HeartSaVioR commented on code in PR #44884:
URL: https://github.com/apache/spark/pull/44884#discussion_r1475138458
##########
sql/core/src/main/scala/org/apache/spark/sql/execution/streaming/TransformWithStateExec.scala:
##########
@@ -155,23 +162,109 @@ case class TransformWithStateExec(
override protected def doExecute(): RDD[InternalRow] = {
metrics // force lazy init at driver
- child.execute().mapPartitionsWithStateStore[InternalRow](
- getStateInfo,
- schemaForKeyRow,
- schemaForValueRow,
- numColsPrefixKey = 0,
- session.sqlContext.sessionState,
- Some(session.sqlContext.streams.stateStoreCoordinator),
- useColumnFamilies = true
- ) {
- case (store: StateStore, singleIterator: Iterator[InternalRow]) =>
- val processorHandle = new StatefulProcessorHandleImpl(store,
getStateInfo.queryRunId,
- keyEncoder)
- assert(processorHandle.getHandleState ==
StatefulProcessorHandleState.CREATED)
- statefulProcessor.init(processorHandle, outputMode)
-
processorHandle.setHandleState(StatefulProcessorHandleState.INITIALIZED)
- val result = processDataWithPartition(singleIterator, store,
processorHandle)
- result
+ if (isStreaming) {
+ child.execute().mapPartitionsWithStateStore[InternalRow](
+ getStateInfo,
+ schemaForKeyRow,
+ schemaForValueRow,
+ numColsPrefixKey = 0,
+ session.sqlContext.sessionState,
+ Some(session.sqlContext.streams.stateStoreCoordinator),
+ useColumnFamilies = true
+ ) {
+ case (store: StateStore, singleIterator: Iterator[InternalRow]) =>
+ processData(store, singleIterator)
+ }
+ } else {
+ // If the query is running in batch mode, we need to create a new
StateStore and instantiate
+ // a temp directory on the executors in mapPartitionsWithIndex.
+ child.execute().mapPartitionsWithIndex[InternalRow](
+ (i, iter) => {
+ val providerId = {
+ // lazy creation to initialize tempDirPath once
+ lazy val tempDirPath = Utils.createTempDir().getAbsolutePath
+ new StateStoreProviderId(
+ StateStoreId(tempDirPath, 0, i), getStateInfo.queryRunId)
+ }
+
+ val sqlConf = new SQLConf()
+ sqlConf.setConfString(SQLConf.STATE_STORE_PROVIDER_CLASS.key,
+ classOf[RocksDBStateStoreProvider].getName)
+ val storeConf = new StateStoreConf(sqlConf)
+
+ // Create StateStoreProvider for this partition
+ val stateStoreProvider = StateStoreProvider.createAndInit(
Review Comment:
You need to clean up provider by yourself if you are creating it by yourself.
##########
sql/core/src/main/scala/org/apache/spark/sql/execution/SparkStrategies.scala:
##########
@@ -723,7 +723,7 @@ abstract class SparkStrategies extends
QueryPlanner[SparkPlan] {
* Strategy to convert [[TransformWithState]] logical operator to physical
operator
* in streaming plans.
*/
- object TransformWithStateStrategy extends Strategy {
+ object TransformWithStateStreamingSrategy extends Strategy {
Review Comment:
nit: TransformWithStateStreamingSrategy ->
TransformWithStateStreamingS`t`rategy
##########
sql/core/src/main/scala/org/apache/spark/sql/execution/streaming/TransformWithStateExec.scala:
##########
@@ -60,6 +66,7 @@ case class TransformWithStateExec(
batchTimestampMs: Option[Long],
eventTimeWatermarkForLateEvents: Option[Long],
eventTimeWatermarkForEviction: Option[Long],
+ isStreaming: Boolean = true,
Review Comment:
Is this allowed? Params with default values should be put to the end.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
