- 進程
進程的概念是需要理解的,進程是操作系統(tǒng)中正在運行的一個程序?qū)嵗?,操作系統(tǒng)通過進程操作原語來對其進行調(diào)度。操作系統(tǒng)得到調(diào)用某個進程指令時,將硬盤上的程序調(diào)入內(nèi)存,分配空間,初始化進程堆棧,然后進程開始運行。有時候我們有同時運行多個程序的需求,如果你的電腦只能做一件事,那是一件很抓狂的事。操作系統(tǒng)通過進程調(diào)度算法調(diào)度進程運行,使計算機看起來同時運行了很多程序。 - python中多進程實現(xiàn)
- fork,fork是linux下創(chuàng)建新進程的機制,通過fork父進程復制出一個相似,通過fork返回值判斷執(zhí)行子進程代碼
得到如下輸出import os def main(): print 'current Process {} start...'.format(os.getpid()) pid = os.fork() if pid < 0: print 'fork error!' exit(1) elif pid == 0: print 'child Process {} starting... , and my parent process is {}'.format(os.getpid(), os.getppid()) else: print 'I({}) created the child({})'.format(os.getpid(), pid) if __name__ == '__main__': main()current Process 5351 start... I(5351) created the child(5352) child Process 5352 starting... , and my parent process is 5351 - 使用multiprocess模塊創(chuàng)建子進程,模塊提供一個Process對象描述進程,創(chuàng)建進程時,只需要傳入一個可調(diào)用的函數(shù),以及函數(shù)運行時的參數(shù)即可
import os import multiprocessing def run_proc(name): print 'child process {}({}) running...'.format(name, os.getpid()) def main(): print 'main process starting... {}'.format(os.getpid()) processes = [] for i in range(5): p = multiprocessing.Process(target=run_proc, args=(str(i),)) processes.append(p) print 'process {} will start' p.start() for p in processes: p.join() print 'processes end' if __name__ == '__main__': main()- 使用進程池限制進程個數(shù)。multiprocessing模塊中的Pool對象,用來表示進程池,Pool對象的apply_async函數(shù)用于創(chuàng)建進程,同樣的給出可調(diào)用的函數(shù)與函數(shù)運行需要的參數(shù)
import os from multiprocessing import Pool def run_proc(name): print 'child process {}({}) running...'.format(name, os.getpid()) def main(): print 'main process starting... {}'.format(os.getpid()) processes = Pool(processes=3) for i in range(5): processes.apply_async(run_proc,(str(i),)) processes.close() processes.join() print 'processes end' if __name__ == '__main__': main()
- 使用進程池限制進程個數(shù)。multiprocessing模塊中的Pool對象,用來表示進程池,Pool對象的apply_async函數(shù)用于創(chuàng)建進程,同樣的給出可調(diào)用的函數(shù)與函數(shù)運行需要的參數(shù)
- 進程間通信
- 通過隊列。
隊列,即multiprocessing模塊中的Queue對象,隊列中有某種資源,可以向隊列中放入數(shù)據(jù),另一個進程從隊列中取出數(shù)據(jù),當無數(shù)據(jù)可用時,消費者應該決定是阻塞等待資源還是返回一個錯誤,當隊列已滿,生產(chǎn)者應決定是阻塞等待可用空間還是返回錯誤。Queue對象有兩個主要方法,get和put,get從隊列中取出數(shù)據(jù),put向隊列中添加數(shù)據(jù)。blocked參數(shù)決定當隊列不滿足條件時是阻塞等待還是返回錯誤,默認為True,表示阻塞等待。timeout指定了隊列阻塞的時間,如果超時,同樣返回異常from multiprocessing import Queue, Process import os, time, random def Proc_writer(q, urls): print 'Process {} is writing...'.format(os.getpid()) for url in urls: q.put(url) print 'put {} to the Queue'.format(url) time.sleep(random.random()) def Proc_reader(q): print 'Process {} is reading...'.format(os.getpid()) while True: url = q.get(True) print 'get the {} from the Queue'.format(url) def main(): print 'main process {} is running...'.format(os.getpid()) q = Queue() process_1 = Process(target=Proc_writer, args=(q,['url_1', 'url_2', 'url_3'])) process_2 = Process(target=Proc_writer, args=(q,['url_4', 'url_5', 'url_6'])) process_3 = Process(target=Proc_reader, args=(q,)) process_1.start() process_2.start() process_3.start() process_1.join() process_2.join() process_3.terminate() print 'done' if __name__ == '__main__': main() - 通過管道
multiprocessing模塊的Pipe方法,返回一個二元組(conn1,conn2),Pipe方法有一個duplex參數(shù),為True時代表管道連接是全雙工的,為False時代表管道連接是單方向的,只能由conn2發(fā)送到conn1。send和recv方法用于發(fā)送與接受消息,如果沒有消息可接受,recv阻塞,如果管道關(guān)閉,recv會拋出EOFErrorimport multiprocessing import os, time, random def proc_send(pipe, urls): print 'process {} is read to send urls'.format(os.getpid()) for url in urls: pipe.send(url) print 'process {}: send {}'.format(os.getpid(), url) time.sleep(random.random()) def proc_recv(pipe): print 'process {} is ready to recv urls'.format(os.getpid()) while True: print 'process {}: recv {}'.format(os.getpid(), pipe.recv()) time.sleep(random.random()) def main(): pipe = multiprocessing.Pipe() process_send = multiprocessing.Process( target=proc_send, args=(pipe[0], ['url_' + str(i) for i in range(10)])) process_recv = multiprocessing.Process( target=proc_recv, args=(pipe[1],) ) process_send.start() process_recv.start() process_send.join() process_recv.join() print 'done' if __name__ == '__main__': main()
- 通過隊列。
- 分布式多進程
分布式也是一個比較重要的概念,通過將負載高的計算分攤到多臺計算機上來提高系統(tǒng)性能。使用python完成分布式計算功能是簡單的。需要用到的一個數(shù)據(jù)結(jié)構(gòu)是隊列,聯(lián)想一下操作系統(tǒng)中的生產(chǎn)者消費者模型,一些進程放入數(shù)據(jù),一些進程取出數(shù)據(jù)。程序開始需要在服務端維護一個網(wǎng)絡隊列管理器,服務端程序注冊操作網(wǎng)絡隊列的方法,隨后使用該方法從網(wǎng)絡上獲取隊列,對該隊列的操作,對網(wǎng)絡上的其他進程是可見的。隊列的put和get方法用于放入取出數(shù)據(jù),注意服務端和客戶端注冊的接口方法需統(tǒng)一。
使用multiprocessing子模塊managers管理網(wǎng)絡隊列,其中的BaseManager類是一個基本的管理器,新建類繼承該類。使用該類的register方法注冊操作隊列的方法,隨后監(jiān)聽信道。如下例程
server
client#!/usr/bin/env python import Queue from multiprocessing.managers import BaseManager # 創(chuàng)建隊列實體 task_queue = Queue.Queue() result_queue = Queue.Queue() class Queuemanager(BaseManager): pass # 注冊方法 print 'register the func' Queuemanager.register('get_task_queue', callable=lambda:task_queue) Queuemanager.register('get_result_queue', callable=lambda:result_queue) # 創(chuàng)建manager對象 print 'initialing the task manager' manager = Queuemanager(address=('192.168.56.1', 8000), authkey='password') # 開始監(jiān)聽 manager.start() # 從網(wǎng)絡得到隊列 print 'get the queue from network...' task = manager.get_task_queue() result = manager.get_result_queue() # 向隊列中放入數(shù)據(jù)等待處理 print 'put urls to the task queue' for url in ['ImageUrl_' + str(i) for i in range(10)]: print 'put {} in task'.format(url) task.put(url) # 從隊列中取出數(shù)據(jù),阻塞等待 for i in range(10): print 'result is {}'.format(result.get()) manager.shutdown()#!/usr/bin/env python from multiprocessing.managers import BaseManager import Queue class Queuemanager(BaseManager): pass Queuemanager.register('get_task_queue') Queuemanager.register('get_result_queue') server = '192.168.56.1' port = 8000 key = 'password' print 'try to connect to {}'.format(server) manager = Queuemanager(address=(server, port), authkey=key) manager.connect() task = manager.get_task_queue() result = manager.get_result_queue() while not task.empty(): image_url = task.get(True, timeout=10) print 'run task download {}'.format(image_url) result.put(image_url + '------>completed!') print 'worker exit!'
- fork,fork是linux下創(chuàng)建新進程的機制,通過fork父進程復制出一個相似,通過fork返回值判斷執(zhí)行子進程代碼
- 線程
線程是一個存在于進程中的概念,用于在進程中并行完成不同的工作。線程與進程的不同另做介紹 - python中的多線程
-
threading推薦使用的多線程模塊
threading中的模塊對象
threading中的常見方法
Thread類
初始化一個thread類來創(chuàng)建一個線程,我們可以
- 初始化Thread類,傳入我們要運行的函數(shù)與參數(shù)
- 初始化Thread類,傳入可調(diào)用對象,比如自定義可調(diào)用類
- 創(chuàng)建類繼承Thread,覆蓋run函數(shù)
-
threading模塊實例
直接使用thread類#!/usr/bin/env python import threading from time import ctime, sleep secLoop = [6, 4] def loop(sec, i): print 'loop', i, 'start at', ctime() sleep(sec) print 'loop', i, 'finished at', ctime() def main(): nloop = range(len(secLoop)) threads = [] for i in nloop: threads.append(threading.Thread(target=loop, args=(secLoop[i], i))) for i in nloop: threads[i].start() for i in nloop: threads[i].join() if __name__ == '__main__': main()自定義可調(diào)用類
import threading from time import ctime, sleep secLoop = [6, 4] def loop(sec, i): print 'loop', i, 'start at', ctime() sleep(sec) print 'loop', i, 'finished at', ctime() class ThreadFunc(object): def __init__(self, func, args): self.func = func self.args = args def __call__(self): apply(self.func,self.args) def main(): nloop = range(len(secLoop)) threads = [] for i in nloop: threads.append(threading.Thread(target=ThreadFunc(loop,(secLoop[i],i)))) for i in nloop: threads[i].start() for i in nloop: threads[i].join() if __name__ == '__main__': main()
-


