rdblue commented on a change in pull request #24129: [SPARK-27190][SQL] add
table capability for streaming
URL: https://github.com/apache/spark/pull/24129#discussion_r267030788
##########
File path:
sql/core/src/main/scala/org/apache/spark/sql/execution/streaming/MicroBatchExecution.scala
##########
@@ -88,31 +88,33 @@ class MicroBatchExecution(
logInfo(s"Using Source [$source] from DataSourceV1 named
'$sourceName' [$dataSourceV1]")
StreamingExecutionRelation(source, output)(sparkSession)
})
- case s @ StreamingRelationV2(ds, dsName, table: SupportsMicroBatchRead,
options, output, _)
- if !disabledSources.contains(ds.getClass.getCanonicalName) =>
- v2ToRelationMap.getOrElseUpdate(s, {
- // Materialize source to avoid creating it in every batch
- val metadataPath = s"$resolvedCheckpointRoot/sources/$nextSourceId"
- nextSourceId += 1
- logInfo(s"Reading table [$table] from DataSourceV2 named '$dsName'
[$ds]")
- // TODO: operator pushdown.
- val scan = table.newScanBuilder(options).build()
- val stream = scan.toMicroBatchStream(metadataPath)
- StreamingDataSourceV2Relation(output, scan, stream)
- })
- case s @ StreamingRelationV2(ds, dsName, _, _, output, v1Relation) =>
- v2ToExecutionRelationMap.getOrElseUpdate(s, {
- // Materialize source to avoid creating it in every batch
- val metadataPath = s"$resolvedCheckpointRoot/sources/$nextSourceId"
- if (v1Relation.isEmpty) {
- throw new UnsupportedOperationException(
- s"Data source $dsName does not support microbatch processing.")
- }
- val source = v1Relation.get.dataSource.createSource(metadataPath)
- nextSourceId += 1
- logInfo(s"Using Source [$source] from DataSourceV2 named '$dsName'
[$ds]")
- StreamingExecutionRelation(source, output)(sparkSession)
- })
+
+ case s @ StreamingRelationV2(src, srcName, table: SupportsRead, options,
output, v1) =>
+ val v2Disabled =
disabledSources.contains(src.getClass.getCanonicalName)
+ if (!v2Disabled && table.supports(TableCapability.MICRO_BATCH_READ)) {
+ v2ToRelationMap.getOrElseUpdate(s, {
+ // Materialize source to avoid creating it in every batch
+ val metadataPath = s"$resolvedCheckpointRoot/sources/$nextSourceId"
+ nextSourceId += 1
+ logInfo(s"Reading table [$table] from DataSourceV2 named
'$srcName' [$src]")
+ // TODO: operator pushdown.
+ val scan = table.newScanBuilder(options).build()
+ val stream = scan.toMicroBatchStream(metadataPath)
+ StreamingDataSourceV2Relation(output, scan, stream)
+ })
+ } else if (v1.isEmpty) {
+ throw new UnsupportedOperationException(
+ s"Data source $srcName does not support microbatch processing.")
+ } else {
+ v2ToExecutionRelationMap.getOrElseUpdate(s, {
+ // Materialize source to avoid creating it in every batch
+ val metadataPath = s"$resolvedCheckpointRoot/sources/$nextSourceId"
+ val source = v1.get.dataSource.createSource(metadataPath)
+ nextSourceId += 1
+ logInfo(s"Using Source [$source] from DataSourceV2 named
'$srcName' [$src]")
Review comment:
I think it would be valuable to include whether v1 or v2 is used when the
source supports both.
----------------------------------------------------------------
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.
For queries about this service, please contact Infrastructure at:
[email protected]
With regards,
Apache Git Services
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
