Matthias J. Sax created KAFKA-16575:
---------------------------------------
Summary: Automatically remove KTable aggregation result when group
becomes empty
Key: KAFKA-16575
URL: https://issues.apache.org/jira/browse/KAFKA-16575
Project: Kafka
Issue Type: Improvement
Components: streams
Reporter: Matthias J. Sax
Using `KTable.groupBy(...).aggregate(...)` can handle updates (inserts,
deletes, actual updates) of the input KTable, by calling the provided `Adder`
and `Subtractor`. However, when all records from the input table (which map to
the same group/row in the result table) get removed, the result entry is not
removed automatically.
For example, if we implement a "count", the count would go to zero for a group
by default, instead of removing the row from the result, if all input record
for this group got deleted.
Users can let their `Subtractor` return `null` for this case, to actually
delete the row, but it's not well documented and it seems it should be a
built-in feature of the table-aggregation to remove "empty groups" from the
result, instead of relying on "correct" behavior of user-code.
(Also the built-in `count()` does not return `null`, but actually zero...)
An internal counter how many elements are in a group should be sufficient. Of
course, there is backward compatibility questions we need to answer.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
