Ketu

A Clojure Apache Kafka client with core.async api

[com.appsflyer/ketu "0.6.0"]

Features

• Channels API: Take kafka data from a channel and send data to kafka through a channel.
• Consumer Source: Polls records from kafka and puts them on a channel.
• Producer Sink: Takes records from a channel and sends them to kafka.
• Shapes: Transform the original objects of the java client to clojure data and back.
• Simple Configuration: Friendly, validated configuration.

Minimal Example

Consume a name string from kafka and produce a greeting string for that name back into kafka, all through channels:

(ns example
  (:require [clojure.core.async :refer [chan close! <!! >!!]]
            [ketu.async.source :as source]
            [ketu.async.sink :as sink]))
  
(let [<names (chan 10)
      source-opts {:name "greeter-consumer"
                   :brokers "broker1:9092"
                   :topic "names"
                   :group-id "greeter"
                   :value-type :string
                   :shape :value}
      source (source/source <names source-opts)

      >greets (chan 10)
      sink-opts {:name "greeter-producer"
                 :brokers "broker2:9091"
                 :topic "greetings"
                 :value-type :string
                 :shape :value}
      sink (sink/sink >greets sink-opts)]

  ;; Consume a name and produce a greeting. You could also do this with e.g. clojure.core.async/pipeline.
  (->> (<!! <names)
       (str "Hi, ")
       (>!! >greets))

  ;; Close the source. It automatically closes the source channel `<names`.
  (source/stop! source)
  ;; Close the sink channel `>greets`. It causes the sink to close itself as a consequence.
  (close! >greets))

Configuration reference

Anything that is not documented is not supported and might change.

Note: int is used for brevity but can also mean long. Don't worry about it.

Common options (both source and sink accept these)

Key	Type	Req?	Notes
:brokers	string	required	Comma separated host:port values e.g "broker1:9092,broker2:9092"
:topic	string	required
:name	string	required	Simple human-readable identifier, used in logs and thread names
:key-type	:string,:byte-array	optional	Default :byte-array, used in configuring key serializer/deserializer
:value-type	:string,:byte-array	optional	Default :byte-array, used in configuring value serializer/deserializer
:internal-config	map	optional	A map of the underlying java client properties, for any extra lower level config

Consumer-source options

Key	Type	Req?	Notes
:group-id	string	required
:shape	:value:, [:vector <fields>],[:map <fields>], or an arity-1 function of ConsumerRecord	optional	If unspecified, channel will contain ConsumerRecord objects. Examples

Producer-sink options

Key	Type	Req?	Notes
:shape	:value, [:vector <fields>],[:map <fields>], or an arity-1 function of the input returning ProducerRecord	optional	If unspecified, you must put ProducerRecord objects on the channel. Examples
:compression-type	"none" "gzip" "snappy" "lz4" "zstd"	optional	Default "none", values are same as "compression.type" of the java producer
:workers	int	optional	Default 1, number of threads that take from the channel and invoke the internal producer

Data shapes

You don't have to deal with ConsumerRecord or ProducerRecord objects.
To get a clojure data structure with any of the ConsumerRecord fields, configure the consumer shape:

; Value only:
{:topic "names"
 :key-type :string
 :value-type :string
 :shape :value}
(<!! consumer-chan)
;=> "v"

; Vector:
{:shape [:vector :key :value :topic]}
(<!! consumer-chan)
;=> ["k" "v" "names"]

; Map
{:shape [:map :key :value :topic]}
(<!! consumer-chan)
;=> {:key "k", :value "v", :topic "names"}

Similarly, to put a clojure data structure on the producer channel:

; Value only:
{:key-type :string
 :value-type :string
 :shape :value}
(>!! producer-chan "v")

; Vector:
{:shape [:vector :key :value]}
(>!! producer-chan ["k" "v"])

; Vector with topic in each message:
{:shape [:vector :key :value :topic]}
(>!! producer-chan ["k1" "v1" "names"])
(>!! producer-chan ["k2" "v2" "events"])

Development & Contribution

We welcome feedback and would love to hear about use-cases other than ours. You can open issues, send pull requests, or contact us at clojurians slack.