Status

Current state: Under DiscussionAccepted

Discussion thread: here

JIRA: KAFKA-6849

...

The following methods would be added to the the Java KTable interface: 

<VR> KTable<K, VR> transformValues(final ValueTransformerWithKeySupplier<? super K, ? super V, ? extends VR> valueTransformerSupplier,
                                   final String... stateStoreNames);
 
<VR> KTable<K, VR> transformValues(final ValueTransformerWithKeySupplier<? super K, ? super V, ? extends VR> valueTransformerSupplier,
                                   final Materialized<K, VR, KeyValueStore<Bytes, byte[]>> materialized,
                                   final String... stateStoreNames);

The following methods would be added to the Scala KTable class.

def transformValues[VR](valueTransformerSupplier: ValueTransformerWithKeySupplier[K, V, VR],
                        stateStoreNames: String*): KTable[K, VR]

def transformValues[VR](valueTransformerSupplier: ValueTransformerWithKeySupplier[K, V, VR],
                        materialized: Materialized[K, VR, KeyValueStore[Bytes, Array[Byte]]],
                        stateStoreNames: String*): KTable[K, VR]

Proposed Changes

The new methods on `KTableImpl` will add a new KTableTransformValues processor node and attach any state stores.  The new KTableTransformValues will be implemented in a similar manner to other processors, instantiating the user supplied transformer once per task.

The `ProcessorContext` passed to the `init()` method of transformer implementations will be restricted. Any : any call to any variant of the `forward()` method will throw a `StreamsException`, otherwise implementations could output . This will stop implementations outputting values with a new key. Any call to `getStateStore` where the requested store is the one passed in via the `materialized` parameter will throw a `StreamsException`, as this state store is managed by the the streams library.

Compatibility, Deprecation, and Migration Plan

...