forked from apache/spark
-
Notifications
You must be signed in to change notification settings - Fork 0
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
[SPARK-50513][SS][SQL] Split EncoderImplicits from SQLImplicits and p…
…rovide helper object within StatefulProcessor to access underlying SQL Encoder related implicit functions ### What changes were proposed in this pull request? Split EncoderImplicits from SQLImplicits and provide helper object within StatefulProcessor to access underlying SQL Encoder related implicit functions ### Why are the changes needed? Without this, we cannot handle the implicit encoder APIs on the executor. We would run into a NPE since the spark session is not available on the executors. One option is to pass the `SparkSession` or `SQLImplicits` directly to the `StatefulProcessor`. However, this risks exposing some methods relying on `SparkSession` again. ``` Caused by: Task 11 in stage 4.0 failed 4 times, most recent failure: Lost task 11.3 in stage 4.0 (TID 57) (10.68.181.194 executor 2): java.lang.NullPointerException: Cannot invoke "org.apache.spark.sql.SparkSession.implicits()" because the return value of "$line94e51c7e8a51415d827a613fcad00cc19.$read$$iw$$iw.spark()" is null at $line94e51c7e8a51415d827a613fcad00cc152.$read$$iw$$iw$$iw$$iw$$iw$$iw$$iw$$iw$$iw$$iw$FruitCountStatefulProcessor.init(command-412580759966543:18) at org.apache.spark.sql.execution.streaming.TransformWithStateExec.processDataWithInitialState(TransformWithStateExec.scala:732) at org.apache.spark.sql.execution.streaming.TransformWithStateExec.$anonfun$doExecute$2(TransformWithStateExec.scala:625) ``` Hence, we do 2 things: - split the encoder related functions that don't rely on `SparkSession` into a separate trait - expose a helper `implicits` object within the `StatefulProcessor` interface that users can import directly, providing access to only the necessary functions ### Does this PR introduce _any_ user-facing change? Yes ### How was this patch tested? Added unit tests ``` [info] Run completed in 9 minutes, 30 seconds. [info] Total number of tests run: 128 [info] Suites: completed 1, aborted 0 [info] Tests: succeeded 128, failed 0, canceled 0, ignored 0, pending 0 [info] All tests passed. ``` ### Was this patch authored or co-authored using generative AI tooling? No Closes apache#49099 from anishshri-db/task/SPARK-50513. Authored-by: Anish Shrigondekar <[email protected]> Signed-off-by: Jungtaek Lim <[email protected]>
- Loading branch information
1 parent
02ebf12
commit 03c5799
Showing
5 changed files
with
357 additions
and
119 deletions.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
211 changes: 211 additions & 0 deletions
211
sql/core/src/test/scala/org/apache/spark/sql/streaming/TransformWithStateClusterSuite.scala
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,211 @@ | ||
| /* | ||
| * Licensed to the Apache Software Foundation (ASF) under one or more | ||
| * contributor license agreements. See the NOTICE file distributed with | ||
| * this work for additional information regarding copyright ownership. | ||
| * The ASF licenses this file to You under the Apache License, Version 2.0 | ||
| * (the "License"); you may not use this file except in compliance with | ||
| * the License. You may obtain a copy of the License at | ||
| * | ||
| * http://www.apache.org/licenses/LICENSE-2.0 | ||
| * | ||
| * Unless required by applicable law or agreed to in writing, software | ||
| * distributed under the License is distributed on an "AS IS" BASIS, | ||
| * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. | ||
| * See the License for the specific language governing permissions and | ||
| * limitations under the License. | ||
| */ | ||
|
|
||
| package org.apache.spark.sql.streaming | ||
|
|
||
| import org.apache.spark.{SparkConf, SparkFunSuite} | ||
| import org.apache.spark.sql.{Dataset, Encoders, Row, SparkSession} | ||
| import org.apache.spark.sql.LocalSparkSession.withSparkSession | ||
| import org.apache.spark.sql.execution.streaming.MemoryStream | ||
| import org.apache.spark.sql.execution.streaming.state.RocksDBStateStoreProvider | ||
| import org.apache.spark.sql.internal.SQLConf | ||
|
|
||
| case class FruitState( | ||
| name: String, | ||
| count: Long, | ||
| family: String | ||
| ) | ||
|
|
||
| class FruitCountStatefulProcessor(useImplicits: Boolean) | ||
| extends StatefulProcessor[String, String, (String, Long, String)] { | ||
| import implicits._ | ||
|
|
||
| @transient protected var _fruitState: ValueState[FruitState] = _ | ||
|
|
||
| override def init(outputMode: OutputMode, timeMode: TimeMode): Unit = { | ||
| if (useImplicits) { | ||
| _fruitState = getHandle.getValueState[FruitState]("fruitState", TTLConfig.NONE) | ||
| } else { | ||
| _fruitState = getHandle.getValueState("fruitState", Encoders.product[FruitState], | ||
| TTLConfig.NONE) | ||
| } | ||
| } | ||
|
|
||
| private def getFamily(fruitName: String): String = { | ||
| if (fruitName == "orange" || fruitName == "lemon" || fruitName == "lime") { | ||
| "citrus" | ||
| } else { | ||
| "non-citrus" | ||
| } | ||
| } | ||
|
|
||
| override def handleInputRows(key: String, inputRows: Iterator[String], timerValues: TimerValues): | ||
| Iterator[(String, Long, String)] = { | ||
| val new_cnt = _fruitState.getOption().map(x => x.count).getOrElse(0L) + inputRows.size | ||
| val family = getFamily(key) | ||
| _fruitState.update(FruitState(key, new_cnt, family)) | ||
| Iterator.single((key, new_cnt, family)) | ||
| } | ||
| } | ||
|
|
||
| class FruitCountStatefulProcessorWithInitialState(useImplicits: Boolean) | ||
| extends StatefulProcessorWithInitialState[String, String, (String, Long, String), String] { | ||
| import implicits._ | ||
|
|
||
| @transient protected var _fruitState: ValueState[FruitState] = _ | ||
|
|
||
| override def init(outputMode: OutputMode, timeMode: TimeMode): Unit = { | ||
| if (useImplicits) { | ||
| _fruitState = getHandle.getValueState[FruitState]("fruitState", TTLConfig.NONE) | ||
| } else { | ||
| _fruitState = getHandle.getValueState("fruitState", Encoders.product[FruitState], | ||
| TTLConfig.NONE) | ||
| } | ||
| } | ||
|
|
||
| private def getFamily(fruitName: String): String = { | ||
| if (fruitName == "orange" || fruitName == "lemon" || fruitName == "lime") { | ||
| "citrus" | ||
| } else { | ||
| "non-citrus" | ||
| } | ||
| } | ||
|
|
||
| override def handleInitialState(key: String, initialState: String, | ||
| timerValues: TimerValues): Unit = { | ||
| val new_cnt = _fruitState.getOption().map(x => x.count).getOrElse(0L) + 1 | ||
| val family = getFamily(key) | ||
| _fruitState.update(FruitState(key, new_cnt, family)) | ||
| } | ||
|
|
||
| override def handleInputRows(key: String, inputRows: Iterator[String], timerValues: TimerValues): | ||
| Iterator[(String, Long, String)] = { | ||
| val new_cnt = _fruitState.getOption().map(x => x.count).getOrElse(0L) + inputRows.size | ||
| val family = getFamily(key) | ||
| _fruitState.update(FruitState(key, new_cnt, family)) | ||
| Iterator.single((key, new_cnt, family)) | ||
| } | ||
| } | ||
|
|
||
| trait TransformWithStateClusterSuiteBase extends SparkFunSuite { | ||
| def getSparkConf(): SparkConf = { | ||
| val conf = new SparkConf() | ||
| .setMaster("local-cluster[2, 2, 1024]") | ||
| .set(SQLConf.STATE_STORE_PROVIDER_CLASS.key, | ||
| classOf[RocksDBStateStoreProvider].getCanonicalName) | ||
| .set(SQLConf.SHUFFLE_PARTITIONS.key, | ||
| TransformWithStateSuiteUtils.NUM_SHUFFLE_PARTITIONS.toString) | ||
| .set(SQLConf.STREAMING_STOP_TIMEOUT, 5000L) | ||
| conf | ||
| } | ||
|
|
||
| // Start a new test with cluster containing two executors and streaming stop timeout set to 5s | ||
| val testSparkConf = getSparkConf() | ||
|
|
||
| protected def testWithAndWithoutImplicitEncoders(name: String) | ||
| (func: (SparkSession, Boolean) => Any): Unit = { | ||
| Seq(false, true).foreach { useImplicits => | ||
| test(s"$name - useImplicits = $useImplicits") { | ||
| withSparkSession(SparkSession.builder().config(testSparkConf).getOrCreate()) { spark => | ||
| func(spark, useImplicits) | ||
| } | ||
| } | ||
| } | ||
| } | ||
| } | ||
|
|
||
| /** | ||
| * Test suite spawning local cluster with multiple executors to test serde of stateful | ||
| * processors along with use of implicit encoders, if applicable in transformWithState operator. | ||
| */ | ||
| class TransformWithStateClusterSuite extends StreamTest with TransformWithStateClusterSuiteBase { | ||
| testWithAndWithoutImplicitEncoders("streaming with transformWithState - " + | ||
| "without initial state") { (spark, useImplicits) => | ||
| import spark.implicits._ | ||
| val input = MemoryStream(Encoders.STRING, spark.sqlContext) | ||
| val agg = input.toDS() | ||
| .groupByKey(x => x) | ||
| .transformWithState(new FruitCountStatefulProcessor(useImplicits), | ||
| TimeMode.None(), | ||
| OutputMode.Update() | ||
| ) | ||
|
|
||
| val query = agg.writeStream | ||
| .format("memory") | ||
| .outputMode("update") | ||
| .queryName("output") | ||
| .start() | ||
|
|
||
| input.addData("apple", "apple", "orange", "orange", "orange") | ||
| query.processAllAvailable() | ||
|
|
||
| checkAnswer(spark.sql("select * from output"), | ||
| Seq(Row("apple", 2, "non-citrus"), | ||
| Row("orange", 3, "citrus"))) | ||
|
|
||
| input.addData("lemon", "lime") | ||
| query.processAllAvailable() | ||
| checkAnswer(spark.sql("select * from output"), | ||
| Seq(Row("apple", 2, "non-citrus"), | ||
| Row("orange", 3, "citrus"), | ||
| Row("lemon", 1, "citrus"), | ||
| Row("lime", 1, "citrus"))) | ||
|
|
||
| query.stop() | ||
| } | ||
|
|
||
| testWithAndWithoutImplicitEncoders("streaming with transformWithState - " + | ||
| "with initial state") { (spark, useImplicits) => | ||
| import spark.implicits._ | ||
|
|
||
| val fruitCountInitialDS: Dataset[String] = Seq( | ||
| "apple", "apple", "orange", "orange", "orange").toDS() | ||
|
|
||
| val fruitCountInitial = fruitCountInitialDS | ||
| .groupByKey(x => x) | ||
|
|
||
| val input = MemoryStream(Encoders.STRING, spark.sqlContext) | ||
| val agg = input.toDS() | ||
| .groupByKey(x => x) | ||
| .transformWithState(new FruitCountStatefulProcessorWithInitialState(useImplicits), | ||
| TimeMode.None(), | ||
| OutputMode.Update(), fruitCountInitial) | ||
|
|
||
| val query = agg.writeStream | ||
| .format("memory") | ||
| .outputMode("update") | ||
| .queryName("output") | ||
| .start() | ||
|
|
||
| input.addData("apple", "apple", "orange", "orange", "orange") | ||
| query.processAllAvailable() | ||
|
|
||
| checkAnswer(spark.sql("select * from output"), | ||
| Seq(Row("apple", 4, "non-citrus"), | ||
| Row("orange", 6, "citrus"))) | ||
|
|
||
| input.addData("lemon", "lime") | ||
| query.processAllAvailable() | ||
| checkAnswer(spark.sql("select * from output"), | ||
| Seq(Row("apple", 4, "non-citrus"), | ||
| Row("orange", 6, "citrus"), | ||
| Row("lemon", 1, "citrus"), | ||
| Row("lime", 1, "citrus"))) | ||
|
|
||
| query.stop() | ||
| } | ||
| } |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Oops, something went wrong.