Pebble 1.1.2

Threading and multiprocessing eye-candy.

Note

Since Pebble 1.0.0 decorators are functions and not classes anymore.

Description

Pebble provides a neat API to manage threads and processes within an application.

Examples

Launch a task in a thread and wait for its results:

from pebble import asynchronous


@asynchronous
def do_job(foo, bar=0):
    return foo + bar


if __name__ == "__main__":
    task = do_job(1, bar=2)
    print task.get()  # it will block until do_job has completed

Launch five tasks in separate processes and handle their results in a callback:

from pebble import concurrent


def task_done(task_id, results):
    print "Task %s has returned %d" % (task_id, results)


@concurrent(callback=task_done)
def do_job(foo, bar=0):
    return foo + bar


if __name__ == "__main__":
    for i in range(0, 5):
        do_job(i)

    raw_input("Press any key to exit.")

Callbacks can be dynamically (re)assigned, useful to set instance methods as callback:

from pebble import concurrent


class Foo(object):
    def __init__(self):
        self.counter = 0
        self.errors = 0
        self.do_job.callback = self.task_done
        self.do_wrong_job.error_callback = self.task_error

    def task_done(self, task_id, results):
        self.counter += results

    def task_error(self, task_id, exception):
        self.errors += 1

    @concurrent
    def do_job():
        return 1

    @concurrent
    def do_wrong_job():
        raise Exception("Ops!")


if __name__ == "__main__":
    foo = Foo()
    tasks = []

    for i in range(0, 5):
        task = foo.do_job()
        tasks.append(task)
        task = foo.do_wrong_job()
        tasks.append(task)

    for task in tasks:
        print "Waiting for %s" % task.id
        try:
            task.get()  # wait for task to finish
        except Exception:
            pass  # exception are re-raised by the get() method

    print foo.counter
    print foo.errors

TODO

A roadmap:

* return traceback in error callbacks
* implement pebbles (pools of workers) and make it possible to bind decorators to them
* memoization as decorator parameter