Github user jerryshao commented on a diff in the pull request:
https://github.com/apache/spark/pull/9256#discussion_r43224039
--- Diff:
streaming/src/main/scala/org/apache/spark/streaming/dstream/TrackeStateDStream.scala
---
@@ -0,0 +1,192 @@
+/*
+ * 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.streaming.dstream
+
+import java.io.{IOException, ObjectOutputStream}
+
+import scala.collection.mutable.ArrayBuffer
+import scala.reflect.ClassTag
+
+import org.apache.spark._
+import org.apache.spark.rdd.{EmptyRDD, RDD}
+import org.apache.spark.storage.StorageLevel
+import org.apache.spark.streaming._
+import org.apache.spark.streaming.util.StateMap
+import org.apache.spark.util.Utils
+
+private[streaming] case class TrackStateRDDRecord[K: ClassTag, S:
ClassTag, T: ClassTag](
+ stateMap: StateMap[K, S], emittedRecords: Seq[T])
+
+
+private[streaming] class TrackStateRDDPartition(
+ idx: Int,
+ @transient private var prevStateRDD: RDD[_],
+ @transient private var partitionedDataRDD: RDD[_]) extends Partition {
+
+ private[dstream] var previousSessionRDDPartition: Partition = null
+ private[dstream] var partitionedDataRDDPartition: Partition = null
+
+ override def index: Int = idx
+ override def hashCode(): Int = idx
+
+ @throws(classOf[IOException])
+ private def writeObject(oos: ObjectOutputStream): Unit =
Utils.tryOrIOException {
+ // Update the reference to parent split at the time of task
serialization
+ previousSessionRDDPartition = prevStateRDD.partitions(index)
+ partitionedDataRDDPartition = partitionedDataRDD.partitions(index)
+ oos.defaultWriteObject()
+ }
+}
+
+private[streaming] class TrackStateRDD[K: ClassTag, V: ClassTag, S:
ClassTag, T: ClassTag](
+ _sc: SparkContext,
+ private var prevStateRDD: RDD[TrackStateRDDRecord[K, S, T]],
+ private var partitionedDataRDD: RDD[(K, V)],
+ trackingFunction: (K, Option[V], State[S]) => Option[T],
+ currentTime: Long, timeoutThresholdTime: Option[Long]
+ ) extends RDD[TrackStateRDDRecord[K, S, T]](
+ _sc,
+ List(
+ new OneToOneDependency[TrackStateRDDRecord[K, S, T]](prevStateRDD),
+ new OneToOneDependency(partitionedDataRDD))
+ ) {
+
+ @volatile private var doFullScan = false
+
+ require(partitionedDataRDD.partitioner.nonEmpty)
+ require(partitionedDataRDD.partitioner == prevStateRDD.partitioner)
+
+ override val partitioner = prevStateRDD.partitioner
+
+ override def checkpoint(): Unit = {
+ super.checkpoint()
+ doFullScan = true
+ }
+
+ override def compute(
+ partition: Partition, context: TaskContext):
Iterator[TrackStateRDDRecord[K, S, T]] = {
+
+ val stateRDDPartition = partition.asInstanceOf[TrackStateRDDPartition]
+ val prevStateRDDIterator = prevStateRDD.iterator(
+ stateRDDPartition.previousSessionRDDPartition, context)
+ val dataIterator = partitionedDataRDD.iterator(
+ stateRDDPartition.partitionedDataRDDPartition, context)
+ if (!prevStateRDDIterator.hasNext) {
+ throw new SparkException(s"Could not find state map in previous
state RDD")
+ }
+
+ val newStateMap = prevStateRDDIterator.next().stateMap.copy()
+ val emittedRecords = new ArrayBuffer[T]
+
+ val wrappedState = new StateImpl[S]()
+
+ dataIterator.foreach { case (key, value) =>
+ wrappedState.wrap(newStateMap.get(key))
+ val emittedRecord = trackingFunction(key, Some(value), wrappedState)
+ if (wrappedState.isRemoved) {
+ newStateMap.remove(key)
+ } else if (wrappedState.isUpdated) {
+ newStateMap.put(key, wrappedState.get(), currentTime)
+ }
+ emittedRecords ++= emittedRecord
--- End diff --
I don't think being an iterator cannot be persisted in memory and get
reused many times. Oppositely many RDD is designed as iterator like
`HadoopRDD`, whereas can be persisted in memory and get used many times.
Actually if RDD is chosen to be persisted in memory, it will materialize the
iterator by using `unroll` in `MemoryStore`, materialization will be happened
at that step, after that this RDD can be reused without recomputing (you can
check the code of `CacheManager`) .
I think the problem here is that it is hard to use the iterator because you
have to return a combo of (StateMap, emitted records), where StateMap is a
materialized one in memory, so being emitted records as iterator is hard to do.
Maybe we could refactor this part of code to make it more iterator friendly,
otherwise this unmanaged ArrayBuffer will potentially be a OOM issue. Just my
thoughts, any misunderstandings please point out :smile: .
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
