Current state: Under Discussion Merged

Voting thread: here

Discussion thread: here [Change the link from the KIP proposal email archive to your own email thread]
JIRA: here [Change the link from KAFKA-1 to your own ticket]

JIRA:

Motivation

Kafka Streams enforces a strict non-null-key policy in the DSL across all key-dependent operations (like aggregations and joins).
For left-joins, it might make makes sense to still accept a `null`, and add the left-hand record with an empty right-hand-side to the result.
Similarly, for outer-joins it makes sense to keep the record.
The defined semantics for left/outer join are: keep the stream record if no matching join record was found.
Thus, Kafka Streams today's behavior is inconsistent with the defined semantics.

Public Interfaces / Proposed Changes

This KIP won't change any public interfaces. However, It it will change the behavior of several stream dsl operators:

• KStream left & outer join
• table KTable LeftJoin: foreign key may be null

Proposed Changes

the following:

Operators

• left join Kstream-Kstream: no longer drop left records with null-key and call ValueJoiner with 'null' for right value.
• outer join Kstream-Kstream: no longer drop left/right records with null-key and call ValueJoiner with 'null' for right/left value.
• left-foreign-key join Ktable-Ktable: no longer drop left records with null-foreign-key returned by the ForeignKeyExtractor and call ValueJoiner with 'null' for right value.
• left join KStream-Ktable: no longer drop left records with null-key and call ValueJoiner with 'null' for right value.
• left join KStream-GlobalTable: no longer drop records when KeyValueMapper returns 'null' and call ValueJoiner with 'null' for right value.

Repartition of null-key records

Currently, repartitioning causes records with 'null'-keys to be dropped.
Describe the new thing you want to do in appropriate detail. This may be fairly extensive and have large subsections of its own. Or it may be a few sentences. Use judgement based on the scope of the change.This behavior needs to be changed.
The above described change in behavior would otherwise not work for topologies with repartitioning.
Going forward, they will be sent to an arbitrary partition.

Compatibility, Deprecation, and Migration Plan

...

• What impact (if any) will there be on existing users?
• If we are changing behavior how will we phase out the older behavior?
• If we need special migration tools, describe them here.
• When will we remove the existing behavior?

Test Plan

Describe in few sentences how the KIP will be tested. We are mostly interested in system tests (since unit-tests are specific to implementation details). How will we know that the implementation works as expected? How will we know nothing broke?

Rejected Alternatives

Keeping the old behavior

Users who want to keep the current behavior can prepend a .filter() operator to the aforementioned operators and filter accordingly.

Examples

//left join Kstream-Kstream
leftStream
    .filter((key, value) -> key != null)
    .leftJoin(rightStream, (lv, rv) -> join(lv, rv), windows);

//outer join Kstream-Kstream
rightStream
    .filter((key, value) -> key != null);
leftStream
    .filter((key, value) -> key != null)
    .outerJoin(rightStream, (lv, rv) -> join(lv, rv), windows);

//left-foreign-key join Ktable-Ktable
Function<String, String> foreignKeyExtractor = leftValue -> ...
leftTable
    .filter((key, value) -> foreignKeyExtractor.apply(value) != null)
    .leftJoin(rightTable, foreignKeyExtractor, (lv, rv) -> join(lv, rv), Named.as("left-foreign-key-table-join"));

//left join Kstream-Ktable
leftStream
    .filter((key, value) -> key != null)
    .leftJoin(ktable, (k, lv, rv) -> join(lv, rv));

//left join KStream-GlobalTable
KeyValueMapper<String, String, String> keyValueMapper = (k, v) -> ...;
leftStream
    .filter((key, value) -> keyValueMapper.apply(key,value) != null)
    .leftJoin(globalTable, keyValueMapper, (lv, rv) -> join(lv, rv));

Remarks

Note that the above list of operator examples is not exhaustive.
E.g. There is only one example with a 'ValueJoinerWithKey'. All other examples pass a 'ValueJoiner' into the operator.

We will hint to the DSL users via Java docs that they have now have the option to distinguish between the following scenarios within a KTable-KTable foreign- key-left-join operator:
"null-key" v.s. "not-null-key but null-value" by passing a 'ValueJoinerWithKey' instead of 'ValueJoiner' to the left join operator.
The remark will be made in the Java docs of all left join methods with a 'ValueJoiner' as a parameter.

As part of this KIP the above information on how to keep the old behavior will also be documented here: https://kafka.apache.org/documentation/streams/upgrade-guide
Plus, https://cwiki.apache.org/confluence/display/KAFKA/Kafka+Streams+Join+Semantics and https://kafka.apache.org/35/documentation/streams/developer-guide/dsl-api.html#joining will be updated accordingly.

Why not make this change opt in?

By default we would like the behavior of a stream defined via DSL to be consistent with the defined semanticsIf there are alternative ways of accomplishing the same thing, what were they? The purpose of this section is to motivate why the design is the way it is and not some other way.