Skip to content

[SPARK-58069][SQL][FOLLOWUP] Handle empty approx_top_k combine buffers #57375

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
cloud-fan wants to merge 2 commits into
base: master
Choose a base branch
from
+56 −6
Open

[SPARK-58069][SQL][FOLLOWUP] Handle empty approx_top_k combine buffers #57375

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
b4295d7
[SPARK-58069][SQL][FOLLOWUP] Handle empty approx_top_k combine buffers
cloud-fan Jul 20, 2026
41eecba
[SPARK-58069][SQL][TEST] Cover empty combine partitions
cloud-fan Jul 22, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
31 changes: 25 additions & 6 deletions ...main/scala/org/apache/spark/sql/catalyst/expressions/aggregate/ApproxTopKAggregates.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -697,9 +697,20 @@ class CombineInternal[T](
* boundaries (SPARK-58069).
*/
def serialize(): Array[Byte] = {
val sketchWithNullCountBytes = sketchWithNullCount.serialize(
ApproxTopK.genSketchSerDe(itemDataType).asInstanceOf[ArrayOfItemsSerDe[T]])
val typeBytes: Array[Byte] = itemDataType.json.getBytes(StandardCharsets.UTF_8)
// An empty partition has a placeholder buffer whose itemDataType has not been initialized.
// It can still be serialized between aggregation stages, so use a default serde for its empty
// sketch and encode the missing type as a zero-length section.
val serDe: ArrayOfItemsSerDe[T] = if (itemDataType == null) {
new ArrayOfStringsSerDe().asInstanceOf[ArrayOfItemsSerDe[T]]
} else {
ApproxTopK.genSketchSerDe(itemDataType).asInstanceOf[ArrayOfItemsSerDe[T]]
}
val sketchWithNullCountBytes = sketchWithNullCount.serialize(serDe)
val typeBytes: Array[Byte] = if (itemDataType == null) {
Array.emptyByteArray
} else {
itemDataType.json.getBytes(StandardCharsets.UTF_8)
}
val byteArray = new Array[Byte](
sketchWithNullCountBytes.length + Integer.BYTES + Integer.BYTES + typeBytes.length)

Expand Down Expand Up @@ -729,12 +740,20 @@ object CombineInternal {
val typeLength = byteBuffer.getInt
val typeBytes = new Array[Byte](typeLength)
byteBuffer.get(typeBytes)
val itemDataType = DataType.fromJson(new String(typeBytes, StandardCharsets.UTF_8))
val itemDataType = if (typeLength == 0) {
null
} else {
DataType.fromJson(new String(typeBytes, StandardCharsets.UTF_8))
}
// read sketchBytes
val sketchBytes = new Array[Byte](buffer.length - Integer.BYTES - Integer.BYTES - typeLength)
byteBuffer.get(sketchBytes)
val sketchWithNullCount = ApproxTopKAggregateBuffer.deserialize(
sketchBytes, ApproxTopK.genSketchSerDe(itemDataType))
val serDe = if (itemDataType == null) {
new ArrayOfStringsSerDe().asInstanceOf[ArrayOfItemsSerDe[Any]]
} else {
ApproxTopK.genSketchSerDe(itemDataType)
}
val sketchWithNullCount = ApproxTopKAggregateBuffer.deserialize(sketchBytes, serDe)
new CombineInternal[Any](sketchWithNullCount, itemDataType, maxItemsTracked)
}
}
Expand Down
31 changes: 31 additions & 0 deletions sql/core/src/test/scala/org/apache/spark/sql/ApproxTopKSuite.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -20,8 +20,12 @@ package org.apache.spark.sql
import java.sql.{Date, Timestamp}
import java.time.{LocalDateTime, LocalTime}

import org.apache.datasketches.frequencies.ItemsSketch

import org.apache.spark.{SparkArithmeticException, SparkRuntimeException}
import org.apache.spark.sql.catalyst.ExtendedAnalysisException
import org.apache.spark.sql.catalyst.expressions.aggregate.{ApproxTopK, ApproxTopKAggregateBuffer,
CombineInternal}
import org.apache.spark.sql.errors.DataTypeErrors.toSQLType
import org.apache.spark.sql.test.SharedSparkSession
import org.apache.spark.sql.types.{BooleanType, ByteType, DataType, DateType, DecimalType, DoubleType, FloatType, IntegerType, LongType, ShortType, StringType, TimestampNTZType, TimestampType, TimeType}
Expand Down Expand Up @@ -575,6 +579,33 @@ class ApproxTopKSuite extends SharedSparkSession {
"CAST('13:00:00.123' AS TIME(3))"))
)

test("SPARK-58069: serialize an empty approx_top_k_combine buffer") {

@uros-b uros-b Jul 21, 2026

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The new regression test is a pure in-JVM round-trip of serialize()/deserialize(); it never drives the reported failure through a real SQL query with a shuffle-forced empty partition (e.g. a REPARTITION or UNION of empty + non-empty approx_top_k_accumulate). It also exercises only the concrete maxItemsTracked = 100 variant, not the VOID_MAX_ITEMS_TRACKED = -1 placeholder that the size-unspecified production path (createAggregationBuffer) actually creates. An end-to-end SQL test (and a VOID variant) would faithfully guard the failure this PR claims to fix.

@cloud-fan cloud-fan Jul 22, 2026

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Addressed in 41eecba. The round-trip test now covers both a concrete maxItemsTracked value and VOID_MAX_ITEMS_TRACKED, and asserts that nullCount is restored. I also added an end-to-end SQL test that repartitions 10 input sketches across 20 partitions before approx_top_k_combine, guaranteeing empty shuffle partitions while exercising the unspecified-size production path.

Seq(100, ApproxTopK.VOID_MAX_ITEMS_TRACKED).foreach { maxItemsTracked =>
val buffer = new CombineInternal[Any](
new ApproxTopKAggregateBuffer[Any](new ItemsSketch[Any](128), 0L),
null,
maxItemsTracked)

val restored = CombineInternal.deserialize(buffer.serialize())

assert(restored.getItemDataType == null)
assert(restored.getMaxItemsTracked == maxItemsTracked)
assert(restored.getSketchWithNullCount.sketch.isEmpty)
assert(restored.getSketchWithNullCount.nullCount == 0L)
}
}

test("SPARK-58069: combine sketches with empty shuffle partitions") {
val sketches = sql(
"SELECT approx_top_k_accumulate(id) AS sketch FROM range(10) GROUP BY id")
.repartition(20)
val combined = sketches.selectExpr("approx_top_k_combine(sketch) AS sketch")

checkAnswer(
combined.selectExpr("inline(approx_top_k_estimate(sketch))"),
(0L until 10L).map(Row(_, 1L)))
}

// positive tests for approx_top_k_combine on every types
gridTest("SPARK-52798: same type, same size, specified combine size - success")(itemsWithTopK) {
case (input, expected) =>
Expand Down