Readme — net.tbt-post/clj-kafka-x 0.7.5

clj-kafka-x

A Clojure library for the Apache Kafka (distributed stream-processing software platform).

Uses KF protocol and does not rely on ZooKeeper.

Still contains zookeeper as a dependency.

Tries to be as lightweigh as possible thus depends only on

org.apache.kafka/kafka_2.12 "3.4.0"
org.apache.kafka/kafka-clients "3.4.0"
org.apache.zookeeper/zookeeper "3.8.1"

but excluding jms,jmx* and logging.

Some builds (for instance of v0.4.x branch) may partially (sometimes even fully) be incompatible with some versions of other libraries that also use NIO! If you’re experiencing build problems and/or your application is unexpectedly crashed on start - try check your project dependencies more deeply, may be you will need to correct existing dependencies version or to add an actual version of full [io.netty/netty-all]

Actual library info:

Installation

Add the following to your Leiningen’s project.clj:

[net.tbt-post/clj-kafka-x "0.7.5"]

Usage

Producer

(require '[clj-kafka-x.producer :as kp])

(with-open [p (kp/producer {"bootstrap.servers" "localhost:9092"}
                           (kp/string-serializer)
                           (kp/string-serializer))]
  @(kp/send p (kp/record "topic-a" "Hi there!")))

Consumer

(require '[clj-kafka-x.consumers.simple :as kc])

(with-open [c (kc/consumer {"bootstrap.servers" "localhost:9092"
                            "group.id" "consumer-id"}
                            (kc/string-deserializer)
                            (kc/string-deserializer))]
  (kc/subscribe c "topic-a")
  (kc/messages c))

When you use multiple partitions per topic it is required to specify them explicitly when subscribing, i.e. (kc/subscribe c [{:topic "topic-a" :partitions #{0 1}} {:topic "topic-b" :partitions #{0 1 2}}])

Real-life (almost) example

(ns buzz.consumer.kafka
  (:require [clj-kafka-x.consumers.simple :as kc]
            [clojure.tools.logging :as log]))

(defn processor [msg schema] msg)
(def schema nil)
(def config {"bootstrap.servers" "localhost:9092"
             "group.id" "consumer-id"})

(defn process-message [msg]
  (let [{:keys [value topic partition offset]} msg
        processor processor ;; choose one by topic name
        schema schema]      ;; choose one by topic name
    (if (fn? processor) (processor value schema) value)))

(defn consume []
  (with-open [c (kc/consumer config
                             (kc/byte-array-deserializer)
                             (kc/byte-array-deserializer))]
    (kc/subscribe c (config/kafka-topics))
    (let [pool (kc/messages c)]
      (doseq [message pool]
        (log/warn (process-message message))))))

you may also use specific timeouts form

(defn- consume [instance process-message]
  (when-let [co (kc/consumer config
                             (kc/byte-array-deserializer)
                             (kc/byte-array-deserializer))
             messages (kc/messages
                        co
                        :timeout (:request-timeout-ms config))]
    (doall (map process-message messages))))

message count per poll execution may be specified by max.poll.records field of configuration

License