Given the celery instance, it inspects all available celery workers to get the information about the queues they serve and tasks they know about.
Then it creates a chain of attributes allowing to execute any task as
queue_name.full_task_name.delay
.
With help of this class you can turn your Celery installation to a set of independent modules, each of which "exposes" its own "Celery API".
To make it more clear, the analogy with a random HTTP-based API available at http://example.com/users/[email protected] can be like:
- Celery object (including broker URL, result backend settings, etc) is the analogue of the protocol (http://)
- Queue name is the analogue of the hostname (example.com)
- Task name is the analogue of the URL path (/users/get)
- Task parameters the the analogue the querystring ([email protected])
Usage example.
If we have a Celery installation with two queues: "download" (knows how to execute "downloader.download_url" task) and "parse" (knows how to execute "parser.parse_html"), we can instantiate API and work with it the following way:
>>> api = celery_api.CeleryApi(celery)
>>> html_page = api.download.downloader.download_html.delay('http://example.com').get()
>>> html_tree = api.parse.parser.parse_html.delay(html_page).get()
Note
Ensure that workers are up and available from clients for inspection.
You may re-discover your installation after object creation by executing
api._discover()
.
Note
If you can't launch the task and get a NotRegistered exception instead,
it's most likely you have the CELERY_ALWAYS_EAGER
config option set to
True
.