ableegoldman commented on code in PR #18194:
URL: https://github.com/apache/kafka/pull/18194#discussion_r1886132059
##########
streams/src/main/java/org/apache/kafka/streams/kstream/internals/KTableImpl.java:
##########
@@ -1278,12 +1278,11 @@ private <VR, KO, VO> KTable<K, VR>
doJoinOnForeignKey(final KTable<KO, VO> forei
final String foreignTableJoinName = renamed
.suffixWithOrElseGet("-foreign-join-subscription", builder,
SUBSCRIPTION_PROCESSOR);
- final StatefulProcessorNode<KO, Change<VO>> foreignTableJoinNode = new
ForeignTableJoinNode<>(
Review Comment:
good question -- basically yes and no. Almost everything has/will be
converted from StatefulProcesorNode to ProcessorGraphNode, but there are a few
exceptions where we still need the StatefulProcessorNode to connect a processor
and state store. For example any processors that access upstream state via a
value getter, or a custom ProcessorSupplier passed into the
`.process`/`.processValues` operator that doesn't implement `#stores` and
manually adds stores by calling `StreamsBuilder#addStateStore` instead.
I actually have the PR for this ready but it was done on top of this so I
was waiting for this one to be merged before calling for review. But it's ready
now if you want to take a look (though note that it still can't be merged until
Almog's TableSuppressNode PR is merged). See
https://github.com/apache/kafka/pull/18195
--
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]
