problem

The Celery daemon needs to be restarted every time an existing task is modified or new tasks are added.

solution

Use only one Celery task that’s generic enough to run arbitrary Python functions with arbitrary arguments and wrap this task in a custom decorator.

As a bonus, the job will still work (synchronously) when Celery/the broker are not running.

The job’s calling API is the same as Celery’s: .s(), .delay() and .apply_async()

example

• in celeryapp.py which is in the same directory as celeryconfig.py, create a Celery app and then import the custom decorator:

import celery
import os
import sys

sys.path.insert(0, os.path.join(os.path.dirname(__file__), '.'))
#os.environ.setdefault('DJANGO_SETTINGS_MODULE', 'settings')
app = celery.Celery()
app.config_from_object('celeryconfig')

from generic_celery_task.decorators import task

• start Celery along with its broker. You no longer need to restart Celery after this.

• in another file, importable from celeryapp.py, create your task:

from celeryapp import task

@task
def job(x, y):
    return 'x + y = %d' % (x + y)

• now use it:

import celery

# direct function call
assert job(1, 2) == 'x + y = 3'

# using the delay() method
res = job.delay(1, 2)
# if the Celery daemon and its backend broker are running, 'res' is an instance of AsyncResult
if isinstance(res, celery.result.AsyncResult):
    assert res.get() == 'x + y = 3'
else:
    # if either one is not running, a warning is issued and the function is executed synchronously
    # with 'res' being the actual function result
    # if you want to silence the warnings, use '@task(quiet=True)'
    assert res == 'x + y = 3'

# using the apply_async() method
res = job.apply_async(args=[1, 2])
# and process the result as above, if you need to

installation

A setup.py is provided. You know what to do with it.

testing

The tests require nose, redis, redis-py and assume that the port 6389 is free.

Run the tests with “python setup.py test” or with “nosetests -v”.

This package was tested with python-2.7.6, python-3.3.4, nose-1.3.0, celery-3.1.10, redis-2.8.7 and redis-py-2.9.1 .

caveats

• the module which holds your custom task will be reloaded. If it contains a class using ‘super’ and its instance, you might run into the problem described here. Apply one of the proposed fixes.

• the state of the Celery daemon and its broker are checked only once, when the first .delay() or .apply_async() method is called on a custom task.

credits

• author: Stefan Talpalaru stefantalpalaru@yahoo.com

• homepage: https://github.com/stefantalpalaru/generic_celery_task