Using a a manager for updating a Queue in a Python multiprocess

问题

I am designing a Python multiprocessing code to work in a queue that might be updated along the processing. The following code sometimes works, or get stuck, or rises an Empty error.

import multiprocessing as mp

def worker(working_queue, output_queue):
    while True:
        if working_queue.empty() is True:
            break    
        else:
            picked = working_queue.get_nowait()
            if picked % 2 == 0: 
                    output_queue.put(picked)
            else:
                working_queue.put(picked+1)
    return

if __name__ == '__main__':
    manager = mp.Manager()
    static_input = xrange(100)    
    working_q = manager.Queue()
    output_q = mp.Queue()
    for i in static_input:
        working_q.put(i)
    processes = [mp.Process(target=worker,args=(working_q, output_q)) for i in range(mp.cpu_count())]
    for proc in processes:
        proc.start()
    for proc in processes:
        proc.join()
    results_bank = []
    while True:
       if output_q.empty() is True:
           break
       results_bank.append(output_q.get_nowait())
    print len(results_bank) # length of this list should be equal to static_input, which is the range used to populate the input queue. In other words, this tells whether all the items placed for processing were actually processed.
    results_bank.sort()
    print results_bank

Should I use a list as a global variable, and lock it, instead of a manager.Queue()?

回答1:

I just added a try: and except Exception: to handle the Empty error. The results seem to be consistent now. Please let me know If you find problems I overlooked in this solution.

import multiprocessing as mp

def worker(working_queue, output_queue):
    while True:
        try:
            if working_queue.empty() is True:
                break  
            else:
                picked = working_queue.get_nowait()
                if picked % 2 == 0: 
                        output_queue.put(picked)
                else:
                    working_queue.put(picked+1)
        except Exception:
            continue

    return

if __name__ == '__main__':
    #Manager seem to be unnecessary.
    #manager = mp.Manager()
    #working_q = manager.Queue()

    working_q = mp.Queue()
    output_q = mp.Queue()
    static_input = xrange(100)     
    for i in static_input:
        working_q.put(i)
    processes = [mp.Process(target=worker,args=(working_q, output_q)) for i in range(mp.cpu_count())]
    for proc in processes:
        proc.start()
    for proc in processes:
        proc.join()
    results_bank = []
    while True:
       if output_q.empty() is True:
           break
       results_bank.append(output_q.get_nowait())
    print len(results_bank) # length of this list should be equal to static_input, which is the range used to populate the input queue. In other words, this tells whether all the items placed for processing were actually processed.
    results_bank.sort()
    print results_bank

回答2:

Just use a lock to protect the access to shared data, it's safer (and will protect you from weird behavior of the process):

import multiprocessing as mp

def worker(working_queue, output_queue, lock):
    while True:
        shouldBeak = False
        lock.acquire()
        if working_queue.empty() is True:
            shouldBeak = True    
        else:

            picked = working_queue.get_nowait()
            if picked % 2 == 0: 
                output_queue.put(picked)
            else:
                working_queue.put(picked+1)
        lock.release()
        if shouldBeak:
            break
    return

if __name__ == '__main__':
    manager = mp.Manager()
    static_input = xrange(1000)    
    working_q = manager.Queue()
    output_q = mp.Queue()
    lock = mp.Lock()
    for i in static_input:
        working_q.put(i)
    processes = [mp.Process(target=worker,args=(working_q, output_q,lock)) for i in range(mp.cpu_count())]
    for proc in processes:
        proc.start()
    for proc in processes:
        proc.join()
    results_bank = []
    while True:
       if output_q.empty() is True:
           break
       results_bank.append(output_q.get_nowait())
    print len(results_bank) # length of this list should be equal to static_input, which is the range used to populate the input queue. In other words, this tells whether all the items placed for processing were actually processed.
    results_bank.sort()
    print results_bank

来源：https://stackoverflow.com/questions/21680903/using-a-a-manager-for-updating-a-queue-in-a-python-multiprocess