Table of Contents

This page is meant as a template for writing a KIP. To create a KIP choose Tools->Copy on this page and modify with your content and replace the heading with the next KIP number and a description of your issue. Replace anything in italics with your own description.

Status

Current state: Under Discussion: Accepted [VOTE] KIP-120: Cleanup Kafka Streams builder API

Discussion thread: [DISCUSS] KIP-120: Cleanup Kafka Streams builder API

JIRA: KAFKA-3856

Released: 1.0.10.30

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

...

Currently, Kafka Streams public API leaks a bunch of internal methods that should not be public. Furthermore, DSL and PAPI abstraction are not completely separated at the moment and TopologyBuilder offers methods that belong to DSL only. Additionally, we want to change the creation of KafkaStreams instances and remove the uncommon constructor pattern.

Public Interfaces

In order to get a clean refactoring, we will deprecate classes

org.apache.kafka.streams.processor.TopologyBuilder
org.apache.kafka.streams.kstream.KStreamBuilder

and add them again ( with same new name but and different package)

org.apache.kafka.streams.TopologyBuilderTopology
org.apache.kafka.streams.KStreamBuilderStreamsBuilder

Furthermore, we deprecate all constructors of class KafkaStreams.add a new interface to get a full description the a topology (to compensate for some internal methods that get removed)

org.apache.kafka.streams.TopologyDescription
- TopologyDescription will have public interfaces:
  - ```
  Subtopology
```
- ```
GlobalStores
```
  - Node
  - Source
  - Sink
  - Processor

The following methods will not be available in the newly added classes:

TopologyBuilder -> Topology: setApplicationId, connectSourceStoreAndTopic, connectProcessors, addInternalTopic, copartitionSources, nodeGroups, build, buildGlobalStateTopology, globalStateStores, topicGroups, earliestResetTopicPattern, latestResetTopicPattern, stateStoreNamesToSourceTopics, copartitioneGroups, sourceTopicPattern, updateSubscriptions
KStreamBuilder -> StreamsBuilder: all of above methods from TopologyBuilder (as KStreamBuilder does not inherit from TopologyBuilder anymore) plus ) except addStateStore and addGlobalStore (which are both added explicitly to StreamsBuilder) plus newName
methods highlighted in TopologyBuilder list, are methods that actually belong to DSL abstraction

...

New org.apache.kafka.streams.TopologyBuilderTopology class:

Code Block

language	java

public final class TopologyBuilderTopology { // make class final compared to old TopologyBuilder

    // existing public API from current TopologyBuilder class that will not be changed

    public enum AutoOffsetReset {
        EARLIEST , LATEST
    }

    public TopologyBuilderTopology();


    public synchronized TopologyBuilderTopology addSource(String name, String... topics);
    public synchronized TopologyBuilderTopology addSource(String name, Pattern topicPattern);

    public synchronized TopologyBuilderTopology addSource(AutoOffsetReset offsetReset, String name,  String... topics);
    public synchronized TopologyBuilderTopology addSource(AutoOffsetReset offsetReset, String name,  Pattern topicPattern);

    public synchronized TopologyBuilderTopology addSource(String name, Deserializer keyDeserializer, Deserializer valDeserializervalueDeserializer, String... topics);
    public synchronized TopologyBuilderTopology addSource(String name, Deserializer keyDeserializer, Deserializer valDeserializervalueDeserializer, Pattern topicPattern);
   
    public synchronized TopologyBuilderTopology addSource(AutoOffsetReset offsetReset, String name, Deserializer keyDeserializer, Deserializer valDeserializervalueDeserializer, String... topics);
    public synchronized TopologyBuilderTopology addSource(AutoOffsetReset offsetReset, String name,  Deserializer keyDeserializer, Deserializer valDeserializervalueDeserializer, Pattern topicPattern);


    public synchronized TopologyBuilderTopology addSink(String name, String topic, String... parentNames);
    public synchronized TopologyBuilderTopology addSink(String name, String topic, Serializer keySerializer, Serializer valSerializervalueSerializer, String... parentNames);
 
    public synchronized TopologyBuilderTopology addSink(String name, String topic, StreamPartitioner partitioner, String... parentNames);
    public synchronized <K, V> TopologyBuilderTopology addSink(String name, String topic, Serializer<K> keySerializer, Serializer<V> valSerializervalueSerializer, StreamPartitioner<? super K, ? super V> partitioner, String... parentNames);


    public synchronized TopologyBuilderTopology addProcessor(String name, ProcessorSupplier supplier, String... parentNames);
 
    public synchronized TopologyBuilderTopology addStateStore(StateStoreSupplier supplier, String... processorNames);
 
    public synchronized TopologyBuilderTopology addGlobalStore(final StateStore store,
                                                       final String sourceName,
                                                       final Deserializer keyDeserializer,
                                                       final Deserializer valueDeserializer,
                                                       final String topic,
                                                       final String processorName,
                                                       final ProcessorSupplier stateUpdateSupplier);

    public synchronized TopologyBuilderTopology connectProcessorAndStateStores(String processorName, String... stateStoreNames);



    // newly added method to supportreveal buildertopology pattern
 
    // should be used instead of KafkaStreams() constructorstructure
    
    // old:
describes the current  //     TopologyBuilder builder = new TopologyBuilder();
    //     KafkaStreams streams = new KafkaStreams(builder, new StreamsConfig(...));
    // new:
    //     TopologyBuilder builder = new TopologyBuilder();
    //     KafkaStreams streams = builder.build(new StreamsConfig(...));
 
Topology, ie, TopologyDescription will not be updated if Topology is modified but #describe() must be called again
    public synchronized KafkaStreamsTopologyDescription builddescribe();

}

New org.apache.kafka.streams.KStreamBuilderStreamsBuilder class:

Code Block

language	java

public final class KStreamBuilderStreamsBuilder {

    // existing public API from current TopologyBuilderKStreamBuilder class that will not be changed
    // small change: adding `synchronized` to all methods to align with TopologyBuilder

    public KStreamBuilderStreamsBuilder();
 

    public synchronized <K, V> KStream<K, V> stream(final String... topics);
    public synchronized <K, V> KStream<K, V> stream(final Pattern topicPattern);

    public synchronized <K, V> KStream<K, V> stream(final AutoOffsetReset offsetReset, final String... topics);
    public synchronized <K, V> KStream<K, V> stream(final AutoOffsetReset offsetReset, final Pattern topicPattern);

    public synchronized <K, V> KStream<K, V> stream(final Serde<K> keySerde, final Serde<V> valSerdevalueSerde, final String... topics);
    public synchronized <K, V> KStream<K, V> stream(final Serde<K> keySerde, final Serde<V> valSerdevalueSerde, final Pattern topicPattern);

    public synchronized <K, V> KStream<K, V> stream(final AutoOffsetReset offsetReset, final Serde<K> keySerde, final Serde<V> valSerdevalueSerde, final String... topics);
    public synchronized <K, V> KStream<K, V> stream(final AutoOffsetReset offsetReset, final Serde<K> keySerde, final Serde<V> valSerdevalueSerde, final Pattern topicPattern);
 

    public synchronized <K, V> KTable<K, V> table(final String topic, final String storeName);
    public synchronized <K, V> KTable<K, V> table(final AutoOffsetReset offsetReset, final String topic, final String storeName);
    public synchronized <K, V> KTable<K, V> table(final Serde<K> keySerde, final Serde<V> valSerdevalueSerde, final String topic, final String storeName);
    public synchronized <K, V> KTable<K, V> table(final AutoOffsetReset offsetReset, final Serde<K> keySerde, final Serde<V> valSerdevalueSerde, final String topic, final String storeName);


    public synchronized <K, V> GlobalKTable<K, V> globalTable(final String topic, final String storeName);
    public synchronized <K, V> GlobalKTable<K, V> globalTable(final Serde<K> keySerde, final Serde<V> valSerdevalueSerde, final String topic, final String storeName);

    public synchronized <K, V> KStream<K, V> merge(final KStream<K, V>... streams);



    // newly added method to support builder pattern
 
    // should be used instead of KafkaStreams() constructor
    // old:
    //     KStreamBuilder builder = new KStreamBuilder();
    //     KafkaStreams streams = new KafkaStreams(builder, new StreamsConfig(...));
    // new:
    //     KStreamBuilder builder = new KStreamBuilder();
    //     KafkaStreams streams = builder.build(new StreamsConfig(...));
 
    public synchronized KafkaStreams build();
}

Proposed Changes

...



    public synchronized KStreamBuilder addStateStore(StateStoreSupplier supplier, String... processorNames);
    public synchronized KStreamBuilder addGlobalStore(StateStore store,
                                                      String sourceName,
                                                      Deserializer keyDeserializer,
                                                      Deserializer valueDeserializer,
                                                      String topic,
                                                      String processorName,
                                                      ProcessorSupplier stateUpdateSupplier);

    public synchronized Topology build();
}

New org.apache.kafka.streams.TopologyDescription class:

Code Block

language	java

public interface TopologyDescription {
    Set<Subtopology> subtopologies();
    Set<GlobalStore> globalStores();

    inferface Subtopology {
        int id();
        Set<Node> nodes();
    }

    interface GlobalStore {
        Source source();
        Processor processor();
    }

    interface Node {
        String name()
        Set<Node> predecessors();
        Set<Node> successors();
    }

    interface Source extends Node {
        String topics(); // can be comma separated list of topic names or pattern (as String)
    }

    interface Processor extends Node {
        Set<String> stores();
    }

    interface Sink extends Node {
        String topic();
    }

}

Proposed Changes

We will add two new internal classes

org.apache.kafka.streams.processor.internals.InternalTopologyBuilder
org.apache.kafka.streams.kstream.internals.InternalKStreamBuilder InternalStreamsBuilder

that offer the methods remove from current API. Thus, both classes are the actual implementation. Old TopologyBuilder and KStreamBuilder are only proxy classes to both classes respectively, for backward compatibility.

The newly added TopologyBuilder uses Topology uses InternalTopologyBuilder as member.

The newly added KStreamBuilder used the new TopologyBuilder StreamsBuilder uses the new Topology as a member (no class hierarchy anymore – using it as member gives a clear separation between PAPI and DSL).

Because the new StreamsBuilder does not inherit from new Topology we need to add StreamsBuilder#build() that returns the actual Topology to be passed into KafkaStreams client.

Note: because of backward compatibility, removed DSL specific classes offered by old TopologyBuilder must be offered by InternalTopologyBuilder for now. However, after both deprecated classes got removed, this cleanup can be done (and does not require a KIP anymore, because it's internal refactoring -- we just need to create a JIRA for this). Thus, this KIP falls short of separating PAPI and DSL completely. But it's a necessary first step to do the separation in a backward compatible way (backward compatibility requires a two step approach).

...

Because no classes/method will be removed but only deprecated, this change will be fully backward compatible
We intend to remove all deprecated classes/methods in 0.1011.41, but we can keep them longer on user request

Test Plan

Test Tests need to be rewritten but no new tests are required. All tests for public API need to be updated to use new TopologyBuilder and KStreamBuilder, and thus will not call deprecated KafkaStreams constructor anymoreTopology and StreamsBuilder. All tests using internal API, need to be rewritten to use InternalTopologyBuilder and InternalKStreamBuilderInternalStreamsBuilder.

Rejected Alternatives

If 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 wayNone.

Space shortcuts

Child pages

Versions Compared

Old Version 4

New Version Current

Key

Status

Public Interfaces

Proposed Changes

Proposed Changes

Test Plan

Rejected Alternatives

Space shortcuts

Child pages

Page History

Versions Compared

Old Version 4

New Version Current

Key

Status

Public Interfaces

Proposed Changes

Proposed Changes

Test Plan

Rejected Alternatives