API info
Relation to most.js
Relation to JAVA (kafka) kafka-streams
Relation to Flink
How to get started
Documentation (is still WIP)
1. most.js
Relation to most
Both KStream and KTable classes depend on StreamDSL
that internally wraps around most.js's
observable/stream
as .stream$ class-member.
A large amount of stream operators that most.js offers are available as functions of KStream & KTable - however they are not always equal and they NEVER return a new KStream or KTable instance -> this is a big difference compared to most.js.
Join and window operatios, like .merge() and .window() actually do return a new KStream or KTable instance as of their nature.
You can find the most.js project here
2. Apache kafka-streams
Relation to kafka-streams
In its core kafka-streams.js tried to adopt the behaviour and API of the original kafa-streams lib that is part of Kafka. Due to the nature of NodeJS and observables to decision was made to simply mimic the API of the original, but not exactly port it 1:1.
Therefore you can expect almost the same functionality under key-components like KStream and KTable but not exactly the same API. E.g. KafkaStreams's role as factory.
You can find the Apache Kafka project here
3. Apache Flink
Relation to Flink
Kafka-streams actually implements the concept of αΊindow
s to enable merge
operations. However the Flink
project (also a stream-processing framework
similar to kafka-streams) has a different and more advanced approach at windows.
kafka-streams.js offers window operations to merge streams, but it also tries
to offer additional windowing features similar to what Flink offers.
You can find the Apache Flink project here
4. How to get started
Getting started
A good place to start using this lib is the Quick Start Tutorial, besides the examples.
The examples
The /examples
should work out of the box, however they require
you to have a local-setup that works with the tests. So take
a quick look at /test/test-config.js
and if you do not yet
have a local Zookeeper + Kafka Broker running, take a look at
/kafka-setup/docker-compose.yml
running ./start.sh
will
start a local combination.