anishshri-db commented on code in PR #44884:
URL: https://github.com/apache/spark/pull/44884#discussion_r1473307363
##########
sql/core/src/main/scala/org/apache/spark/sql/execution/streaming/TransformWithStateExec.scala:
##########
@@ -152,22 +159,116 @@ 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)
- assert(processorHandle.getHandleState ==
StatefulProcessorHandleState.CREATED)
- statefulProcessor.init(processorHandle, outputMode)
-
processorHandle.setHandleState(StatefulProcessorHandleState.INITIALIZED)
- val result = processDataWithPartition(singleIterator, store,
processorHandle)
- result
+ // If the query is running in batch mode, we need to create a new
StateStore and instantiate
+ // a temp directory on the executors in mapPartitions
+ 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 {
+ child.execute().mapPartitionsWithIndex[InternalRow](
+ (i, iter) => {
+ val providerId = new StateStoreProviderId(
+ StateStoreId(Utils.createTempDir().getAbsolutePath,
+ i, 0), getStateInfo.queryRunId)
+
+ // Create StateStoreProvider for this partition
+ val stateStoreProvider = StateStoreProvider.createAndInit(
+ providerId,
+ schemaForKeyRow,
+ schemaForValueRow,
+ numColsPrefixKey = 0,
+ useColumnFamilies = true,
+ storeConf = StateStoreConf(getDefaultStateStoreSQLConf),
Review Comment:
Instead of trying to copy the conf here, could we do one of the following:
- Create the RocksDB provider explicitly and then call init. Something like
this
```
val provider = new RocksDBStateStoreProvider()
val sqlConf = new SQLConf()
val storeConf = new StateStoreConf(sqlConf)
provider.init(
storeId, keySchema, valueSchema, 0, useColumnFamilies = true,
storeConf, new Configuration)
```
- Pass the SQL conf setting and then call createAndInit
```
val sqlConf = new SQLConf()
sqlConf.setConfString("<provider class>", "<RocksDBStateStoreProvider>")
val storeConf = new StateStoreConf(sqlConf)
StateStoreProvider.createAndInit(
storeId, keySchema, valueSchema, 0, useColumnFamilies = true,
storeConf, new Configuration)
```
--
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]
