Skip to content

fix: copy rows retained by identity broadcasts #7260

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

Merged
sunchao merged 1 commit into from
Oct 27, 2026
Merged

fix: copy rows retained by identity broadcasts #7260

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
10 changes: 8 additions & 2 deletions spark/src/main/scala/org/apache/spark/sql/comet/CometColumnarToRowExec.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -32,7 +32,7 @@ import org.apache.spark.sql.catalyst.InternalRow
import org.apache.spark.sql.catalyst.expressions.{Attribute, SortOrder}
import org.apache.spark.sql.catalyst.expressions.codegen._
import org.apache.spark.sql.catalyst.expressions.codegen.Block._
import org.apache.spark.sql.catalyst.plans.physical.Partitioning
import org.apache.spark.sql.catalyst.plans.physical.{IdentityBroadcastMode, Partitioning}
import org.apache.spark.sql.comet.util.{Utils => CometUtils}
import org.apache.spark.sql.errors.QueryExecutionErrors
import org.apache.spark.sql.execution.{CodegenSupport, ColumnarToRowTransition, SparkPlan, SQLExecution}
Expand Down Expand Up @@ -137,7 +137,13 @@ case class CometColumnarToRowExec(child: SparkPlan)
numOutputRows)

val mode = cometBroadcastExchange.get.mode
val relation = mode.transform(rows, Some(numOutputRows.value))
// Identity broadcasts retain rows, but both JVM conversion paths reuse the returned row.
// Hashed relations copy row bytes into their own storage while consuming the iterator.
val broadcastRows = mode match {
case IdentityBroadcastMode => rows.map(_.copy())
case _ => rows
}
val relation = mode.transform(broadcastRows, Some(numOutputRows.value))
val broadcasted = sparkContext.broadcastInternal(relation, serializedOnly = true)
val executionId = sparkContext.getLocalProperty(SQLExecution.EXECUTION_ID_KEY)
SQLMetrics.postDriverMetricUpdates(sparkContext, executionId, metrics.values.toSeq)
Expand Down
29 changes: 27 additions & 2 deletions spark/src/test/scala/org/apache/comet/rules/CometExecRuleSuite.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -28,13 +28,13 @@ import org.scalatest.PrivateMethodTester._

import org.apache.logging.log4j.Level
import org.apache.spark.sql._
import org.apache.spark.sql.catalyst.FunctionIdentifier
import org.apache.spark.sql.catalyst.{FunctionIdentifier, InternalRow}
import org.apache.spark.sql.catalyst.expressions.{Alias, Attribute, AttributeReference, Descending, Expression, ExpressionInfo, GreaterThan, In, InSet, KnownFloatingPointNormalized, Literal, Not, RuntimeReplaceable, SortOrder}
import org.apache.spark.sql.catalyst.expressions.aggregate.{AggregateExpression, BloomFilterAggregate, Final, Min, Partial, PartialMerge}
import org.apache.spark.sql.catalyst.expressions.objects.{Invoke, StaticInvoke}
import org.apache.spark.sql.catalyst.optimizer.NormalizeNaNAndZero
import org.apache.spark.sql.catalyst.plans.logical.{LocalRelation, LogicalPlan}
import org.apache.spark.sql.catalyst.plans.physical.{HashPartitioning, RoundRobinPartitioning}
import org.apache.spark.sql.catalyst.plans.physical.{HashPartitioning, IdentityBroadcastMode, RoundRobinPartitioning}
import org.apache.spark.sql.catalyst.rules.Rule
import org.apache.spark.sql.comet._
import org.apache.spark.sql.comet.CometExec.nativeLeaves
Expand Down Expand Up @@ -98,6 +98,31 @@ class CometExecRuleSuite extends CometTestBase {
CometExecRule(spark).apply(stripAQEPlan(plan))
}

test("columnar-to-row identity broadcasts retain every build row") {
for (direct <- Seq(false, true)) {
withSQLConf(
SQLConf.ADAPTIVE_EXECUTION_ENABLED.key -> "false",
SQLConf.WHOLESTAGE_CODEGEN_ENABLED.key -> "false",
CometConf.COMET_EXEC_RANGE_ENABLED.key -> "true",
CometConf.COMET_DIRECT_COLUMNAR_TO_ROW_ENABLED.key -> direct.toString,
CometConf.COMET_DIRECT_COLUMNAR_TO_ROW_MIN_BATCH_SIZE.key -> "1") {
val build = applyCometExecRule(
createSparkPlan(spark, "SELECT CAST(id AS INT) AS y FROM range(2024, 2028, 1, 2)"))
assert(build.isInstanceOf[CometNativeExec], build.treeString)
val mode = IdentityBroadcastMode
val original = BroadcastExchangeExec(mode, build)
val exchange = CometBroadcastExchangeExec(original, build.output, mode, build)
val rows = CometColumnarToRowExec(exchange)
.executeBroadcast[Array[InternalRow]]()
.value
// Two rows per partition expose row reuse in both JVM conversion paths.
withClue(s"direct=$direct: ") {
assert(rows.map(_.getInt(0)).sorted.toSeq == Seq(2024, 2025, 2026, 2027))
}
}
}
}

/** Create a test data frame that is used in all tests */
private def createTestDataFrame = {
val testSchema = new StructType(
Expand Down
Loading