[
https://issues.apache.org/jira/browse/KAFKA-3839?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15336310#comment-15336310
]
Jeyhun Karimov commented on KAFKA-3839:
---------------------------------------
It seems that groupby operator is gone from KTable. Nevertheless, I checked its
behaviour with null keys:
KTable<Long, String> source = builder.table(longSerde, stringSerde,
"topic1");
KTable<String,Long> counts = source.
groupBy(new KeyValueMapper<Long, String, KeyValue<String,
String>>() {
@Override
public KeyValue<String, String> apply(Long key, String value) {
// TODO Auto-generated method stub
return KeyValue.pair(null, value);
}
},Serdes.String(), Serdes.String()).count("count");
counts.to(stringSerde,longSerde,"topic2");
If I run this code, no exception/error is got and nothing is outputted to topic
(as expected). I check the class KTableRepartitionMap.KTableMapProcessor and I
think the required null checks are there in process() method.
> Handling null keys in KTable.groupBy
> ------------------------------------
>
> Key: KAFKA-3839
> URL: https://issues.apache.org/jira/browse/KAFKA-3839
> Project: Kafka
> Issue Type: Bug
> Components: streams
> Reporter: Guozhang Wang
> Assignee: Jeyhun Karimov
> Labels: architecture, newbie
>
> This is first reported by [~jeff.klu...@gmail.com].
> Similarly as we handle null keys in KStream join and aggregations in
> KAFKA-3561 (https://github.com/apache/kafka/pull/1472), for KTable.groupBy
> operators we should also gracefully handle null keys by filter them out since
> they will not be participating in the aggregation.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
