Skip to content

[SPARK-56633][SQL][TESTS] Add comprehensive Parquet vectorized-reader benchmark coverage #55558

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

Draft
LuciferYang wants to merge 2 commits into
base: master
Choose a base branch
from
Draft

[SPARK-56633][SQL][TESTS] Add comprehensive Parquet vectorized-reader benchmark coverage #55558

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
e30ce3e
[SPARK-XXXXX][SQL][TESTS] Add comprehensive Parquet vectorized-reader…
LuciferYang Apr 27, 2026
1fef57f
Benchmark results for org.apache.spark.sql.execution.datasources.parq…
LuciferYang Apr 27, 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
84 changes: 84 additions & 0 deletions sql/core/benchmarks/ParquetVectorUpdaterBenchmark-results.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,84 @@
================================================================================================
Identity Updaters
================================================================================================

OpenJDK 64-Bit Server VM 17.0.19+10-LTS on Linux 6.17.0-1010-azure
AMD EPYC 9V74 80-Core Processor
Identity Updaters: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
BooleanUpdater 0 0 0 14698.7 0.1 1.0X
ByteUpdater (INT32 -> Byte) 0 0 0 3589.1 0.3 0.2X
ShortUpdater (INT32 -> Short) 1 1 0 1824.9 0.5 0.1X
IntegerUpdater 0 0 0 8168.8 0.1 0.6X
LongUpdater 0 0 0 4059.2 0.2 0.3X
FloatUpdater 0 0 0 8130.7 0.1 0.6X
DoubleUpdater 0 0 0 4062.2 0.2 0.3X
BinaryUpdater 18 19 0 57.3 17.5 0.0X


================================================================================================
Type-converting Updaters
================================================================================================

OpenJDK 64-Bit Server VM 17.0.19+10-LTS on Linux 6.17.0-1010-azure
AMD EPYC 9V74 80-Core Processor
Type-converting Updaters: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
---------------------------------------------------------------------------------------------------------------------------
IntegerToLongUpdater 2 2 0 446.0 2.2 1.0X
IntegerToDoubleUpdater 2 2 0 456.5 2.2 1.0X
FloatToDoubleUpdater 2 2 0 474.8 2.1 1.1X
DateToTimestampNTZUpdater 41 41 1 25.5 39.2 0.1X
DowncastLongUpdater (INT64 -> Decimal(9,2)) 2 2 0 472.5 2.1 1.1X


================================================================================================
Rebase Updaters
================================================================================================

OpenJDK 64-Bit Server VM 17.0.19+10-LTS on Linux 6.17.0-1010-azure
AMD EPYC 9V74 80-Core Processor
Rebase Updaters: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
-------------------------------------------------------------------------------------------------------------------------------
IntegerWithRebaseUpdater (DATE legacy) 0 0 0 2174.1 0.5 1.0X
LongWithRebaseUpdater (TIMESTAMP_MICROS legacy) 1 1 0 1733.7 0.6 0.8X
LongAsMicrosUpdater (TIMESTAMP_MILLIS) 2 3 0 420.7 2.4 0.2X


================================================================================================
Unsigned Updaters
================================================================================================

OpenJDK 64-Bit Server VM 17.0.19+10-LTS on Linux 6.17.0-1010-azure
AMD EPYC 9V74 80-Core Processor
Unsigned Updaters: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
-----------------------------------------------------------------------------------------------------------------------------
UnsignedIntegerUpdater (UINT32 -> Long) 1 1 0 964.4 1.0 1.0X
UnsignedLongUpdater (UINT64 -> Decimal(20,0)) 18 18 0 58.6 17.1 0.1X


================================================================================================
Decimal Updaters
================================================================================================

OpenJDK 64-Bit Server VM 17.0.19+10-LTS on Linux 6.17.0-1010-azure
AMD EPYC 9V74 80-Core Processor
Decimal Updaters: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
IntegerToDecimalUpdater 0 0 0 8133.9 0.1 1.0X
LongToDecimalUpdater 0 0 0 4047.6 0.2 0.5X
FixedLenByteArrayToDecimalUpdater 24 24 0 43.8 22.8 0.0X


================================================================================================
FixedLenByteArray Updaters
================================================================================================

OpenJDK 64-Bit Server VM 17.0.19+10-LTS on Linux 6.17.0-1010-azure
AMD EPYC 9V74 80-Core Processor
FixedLenByteArray Updaters: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
---------------------------------------------------------------------------------------------------------------------------------------
FixedLenByteArrayUpdater (len=16 -> Binary) 22 23 0 46.9 21.3 1.0X
FixedLenByteArrayAsIntUpdater (len=4 -> Decimal(9,2)) 6 6 0 166.3 6.0 3.5X
FixedLenByteArrayAsLongUpdater (len=8 -> Decimal(18,4)) 8 8 0 125.1 8.0 2.7X


58 changes: 52 additions & 6 deletions .../parquet/ParquetReadStateTestAccess.scala → ...tasources/parquet/ParquetTestAccess.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -18,24 +18,35 @@
package org.apache.spark.sql.execution.datasources.parquet

import java.lang.reflect.{InvocationTargetException, Method}
import java.time.ZoneId
import java.util.PrimitiveIterator

import org.apache.parquet.column.ColumnDescriptor
import org.apache.parquet.schema.LogicalTypeAnnotation

import org.apache.spark.sql.execution.vectorized.WritableColumnVector
import org.apache.spark.util.SparkClassUtils

/**
* Reflective bridge to the package-private `ParquetReadState`. Under `spark-submit --jars`,
* test and main classes load from different classloaders, blocking package-private access.
* Reflection with `setAccessible` sidesteps the check without widening production visibility.
* Reflective bridge to package-private classes in
* `org.apache.spark.sql.execution.datasources.parquet`. Under `spark-submit --jars`, test
* and main classes load from different classloaders, blocking package-private access despite
* the matching package name. Reflection with `setAccessible(true)` sidesteps the check
* without widening production visibility.
*
* Currently bridges:
* - `ParquetReadState` (constructor + `resetForNewBatch` + `resetForNewPage`)
* - `VectorizedRleValuesReader.readBatch` (5-arg overload not exposed publicly)
* - `ParquetVectorUpdaterFactory` (constructor)
*/
object ParquetReadStateTestAccess {
object ParquetTestAccess {

// -------- ParquetReadState --------

private val stateCls = SparkClassUtils.classForName[Any](
"org.apache.spark.sql.execution.datasources.parquet.ParquetReadState")

private val ctor = {
private val stateCtor = {
val c = stateCls.getDeclaredConstructor(
classOf[ColumnDescriptor],
java.lang.Boolean.TYPE,
Expand Down Expand Up @@ -71,7 +82,7 @@ object ParquetReadStateTestAccess {
isRequired: Boolean,
rowIndexes: PrimitiveIterator.OfLong = null): AnyRef = {
try {
ctor.newInstance(
stateCtor.newInstance(
descriptor,
Boolean.box(isRequired),
rowIndexes).asInstanceOf[AnyRef]
Expand Down Expand Up @@ -105,6 +116,41 @@ object ParquetReadStateTestAccess {
reader, state, values, defLevels, valueReader, updater)
} catch { case e: ReflectiveOperationException => throw rethrow(e) }

// -------- ParquetVectorUpdaterFactory --------

private val factoryCtor = {
val cls = SparkClassUtils.classForName[Any](
"org.apache.spark.sql.execution.datasources.parquet.ParquetVectorUpdaterFactory")
val c = cls.getDeclaredConstructor(
classOf[LogicalTypeAnnotation],
classOf[ZoneId],
classOf[String],
classOf[String],
classOf[String],
classOf[String])
c.setAccessible(true)
c
}

def newFactory(
logicalTypeAnnotation: LogicalTypeAnnotation,
convertTz: ZoneId,
datetimeRebaseMode: String,
datetimeRebaseTz: String,
int96RebaseMode: String,
int96RebaseTz: String): ParquetVectorUpdaterFactory = {
try {
factoryCtor.newInstance(
logicalTypeAnnotation, convertTz,
datetimeRebaseMode, datetimeRebaseTz,
int96RebaseMode, int96RebaseTz).asInstanceOf[ParquetVectorUpdaterFactory]
} catch {
case e: ReflectiveOperationException => throw rethrow(e)
}
}

// -------- shared helper --------

private def rethrow(e: ReflectiveOperationException): RuntimeException = {
val cause = e match {
case ite: InvocationTargetException => ite.getCause
Expand Down
Loading