Python Custom Process Pool Instance Analysis [Producer and Consumer Model Issues]

This article analyzes the Python custom process pool through examples. Share it with everyone for your reference, the details are as follows:

The code explains everything:

#encoding=utf-8
#author: walker
#date: 2014-05-21
#function: 自定义进程池遍历目录下文件
from multiprocessing import Process, Queue, Lock
import time, os
#消费者
class Consumer(Process):
  def __init__(self, queue, ioLock):
    super(Consumer, self).__init__()
    self.queue = queue
    self.ioLock = ioLock
  def run(self):
    while True:
      task = self.queue.get()  #队列中无任务时，会阻塞进程
      if isinstance(task, str) and task == 'quit':
        break;
      time.sleep(1)  #假定任务处理需要1秒钟
      self.ioLock.acquire()
      print( str(os.getpid()) + ' ' + task)
      self.ioLock.release()
    self.ioLock.acquire()
    print 'Bye-bye'
    self.ioLock.release()
#生产者
def Producer():
  queue = Queue()  #这个队列是进程/线程安全的
  ioLock = Lock()
  subNum = 4  #子进程数量
  workers = build_worker_pool(queue, ioLock, subNum)
  start_time = time.time()
  for parent, dirnames, filenames in os.walk(r'D:\test'):
    for filename in filenames:
      queue.put(filename)
      ioLock.acquire()
      print('qsize:' + str(queue.qsize()))
      ioLock.release()
      while queue.qsize() > subNum * 10: #控制队列中任务数量
        time.sleep(1)
  for worker in workers:
    queue.put('quit')
  for worker in workers:
    worker.join()
  ioLock.acquire()
  print('Done! Time taken: {}'.format(time.time() - start_time))
  ioLock.release()
#创建进程池
def build_worker_pool(queue, ioLock, size):
  workers = []
  for _ in range(size):
    worker = Consumer(queue, ioLock)
    worker.start()
    workers.append(worker)
  return workers
if __name__ == '__main__':
  Producer()

ps:

self.ioLock.acquire()
...
self.ioLock.release()

Available:

with self.ioLock:
  ...

Alternative.

Another fun example:

#encoding=utf-8
#author: walker
#date: 2016-01-06
#function: 一个多进程的好玩例子
import os, sys, time
from multiprocessing import Pool
cur_dir_fullpath = os.path.dirname(os.path.abspath(__file__))
g_List = ['a']
#修改全局变量g_List
def ModifyDict_1():
  global g_List
  g_List.append('b')
#修改全局变量g_List
def ModifyDict_2():
  global g_List
  g_List.append('c')
#处理一个
def ProcOne(num):
  print('ProcOne ' + str(num) + ', g_List:' + repr(g_List))
#处理所有
def ProcAll():
  pool = Pool(processes = 4)
  for i in range(1, 20):
    #ProcOne(i)
    #pool.apply(ProcOne, (i,))
    pool.apply_async(ProcOne, (i,))
  pool.close()
  pool.join()
ModifyDict_1() #修改全局变量g_List
if __name__ == '__main__':
  ModifyDict_2() #修改全局变量g_List
  print('In main g_List :' + repr(g_List))
  ProcAll()

The result of running under Windows 7:

λ python3 demo.py
In main g_List :['a', 'b', 'c']
ProcOne 1, g_List:['a', 'b']
ProcOne 2, g_List:['a', 'b']
ProcOne 3, g_List:['a', 'b']
ProcOne 4, g_List:['a', 'b']
ProcOne 5, g_List:['a', 'b']
ProcOne 6, g_List:['a', 'b']
ProcOne 7, g_List:['a', 'b']
ProcOne 8, g_List:['a', 'b']
ProcOne 9, g_List:['a', 'b']
ProcOne 10, g_List:['a', 'b']
ProcOne 11, g_List:['a', 'b']
ProcOne 12, g_List:['a', 'b']
ProcOne 13, g_List:['a', 'b']
ProcOne 14, g_List:['a', 'b']
ProcOne 15, g_List:['a', 'b']
ProcOne 16, g_List:['a', 'b']
ProcOne 17, g_List:['a', 'b']
ProcOne 18, g_List:['a', 'b']
ProcOne 19, g_List:['a', 'b']

The result of running under Ubuntu 14.04:

In main g_List :['a', 'b', 'c']
ProcOne 1, g_List:['a', 'b', 'c']
ProcOne 2, g_List:['a', 'b', 'c']
ProcOne 3, g_List:['a', 'b', 'c']
ProcOne 5, g_List:['a', 'b', 'c']
ProcOne 4, g_List:['a', 'b', 'c']
ProcOne 8, g_List:['a', 'b', 'c']
ProcOne 9, g_List:['a', 'b', 'c']
ProcOne 7, g_List:['a', 'b', 'c']
ProcOne 11, g_List:['a', 'b', 'c']
ProcOne 6, g_List:['a', 'b', 'c']
ProcOne 12, g_List:['a', 'b', 'c']
ProcOne 13, g_List:['a', 'b', 'c']
ProcOne 10, g_List:['a', 'b', 'c']
ProcOne 14, g_List:['a', 'b', 'c']
ProcOne 15, g_List:['a', 'b', 'c']
ProcOne 16, g_List:['a', 'b', 'c']
ProcOne 17, g_List:['a', 'b', 'c']
ProcOne 18, g_List:['a', 'b', 'c']
ProcOne 19, g_List:['a', 'b', 'c']

It can be seen that the second modification under Windows 7 was not successful, but the modification under Ubuntu was successful. According to limodou, the author of uliweb, the reason is that under Windows, the child process is implemented by restarting; under Linux, it is implemented by fork.

For more Python custom process pool instance analysis [producer and consumer model issues] related articles, please pay attention to the PHP Chinese website!