Serialization problems using Function implementations with Spark

Question

I'm having troubles understanding Spark Function implementations in Java. The documentation gives three ways to use functions in map and reduce :

1. via lambda
2. via inline classes implementing Function and Function2
3. via inner classes implementing Function and Function2

The trouble is that I can't manage to make 2. and 3. work. For instance, this code :

public int countInline(String path) {

    String master = "local";
    SparkConf conf = new SparkConf().setAppName("charCounterInLine")
            .setMaster(master);
    JavaSparkContext sc = new JavaSparkContext(conf);
    JavaRDD<String> lines = sc.textFile(path);

    JavaRDD<Integer> lineLengths = lines
            .map(new Function<String, Integer>() {
                public Integer call(String s) {
                    return s.length();
                }
            });
    return lineLengths.reduce(new Function2<Integer, Integer, Integer>() {
        public Integer call(Integer a, Integer b) {
            return a + b;
        }
    }); // the line causing the error 
}

gives me this mistake :

14/07/09 11:23:20 INFO DAGScheduler: Failed to run reduce at CharCounter.java:42
[WARNING]
java.lang.reflect.InvocationTargetException
        at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
        at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:62)
        at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)
        at java.lang.reflect.Method.invoke(Method.java:483)
        at org.codehaus.mojo.exec.ExecJavaMojo$1.run(ExecJavaMojo.java:297)
        at java.lang.Thread.run(Thread.java:745)
Caused by: org.apache.spark.SparkException: Job aborted due to stage failure: Task not serializable: java.io.NotSerializableException: Hadoop.Spark.basique.CharCounter
        at org.apache.spark.scheduler.DAGScheduler.org$apache$spark$scheduler$DAGScheduler$$failJobAndIndependentStages(DAGScheduler.scala:1033)
        at org.apache.spark.scheduler.DAGScheduler$$anonfun$abortStage$1.apply(DAGScheduler.scala:1017)
        at org.apache.spark.scheduler.DAGScheduler$$anonfun$abortStage$1.apply(DAGScheduler.scala:1015)
        at scala.collection.mutable.ResizableArray$class.foreach(ResizableArray.scala:59)
        at scala.collection.mutable.ArrayBuffer.foreach(ArrayBuffer.scala:47)
        at org.apache.spark.scheduler.DAGScheduler.abortStage(DAGScheduler.scala:1015)
        at org.apache.spark.scheduler.DAGScheduler.org$apache$spark$scheduler$DAGScheduler$$submitMissingTasks(DAGScheduler.scala:770)
        at org.apache.spark.scheduler.DAGScheduler.org$apache$spark$scheduler$DAGScheduler$$submitStage(DAGScheduler.scala:713)
        at org.apache.spark.scheduler.DAGScheduler.handleJobSubmitted(DAGScheduler.scala:697)
        at org.apache.spark.scheduler.DAGSchedulerEventProcessActor$$anonfun$receive$2.applyOrElse(DAGScheduler.scala:1176)
        at akka.actor.ActorCell.receiveMessage(ActorCell.scala:498)
        at akka.actor.ActorCell.invoke(ActorCell.scala:456)
        at akka.dispatch.Mailbox.processMailbox(Mailbox.scala:237)
        at akka.dispatch.Mailbox.run(Mailbox.scala:219)
        at akka.dispatch.ForkJoinExecutorConfigurator$AkkaForkJoinTask.exec(AbstractDispatcher.scala:386)
        at scala.concurrent.forkjoin.ForkJoinTask.doExec(ForkJoinTask.java:260)
        at scala.concurrent.forkjoin.ForkJoinPool$WorkQueue.runTask(ForkJoinPool.java:1339)
        at scala.concurrent.forkjoin.ForkJoinPool.runWorker(ForkJoinPool.java:1979)
        at scala.concurrent.forkjoin.ForkJoinWorkerThread.run(ForkJoinWorkerThread.java:107)

Right now, I can avoid this problem by implementing Function and Function2 in public outer classes. However, it was more a lucky guess than a well-thought decision. Moreover, since I can't manage to make documentation examples work, I guess there are things I don't understand.

To conclude, my questions are :

• how is it possible to make 2. and 3. work ?
• why is only the lambda working ?
• are there other way to use functions ?

Answer 1

The relevant part of this stracktrace is:

Task not serializable: java.io.NotSerializableException: Hadoop.Spark.basique.CharCounter

When you defined your functions as inner classes, their enclosing object is being pulled into the function closure and serialized. If this class is non-serializable or contains a non-serializable field, then you'll run into this error.

You have a few options here:

• Mark the non-serializable fields of the enclosing object as transient.
• Define your functions as outer-classes.
• Define your functions as static nested classes.

Answer 2

Adding "implements Serializable" for enclosing class can solve the problem. It is serializing enclosing class because inner class is a member of that, but enclosing class is not serializable it seems.