序.multiprocessing
python中的多线程其实并不是真正的多线程,如果想要充分地使用多核CPU的资源,在python中大部分情况需要使用多进程。Python提供了非常好用的多进程包multiprocessing,只需要定义一个函数,Python会完成其他所有事情。借助这个包,可以轻松完成从单进程到并发执行的转换。multiprocessing支持子进程、通信和共享数据、执行不同形式的同步,提供了Process、Queue、Pipe、Lock等组件。
1.Process
创建进程的类:Process([group [, target [, name [, args [, kwargs]]]]]),target表示调用对象,args表示调用对象的位置参数元组。kwargs表示调用对象的字典。name为别名。group实质上不使用。
方法:is_alive()、join([timeout])、run()、start()、terminate()。其中,Process以start()启动某个进程。
属性:authkey、daemon(要通过start()设置)、exitcode(进程在运行时为None、如果为–N,表示被信号N结束)、name、pid。其中daemon是父进程终止后自动终止,且自己不能产生新进程,必须在start()之前设置。
例1.1:创建函数并将其作为单个进程
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
|
import
multiprocessing
import
time
def
worker
(
interval
)
:
n
=
5
while
n
>
0
:
print
(
"The time is {0}"
.
format
(
time
.
ctime
(
)
)
)
time
.
sleep
(
interval
)
n
-=
1
if
__name__
==
"__main__"
:
p
=
multiprocessing
.
Process
(
target
=
worker
,
args
=
(
3
,
)
)
p
.
start
(
)
print
"p.pid:"
,
p
.
pid
print
"p.name:"
,
p
.
name
print
"p.is_alive:"
,
p
.
is_alive
(
)
|
结果
1
2
3
4
5
6
7
8
|
p
.
pid
:
8736
p
.
name
:
Process
-
1
p
.
is_alive
:
True
The
time
is
Tue
Apr
21
20
:
55
:
12
2015
The
time
is
Tue
Apr
21
20
:
55
:
15
2015
The
time
is
Tue
Apr
21
20
:
55
:
18
2015
The
time
is
Tue
Apr
21
20
:
55
:
21
2015
The
time
is
Tue
Apr
21
20
:
55
:
24
2015
|
例1.2:创建函数并将其作为多个进程
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
|
import
multiprocessing
import
time
def
worker_1
(
interval
)
:
print
"worker_1"
time
.
sleep
(
interval
)
print
"end worker_1"
def
worker_2
(
interval
)
:
print
"worker_2"
time
.
sleep
(
interval
)
print
"end worker_2"
def
worker_3
(
interval
)
:
print
"worker_3"
time
.
sleep
(
interval
)
print
"end worker_3"
if
__name__
==
"__main__"
:
p1
=
multiprocessing
.
Process
(
target
=
worker_1
,
args
=
(
2
,
)
)
p2
=
multiprocessing
.
Process
(
target
=
worker_2
,
args
=
(
3
,
)
)
p3
=
multiprocessing
.
Process
(
target
=
worker_3
,
args
=
(
4
,
)
)
p1
.
start
(
)
p2
.
start
(
)
p3
.
start
(
)
print
(
"The number of CPU is:"
+
str
(
multiprocessing
.
cpu_count
(
)
)
)
for
p
in
multiprocessing
.
active_children
(
)
:
print
(
"child p.name:"
+
p
.
name
+
"\tp.id"
+
str
(
p
.
pid
)
)
print
"END!!!!!!!!!!!!!!!!!"
|
结果
1
2
3
4
5
6
7
8
9
10
11
|
The
number
of
CPU
is
:
4
child
p
.
name
:
Process
-
3
p
.
id7992
child
p
.
name
:
Process
-
2
p
.
id4204
child
p
.
name
:
Process
-
1
p
.
id6380
END
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
worker_1
worker_3
worker_2
end
worker_1
end
worker_2
end
worker_3
|
例1.3:将进程定义为类
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
|
import
multiprocessing
import
time
class
ClockProcess
(
multiprocessing
.
Process
)
:
def
__init__
(
self
,
interval
)
:
multiprocessing
.
Process
.
__init__
(
self
)
self
.
interval
=
interval
def
run
(
self
)
:
n
=
5
while
n
>
0
:
print
(
"the time is {0}"
.
format
(
time
.
ctime
(
)
)
)
time
.
sleep
(
self
.
interval
)
n
-=
1
if
__name__
==
'__main__'
:
p
=
ClockProcess
(
3
)
p
.
start
(
)
|
注:进程p调用start()时,自动调用run()
结果
1
2
3
4
5
|
the
time
is
Tue
Apr
21
20
:
31
:
30
2015
the
time
is
Tue
Apr
21
20
:
31
:
33
2015
the
time
is
Tue
Apr
21
20
:
31
:
36
2015
the
time
is
Tue
Apr
21
20
:
31
:
39
2015
the
time
is
Tue
Apr
21
20
:
31
:
42
2015
|
例1.4:daemon程序对比结果
#1.4-1 不加daemon属性
1
2
3
4
5
6
7
8
9
10
11
12
|
import
multiprocessing
import
time
def
worker
(
interval
)
:
print
(
"work start:{0}"
.
format
(
time
.
ctime
(
)
)
)
;
time
.
sleep
(
interval
)
print
(
"work end:{0}"
.
format
(
time
.
ctime
(
)
)
)
;
if
__name__
==
"__main__"
:
p
=
multiprocessing
.
Process
(
target
=
worker
,
args
=
(
3
,
)
)
p
.
start
(
)
print
"end!"
|
结果
1
2
3
|
end
!
work
start
:
Tue
Apr
21
21
:
29
:
10
2015
work
end
:
Tue
Apr
21
21
:
29
:
13
2015
|
#1.4-2 加上daemon属性
1
2
3
4
5
6
7
8
9
10
11
12
13
|
import
multiprocessing
import
time
def
worker
(
interval
)
:
print
(
"work start:{0}"
.
format
(
time
.
ctime
(
)
)
)
;
time
.
sleep
(
interval
)
print
(
"work end:{0}"
.
format
(
time
.
ctime
(
)
)
)
;
if
__name__
==
"__main__"
:
p
=
multiprocessing
.
Process
(
target
=
worker
,
args
=
(
3
,
)
)
p
.
daemon
=
True
p
.
start
(
)
print
"end!"
|
结果
1
|
end
!
|
注:因子进程设置了daemon属性,主进程结束,它们就随着结束了。
#1.4-3 设置daemon执行完结束的方法
1
2
3
4
5
6
7
8
9
10
11
12
13
14
|
import
multiprocessing
import
time
def
worker
(
interval
)
:
print
(
"work start:{0}"
.
format
(
time
.
ctime
(
)
)
)
;
time
.
sleep
(
interval
)
print
(
"work end:{0}"
.
format
(
time
.
ctime
(
)
)
)
;
if
__name__
==
"__main__"
:
p
=
multiprocessing
.
Process
(
target
=
worker
,
args
=
(
3
,
)
)
p
.
daemon
=
True
p
.
start
(
)
p
.
join
(
)
print
"end!"
|
结果
1
2
3
|
work
start
:
Tue
Apr
21
22
:
16
:
32
2015
work
end
:
Tue
Apr
21
22
:
16
:
35
2015
end
!
|
2.Lock
当多个进程需要访问共享资源的时候,Lock可以用来避免访问的冲突。
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
|
import
multiprocessing
import
sys
def
worker_with
(
lock
,
f
)
:
with
lock
:
fs
=
open
(
f
,
'a+'
)
n
=
10
while
n
>
1
:
fs
.
write
(
"Lockd acquired via with\n"
)
n
-=
1
fs
.
close
(
)
def
worker_no_with
(
lock
,
f
)
:
lock
.
acquire
(
)
try
:
fs
=
open
(
f
,
'a+'
)
n
=
10
while
n
>
1
:
fs
.
write
(
"Lock acquired directly\n"
)
n
-=
1
fs
.
close
(
)
finally
:
lock
.
release
(
)
if
__name__
==
"__main__"
:
lock
=
multiprocessing
.
Lock
(
)
f
=
"file.txt"
w
=
multiprocessing
.
Process
(
target
=
worker_with
,
args
=
(
lock
,
f
)
)
nw
=
multiprocessing
.
Process
(
target
=
worker_no_with
,
args
=
(
lock
,
f
)
)
w
.
start
(
)
nw
.
start
(
)
print
"end"
|
结果(输出文件)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
|
Lockd
acquired
via
with
Lockd
acquired
via
with
Lockd
acquired
via
with
Lockd
acquired
via
with
Lockd
acquired
via
with
Lockd
acquired
via
with
Lockd
acquired
via
with
Lockd
acquired
via
with
Lockd
acquired
via
with
Lock
acquired
directly
Lock
acquired
directly
Lock
acquired
directly
Lock
acquired
directly
Lock
acquired
directly
Lock
acquired
directly
Lock
acquired
directly
Lock
acquired
directly
Lock
acquired
directly
|
3.Semaphore
Semaphore用来控制对共享资源的访问数量,例如池的最大连接数。
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
|
import
multiprocessing
import
time
def
worker
(
s
,
i
)
:
s
.
acquire
(
)
print
(
multiprocessing
.
current_process
(
)
.
name
+
"acquire"
)
;
time
.
sleep
(
i
)
print
(
multiprocessing
.
current_process
(
)
.
name
+
"release\n"
)
;
s
.
release
(
)
if
__name__
==
"__main__"
:
s
=
multiprocessing
.
Semaphore
(
2
)
for
i
in
range
(
5
)
:
p
=
multiprocessing
.
Process
(
target
=
worker
,
args
=
(
s
,
i
*
2
)
)
p
.
start
(
)
|
结果
1
2
3
4
5
6
7
8
9
10
11
12
13
14
|
Process
-
1acquire
Process
-
1release
Process
-
2acquire
Process
-
3acquire
Process
-
2release
Process
-
5acquire
Process
-
3release
Process
-
4acquire
Process
-
5release
Process
-
4release
|
4.Event
Event用来实现进程间同步通信。
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
|
import
multiprocessing
import
time
def
wait_for_event
(
e
)
:
print
(
"wait_for_event: starting"
)
e
.
wait
(
)
print
(
"wairt_for_event: e.is_set()->"
+
str
(
e
.
is_set
(
)
)
)
def
wait_for_event_timeout
(
e
,
t
)
:
print
(
"wait_for_event_timeout:starting"
)
e
.
wait
(
t
)
print
(
"wait_for_event_timeout:e.is_set->"
+
str
(
e
.
is_set
(
)
)
)
if
__name__
==
"__main__"
:
e
=
multiprocessing
.
Event
(
)
w1
=
multiprocessing
.
Process
(
name
=
"block"
,
target
=
wait_for_event
,
args
=
(
e
,
)
)
w2
=
multiprocessing
.
Process
(
name
=
"non-block"
,
target
=
wait_for_event_timeout
,
args
=
(
e
,
2
)
)
w1
.
start
(
)
w2
.
start
(
)
time
.
sleep
(
3
)
e
.
set
(
)
print
(
"main: event is set"
)
|
结果
1
2
3
4
5
|
wait_for_event
:
starting
wait_for_event_timeout
:
starting
wait_for_event_timeout
:
e
.
is_set
->
False
main
:
event
is
set
wairt_for_event
:
e
.
is_set
(
)
->
True
|
5.Queue
Queue是多进程安全的队列,可以使用Queue实现多进程之间的数据传递。put方法用以插入数据到队列中,put方法还有两个可选参数:blocked和timeout。如果blocked为True(默认值),并且timeout为正值,该方法会阻塞timeout指定的时间,直到该队列有剩余的空间。如果超时,会抛出Queue.Full异常。如果blocked为False,但该Queue已满,会立即抛出Queue.Full异常。
get方法可以从队列读取并且删除一个元素。同样,get方法有两个可选参数:blocked和timeout。如果blocked为True(默认值),并且timeout为正值,那么在等待时间内没有取到任何元素,会抛出Queue.Empty异常。如果blocked为False,有两种情况存在,如果Queue有一个值可用,则立即返回该值,否则,如果队列为空,则立即抛出Queue.Empty异常。Queue的一段示例代码:
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
|
import
multiprocessing
def
writer_proc
(
q
)
:
try
:
q
.
put
(
1
,
block
=
False
)
except
:
pass
def
reader_proc
(
q
)
:
try
:
print
q
.
get
(
block
=
False
)
except
:
pass
if
__name__
==
"__main__"
:
q
=
multiprocessing
.
Queue
(
)
writer
=
multiprocessing
.
Process
(
target
=
writer_proc
,
args
=
(
q
,
)
)
writer
.
start
(
)
reader
=
multiprocessing
.
Process
(
target
=
reader_proc
,
args
=
(
q
,
)
)
reader
.
start
(
)
reader
.
join
(
)
writer
.
join
(
)
|
结果
1
|
1
|
6.Pipe
Pipe方法返回(conn1, conn2)代表一个管道的两个端。Pipe方法有duplex参数,如果duplex参数为True(默认值),那么这个管道是全双工模式,也就是说conn1和conn2均可收发。duplex为False,conn1只负责接受消息,conn2只负责发送消息。
send和recv方法分别是发送和接受消息的方法。例如,在全双工模式下,可以调用conn1.send发送消息,conn1.recv接收消息。如果没有消息可接收,recv方法会一直阻塞。如果管道已经被关闭,那么recv方法会抛出EOFError。
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
|
import
multiprocessing
import
time
def
proc1
(
pipe
)
:
while
True
:
for
i
in
xrange
(
10000
)
:
print
"send: %s"
%
(
i
)
pipe
.
send
(
i
)
time
.
sleep
(
1
)
def
proc2
(
pipe
)
:
while
True
:
print
"proc2 rev:"
,
pipe
.
recv
(
)
time
.
sleep
(
1
)
def
proc3
(
pipe
)
:
while
True
:
print
"PROC3 rev:"
,
pipe
.
recv
(
)
time
.
sleep
(
1
)
if
__name__
==
"__main__"
:
pipe
=
multiprocessing
.
Pipe
(
)
p1
=
multiprocessing
.
Process
(
target
=
proc1
,
args
=
(
pipe
[
0
]
,
)
)
p2
=
multiprocessing
.
Process
(
target
=
proc2
,
args
=
(
pipe
[
1
]
,
)
)
#p3 = multiprocessing.Process(target=proc3, args=(pipe[1],))
p1
.
start
(
)
p2
.
start
(
)
#p3.start()
p1
.
join
(
)
p2
.
join
(
)
#p3.join()
|
结果
7.Pool
在利用Python进行系统管理的时候,特别是同时操作多个文件目录,或者远程控制多台主机,并行操作可以节约大量的时间。当被操作对象数目不大时,可以直接利用multiprocessing中的Process动态成生多个进程,十几个还好,但如果是上百个,上千个目标,手动的去限制进程数量却又太过繁琐,此时可以发挥进程池的功效。
Pool可以提供指定数量的进程,供用户调用,当有新的请求提交到pool中时,如果池还没有满,那么就会创建一个新的进程用来执行该请求;但如果池中的进程数已经达到规定最大值,那么该请求就会等待,直到池中有进程结束,才会创建新的进程来它。
例7.1:使用进程池
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
|
#coding: utf-8
import
multiprocessing
import
time
def
func
(
msg
)
:
print
"msg:"
,
msg
time
.
sleep
(
3
)
print
"end"
if
__name__
==
"__main__"
:
pool
=
multiprocessing.Pool
(
processes
=
3
)
for
i
in
xrange
(
4
)
:
msg
=
"hello %d"
%
(
i
)
pool
.
apply_async
(
func
,
(
msg
,
)
)
#维持执行的进程总数为processes,当一个进程执行完毕后会添加新的进程进去
print
"Mark~ Mark~ Mark~~~~~~~~~~~~~~~~~~~~~~"
pool
.
close
(
)
pool
.
join
(
)
#调用join之前,先调用close函数,否则会出错。执行完close后不会有新的进程加入到pool,join函数等待所有子进程结束
print
"Sub-process(es) done."
|
一次执行结果
1
2
3
4
5
6
7
8
9
10
|
mMsg
:
hark
~
Mark
~
Mark
~
~
~
~
~
~
~
~
~
~
~
~
~
~
~
~
~
~
~
~
~
~
ello
0
msg
:
hello
1
msg
:
hello
2
end
msg
:
hello
3
end
end
end
Sub
-
process
(
es
)
done
.
|
函数解释:
- apply_async(func[, args[, kwds[, callback]]]) 它是非阻塞,apply(func[, args[, kwds]])是阻塞的(理解区别,看例1例2结果区别)
- close() 关闭pool,使其不在接受新的任务。
- terminate() 结束工作进程,不在处理未完成的任务。
- join() 主进程阻塞,等待子进程的退出, join方法要在close或terminate之后使用。
执行说明:创建一个进程池pool,并设定进程的数量为3,xrange(4)会相继产生四个对象[0, 1, 2, 4],四个对象被提交到pool中,因pool指定进程数为3,所以0、1、2会直接送到进程中执行,当其中一个执行完事后才空出一个进程处理对象3,所以会出现输出“msg: hello 3”出现在”end”后。因为为非阻塞,主函数会自己执行自个的,不搭理进程的执行,所以运行完for循环后直接输出“mMsg: hark~ Mark~ Mark~~~~~~~~~~~~~~~~~~~~~~”,主程序在pool.join()处等待各个进程的结束。
例7.2:使用进程池(阻塞)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
|
#coding: utf-8
import
multiprocessing
import
time
def
func
(
msg
)
:
print
"msg:"
,
msg
time
.
sleep
(
3
)
print
"end"
if
__name__
==
"__main__"
:
pool
=
multiprocessing.Pool
(
processes
=
3
)
for
i
in
xrange
(
4
)
:
msg
=
"hello %d"
%
(
i
)
pool
.
apply
(
func
,
(
msg
,
)
)
#维持执行的进程总数为processes,当一个进程执行完毕后会添加新的进程进去
print
"Mark~ Mark~ Mark~~~~~~~~~~~~~~~~~~~~~~"
pool
.
close
(
)
pool
.
join
(
)
#调用join之前,先调用close函数,否则会出错。执行完close后不会有新的进程加入到pool,join函数等待所有子进程结束
print
"Sub-process(es) done."
|
一次执行的结果
1
2
3
4
5
6
7
8
9
10
|
msg
:
hello
0
end
msg
:
hello
1
end
msg
:
hello
2
end
msg
:
hello
3
end
Mark
~
Mark
~
Mark
~
~
~
~
~
~
~
~
~
~
~
~
~
~
~
~
~
~
~
~
~
~
Sub
-
process
(
es
)
done
.
|
例7.3:使用进程池,并关注结果
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
|
import
multiprocessing
import
time
def
func
(
msg
)
:
print
"msg:"
,
msg
time
.
sleep
(
3
)
print
"end"
return
"done"
+
msg
if
__name__
==
"__main__"
:
pool
=
multiprocessing.Pool
(
processes
=
4
)
result
=
[
]
for
i
in
xrange
(
3
)
:
msg
=
"hello %d"
%
(
i
)
result
.
append
(
pool
.
apply_async
(
func
,
(
msg
,
)
)
)
pool
.
close
(
)
pool
.
join
(
)
for
res
in
result
:
print
":::"
,
res
.
get
(
)
print
"Sub-process(es) done."
|
一次执行结果
1
2
3
4
5
6
7
8
9
10
|
msg
:
hello
0
msg
:
hello
1
msg
:
hello
2
end
end
end
::
:
donehello
0
::
:
donehello
1
::
:
donehello
2
Sub
-
process
(
es
)
done
.
|
例7.4:使用多个进程池
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
|
#coding: utf-8
import
multiprocessing
import
os
,
time
,
random
def
Lee
(
)
:
print
"\nRun task Lee-%s"
%
(
os
.
getpid
(
)
)
#os.getpid()获取当前的进程的ID
start
=
time
.
time
(
)
time
.
sleep
(
random
.
random
(
)
*
10
)
#random.random()随机生成0-1之间的小数
end
=
time
.
time
(
)
print
'Task Lee, runs %0.2f seconds.'
%
(
end
-
start
)
def
Marlon
(
)
:
print
"\nRun task Marlon-%s"
%
(
os
.
getpid
(
)
)
start
=
time
.
time
(
)
time
.
sleep
(
random
.
random
(
)
*
40
)
end
=
time
.
time
(
)
print
'Task Marlon runs %0.2f seconds.'
%
(
end
-
start
)
def
Allen
(
)
:
print
"\nRun task Allen-%s"
%
(
os
.
getpid
(
)
)
start
=
time
.
time
(
)
time
.
sleep
(
random
.
random
(
)
*
30
)
end
=
time
.
time
(
)
print
'Task Allen runs %0.2f seconds.'
%
(
end
-
start
)
def
Frank
(
)
:
print
"\nRun task Frank-%s"
%
(
os
.
getpid
(
)
)
start
=
time
.
time
(
)
time
.
sleep
(
random
.
random
(
)
*
20
)
end
=
time
.
time
(
)
print
'Task Frank runs %0.2f seconds.'
%
(
end
-
start
)
if
__name__
==
'__main__'
:
function_list
=
[
Lee
,
Marlon
,
Allen
,
Frank
]
print
"parent process %s"
%
(
os
.
getpid
(
)
)
pool
=
multiprocessing.Pool
(
4
)
for
func
in
function_list
:
pool
.
apply_async
(
func
)
#Pool执行函数,apply执行函数,当有一个进程执行完毕后,会添加一个新的进程到pool中
print
'Waiting for all subprocesses done...'
pool
.
close
(
)
pool
.
join
(
)
#调用join之前,一定要先调用close() 函数,否则会出错, close()执行后不会有新的进程加入到pool,join函数等待素有子进程结束
print
'All subprocesses done.'
|
一次执行结果
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
|
parent
process
7704
Waiting
for
all
subprocesses
done
.
.
.
Run
task
Lee
-
6948
Run
task
Marlon
-
2896
Run
task
Allen
-
7304
Run
task
Frank
-
3052
Task
Lee
,
runs
1.59
seconds
.
Task
Marlon
runs
8.48
seconds
.
Task
Frank
runs
15.68
seconds
.
Task
Allen
runs
18.08
seconds
.
All
subprocesses
done
.
|