Pull request #4: Spark 3.5 support

Merge in PDS-GTB/elasticsearch-hadoop from spark-35 to db-feature/spark-35

* commit 'fc4f33b6c3d609f18b820b25cc1435a2c4c5ead8':
  Spark 3.5 support

Author: nikolay-voskresenskiy-db

buildSrc/src/main/groovy/org/elasticsearch/hadoop/gradle/BuildPlugin.groovy

@@ -594,7 +594,6 @@ class BuildPlugin implements Plugin<Project> {
 
             // Main variant needs the least configuration on its own, since it is the default publication created above.
             sparkVariants.defaultVariant { SparkVariant variant ->
-                project.publishing.publications.main.setAlias(true)
                 updateVariantArtifactId(project, project.publishing.publications.main, variant)
             }
 
@@ -651,6 +650,7 @@ class BuildPlugin implements Plugin<Project> {
                         }
                         configurePom(project, variantPublication)
                         updateVariantArtifactId(project, variantPublication, variant)
+                        variantPublication.setAlias(true)
                     }
                 }
                 if (signingKey.isPresent()) {

buildSrc/src/main/java/org/elasticsearch/hadoop/gradle/scala/SparkVariantPlugin.java

@@ -72,6 +72,8 @@
 
 public class SparkVariantPlugin implements Plugin<Project> {
 
+    public static final String ITEST_SOURCE_SET_NAME = "itest";
+
     public static class SparkVariant {
 
         private final CharSequence name;
@@ -311,6 +313,23 @@ private static void configureDefaultVariant(Project project, SparkVariant sparkV
         runtimeElements.getOutgoing().capability(capability);
 
         configureScalaJarClassifiers(project, sparkVariant);
+        // Extend main and test source set for the main variant - this enables the possibility of having diverging code between variants
+        SourceSetContainer sourceSets = javaPluginExtension.getSourceSets();
+        ScalaSourceDirectorySet scalaSourceSet = getScalaSourceSet(sourceSets.getByName(MAIN_SOURCE_SET_NAME));
+        scalaSourceSet.setSrcDirs(Arrays.asList(
+                "src/" + MAIN_SOURCE_SET_NAME + "/scala",
+                "src/" + MAIN_SOURCE_SET_NAME + "/" + sparkVariant.getName()
+        ));
+        ScalaSourceDirectorySet scalaTestSourceSet = getScalaSourceSet(sourceSets.getByName(TEST_SOURCE_SET_NAME));
+        scalaTestSourceSet.setSrcDirs(Arrays.asList(
+                "src/" + TEST_SOURCE_SET_NAME + "/scala",
+                "src/" + TEST_SOURCE_SET_NAME + "/" + sparkVariant.getName()
+        ));
+        ScalaSourceDirectorySet scalaITestSourceSet = getScalaSourceSet(sourceSets.getByName(ITEST_SOURCE_SET_NAME));
+        scalaITestSourceSet.setSrcDirs(Arrays.asList(
+                "src/" + ITEST_SOURCE_SET_NAME + "/scala",
+                "src/" + ITEST_SOURCE_SET_NAME + "/" + sparkVariant.getName()
+        ));
     }
 
     private static void configureVariant(Project project, SparkVariant sparkVariant, JavaPluginExtension javaPluginExtension) {

gradle.properties

@@ -30,6 +30,7 @@ spark20Version = 2.3.0
 spark22Version = 2.2.3
 spark24Version = 2.4.4
 spark30Version = 3.4.3
+spark35Version = 3.5.6
 
 # same as Spark's
 scala210Version = 2.10.7

spark/core/build.gradle

@@ -14,6 +14,8 @@ sparkVariants {
     // We should maybe move these to a separate config file that can be read from both this file and the pipeline script in the future if it creates issues
     setCoreDefaultVariant "spark30scala213", spark30Version, scala213Version
     addCoreFeatureVariant "spark30scala212", spark30Version, scala212Version
+    addCoreFeatureVariant "spark35scala212", spark35Version, scala212Version
+    addCoreFeatureVariant "spark35scala213", spark35Version, scala213Version
 
     all { SparkVariantPlugin.SparkVariant variant ->
 
@@ -44,6 +46,11 @@ sparkVariants {
             add(variant.configuration('api'), "org.apache.spark:spark-core_${variant.scalaMajorVersion}:${variant.sparkVersion}") {
                 exclude group: 'org.apache.hadoop'
             }
+            if (variant.sparkVersion >= "3.5.0") {
+                add(variant.configuration('implementation'), "org.apache.spark:spark-common-utils_${variant.scalaMajorVersion}:$variant.sparkVersion") {
+                    exclude group: 'org.apache.hadoop'
+                }
+            }
 
             add(variant.configuration('implementation'), project(":elasticsearch-hadoop-mr"))
             add(variant.configuration('implementation'), "commons-logging:commons-logging:1.1.1")
@@ -126,3 +133,23 @@ tasks.withType(ScalaCompile) { ScalaCompile task ->
     task.targetCompatibility = project.ext.minimumRuntimeVersion
     task.options.forkOptions.executable = new File(project.ext.runtimeJavaHome, 'bin/java').absolutePath
 }
+
+tasks.register('copyPoms', Copy) {
+    from(tasks.named('generatePomFileForMainPublication')) {
+        rename 'pom-default.xml', "elasticsearch-spark-30_2.13-${project.getVersion()}.pom"
+    }
+    from(tasks.named('generatePomFileForSpark30scala212Publication')) {
+        rename 'pom-default.xml', "elasticsearch-spark-30_2.12-${project.getVersion()}.pom"
+    }
+    from(tasks.named('generatePomFileForSpark35scala212Publication')) {
+        rename 'pom-default.xml', "elasticsearch-spark-35_2.12-${project.getVersion()}.pom"
+    }
+    from(tasks.named('generatePomFileForSpark35scala213Publication')) {
+        rename 'pom-default.xml', "elasticsearch-spark-35_2.13-${project.getVersion()}.pom"
+    }
+    into(new File(project.buildDir, 'distributions'))
+}
+
+tasks.named('distribution').configure {
+    dependsOn 'copyPoms'
+}

spark/sql-30/build.gradle

@@ -9,8 +9,10 @@ apply plugin: 'spark.variants'
 
 sparkVariants {
     capabilityGroup 'org.elasticsearch.spark.sql.variant'
-    setDefaultVariant "spark30scala213", spark30Version, scala213Version
-    addFeatureVariant "spark30scala212", spark30Version, scala212Version
+    setCoreDefaultVariant "spark30scala213", spark30Version, scala213Version
+    addCoreFeatureVariant "spark30scala212", spark30Version, scala212Version
+    addCoreFeatureVariant "spark35scala213", spark35Version, scala213Version
+    addCoreFeatureVariant "spark35scala212", spark35Version, scala212Version
 
     all { SparkVariantPlugin.SparkVariant variant ->
         String scalaCompileTaskName = project.sourceSets
@@ -58,6 +60,14 @@ sparkVariants {
                 exclude group: 'javax.servlet'
                 exclude group: 'org.apache.hadoop'
             }
+            if (variant.sparkVersion >= "3.5.0") {
+                add(variant.configuration('implementation'), "org.apache.spark:spark-common-utils_${variant.scalaMajorVersion}:$variant.sparkVersion") {
+                    exclude group: 'org.apache.hadoop'
+                }
+                add(variant.configuration('implementation'), "org.apache.spark:spark-sql-api_${variant.scalaMajorVersion}:$variant.sparkVersion") {
+                    exclude group: 'org.apache.hadoop'
+                }
+            }
 
             add(variant.configuration('implementation'), "org.apache.spark:spark-sql_${variant.scalaMajorVersion}:$variant.sparkVersion") {
                 exclude group: 'org.apache.hadoop'
@@ -198,6 +208,12 @@ tasks.register('copyPoms', Copy) {
     from(tasks.named('generatePomFileForSpark30scala212Publication')) {
         rename 'pom-default.xml', "elasticsearch-spark-30_2.12-${project.getVersion()}.pom"
     }
+    from(tasks.named('generatePomFileForSpark35scala212Publication')) {
+        rename 'pom-default.xml', "elasticsearch-spark-35_2.12-${project.getVersion()}.pom"
+    }
+    from(tasks.named('generatePomFileForSpark35scala213Publication')) {
+        rename 'pom-default.xml', "elasticsearch-spark-35_2.13-${project.getVersion()}.pom"
+    }
     into(new File(project.buildDir, 'distributions'))
 }
 

spark/sql-30/src/itest/scala/org/elasticsearch/spark/sql/streaming/StreamingQueryLifecycleListener.scala

@@ -0,0 +1,127 @@
+package org.elasticsearch.spark.sql.streaming
+
+import org.apache.spark.sql.streaming.StreamingQueryListener
+import org.elasticsearch.hadoop.util.unit.TimeValue
+import org.junit.Assert
+
+import java.util.UUID
+import java.util.concurrent.{CountDownLatch, TimeUnit}
+
+// Listener to 1) ensure no more than a single stream is running at a time, 2) know when we're done processing inputs
+// and 3) to capture any Exceptions encountered during the execution of the stream.
+class StreamingQueryLifecycleListener extends StreamingQueryListener {
+
+  private var uuid: Option[UUID] = None
+
+  private var inputsRequired = 0L
+  private var inputsSeen = 0L
+
+  private var expectingToThrow: Option[Class[_]] = None
+  private var foundExpectedException: Boolean = false
+  private var encounteredException: Option[String] = None
+
+  private var latch = new CountDownLatch(1) // expects just a single batch
+
+  def incrementExpected(): Unit = inputsRequired = inputsRequired + 1
+
+  def setExpectedException(clazz: Class[_]): Unit = {
+    expectingToThrow match {
+      case Some(cls) => throw new IllegalArgumentException(s"Already expecting exception of type [$cls]!")
+      case None => expectingToThrow = Some(clazz)
+    }
+  }
+
+  // Make sure we only ever watch one query at a time.
+  private def captureQueryID(eventId: UUID): Unit = {
+    uuid match {
+      case Some(id) if eventId != id => throw new IllegalStateException("Multiple queries are not supported")
+      case None => uuid = Some(eventId)
+      case _ => // No problem for now
+    }
+  }
+
+  override def onQueryStarted(event: StreamingQueryListener.QueryStartedEvent): Unit = {
+    captureQueryID(event.id)
+  }
+
+  override def onQueryProgress(event: StreamingQueryListener.QueryProgressEvent): Unit = {
+    captureQueryID(event.progress.id)
+
+    // keep track of the number of input rows seen. When we reach the number of expected inputs,
+    // wait for two 0 values to pass before signaling to close
+
+    val rows = event.progress.numInputRows
+    inputsSeen = inputsSeen + rows
+
+    if (inputsSeen == inputsRequired) {
+      if (rows == 0) {
+        // Don't close after meeting the first input level. Wait to make sure we get
+        // one last pass with no new rows processed before signalling.
+        latch.countDown()
+      }
+    } else if (inputsSeen > inputsRequired) {
+      throw new IllegalStateException("Too many inputs encountered. Expected [" + inputsRequired +
+        "] but found [" + inputsSeen + "]")
+    }
+  }
+
+  protected def onQueryIdle(eventId: UUID): Unit = {
+    captureQueryID(eventId)
+    if (inputsSeen == inputsRequired) {
+      latch.countDown()
+    }
+  }
+
+  override def onQueryTerminated(event: StreamingQueryListener.QueryTerminatedEvent): Unit = {
+    try {
+      captureQueryID(event.id)
+
+      encounteredException = event.exception match {
+        case Some(value) =>
+          // This is a whole trace, get everything after the enclosing SparkException (i.e. the original exception + trace)
+          val messageParts = value.split("\\): ")
+          if (messageParts.size > 1) {
+            val nonSparkMessage = messageParts(1)
+            // Ditch the full trace after the first newline
+            val removedNewLine = nonSparkMessage.substring(0, nonSparkMessage.indexOf("\n"))
+            // Split the class name from the exception message and take the class name
+            Some(removedNewLine.substring(0, removedNewLine.indexOf(":")))
+          } else {
+            // Return the original framework error
+            Some(value.substring(0, value.indexOf(":")))
+          }
+        case None => None
+      }
+
+      val expectedExceptionName = expectingToThrow.map(_.getCanonicalName).getOrElse("None")
+
+      foundExpectedException = encounteredException.exists(_.equals(expectedExceptionName))
+    } finally {
+      // signal no matter what to avoid deadlock
+      latch.countDown()
+    }
+  }
+
+  def waitOnComplete(timeValue: TimeValue): Boolean = latch.await(timeValue.millis, TimeUnit.MILLISECONDS)
+
+  def expectAnotherBatch(): Unit = {
+    latch = new CountDownLatch(1)
+  }
+
+  def assertExpectedExceptions(message: Option[String]): Unit = {
+    expectingToThrow match {
+      case Some(exceptionClass) =>
+        if (!foundExpectedException) {
+          encounteredException match {
+            case Some(s) => Assert.fail(s"Expected ${exceptionClass.getCanonicalName} but got $s")
+            case None => Assert.fail(message.getOrElse(s"Expected ${exceptionClass.getCanonicalName} but no Exceptions were thrown"))
+          }
+        }
+      case None =>
+        encounteredException match {
+          case Some(exception) => Assert.fail(s"Expected no exceptions but got $exception")
+          case None => ()
+        }
+    }
+  }
+}