I would like to which is the best way to use celery with tasks registered at runtime. My workfflow is as follows:
The way I have accomplished is the based on the "plugin" concept based on the same idea that the click package has with custom subcommands has.
The app structure (based on python 3):
.
├── dynamic_tasks.py
├── run.py
└── tasks
└── get_rate.py
The celery task dynamic_tasks.py is defined as following:
import os
import celery
app = celery.Celery('dynamic_tasks', broker='amqp://guest@192.168.169.1/', backend='rpc://')
PLUGIN_FOLDER = os.path.join(os.path.dirname(__file__), 'tasks')
def _absolutepath(filename):
""" Return the absolute path to the filename"""
return os.path.join(PLUGIN_FOLDER, filename)
@app.task
def tasks(funcname, *args, **kwargs):
try:
funcname = funcname.replace('-', '_')
funcname += '.py'
func = _absolutepath(funcname)
ns = {}
with open(func) as f:
code = compile(f.read(), func, 'exec')
eval(code, ns, ns)
return ns['task'](*args, **kwargs)
except IOError as e:
# Manage IOError
raise e
The plugable task example tasks/get_rate.py:
""" This task get the currency rate between a pair of currencies """
import urllib.request
URL = 'http://finance.yahoo.com/d/quotes.csv?s={}=X&f=p'
def task(pair='EURSEK', url_tmplt=URL):
with urllib.request.urlopen(url_tmplt.format(pair)) as res:
body = res.read()
return (pair, float(body.strip()))
And, simply, to run the example from run.py:
from dynamic_tasks import tasks
print(tasks.delay('get_rate', 'EURSEK').get())
EDITED Since celery runs on differents machine it is not possible to rely on the local filesystem. My new approach is to send the function to execute as string:
@app.task
def dynamic_tasks(funcname, funccode, *args, **kwargs):
try:
ns = {}
code = compile(funccode, funcname, 'exec')
eval(code, ns, ns)
logger.info('execute %r with args %r, %r', funcname, args, kwargs)
return ns['task'](*args, **kwargs)
except IOError:
logger.error("Error loading the dynamic function from text %s", funcname)