Status

Current state: "under discussion"

Discussion thread: HERE

JIRA: K AFKA-4218, KAFKA-4726 , KAFKA-3745

Please keep the discussion on the mailing list rather than commenting on the wiki (wiki discussions get unwieldy fast).

The PR can be found here.

Motivation

(taken from JIRA descriptions)

Key access to ValueTransformer: While transforming values via KStream.transformValues and ValueTransformer, the key associated with the value may be needed, even if it is not changed. For instance, it may be used to access stores.
As of now, the key is not available within these methods and interfaces, leading to the use of KStream.transform and Transformer, and the unnecessary creation of new KeyValue objects.
Key access to ValueMapper: ValueMapper should have read-only access to the key for the value it is mapping. Sometimes the value transformation will depend on the key.
It is possible to do this with a full blown KeyValueMapper but that loses the promise that you won't change the key – so you might introduce a re-keying phase that is totally unnecessary.
Key access to ValueJoiner interface: In working with Kafka Stream joining, it's sometimes the case that a join key is not actually present in the values of the joins themselves (if, for example, a previous transform generated an ephemeral join key.) In such cases, the actual key of the join is not available in the ValueJoiner implementation to be used to construct the final joined value. This can be worked around by explicitly threading the join key into the value if needed, but it seems like extending the interface to pass the join key along as well would be helpful
Apart from key accesses providing the Rich versions interfaces including ValueJoiner, ValueMapper, Initializer, Aggregator is needed in various use-cases.

Public Changes

KStream interface:

<GK, GV, RV> KStream<K, RV> leftJoin(final GlobalKTable<GK, GV> globalKTable,
                                     final KeyValueMapper<? super K, ? super V, ? extends GK> keyValueMapper,
                                     final ValueJoinerWithKey<? super K, ? super V, ? super GV, ? extends RV> valueJoinerWithKey);

<GK, GV, RV> KStream<K, RV> join(final GlobalKTable<GK, GV> globalKTable,
                                 final KeyValueMapper<? super K, ? super V, ? extends GK> keyValueMapper,
                                 final ValueJoinerWithKey<? super K, ? super V, ? super GV, ? extends RV> valueJoinerWithKey);
 
<VT, VR> KStream<K, VR> leftJoin(final KTable<K, VT> table,
                                 final ValueJoinerWithKey<? super K, ? super V, ? super VT, ? extends VR> valueJoinerWithKey,
                                 final Serde<K> keySerde,
                                 final Serde<V> valSerde);

<VT, VR> KStream<K, VR> leftJoin(final KTable<K, VT> table,
                                 final ValueJoinerWithKey<? super K, ? super V, ? super VT, ? extends VR> valueJoinerWithKey);

<VT, VR> KStream<K, VR> join(final KTable<K, VT> table,
                             final ValueJoinerWithKey<? super K, ? super V, ? super VT, ? extends VR> valueJoinerWithKey,
                             final Serde<K> keySerde,
                             final Serde<V> valSerde);

<VT, VR> KStream<K, VR> join(final KTable<K, VT> table,
                             final ValueJoinerWithKey<? super K, ? super V, ? super VT, ? extends VR> valueJoinerWithKey);

<VO, VR> KStream<K, VR> outerJoin(final KStream<K, VO> otherStream,
                                  final ValueJoinerWithKey<? super K, ? super V, ? super VO, ? extends VR> valueJoinerWithKey,
                                  final JoinWindows windows,
                                  final Serde<K> keySerde,
                                  final Serde<V> thisValueSerde,
                                  final Serde<VO> otherValueSerde);

<VO, VR> KStream<K, VR> outerJoin(final KStream<K, VO> otherStream,
                                  final ValueJoinerWithKey<? super K, ? super V, ? super VO, ? extends VR> valueJoinerWithKey,
                                  final JoinWindows windows);

 
<VO, VR> KStream<K, VR> leftJoin(final KStream<K, VO> otherStream,
                                 final ValueJoinerWithKey<? super K, ? super V, ? super VO, ? extends VR> valueJoinerWithKey,
                                 final JoinWindows windows,
                                 final Serde<K> keySerde,
                                 final Serde<V> thisValSerde,
                                 final Serde<VO> otherValueSerde);

<VO, VR> KStream<K, VR> leftJoin(final KStream<K, VO> otherStream,
                                 final ValueJoinerWithKey<? super K, ? super V, ? super VO, ? extends VR> valueJoinerWithKey,
                                 final JoinWindows windows);

<VO, VR> KStream<K, VR> join(final KStream<K, VO> otherStream,
                             final ValueJoinerWithKey<? super K, ? super V, ? super VO, ? extends VR> valueJoinerWithKey,
                             final JoinWindows windows,
                             final Serde<K> keySerde,
                             final Serde<V> thisValueSerde,
                             final Serde<VO> otherValueSerde);

<VO, VR> KStream<K, VR> join(final KStream<K, VO> otherStream,
                             final ValueJoinerWithKey<? super K, ? super V, ? super VO, ? extends VR> valueJoinerWithKey,
                             final JoinWindows windows);

<VR> KStream<K, VR> transformValues(final ValueTransformerWithKeySupplier<? super K, ? super V, ? extends VR> valueTransformerWithKeySupplier,
                                    final String... stateStoreNames);

<VR> KStream<K, VR> mapValues(ValueMapperWithKey<? super K, ? super V, ? extends VR> mapperWithKey);

Proposed Changes

Handling lambdas

For Initializer, Aggregator, ValueMapper, ValueJoiner and their "withKey" interfaces we support lambdas. For ValueTransformer interface we don't need lambdas by the core definition of the class.

For all above interfaces we support rich functions.

To support lambdas, we separate withKey interface from original ones, meaning we don't inherit or extend from one to another.

```
RichValueMapper
```

public interface RichValueMapper<K, V, VR>  implements ValueMapperWithKey<K, V, VR>, RichFunction {
} 

public interface ValueMapperWithKey<K, V, VR> {
    VR apply(final K key, final V value);
}

public interface RichFunction {
    void init();
    void close();
}

```
RichValueJoiner
```

public interface RichValueJoiner<K, V1, V2, VR>  extends ValueJoinerWithKey<K, V1, V2, VR>, RichFunction {
}
 
public interface ValueJoinerWithKey<K, V1, V2, VR> {
    VR apply(final K key, final V1 value1, final V2 value2);
}

```
RichInitializer
```

public interface RichInitializer<VA> extends Initializer<VA>, RichFunction {
}

```
RichAggregator
```

public interface RichAggregator<K, V, VA> extends Aggregator<K, V, VA>, RichFunction {
}

```
ValueTransformerWithKeySupplier
```

public interface ValueTransformerWithKeySupplier<K, V, VR> {
    ValueTransformerWithKey<K, V, VR> get();
}

Handling rich functions while building the topology

In general, we change the constructors of all related backend Processors to be Rich types as we can easily convert lambdas/withKey-lambdas to Rich functions.

```
RichValueMapper
```

@Override
public <V1> KStream<K, V1> mapValues(final ValueMapperWithKey<? super K, ? super V, ? extends V1> mapperWithKey) {
    Objects.requireNonNull(mapperWithKey, "mapperWithKey can't be null");
    String name = topology.newName(MAPVALUES_NAME);
    final RichValueMapper<K, V, V1> richValueMapper;
    if (mapperWithKey instanceof RichValueMapper) {
        richValueMapper = (RichValueMapper<K, V, V1>) mapperWithKey;
    } else {
        richValueMapper = new RichValueMapper<K, V, V1>() {
            @Override
            public void init() {}

            @Override
            public void close() {}

            @Override
            public V1 apply(K key, V value) {
                return mapperWithKey.apply(key, value);
            }
        };
    }
    topology.addProcessor(name, new KStreamMapValues<>(richValueMapper), this.name);
    return new KStreamImpl<>(topology, name, sourceNodes, this.repartitionRequired);
}

```
RichValueJoiner
```

static <K, T1, T2, R> ValueJoinerWithKey<K, T1, T2, R> convertToValueJoinerWithKey(final ValueJoiner<T1, T2, R> valueJoiner) {
    Objects.requireNonNull(valueJoiner, "valueJoiner can't be null");
    return new ValueJoinerWithKey<K, T1, T2, R>() {
        @Override
        public R apply(K key, T1 value1, T2 value2) {
            return valueJoiner.apply(value1, value2);
        }
    };
}


public <V1, R> KStream<K, R> leftJoin(
    final KStream<K, V1> other,
    final ValueJoiner<? super V, ? super V1, ? extends R> joiner,
    final JoinWindows windows,
    final Serde<K> keySerde,
    final Serde<V> thisValSerde,
    final Serde<V1> otherValueSerde) {

    return doJoin(other,
        convertToValueJoinerWithKey(joiner),   // doJoin and join methods accept ValueJoinerWithKey type and corresponding Processors accept only Rich functions in their constructor.
        windows,
        keySerde,
        thisValSerde,
        otherValueSerde,
        new KStreamImplJoin(true, false));
}

Converters between Rich/WithoutKey/WithKey types

static <K, T1, T2, R> RichValueJoiner<K, T1, T2, R> convertToRichValueJoiner(final ValueJoinerWithKey<K, T1, T2, R> valueJoinerWithKey) {
    Objects.requireNonNull(valueJoinerWithKey, "valueJoiner can't be null");
    if (valueJoinerWithKey instanceof RichValueJoiner) {
        return (RichValueJoiner<K, T1, T2, R>) valueJoinerWithKey;
    } else {
        return new RichValueJoiner<K, T1, T2, R>() {
            @Override
            public void init() {}

            @Override
            public void close() {}

            @Override
            public R apply(K key, T1 value1, T2 value2) {
                return valueJoinerWithKey.apply(key, value1, value2);
            }
        };
    }
}

static <K, T1, T2, R> ValueJoinerWithKey<K, T1, T2, R> convertToValueJoinerWithKey(final ValueJoiner<T1, T2, R> valueJoiner) {
    Objects.requireNonNull(valueJoiner, "valueJoiner can't be null");
    return new ValueJoinerWithKey<K, T1, T2, R>() {
        @Override
        public R apply(K key, T1 value1, T2 value2) {
            return valueJoiner.apply(value1, value2);
        }
    };
}

static <VA> RichInitializer<VA> checkAndMaybeConvertToRichInitializer(final Initializer<VA> initializer) {
    Objects.requireNonNull(initializer, "initializer can't be null");
    if (initializer instanceof RichInitializer) {
        return (RichInitializer<VA>) initializer;
    } else {
        return new RichInitializer<VA>() {
            @Override
            public VA apply() {
                return initializer.apply();
            }

            @Override
            public void init() {}

            @Override
            public void close() {}
        };
    }
}

static <K, V, VA> RichAggregator<K, V, VA> checkAndMaybeConvertToRichAggregator(final Aggregator<K, V, VA> aggregator) {
    Objects.requireNonNull(aggregator, "aggregator can't be null");
    if (aggregator instanceof RichAggregator) {
        return (RichAggregator<K, V, VA>) aggregator;
    } else {
        return new RichAggregator<K, V, VA>() {
            @Override
            public VA apply(K key, V value, VA aggregate) {
                return aggregator.apply(key, value, aggregate);
            }

            @Override
            public void init() {}

            @Override
            public void close() {}
        };
    }
}

Test Plan

The unit tests are changed accordingly to support the changes in core classes.

Rejected Alternatives

Lambdas are not supported

This document is proposed with ValueMapper example but it can be applied to other interfaces as well. Rich functions are proposed:

public interface RichFunction {
    void init(final ProcessorContext context);

 void close();
}

public abstract class AbstractRichFunction implements RichFunction {
    @Override
  public void init(final ProcessorContext context) {}

    @Override
  public void close() {}
}

public abstract class RichValueJoiner<K, V1, V2, VR> extends AbstractRichFunction implements ValueJoiner<V1, V2, VR>  {
    @Override
  public final VR apply(final V1 value1, final V2 value2) {
        return apply(null, value1, value2);
  }

    public abstract VR apply(final K key, final V1 value1, final V2 value2);
}

Inside processor, we check if the instance (for example ValueMapper instance) is rich (for example RichValueMapper):

KStreamFlatMapValues(ValueMapper<? super V, ? extends Iterable<? extends V1>> mapper) {
    this.mapper = mapper;
  isRichFunction = mapper instanceof RichValueMapper ? true : false;
}

@Override
public void process(K key, V value) {
    Iterable<? extends V1> newValues;
 if (isRichFunction) {
        newValues = ((RichValueMapper<? super K, ? super V, ? extends Iterable<? extends V1>>) mapper).apply(key, value);
  } else {
        newValues = mapper.apply(value);
  }
    for (V1 v : newValues) {
        context().forward(key, v);
  }
}

Not backward-compatible

We propose adding key information for ValueJoiner, ValueTransformer, and ValueMapper classes and their apply(...) methods.

As a result, we perform the following public changes (and their overloaded versions)

Class	Old	New
KStream	<VR> KStream<K, VR> mapValues(ValueMapper<? super V, ? extends VR> mapper);	<VR> KStream<K, VR> mapValues(ValueMapper<? super K, ? super V, ? extends VR> mapper);
KStream	<VR> KStream<K, VR> transformValues(final ValueTransformerSupplier<? super V, ? extends VR> valueTransformerSupplier, final String... stateStoreNames);	<VR> KStream<K, VR> transformValues(final ValueTransformerSupplier<? super K, ? super V, ? extends VR> valueTransformerSupplier,final String... stateStoreNames);
KStream	<VO, VR> KStream<K, VR> join(final KStream<K, VO> otherStream, final ValueJoiner<? super V, ? super VO, ? extends VR> joiner, final JoinWindows windows);	<VO, VR> KStream<K, VR> join(final KStream<K, VO> otherStream, final ValueJoiner<? super K, ? super V, ? super VO, ? extends VR> joiner, final JoinWindows windows);
KTable	<VR> KTable<K, VR> mapValues(final ValueMapper<? super V, ? extends VR> mapper);	<VR> KTable<K, VR> mapValues(final ValueMapper<? super K, ? super V, ? extends VR> mapper);
KTable	<VO, VR> KTable<K, VR> join(final KTable<K, VO> other, final ValueJoiner<? super V, ? super VO, ? extends VR> joiner);	<VO, VR> KTable<K, VR> join(final KTable<K, VO> other, final ValueJoiner<? super K, ? super V, ? super VO, ? extends VR> joiner);

Lacking performance because deep-copy and need for RichFunctions

We extend the target interfaces ValueJoiner, ValueTransformer, and ValueMapper as ValueJoinerWithKey, ValueTransformerWithKey, and ValueMapperWithKey. In extended abstract classes we have an access to keys.

In Processor we check the actual instance of object:

this.valueTransformer = valueTransformer;
if (valueTransformer instanceof ValueTransformerWithKey) {
isTransformerWithKey = true;
} else {
isTransformerWithKey = false;
}

..............

@Override
public void process(K key, V value) {
if (isTransformerWithKey) {
K keyCopy = (K) Utils.deepCopy(key);
  context.forward(key, ((ValueTransformerWithKey<K, V, R>) valueTransformer).transform(keyCopy, value));
  } else {
context.forward(key, valueTransformer.transform(value));
  }
}

As we can see from the above code snippet, we can guard the key change in Processors by deeply copying the object before calling the apply() method.

Space shortcuts

Child pages

Motivation

Public Changes

`KStream` interface:

Handling lambdas

Handling rich functions while building the topology

Test Plan

Rejected Alternatives

Lambdas are not supported

Not backward-compatible

Lacking performance because deep-copy and need for RichFunctions

Space shortcuts

Child pages

KIP-149: Enabling key access in ValueTransformer, ValueMapper, and ValueJoiner

Motivation

Public Changes

KStream interface:

Handling lambdas

Handling rich functions while building the topology

Test Plan

Rejected Alternatives

Lambdas are not supported

Not backward-compatible

Lacking performance because deep-copy and need for RichFunctions

`KStream` interface: