Hi Ties, You're on the right track. You need to use `KTable.map` ahead of the join to select the new key. This will allow Streams to make sure the data is correctly partitioned to perform the join.
Thanks, -John On Mon, Jul 15, 2019 at 10:07 AM Ven, Ties Jens van de <ties.van.de....@alliander.com> wrote: > > I recently started working with kafka streams and I noticed some odd behavior. > > I was using a KTable left join with a null key, and ofcourse this will not > work, since it will join based on keys. > But I also supplied a KeyValueMapper, which takes a property from the value > and returns this as key, and uses this value to join. > > It turns out that in the code, it firsts checks if there is a null key, and > if so, it skips. > Would it be more logical to check the result of the keyMapper for null > instead of the actual key? > > https://github.com/apache/kafka/blob/trunk/streams/src/main/java/org/apache/kafka/streams/kstream/internals/KStreamKTableJoinProcessor.java > > Kind regards > > Ties
