Thanks John. I'm looking forward to any feedback on this. In the meantime I will work on the unit tests to ensure that we have well-defined and readable coverage.
At the moment I cannot see a way around emitting (k,null) whenever we emit an event that lacks a matching foreign key on the RHS, except in the (k,null) -> (k,fk) case. If this LHS oldValue=null, we know we would have emitted a deletion and so (k,null) would be emitted out of the join. In this case we don't need to send another null. Adam On Wed, Jun 26, 2019 at 11:53 AM John Roesler <j...@confluent.io> wrote: > Hi Adam, > > Thanks for the proposed revision to your KIP > ( > https://cwiki.apache.org/confluence/pages/diffpagesbyversion.action?pageId=74684836&selectedPageVersions=77&selectedPageVersions=74 > ) > > in response to the concern pointed out during code review > (https://github.com/apache/kafka/pull/5527#issuecomment-505137962) > > We should have a brief discussion thread (here) in the mailing list to > make sure everyone who wants to gets a chance to consider the > modification to the design. > > Thanks, > -John >