 21 08:12:23 2026 +0800
[SPARK-58916][SQL] Handle infinite relativeError in approxQuantile
### What changes were proposed in this pull request?
Clamp the integer accuracy derived by
`DataFrameStatFunctions.approxQuantile` to at least 1 when delegating to
`approx_percentile`.
### Why are the changes needed?
After SPARK-51816, `approxQuantile` computes `approx_percentile` accuracy
from `relativeError` as `ceil(1.0 / relativeError)`. `Double.PositiveInfinity`
is accepted by the existing non-negative relative-error check, but that formula
produces accuracy 0, which is invalid for `approx_percentile`. Clamping to 1
keeps the coarsest valid approximation for this edge case.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
Added `Double.PositiveInfinity` to the existing `DataFrameStatSuite`
coverage for `relativeError > 1`.
Not run locally.
Closes #58173 from
zhengruifeng/zhengruifeng/approx-quantile-infinite-relative-error.
Authored-by: Ruifeng Zheng <[email protected]>
Signed-off-by: Ruifeng Zheng <[email protected]>
---
.../main/scala/org/apache/spark/sql/execution/stat/StatFunctions.scala | 3 ++-
sql/core/src/test/scala/org/apache/spark/sql/DataFrameStatSuite.scala | 2 +-
2 files changed, 3 insertions(+), 2 deletions(-)
diff --git
a/sql/core/src/main/scala/org/apache/spark/sql/execution/stat/StatFunctions.scala
b/sql/core/src/main/scala/org/apache/spark/sql/execution/stat/StatFunctions.scala
index e812af229524..57b7bf7ac41b 100644
---
a/sql/core/src/main/scala/org/apache/spark/sql/execution/stat/StatFunctions.scala
+++
b/sql/core/src/main/scala/org/apache/spark/sql/execution/stat/StatFunctions.scala
@@ -81,7 +81,8 @@ object StatFunctions extends Logging {
val accuracy = if (relativeError == 0.0) {
Int.MaxValue
} else {
- math.min(Int.MaxValue, (1.0 / relativeError).ceil.toLong).toInt
+ val raw = (1.0 / relativeError).ceil.toLong
+ math.max(1, math.min(Int.MaxValue, raw)).toInt
}
val results = Array.fill(cols.size)(Seq.empty[Double])
diff --git
a/sql/core/src/test/scala/org/apache/spark/sql/DataFrameStatSuite.scala
b/sql/core/src/test/scala/org/apache/spark/sql/DataFrameStatSuite.scala
index 99d4a66195c8..1154863a2dcd 100644
--- a/sql/core/src/test/scala/org/apache/spark/sql/DataFrameStatSuite.scala
+++ b/sql/core/src/test/scala/org/apache/spark/sql/DataFrameStatSuite.scala
@@ -264,7 +264,7 @@ class DataFrameStatSuite extends SharedSparkSession {
val q1 = 0.5
val q2 = 0.8
- val epsilons = List(2.0, 5.0, 100.0)
+ val epsilons = List(2.0, 5.0, 100.0, Double.PositiveInfinity)
val Array(single1_1) = df.stat.approxQuantile("singles", Array(q1), 1.0)
val Array(s1_1, s2_1) = df.stat.approxQuantile("singles", Array(q1, q2),
1.0)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
