Liking cljdoc? Tell your friends :D

twarc

Travis status

A Simple Clojure wrapper around Quartz Scheduler.

Artifacts

Clojars Project

Why?

There are a few Clojure libraries for working with Quartz, but each one has a fatal flaw (at least one):

  • Quartzite
    • Global state, thread bindings
    • You can't explicitly pass some type of state (e.g., a db connection) to jobs
    • Complex API
  • Immutant Scheduling
    • You can't explicity pass some type of state (e.g., a db connection) to jobs
    • You can't use it with persistent JobStores and in Clustered environments
    • Complex dependencies

The main disadvantage of these libs is that you can't use it in the right way, i.e. without any global state but with strong dependency management (see stuartsierra/component)

Features

  • No global state
  • You can pass any instance-aware context to jobs
  • Data structure-centric API (see examples)
  • Jobs are usual vars with function (i.e. defn)
  • Stateful jobs
  • No magic
  • stuartsierra.component support out of the box
  • Quartz' Listeners support via core.async channels

Usage

Basic config (see Quartz Configuration Reference):


(require '[twarc.core :as twarc])

(def props {:threadPool.class "org.quartz.simpl.SimpleThreadPool"
            :threadPool.threadCount 1
            :plugin.triggHistory.class "org.quartz.plugins.history.LoggingTriggerHistoryPlugin"
            :plugin.jobHistory.class "org.quartz.plugins.history.LoggingJobHistoryPlugin"})

;; Scheduler supports component/Lifecycle protocol and clojure.lang.Associative (its
;; Clojure record), so you can simply drop it into your system map. Or use some other DI
;; system.

(def sched (-> (twarc/make-scheduler props) (twarc/start)))

defjob macro defines two functions, in this case test-job and test-job*. test-job* is an actual job with the body provided by you. It executes in Quartz' thread pool. Generated test-job is a helper function that can be used for scheduling jobs.

Job function accepts scheduler instance as the first argument, and the rest of the arguments are passed on to job scheduling.

(twarc/defjob test-job
  [scheduler name message]
  (prn "Message for!" name message))

Let's run it!

;; If you use cider, note that Quartz threads know nothing about repl's stdout. So keep an eye on
;; messages in nrepl-server buffer

(test-job sched ["Andrew" "Hello world"])

That's all. The first argument is a scheduler instance; the second one is a vector of arguments, and optional tail arguments are options for schedule-job function (job and trigger params actually – see Quartz documentation for details).

You can schedule execution of any defn without a helper:

(defn test-job2
  [scheduler name message]
  (prn "Message  from!" name message))

(twarc/schedule-job sched #'test-job2 ["Petr" "Hi world!"])

Define simple or cron trigger via map:

(test-job sched ["Andrew" "Hello world"] :trigger {:simple {:repeat 5 :interval 1000}})

(test-job sched ["Andrew" "Hello world"]
          :job {:identity "eternal job"}
          :trigger {:cron "*/10 * * * * ?"})

(twarc/delete-job sched "eternal job")

Persistent JobStore

You can persist your jobs and triggers in JDBC-store.

First of all, you need to create tables, see these scripts – https://github.com/quartz-scheduler/quartz/tree/v2.3.2/quartz-core/src/main/resources/org/quartz/impl/jdbcjobstore

Secondly, configure Quartz for your store. You should also pick a well-defined name for your scheduler:

(def persistent-props
  (assoc props
    :jobStore.class "org.quartz.impl.jdbcjobstore.JobStoreTX"
    :jobStore.driverDelegateClass "org.quartz.impl.jdbcjobstore.PostgreSQLDelegate"
    :jobStore.tablePrefix "QRTZ_"
    :jobStore.dataSource "db"
    :dataSource.db.driver "org.postgresql.Driver"
    :dataSource.db.URL "jdbc:postgresql://localhost:5432/db_name"
    :dataSource.db.user "user"
    :dataSource.db.password "pass"))

(def persistent-sched (-> (twarc/make-scheduler persistent-props {:name "main-sched"})
                          (twarc/start)))

In this example we can also see how to configure a job and a trigger. With the :state param provided, a job becomes a Stateful job, and the job function accepts state as the second argument and should return updated state.

(twarc/defjob test-statefull-job
  [scheduler state i]
  (prn "State!" state)
  (update-in state [:counter] + i))

(test-statefull-job persistent-sched [4]
                    :job {:state {:counter 1}}
                    :trigger {:simple {:repeat :inf :interval 1000}})

And now stop and start a new scheduler without scheduling a task. Our previously scheduled task will continue executing.

(twarc/stop persistent-sched)
(def persistent-sched2 (-> (twarc/make-scheduler persistent-props {:name "main-sched"})
                           (twarc/start)))

Listeners

You can define listeners of some events with core.async channels.

(require '[clojure.core.async :as a])
(def executed (twarc/add-listener persistent-sched2 {:everything true} :was-executed))

(loop []
  (prn "--EXECUTED!" (->  (a/<!! executed) .getJobDetail .getJobDataMap (get "state")))
  (recur))

License

Copyright © 2015 Andrew Rudenko

Distributed under the Eclipse Public License either version 1.0 or (at your option) any later version.

Can you improve this documentation? These fine people already did:
Andrew Rudenko & Lyosha Kuleshov
Edit on GitHub

cljdoc is a website building & hosting documentation for Clojure/Script libraries

× close