This is an automated email from the ASF dual-hosted git repository.
dtenedor 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 6939fcd4a313 [SPARK-43986][SQL][FOLLOWUP] Catch
ArrayIndexOutOfBoundsException in Datasketches HLL functions
6939fcd4a313 is described below
commit 6939fcd4a313717e0ac425a3bfed7a784908adca
Author: Daniel Tenedorio <[email protected]>
AuthorDate: Fri Jan 16 10:20:15 2026 -0800
[SPARK-43986][SQL][FOLLOWUP] Catch ArrayIndexOutOfBoundsException in
Datasketches HLL functions
### What changes were proposed in this pull request?
This PR fixes error handling in `HllUnionAgg` to properly catch
`ArrayIndexOutOfBoundsException` and convert it to the user-friendly
`HLL_INVALID_INPUT_SKETCH_BUFFER` error.
When an invalid HLL sketch binary is passed to `HllUnionAgg`, the
DataSketches library may throw an `ArrayIndexOutOfBoundsException` during
parsing. For example, if the `curMode` ordinal in the HLL preamble is set to an
invalid value (e.g., 3 when only 0, 1, 2 are valid), `CurMode.fromOrdinal()`
throws this exception. Previously, this exception would propagate to the user
as an unhelpful `ArrayIndexOutOfBoundsException`.
The fix adds `ArrayIndexOutOfBoundsException` to the existing catch block
that already handles `SketchesArgumentException` and `java.lang.Error`,
ensuring all invalid sketch buffer errors are converted to the clear
`HLL_INVALID_INPUT_SKETCH_BUFFER` error message.
### Why are the changes needed?
Without this fix, users see a confusing `ArrayIndexOutOfBoundsException`
when passing malformed binary data to HLL union functions. This provides a poor
user experience as it doesn't indicate what went wrong or how to fix it. The
`HLL_INVALID_INPUT_SKETCH_BUFFER` error clearly indicates that the input binary
is not a valid HLL sketch.
### Does this PR introduce _any_ user-facing change?
Yes. Users who pass invalid binary data to `hll_union_agg` will now see a
clear error message (`HLL_INVALID_INPUT_SKETCH_BUFFER`) instead of an
`ArrayIndexOutOfBoundsException`.
### How was this patch tested?
Added a new unit test `"HllUnionAgg throws proper error for invalid binary
input causing ArrayIndexOutOfBounds"` in `DatasketchesHllSketchSuite` that:
1. Crafts a byte array with a valid HLL preamble structure but an invalid
`curMode` ordinal (3 instead of 0, 1, or 2)
2. Verifies that the exception is NOT an `ArrayIndexOutOfBoundsException`
3. Verifies that the error message contains
`HLL_INVALID_INPUT_SKETCH_BUFFER`
### Was this patch authored or co-authored using generative AI tooling?
Yes, `claude-4.5-opus-high` plus manual review and editing.
Closes #53823 from dtenedor/fix-kll-quantiles-errors.
Authored-by: Daniel Tenedorio <[email protected]>
Signed-off-by: Daniel Tenedorio <[email protected]>
---
.../aggregate/datasketchesAggregates.scala | 3 +-
.../expressions/datasketchesExpressions.scala | 9 +++--
.../aggregate/DatasketchesHllSketchSuite.scala | 45 ++++++++++++++++++++++
3 files changed, 53 insertions(+), 4 deletions(-)
diff --git
a/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/aggregate/datasketchesAggregates.scala
b/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/aggregate/datasketchesAggregates.scala
index 8ae9b8fddde7..8fb1bf51319c 100644
---
a/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/aggregate/datasketchesAggregates.scala
+++
b/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/aggregate/datasketchesAggregates.scala
@@ -338,7 +338,8 @@ case class HllUnionAgg(
union.update(sketch)
Some(union)
} catch {
- case _: SketchesArgumentException | _: java.lang.Error =>
+ case _: SketchesArgumentException | _: java.lang.Error
+ | _: ArrayIndexOutOfBoundsException =>
throw
QueryExecutionErrors.hllInvalidInputSketchBuffer(prettyName)
}
case _ =>
diff --git
a/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/datasketchesExpressions.scala
b/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/datasketchesExpressions.scala
index a4ac0bdbb11d..1880d71e7d54 100644
---
a/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/datasketchesExpressions.scala
+++
b/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/datasketchesExpressions.scala
@@ -56,7 +56,8 @@ case class HllSketchEstimate(child: Expression)
try {
Math.round(HllSketch.heapify(Memory.wrap(buffer)).getEstimate)
} catch {
- case _: SketchesArgumentException | _: java.lang.Error =>
+ case _: SketchesArgumentException | _: java.lang.Error
+ | _: ArrayIndexOutOfBoundsException =>
throw QueryExecutionErrors.hllInvalidInputSketchBuffer(prettyName)
}
}
@@ -108,13 +109,15 @@ case class HllUnion(first: Expression, second:
Expression, third: Expression)
val sketch1 = try {
HllSketch.heapify(Memory.wrap(value1.asInstanceOf[Array[Byte]]))
} catch {
- case _: SketchesArgumentException | _: java.lang.Error =>
+ case _: SketchesArgumentException | _: java.lang.Error
+ | _: ArrayIndexOutOfBoundsException =>
throw QueryExecutionErrors.hllInvalidInputSketchBuffer(prettyName)
}
val sketch2 = try {
HllSketch.heapify(Memory.wrap(value2.asInstanceOf[Array[Byte]]))
} catch {
- case _: SketchesArgumentException | _: java.lang.Error =>
+ case _: SketchesArgumentException | _: java.lang.Error
+ | _: ArrayIndexOutOfBoundsException =>
throw QueryExecutionErrors.hllInvalidInputSketchBuffer(prettyName)
}
val allowDifferentLgConfigK = value3.asInstanceOf[Boolean]
diff --git
a/sql/catalyst/src/test/scala/org/apache/spark/sql/catalyst/expressions/aggregate/DatasketchesHllSketchSuite.scala
b/sql/catalyst/src/test/scala/org/apache/spark/sql/catalyst/expressions/aggregate/DatasketchesHllSketchSuite.scala
index 0841702cc518..0f7f5ca54be0 100644
---
a/sql/catalyst/src/test/scala/org/apache/spark/sql/catalyst/expressions/aggregate/DatasketchesHllSketchSuite.scala
+++
b/sql/catalyst/src/test/scala/org/apache/spark/sql/catalyst/expressions/aggregate/DatasketchesHllSketchSuite.scala
@@ -108,4 +108,49 @@ class DatasketchesHllSketchSuite extends SparkFunSuite {
assert(HllSketch.heapify(Memory.wrap(binary3.asInstanceOf[Array[Byte]])).getLgConfigK
== 12)
}
+
+ test("HllUnionAgg throws proper error for invalid binary input causing
ArrayIndexOutOfBounds") {
+ val aggFunc = new HllUnionAgg(BoundReference(0, BinaryType, nullable =
true), true)
+ val union = aggFunc.createAggregationBuffer()
+
+ // Craft a byte array that passes initial size checks but has an invalid
CurMode ordinal.
+ // HLL preamble layout:
+ // Byte 0: preInts (preamble size in ints)
+ // Byte 1: serVer (must be 1)
+ // Byte 2: famId (must be 7 for HLL)
+ // Byte 3: lgK (4-21)
+ // Byte 5: flags
+ // Byte 7: modeByte - bits 0-1 contain curMode ordinal (0=LIST, 1=SET,
2=HLL)
+ //
+ // Setting bits 0-1 of byte 7 to 0b11 (=3) causes CurMode.fromOrdinal(3)
to throw
+ // ArrayIndexOutOfBoundsException since CurMode only has ordinals 0, 1, 2.
+ // This happens in PreambleUtil.extractCurMode() before other validations
run.
+ val invalidBinary = Array[Byte](
+ 2, // byte 0: preInts = 2 (LIST_PREINTS, passes check)
+ 1, // byte 1: serVer = 1 (valid)
+ 7, // byte 2: famId = 7 (HLL family)
+ 12, // byte 3: lgK = 12 (valid range 4-21)
+ 0, // byte 4: unused
+ 0, // byte 5: flags = 0
+ 0, // byte 6: unused
+ 3 // byte 7: modeByte with bits 0-1 = 0b11 = 3 (INVALID curMode
ordinal!)
+ )
+
+ val exception = intercept[Exception] {
+ aggFunc.update(union, InternalRow(invalidBinary))
+ }
+
+ // Verify that ArrayIndexOutOfBoundsException is properly caught and
converted
+ // to the user-friendly HLL_INVALID_INPUT_SKETCH_BUFFER error
+ assert(
+ !exception.isInstanceOf[ArrayIndexOutOfBoundsException],
+ s"ArrayIndexOutOfBoundsException should be caught and converted to " +
+ s"HLL_INVALID_INPUT_SKETCH_BUFFER error, but got:
${exception.getClass.getName}"
+ )
+ assert(
+ exception.getMessage.contains("HLL_INVALID_INPUT_SKETCH_BUFFER"),
+ s"Expected HLL_INVALID_INPUT_SKETCH_BUFFER error, " +
+ s"but got: ${exception.getClass.getName}: ${exception.getMessage}"
+ )
+ }
}
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
