JarCastingcelery-java
Java implementation of Celery client and worker. Quoting from the project website:
Celery is an asynchronous task queue/job queue based on distributed message passing. It is focused on real-time operation, but supports scheduling as well.
The execution units, called tasks, are executed concurrently on a single or more worker servers using multiprocessing, Eventlet, or gevent. Tasks can execute asynchronously (in the background) or synchronously (wait until ready).
Celery is used in production systems to process millions of tasks a day.
The aim is to be compatible with existing Python Celery implementation. That means you should be able to run a Java client with a Python worker or vice-versa. Tested with Python Celery 4.1.
At the moment, this is a very alpha version. It can
- execute a task
- report result
- report failure (and throw it as exception on the client)
What's missing:
- advanced features of Celery protocol
- retries
- chords
- groups
- chains
Patches providing any of these are welcome.
Maven dependency
Releases are available from Maven Central. Latest version:
<dependency>
<groupId>org.sedlakovi.celery</groupId>
<artifactId>celery-java</artifactId>
<version>...</version>
</dependency>
Snapshots are available from Sonatype OSRH:
<repository>
<id>sonatype</id>
<url>https://oss.sonatype.org/content/groups/public</url>
<snapshots>
<enabled>true</enabled>
<updatePolicy>always</updatePolicy>
</snapshots>
</repository>
Javadoc
Check out generated Javadoc at http://crabhi.github.io/celery-java/apidocs/.
Calling a Java task from Python
-
Annotate your class that does something useful as a
@CeleryTask.import com.geneea.celery.CeleryTask; @CeleryTask public class TestTask { public int sum(int x, int y) { return x + y; } }
-
Run
Workerwith your tasks on classpath. You can directly use theWorkerclass or embed it into yourmainfunction.import com.geneea.celery.CeleryWorker; public class MyWorker { public static void main(String[] args) throws Exception { CeleryWorker.main(args); } }
-
From the Python side, call the task by the class name hash (
#) method name.In [1]: import celery In [2]: app = celery.Celery(broker="amqp://localhost/", backend="rpc://localhost") In [3]: app.signature("com.geneea.celery.examples.TestTask#sum", [1, 2]).delay().get() Out[3]: 3 In [4]: %%timeit ...: app.signature("com.geneea.celery.examples.TestTask#sum", [1, 2]).delay().get() ...: 2.1 ms ± 170 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)
Calling Python task from Java
-
Start a celery worker as described in First Steps with Celery.
-
Call the task by name.
Celery client = Celery.builder()
.brokerUri("amqp://localhost/%2F")
.backendUri("rpc://localhost/%2F")
.build();
System.out.println(client.submit("tasks.add", new Object[]{1, 2}).get());
Calling Java task from Java
The @CeleryTask annotation on a class MyClass causes MyClassProxy and MyClassLoader to be generated. MyClassLoader registers the task into the worker and MyClassProxy has all the task methods tweaked so they now return a Future<...> instead of the original type.
To use the proxy, you need a Celery Client.
Celery client = Celery.builder()
.brokerUri("amqp://localhost/%2F")
.backendUri("rpc://localhost/%2F")
.build();
Integer result = TestTaskProxy.with(client).sum(1, 7).get();
Development
Local build
Build with mvn -Dgpg.skip to avoid the signing step.
Releasing
mvn release:clean release:prepare
mvn release:perform
Tests
Unit tests are part of the celery-java module. Integration tests are part of the examples module and are based on the example tasks. They start the queue in backend automatically via Docker. You need to have Docker configured on the machine running the tests of the examples module.
Relase notes
-
1.2 - Moved the package from
org.sedlakovitocom.geneea. No functionality changes. -
1.1 - Lots of breaking changes (renames, API improvements, the Client can be constructed without an existing connection...). If you're already using this library, please get in touch. The next release should have much fewer breaking changes and they will be listed explicitly.
-
1.0 - Initial release. Don't expect it to be stable.