Merge remote-tracking branch 'upstream/master' into dt-opt1

Author: jkbradley

.rat-excludes

@@ -25,6 +25,7 @@ log4j-defaults.properties
 bootstrap-tooltip.js
 jquery-1.11.1.min.js
 sorttable.js
+.*avsc
 .*txt
 .*json
 .*data

core/src/main/scala/org/apache/spark/InterruptibleIterator.scala

@@ -33,7 +33,7 @@ class InterruptibleIterator[+T](val context: TaskContext, val delegate: Iterator
     // is allowed. The assumption is that Thread.interrupted does not have a memory fence in read
     // (just a volatile field in C), while context.interrupted is a volatile in the JVM, which
     // introduces an expensive read fence.
-    if (context.interrupted) {
+    if (context.isInterrupted) {
       throw new TaskKilledException
     } else {
       delegate.hasNext

core/src/main/scala/org/apache/spark/TaskContext.scala

@@ -21,10 +21,18 @@ import scala.collection.mutable.ArrayBuffer
 
 import org.apache.spark.annotation.DeveloperApi
 import org.apache.spark.executor.TaskMetrics
+import org.apache.spark.util.TaskCompletionListener
+
 
 /**
  * :: DeveloperApi ::
  * Contextual information about a task which can be read or mutated during execution.
+ *
+ * @param stageId stage id
+ * @param partitionId index of the partition
+ * @param attemptId the number of attempts to execute this task
+ * @param runningLocally whether the task is running locally in the driver JVM
+ * @param taskMetrics performance metrics of the task
  */
 @DeveloperApi
 class TaskContext(
@@ -39,27 +47,68 @@ class TaskContext(
   def splitId = partitionId
 
   // List of callback functions to execute when the task completes.
-  @transient private val onCompleteCallbacks = new ArrayBuffer[() => Unit]
+  @transient private val onCompleteCallbacks = new ArrayBuffer[TaskCompletionListener]
 
   // Whether the corresponding task has been killed.
-  @volatile var interrupted: Boolean = false
+  @volatile private var interrupted: Boolean = false
+
+  // Whether the task has completed.
+  @volatile private var completed: Boolean = false
+
+  /** Checks whether the task has completed. */
+  def isCompleted: Boolean = completed
 
-  // Whether the task has completed, before the onCompleteCallbacks are executed.
-  @volatile var completed: Boolean = false
+  /** Checks whether the task has been killed. */
+  def isInterrupted: Boolean = interrupted
+
+  // TODO: Also track whether the task has completed successfully or with exception.
+
+  /**
+   * Add a (Java friendly) listener to be executed on task completion.
+   * This will be called in all situation - success, failure, or cancellation.
+   *
+   * An example use is for HadoopRDD to register a callback to close the input stream.
+   */
+  def addTaskCompletionListener(listener: TaskCompletionListener): this.type = {
+    onCompleteCallbacks += listener
+    this
+  }
+
+  /**
+   * Add a listener in the form of a Scala closure to be executed on task completion.
+   * This will be called in all situation - success, failure, or cancellation.
+   *
+   * An example use is for HadoopRDD to register a callback to close the input stream.
+   */
+  def addTaskCompletionListener(f: TaskContext => Unit): this.type = {
+    onCompleteCallbacks += new TaskCompletionListener {
+      override def onTaskCompletion(context: TaskContext): Unit = f(context)
+    }
+    this
+  }
 
   /**
    * Add a callback function to be executed on task completion. An example use
    * is for HadoopRDD to register a callback to close the input stream.
    * Will be called in any situation - success, failure, or cancellation.
    * @param f Callback function.
    */
+  @deprecated("use addTaskCompletionListener", "1.1.0")
   def addOnCompleteCallback(f: () => Unit) {
-    onCompleteCallbacks += f
+    onCompleteCallbacks += new TaskCompletionListener {
+      override def onTaskCompletion(context: TaskContext): Unit = f()
+    }
   }
 
-  def executeOnCompleteCallbacks() {
+  /** Marks the task as completed and triggers the listeners. */
+  private[spark] def markTaskCompleted(): Unit = {
     completed = true
     // Process complete callbacks in the reverse order of registration
-    onCompleteCallbacks.reverse.foreach { _() }
+    onCompleteCallbacks.reverse.foreach { _.onTaskCompletion(this) }
+  }
+
+  /** Marks the task for interruption, i.e. cancellation. */
+  private[spark] def markInterrupted(): Unit = {
+    interrupted = true
   }
 }

core/src/main/scala/org/apache/spark/api/python/PythonHadoopUtil.scala

@@ -19,6 +19,7 @@ package org.apache.spark.api.python
 
 import org.apache.spark.broadcast.Broadcast
 import org.apache.spark.rdd.RDD
+import org.apache.spark.util.Utils
 import org.apache.spark.{Logging, SerializableWritable, SparkException}
 import org.apache.hadoop.conf.Configuration
 import org.apache.hadoop.io._
@@ -42,7 +43,7 @@ private[python] object Converter extends Logging {
                   defaultConverter: Converter[Any, Any]): Converter[Any, Any] = {
     converterClass.map { cc =>
       Try {
-        val c = Class.forName(cc).newInstance().asInstanceOf[Converter[Any, Any]]
+        val c = Utils.classForName(cc).newInstance().asInstanceOf[Converter[Any, Any]]
         logInfo(s"Loaded converter: $cc")
         c
       } match {

core/src/main/scala/org/apache/spark/api/python/PythonRDD.scala

@@ -68,7 +68,7 @@ private[spark] class PythonRDD(
     // Start a thread to feed the process input from our parent's iterator
     val writerThread = new WriterThread(env, worker, split, context)
 
-    context.addOnCompleteCallback { () =>
+    context.addTaskCompletionListener { context =>
       writerThread.shutdownOnTaskCompletion()
 
       // Cleanup the worker socket. This will also cause the Python worker to exit.
@@ -137,7 +137,7 @@ private[spark] class PythonRDD(
           }
         } catch {
 
-          case e: Exception if context.interrupted =>
+          case e: Exception if context.isInterrupted =>
             logDebug("Exception thrown after task interruption", e)
             throw new TaskKilledException
 
@@ -176,7 +176,7 @@ private[spark] class PythonRDD(
 
     /** Terminates the writer thread, ignoring any exceptions that may occur due to cleanup. */
     def shutdownOnTaskCompletion() {
-      assert(context.completed)
+      assert(context.isCompleted)
       this.interrupt()
     }
 
@@ -209,7 +209,7 @@ private[spark] class PythonRDD(
         PythonRDD.writeIteratorToStream(parent.iterator(split, context), dataOut)
         dataOut.flush()
       } catch {
-        case e: Exception if context.completed || context.interrupted =>
+        case e: Exception if context.isCompleted || context.isInterrupted =>
           logDebug("Exception thrown after task completion (likely due to cleanup)", e)
 
         case e: Exception =>
@@ -235,10 +235,10 @@ private[spark] class PythonRDD(
     override def run() {
       // Kill the worker if it is interrupted, checking until task completion.
       // TODO: This has a race condition if interruption occurs, as completed may still become true.
-      while (!context.interrupted && !context.completed) {
+      while (!context.isInterrupted && !context.isCompleted) {
         Thread.sleep(2000)
       }
-      if (!context.completed) {
+      if (!context.isCompleted) {
         try {
           logWarning("Incomplete task interrupted: Attempting to kill Python Worker")
           env.destroyPythonWorker(pythonExec, envVars.toMap, worker)
@@ -372,8 +372,8 @@ private[spark] object PythonRDD extends Logging {
       batchSize: Int) = {
     val keyClass = Option(keyClassMaybeNull).getOrElse("org.apache.hadoop.io.Text")
     val valueClass = Option(valueClassMaybeNull).getOrElse("org.apache.hadoop.io.Text")
-    val kc = Class.forName(keyClass).asInstanceOf[Class[K]]
-    val vc = Class.forName(valueClass).asInstanceOf[Class[V]]
+    val kc = Utils.classForName(keyClass).asInstanceOf[Class[K]]
+    val vc = Utils.classForName(valueClass).asInstanceOf[Class[V]]
     val rdd = sc.sc.sequenceFile[K, V](path, kc, vc, minSplits)
     val confBroadcasted = sc.sc.broadcast(new SerializableWritable(sc.hadoopConfiguration()))
     val converted = convertRDD(rdd, keyConverterClass, valueConverterClass,
@@ -440,9 +440,9 @@ private[spark] object PythonRDD extends Logging {
       keyClass: String,
       valueClass: String,
       conf: Configuration) = {
-    val kc = Class.forName(keyClass).asInstanceOf[Class[K]]
-    val vc = Class.forName(valueClass).asInstanceOf[Class[V]]
-    val fc = Class.forName(inputFormatClass).asInstanceOf[Class[F]]
+    val kc = Utils.classForName(keyClass).asInstanceOf[Class[K]]
+    val vc = Utils.classForName(valueClass).asInstanceOf[Class[V]]
+    val fc = Utils.classForName(inputFormatClass).asInstanceOf[Class[F]]
     if (path.isDefined) {
       sc.sc.newAPIHadoopFile[K, V, F](path.get, fc, kc, vc, conf)
     } else {
@@ -509,9 +509,9 @@ private[spark] object PythonRDD extends Logging {
       keyClass: String,
       valueClass: String,
       conf: Configuration) = {
-    val kc = Class.forName(keyClass).asInstanceOf[Class[K]]
-    val vc = Class.forName(valueClass).asInstanceOf[Class[V]]
-    val fc = Class.forName(inputFormatClass).asInstanceOf[Class[F]]
+    val kc = Utils.classForName(keyClass).asInstanceOf[Class[K]]
+    val vc = Utils.classForName(valueClass).asInstanceOf[Class[V]]
+    val fc = Utils.classForName(inputFormatClass).asInstanceOf[Class[F]]
     if (path.isDefined) {
       sc.sc.hadoopFile(path.get, fc, kc, vc)
     } else {
@@ -558,7 +558,7 @@ private[spark] object PythonRDD extends Logging {
     for {
       k <- Option(keyClass)
       v <- Option(valueClass)
-    } yield (Class.forName(k), Class.forName(v))
+    } yield (Utils.classForName(k), Utils.classForName(v))
   }
 
   private def getKeyValueConverters(keyConverterClass: String, valueConverterClass: String,
@@ -621,10 +621,10 @@ private[spark] object PythonRDD extends Logging {
     val (kc, vc) = getKeyValueTypes(keyClass, valueClass).getOrElse(
       inferKeyValueTypes(rdd, keyConverterClass, valueConverterClass))
     val mergedConf = getMergedConf(confAsMap, pyRDD.context.hadoopConfiguration)
-    val codec = Option(compressionCodecClass).map(Class.forName(_).asInstanceOf[Class[C]])
+    val codec = Option(compressionCodecClass).map(Utils.classForName(_).asInstanceOf[Class[C]])
     val converted = convertRDD(rdd, keyConverterClass, valueConverterClass,
       new JavaToWritableConverter)
-    val fc = Class.forName(outputFormatClass).asInstanceOf[Class[F]]
+    val fc = Utils.classForName(outputFormatClass).asInstanceOf[Class[F]]
     converted.saveAsHadoopFile(path, kc, vc, fc, new JobConf(mergedConf), codec=codec)
   }
 
@@ -653,7 +653,7 @@ private[spark] object PythonRDD extends Logging {
     val mergedConf = getMergedConf(confAsMap, pyRDD.context.hadoopConfiguration)
     val converted = convertRDD(rdd, keyConverterClass, valueConverterClass,
       new JavaToWritableConverter)
-    val fc = Class.forName(outputFormatClass).asInstanceOf[Class[F]]
+    val fc = Utils.classForName(outputFormatClass).asInstanceOf[Class[F]]
     converted.saveAsNewAPIHadoopFile(path, kc, vc, fc, mergedConf)
   }
 

core/src/main/scala/org/apache/spark/deploy/master/ApplicationInfo.scala

@@ -46,6 +46,11 @@ private[spark] class ApplicationInfo(
 
   init()
 
+  private def readObject(in: java.io.ObjectInputStream): Unit = {
+    in.defaultReadObject()
+    init()
+  }
+
   private def init() {
     state = ApplicationState.WAITING
     executors = new mutable.HashMap[Int, ExecutorInfo]

core/src/main/scala/org/apache/spark/rdd/CheckpointRDD.scala

@@ -141,7 +141,7 @@ private[spark] object CheckpointRDD extends Logging {
     val deserializeStream = serializer.deserializeStream(fileInputStream)
 
     // Register an on-task-completion callback to close the input stream.
-    context.addOnCompleteCallback(() => deserializeStream.close())
+    context.addTaskCompletionListener(context => deserializeStream.close())
 
     deserializeStream.asIterator.asInstanceOf[Iterator[T]]
   }

core/src/main/scala/org/apache/spark/rdd/HadoopRDD.scala

@@ -197,7 +197,7 @@ class HadoopRDD[K, V](
       reader = inputFormat.getRecordReader(split.inputSplit.value, jobConf, Reporter.NULL)
 
       // Register an on-task-completion callback to close the input stream.
-      context.addOnCompleteCallback{ () => closeIfNeeded() }
+      context.addTaskCompletionListener{ context => closeIfNeeded() }
       val key: K = reader.createKey()
       val value: V = reader.createValue()
 

core/src/main/scala/org/apache/spark/rdd/JdbcRDD.scala

@@ -68,7 +68,7 @@ class JdbcRDD[T: ClassTag](
   }
 
   override def compute(thePart: Partition, context: TaskContext) = new NextIterator[T] {
-    context.addOnCompleteCallback{ () => closeIfNeeded() }
+    context.addTaskCompletionListener{ context => closeIfNeeded() }
     val part = thePart.asInstanceOf[JdbcPartition]
     val conn = getConnection()
     val stmt = conn.prepareStatement(sql, ResultSet.TYPE_FORWARD_ONLY, ResultSet.CONCUR_READ_ONLY)

core/src/main/scala/org/apache/spark/rdd/NewHadoopRDD.scala

@@ -129,7 +129,7 @@ class NewHadoopRDD[K, V](
       context.taskMetrics.inputMetrics = Some(inputMetrics)
 
       // Register an on-task-completion callback to close the input stream.
-      context.addOnCompleteCallback(() => close())
+      context.addTaskCompletionListener(context => close())
       var havePair = false
       var finished = false