因为GIL(全局解释器锁)的限制(GIL是用来保证在任意时刻只能有一个控制线程在执行),所以python中的多线程并非真正的多线程。只有python程序是I/O密集型应用时,多线程才会对运行效率有显著提高(因在等待I/O的时,会释放GIL允许其他线程继续执行),而在计算密集型应用中,多线程并没有什么用处。考虑到要充分利用多核CPU的资源,允许python可以并行处理一些任务,这里就用到了python多进程编程了。multiprocessing是python中的多进程模块,使用这个模块可以方便地进行多进程应用程序开发。multiprocessing模块中提供了:Process、Pool、Queue、Manager等组件。
1 Process类
1.1 构造方法
def __init__(self, group=None, target=None, name=None, args=(), kwargs={})
group:进程所属组,基本不用
target:进程调用对象(可以是一个函数名,也可以是一个可调用的对象(实现了__call__方法的类))
args:调用对象的位置参数元组
name:别名
kwargs:调用对象的关键字参数字典
1.2 实例方法
is_alive():返回进程是否在运行
start():启动进程,等待CPU调度
join([timeout]):阻塞当前上下文环境,直到调用此方法的进程终止或者到达指定timeout
terminate():不管任务是否完成,立即停止该进程
run():start()调用该方法,当实例进程没有传入target参数,stat()将执行默认的run()方法
1.3 属性
authkey:
daemon:守护进程标识,在start()调用之前可以对其进行修改
exitcode:进程的退出状态码
name:进程名
pid:进程id
1.4 实例
实例一:传入的target为一个函数
#!/usr/bin/python
#coding=utf-8
import time
import random
from multiprocessing import Process
def foo(i):
print time.ctime(), "process the %d begin ......" %i
time.sleep(random.uniform(1,3))
print time.ctime(), "process the %d end !!!!!!" %i
if __name__ == "__main__":
print time.ctime(), "process begin......"
p_lst = list()
for i in range(4):
p_lst.append(Process(target=foo, args=(i,))) #创建4个子进程
#启动子进程
for p in p_lst:
p.start()
#等待子进程全部结束
for p in p_lst:
p.join()
print time.ctime(), "process end!!!!!"
实例二:传入的target为一个可调用对象
#!/usr/bin/python
#coding=utf-8
import time
import random
from multiprocessing import Process
class Foo(object):
def __init__(self, i):
self.i = i
def __call__(self):
'''
使Foo的实例对象成为可调用对象
'''
print time.ctime(), "process the %d begin ......" %self.i
time.sleep(random.uniform(1,3))
print time.ctime(), "process the %d end !!!!!!" %self.i
if __name__ == "__main__":
print time.ctime(), "process begin......"
p_lst = list()
for i in range(4):
p_lst.append(Process(target=Foo(i))) #创建4个子进程
#启动子进程
for p in p_lst:
p.start()
#等待子进程全部结束
for p in p_lst:
p.join()
print time.ctime(), "process end!!!!!"
实例三:派生Process子类,并创建子类的实例
#!/usr/bin/python
#coding=utf-8
import time
import random
from multiprocessing import Process
class MyProcess(Process):
def __init__(self, i):
Process.__init__(self)
self.i = i
def run(self):
'''
#重写run方法,当调用start方法时,就会调用当前重写的run方法中的程序
'''
print time.ctime(), "process the %d begin ......" %self.i
time.sleep(random.uniform(1,3))
print time.ctime(), "process the %d end !!!!!!" %self.i
if __name__ == "__main__":
print time.ctime(), "process begin......"
p_lst = list()
for i in range(4):
p_lst.append(MyProcess(i)) #创建4个子进程
#启动子进程
for p in p_lst:
p.start()
#等待子进程全部结束
for p in p_lst:
p.join()
print time.ctime(), "process end!!!!!"
2 Pool类
当使用Process类管理非常多(几十上百个)的进程时,就会显得比较繁琐,这是就可以使用Pool(进程池)来对进程进行统一管理。当池中进程已满时,有新进程请求执行时,就会被阻塞,直到池中有进程执行结束,新的进程请求才会被放入池中并执行。
2.1 构造方法
def __init__(self, processes=None, initializer=None, initargs=(), maxtasksperchild=None)
processes:池中可容纳的工作进程数量,默认情况使用os.cpu_count()返回的数值,一般默认即可
其他参数暂不清楚有什么用处......
2.2 实例方法
apply(self, func, args=(), kwds={}):阻塞型进程池,会阻塞主进程,直到工作进程全部退出,一般不用这个
apply_async(self, func, args=(), kwds={}, callback=None):非阻塞型进程池
map(self, func, iterable, chunksize=None):与内置map行为一致,它会阻塞主进程,直到map运行结束
map_async(self, func, iterable, chunksize=None, callback=None):非阻塞版本的map
close():关闭进程池,不在接受新任务
terminate():结束工作进程
join():阻塞主进程等待子进程退出,该方法必须在close或terminate之后执行
2.3 实例
#!/usr/bin/python
#coding=utf-8
import time
import random
from multiprocessing import Pool
def foo(i):
print time.ctime(), "process the %d begin ......" %i
time.sleep(random.uniform(1,3))
print time.ctime(), "process the %d end !!!!!!" %i
if __name__ == "__main__":
print time.ctime(), "process begin......"
pool = Pool(processes = 2) #设置进程池中最大并行工作进程数为2
for i in range(4):
pool.apply_async(foo, args=(i,)) #提交4个子进程任务
pool.close()
pool.join()
print time.ctime(), "process end!!!!!"
结果:
Fri Nov 18 13:57:22 2016 process begin......
Fri Nov 18 13:57:22 2016 process the 0 begin ......
Fri Nov 18 13:57:22 2016 process the 1 begin ......
Fri Nov 18 13:57:23 2016 process the 1 end !!!!!!
Fri Nov 18 13:57:23 2016 process the 2 begin ......
Fri Nov 18 13:57:24 2016 process the 0 end !!!!!!
Fri Nov 18 13:57:24 2016 process the 3 begin ......
Fri Nov 18 13:57:25 2016 process the 2 end !!!!!!
Fri Nov 18 13:57:25 2016 process the 3 end !!!!!!
Fri Nov 18 13:57:25 2016 process end!!!!!
3 Queue类
Queue主要提供进程间通信以及共享数据等功能。除Queue外还可以使用Pipes实现进程间通信(Pipes是两个进程间进行通信)
3.1 构造方法
def __init__(self, maxsize=0)
maxsize:用于设置队列最大长度,当为maxsize<=0时,队列的最大长度会被设置为一个非常大的值(我的系统中队列最大长度被设置为2147483647)
3.2 实例方法
put(self, obj, block=True, timeout=None)
1、block为True,若队列已满,并且timeout为正值,该方法会阻塞timeout指定的时间,直到队列中有出现剩余空间,如果超时,会抛出Queue.Full异常
2、block为False,若队列已满,立即抛出Queue.Full异常
get(self, block=True, timeout=None)
block为True,若队列为空,并且timeout为正值,该方法会阻塞timeout指定的时间,直到队列中有出现新的数据,如果超时,会抛出Queue.Empty异常
block为False,若队列为空,立即抛出Queue.Empty异常
3.3 实例
#!/usr/bin/python
#coding=utf-8
import time
import random
from multiprocessing import Process, Queue
def write(q):
for value in "abcd":
print time.ctime(), "put %s to queue" %value
q.put(value)
time.sleep(random.random())
def read(q):
while True:
value = q.get()
print time.ctime(), "get %s from queue" %value
if __name__ == "__main__":
#主进程创建Queue,并作为参数传递给子进程
q = Queue()
pw = Process(target=write, args=(q,))
pr = Process(target=read, args=(q,))
#启动子进程pw,往Queue中写入
pw.start()
#启动子进程pr,从Queue中读取
pr.start()
#等待写进程执行结束
pw.join()
#终止读取进程
pr.terminate()
运行结果:
Fri Nov 18 15:04:13 2016 put a to queue
Fri Nov 18 15:04:13 2016 get a from queue
Fri Nov 18 15:04:13 2016 put b to queue
Fri Nov 18 15:04:13 2016 get b from queue
Fri Nov 18 15:04:13 2016 put c to queue
Fri Nov 18 15:04:13 2016 get c from queue
Fri Nov 18 15:04:13 2016 put d to queue
Fri Nov 18 15:04:13 2016 get d from queue
4 Manager类
Manager是进程间数据共享的高级接口。
Manager()返回的manager对象控制了一个server进程,此进程包含的python对象可以被其他的进程通过proxies来访问。从而达到多进程间数据通信且安全。Manager支持的类型有list, dict, Namespace, Lock, RLock, Semaphore, BoundedSemaphore, Condition, Event, Queue, Value和Array。
如下是使用Manager管理一个用于多进程共享的dict数据
#!/usr/bin/python
#coding=utf-8
import time
import random
from multiprocessing import Manager, Pool
def worker(d, key, value):
print time.ctime(), "insert the k-v pair to dict begin: {%d: %d}" %(key, value)
time.sleep(random.uniform(1,2))
d[key] = value #访问共享数据
print time.ctime(), "insert the k-v pair to dict end: {%d: %d}" %(key, value)
if __name__ == "__main__":
print time.ctime(), "process for manager begin"
mgr = Manager()
d = mgr.dict()
pool = Pool(processes=4)
for i in range(10):
pool.apply_async(worker, args=(d, i, i*i))
pool.close()
pool.join()
print "Result:"
print d
print time.ctime(), "process for manager end"
运行结果
Fri Nov 18 16:36:19 2016 process for manager begin
Fri Nov 18 16:36:19 2016 insert the k-v pair to dict begin: {0: 0}
Fri Nov 18 16:36:19 2016 insert the k-v pair to dict begin: {1: 1}
Fri Nov 18 16:36:19 2016 insert the k-v pair to dict begin: {2: 4}
Fri Nov 18 16:36:19 2016 insert the k-v pair to dict begin: {3: 9}
Fri Nov 18 16:36:20 2016 insert the k-v pair to dict end: {3: 9}
Fri Nov 18 16:36:20 2016 insert the k-v pair to dict begin: {4: 16}
Fri Nov 18 16:36:20 2016 insert the k-v pair to dict end: {0: 0}
Fri Nov 18 16:36:20 2016 insert the k-v pair to dict begin: {5: 25}
Fri Nov 18 16:36:21 2016 insert the k-v pair to dict end: {2: 4}
Fri Nov 18 16:36:21 2016 insert the k-v pair to dict begin: {6: 36}
Fri Nov 18 16:36:21 2016 insert the k-v pair to dict end: {1: 1}
Fri Nov 18 16:36:21 2016 insert the k-v pair to dict begin: {7: 49}
Fri Nov 18 16:36:21 2016 insert the k-v pair to dict end: {5: 25}
Fri Nov 18 16:36:21 2016 insert the k-v pair to dict begin: {8: 64}
Fri Nov 18 16:36:22 2016 insert the k-v pair to dict end: {4: 16}
Fri Nov 18 16:36:22 2016 insert the k-v pair to dict begin: {9: 81}
Fri Nov 18 16:36:23 2016 insert the k-v pair to dict end: {8: 64}
Fri Nov 18 16:36:23 2016 insert the k-v pair to dict end: {6: 36}
Fri Nov 18 16:36:23 2016 insert the k-v pair to dict end: {7: 49}
Fri Nov 18 16:36:23 2016 insert the k-v pair to dict end: {9: 81}
Result:
{0: 0, 1: 1, 2: 4, 3: 9, 4: 16, 5: 25, 6: 36, 7: 49, 8: 64, 9: 81}
Fri Nov 18 16:36:23 2016 process for manager end