python多处理回调

Question

i have a list of functions that does some job like download html from a url(each function is very different so i can't make a single function to accept url and downlaod). i have used multiprocessing to speed up the task. below is my code

def runInParallel(list_of_functions):
  for fn in list_of_functions:
    proc = [Process(target=fn[1]).start() for fn in list_of_functions]
  for p in proc:
    p.join()

what i want is how to store the result that each function returns? each function returns a dict that i need to parse and store in database and i dont want to repeat these steps in each function so what i want is some sort of callback that can be passed with results returned from fucntions. how can i achieve that?

EDIT: using pool but throws error. i have following for list_of_functions :

[('f1', <function f1 at 0x7f34c11c9ed8>), ('f2', <function f2 at 0x7f34c11c9f50>)]


def runInParallel(list_of_functions):
    import multiprocessing
    pool = multiprocessing.Pool(processes = 3)
    x = pool.map(lambda f: f(), list_of_functions)
    print x




File "main.py", line 31, in <module>
    runInParallel(all_functions)
  File "main.py", line 11, in runInParallel
    x = pool.map(lambda f: f(), list_of_functions)
  File "/usr/lib/python2.7/multiprocessing/pool.py", line 251, in map
    return self.map_async(func, iterable, chunksize).get()
  File "/usr/lib/python2.7/multiprocessing/pool.py", line 558, in get
    raise self._value
cPickle.PicklingError: Can't pickle <type 'function'>: attribute lookup __builtin__.function failed

Answer 1

As mentioned in the comments above: if you use Process directly you need to set up a queue where the processes put into, so you can get from the parent process:

from multiprocessing import Process, Queue
from time import sleep

def f1(queue):
    sleep(1) # get url, "simulated" by sleep
    queue.put(dict(iam="type 1"))

def f2(queue):
    sleep(1.5)
    queue.put(dict(iam="type 2"))

def f3(queue):
    sleep(0.5)
    queue.put(dict(iam="type 3"))


def runInParallel(list_of_functions):
    queue = Queue()
    for fn in list_of_functions:
        proc = [Process(target=fn[1], args=(queue,)) for fn in list_of_functions]
    for p in proc:
        p.start()
    res = []
    for p in proc:
        p.join()
        res.append(queue.get())
    return res

if __name__ == '__main__':
    list_of_functions = [("f1", f1), ("f2", f2), ("f3", f3)]
    for d in runInParallel(list_of_functions):
        print d

Prints:

{'iam': 'type 3'}
{'iam': 'type f1'}
{'iam': 'type f2'}

If your functions basically do all the same (fetching urls and process the html in some way), then merging your functions into one with some if / elif logic allows you to use map and you would not need any queue:

from multiprocessing import Pool
from time import sleep

def f(arg):
    url, typ = arg
    if typ == 'a':
        sleep(1) # instead you would do something with `url` here
        return dict(iam="type 1", url=url)
    elif typ == 'b':
        sleep(1.5)
        return dict(iam="type 2", url=url)
    elif typ == 'c':
        sleep(0.5)
        return dict(iam="type 3", url=url)

def runInParallel(work):
    p = Pool(3)
    return p.map(f, work)

if __name__ == '__main__':
    work = [('http://url1', 'a'),
        ('http://url2', 'b'),
        ('http://url3', 'c'),
        ]
    for d in runInParallel(work):
        print d

Prints:

{'url': 'http://url1', 'iam': 'type 1'}
{'url': 'http://url2', 'iam': 'type 2'}
{'url': 'http://url3', 'iam': 'type 3'}

Both scripts work both on Windows as in Unix environments (tried it on OSX)