多线程修改同一个数据
个人理解：
GIL：存在于Cpython中，称为全局解释器锁，在同一时间只能一个python线程在跑，但这并不是说是串行运行的，他还是“并行”的，CPU在不断的分配cpu时间给每个线程去运行，只是同一时间片刻或同一个cpu时间片刻只有一个线程在跑。
线程锁：只让一个线程运行加锁的那段代码。
示例：
1 print(n) 2 n += 1 3 print(n) 4 # 这里三条语句需要三次获取n的值，第一次print(n)和 n += 1时拿到的n 不一定相同，第二次print(n)也不一定是 n += 1的结果。 但 n += 1的操作是原子性的，执行n += 1过程中，从拿到n 再 n + 1再重新赋值给n的这个过程中，其他线程不能修改，且也不能访问n的值。
简单示例1：
import threading
import time
def f(i):
global n
time.sleep(1)
print("begin += 1: i=%s,n=%s"%(i,n))
n += 1
print("end +=1: i=%s,n=%s"%(i,n))
if __name__ == "__main__":
t_list = []
n = 0
for i in range(100):
t = threading.Thread(target=f,args=(i,))
t.start()
t_list.append(t)
for i in t_list:
i.join()
执行结果：
begin += 1: i=2,n=0 end +=1: i=2,n=1 begin += 1: i=1,n=1 end +=1: i=1,n=2 begin += 1: i=0,n=2 end +=1: i=0,n=3 begin += 1: i=4,n=3 end +=1: i=4,n=4 begin += 1: i=3,n=4 end +=1: i=3,n=5 begin += 1: i=6,n=5 end +=1: i=6,n=6 begin += 1: i=5,n=6 end +=1: i=5,n=7 begin += 1: i=9,n=7 begin += 1: i=8,n=7 begin += 1: i=7,n=7 end +=1: i=7,n=8 end +=1: i=8,n=9 end +=1: i=9,n=10 begin += 1: i=13,n=10 end +=1: i=13,n=11 begin += 1: i=15,n=11 begin += 1: i=12,n=11 begin += 1: i=11,n=11 end +=1: i=12,n=12 begin += 1: i=14,n=12 end +=1: i=14,n=13 begin += 1: i=10,n=13 end +=1: i=11,n=14 begin += 1: i=16,n=14 end +=1: i=16,n=15 end +=1: i=10,n=16 end +=1: i=15,n=17 begin += 1: i=19,n=17 end +=1: i=19,n=18 begin += 1: i=18,n=18 begin += 1: i=17,n=18 end +=1: i=18,n=19 end +=1: i=17,n=20 begin += 1: i=21,n=20 end +=1: i=21,n=21 begin += 1: i=22,n=20 end +=1: i=22,n=22 begin += 1: i=20,n=22 end +=1: i=20,n=23 begin += 1: i=24,n=23 end +=1: i=24,n=24 begin += 1: i=23,n=24 end +=1: i=23,n=25 begin += 1: i=25,n=24 end +=1: i=25,n=26 begin += 1: i=28,n=26 end +=1: i=28,n=27 begin += 1: i=27,n=27 begin += 1: i=26,n=27 end +=1: i=26,n=28 end +=1: i=27,n=29 begin += 1: i=29,n=29 end +=1: i=29,n=30 begin += 1: i=37,n=30 begin += 1: i=36,n=30 end +=1: i=36,n=31 begin += 1: i=34,n=31 end +=1: i=34,n=32 begin += 1: i=33,n=32 end +=1: i=33,n=33 begin += 1: i=31,n=33 end +=1: i=31,n=34 begin += 1: i=30,n=34 begin += 1: i=35,n=34 end +=1: i=35,n=35 end +=1: i=30,n=36 end +=1: i=37,n=37 begin += 1: i=40,n=37 end +=1: i=40,n=38 begin += 1: i=32,n=38 end +=1: i=32,n=39 begin += 1: i=39,n=39 end +=1: i=39,n=40 begin += 1: i=38,n=40 end +=1: i=38,n=41 begin += 1: i=42,n=41 begin += 1: i=41,n=42 end +=1: i=41,n=43 end +=1: i=42,n=42 begin += 1: i=43,n=43 end +=1: i=43,n=44 begin += 1: i=46,n=44 end +=1: i=46,n=45 begin += 1: i=45,n=45 end +=1: i=45,n=46 begin += 1: i=44,n=46 end +=1: i=44,n=47 begin += 1: i=48,n=47 end +=1: i=48,n=48 begin += 1: i=47,n=48 end +=1: i=47,n=49 begin += 1: i=50,n=49 end +=1: i=50,n=50 begin += 1: i=51,n=50 end +=1: i=51,n=51 begin += 1: i=49,n=51 end +=1: i=49,n=52 begin += 1: i=53,n=52 begin += 1: i=52,n=52 end +=1: i=53,n=53 end +=1: i=52,n=54 begin += 1: i=54,n=54 end +=1: i=54,n=55 begin += 1: i=55,n=55 end +=1: i=55,n=56 begin += 1: i=56,n=56 end +=1: i=56,n=57 begin += 1: i=57,n=57 end +=1: i=57,n=58 begin += 1: i=59,n=58 end +=1: i=59,n=59 begin += 1: i=58,n=59 end +=1: i=58,n=60 begin += 1: i=60,n=60 end +=1: i=60,n=61 begin += 1: i=62,n=61 end +=1: i=62,n=62 begin += 1: i=61,n=62 end +=1: i=61,n=63 begin += 1: i=65,n=63 begin += 1: i=64,n=63 end +=1: i=64,n=64 end +=1: i=65,n=65 begin += 1: i=63,n=65 end +=1: i=63,n=66 begin += 1: i=67,n=66 end +=1: i=67,n=67 begin += 1: i=68,n=67 end +=1: i=68,n=68 begin += 1: i=66,n=68 end +=1: i=66,n=69 begin += 1: i=70,n=69 end +=1: i=70,n=70 begin += 1: i=69,n=70 end +=1: i=69,n=71 begin += 1: i=73,n=71 begin += 1: i=72,n=71 end +=1: i=72,n=72 end +=1: i=73,n=73 begin += 1: i=71,n=72 end +=1: i=71,n=74 begin += 1: i=74,n=74 end +=1: i=74,n=75 begin += 1: i=75,n=75 end +=1: i=75,n=76 begin += 1: i=76,n=76 end +=1: i=76,n=77 begin += 1: i=77,n=77 end +=1: i=77,n=78 begin += 1: i=78,n=78 end +=1: i=78,n=79 begin += 1: i=79,n=79 begin += 1: i=80,n=79 end +=1: i=80,n=80 end +=1: i=79,n=81 begin += 1: i=82,n=81 end +=1: i=82,n=82 begin += 1: i=81,n=82 end +=1: i=81,n=83 begin += 1: i=84,n=83 begin += 1: i=85,n=83 end +=1: i=85,n=84 end +=1: i=84,n=85 begin += 1: i=83,n=85 end +=1: i=83,n=86 begin += 1: i=86,n=86 end +=1: i=86,n=87 begin += 1: i=87,n=87 begin += 1: i=88,n=87 end +=1: i=87,n=88 end +=1: i=88,n=89 begin += 1: i=90,n=89 end +=1: i=90,n=90 begin += 1: i=89,n=90 end +=1: i=89,n=91 begin += 1: i=93,n=91 end +=1: i=93,n=92 begin += 1: i=91,n=92 end +=1: i=91,n=93 begin += 1: i=92,n=93 end +=1: i=92,n=94 begin += 1: i=94,n=94 end +=1: i=94,n=95 begin += 1: i=95,n=95 end +=1: i=95,n=96 begin += 1: i=96,n=96 begin += 1: i=97,n=96 end +=1: i=97,n=97 end +=1: i=96,n=98 begin += 1: i=99,n=98 begin += 1: i=98,n=98 end +=1: i=98,n=99 end +=1: i=99,n=100
这里可以看到第99次中第一次print(n)获取的值为98，第二次print(n)的值为100。说明执行n += 1时（即n = n + 1）右边的n的值不一定是98。
简单示例2：
import threading
import time
def f(i):
global n
time.sleep(1)
print("begin += 1: i=%s,n=%s,d=%s"%(i,n,d))
d[n] ,n = n+1,n+1
print("end +=1: i=%s,n=%s,d=%s"%(i,n,d))
if __name__ == "__main__":
t_list = []
n = 0
d = {}
for i in range(10):
t = threading.Thread(target=f,args=(i,))
t.start()
t_list.append(t)
for i in t_list:
i.join()
运行结果：
begin += 1: i=0,n=0,d={}
begin += 1: i=1,n=0,d={}
end +=1: i=1,n=1,d={0: 1}
end +=1: i=0,n=2,d={0: 1, 1: 2}
begin += 1: i=5,n=2,d={0: 1, 1: 2}
begin += 1: i=4,n=2,d={0: 1, 1: 2}
end +=1: i=4,n=3,d={0: 1, 1: 2, 2: 3}
end +=1: i=5,n=4,d={0: 1, 1: 2, 2: 3, 3: 4}
begin += 1: i=2,n=4,d={0: 1, 1: 2, 2: 3, 3: 4}
end +=1: i=2,n=5,d={0: 1, 1: 2, 2: 3, 3: 4, 4: 5}
begin += 1: i=3,n=5,d={0: 1, 1: 2, 2: 3, 3: 4, 4: 5}
end +=1: i=3,n=6,d={0: 1, 1: 2, 2: 3, 3: 4, 4: 5, 5: 6}
begin += 1: i=6,n=6,d={0: 1, 1: 2, 2: 3, 3: 4, 4: 5, 5: 6}
end +=1: i=6,n=7,d={0: 1, 1: 2, 2: 3, 3: 4, 4: 5, 5: 6, 6: 7}
begin += 1: i=9,n=7,d={0: 1, 1: 2, 2: 3, 3: 4, 4: 5, 5: 6, 6: 7}
begin += 1: i=8,n=7,d={0: 1, 1: 2, 2: 3, 3: 4, 4: 5, 5: 6, 6: 7}
end +=1: i=8,n=8,d={0: 1, 1: 2, 2: 3, 3: 4, 4: 5, 5: 6, 6: 7, 7: 8}
begin += 1: i=7,n=8,d={0: 1, 1: 2, 2: 3, 3: 4, 4: 5, 5: 6, 6: 7, 7: 8}
end +=1: i=7,n=9,d={0: 1, 1: 2, 2: 3, 3: 4, 4: 5, 5: 6, 6: 7, 7: 8, 8: 9}
end +=1: i=9,n=10,d={0: 1, 1: 2, 2: 3, 3: 4, 4: 5, 5: 6, 6: 7, 7: 8, 8: 9, 9: 10}
这里可以看到d[n], n = n+1, n+1是原子性的，这里先是d[n]和右边的两个n获取值，然后分别赋值给d[n]和n，这里不可能出现d[n]和右边的两个n获取的值是8，而在赋值d[n]和n时赋的值是10.
python2.7:
不加线程锁时，n += 1操作时，多个线程能同时访问n获取n的值，这样就会导致多个线程运行n += 1后n的值并不是 n + 1的次数。
import threading
import time
def f(i):
global n
time.sleep(1)
# print("begin += 1: i=%s,n=%s"%(i,n))
d[i] , n = "%s:%s"%(n,n +1),n+1
# print("end +=1: i=%s,n=%s"%(i,n))
if __name__ == "__main__":
t_list = []
n = 0
d= {}
for i in range(100):
t = threading.Thread(target=f,args=(i,))
t.start()
t_list.append(t)
for i in t_list:
i.join()
time.sleep(2)
for i,v in d.items():
print(i,v)
print("n=",n)
运行结果：
(0, '1:2')
(1, '0:1')
(2, '3:4')
(3, '2:3')
(4, '5:6')
(5, '4:5')
(6, '7:8')
(7, '6:7')
(8, '6:7')
(9, '8:9')
(10, '7:8')
(11, '10:11')
(12, '9:10')
(13, '11:12')
(14, '12:13')
(15, '13:14')
(16, '15:16')
(17, '14:15')
(18, '17:18')
(19, '16:17')
(20, '19:20')
(21, '18:19')
(22, '21:22')
(23, '20:21')
(24, '22:23')
(25, '23:24')
(26, '26:27')
(27, '25:26')
(28, '24:25')
(29, '27:28')
(30, '27:28')
(31, '31:32')
(32, '30:31')
(33, '29:30')
(34, '32:33')
(35, '34:35')
(36, '33:34')
(37, '36:37')
(38, '35:36')
(39, '38:39')
(40, '37:38')
(41, '40:41')
(42, '40:41')
(43, '39:40')
(44, '43:44')
(45, '42:43')
(46, '41:42')
(47, '44:45')
(48, '46:47')
(49, '45:46')
(50, '47:48')
(51, '49:50')
(52, '48:49')
(53, '50:51')
(54, '51:52')
(55, '52:53')
(56, '54:55')
(57, '53:54')
(58, '56:57')
(59, '55:56')
(60, '57:58')
(61, '57:58')
(62, '60:61')
(63, '59:60')
(64, '61:62')
(65, '62:63')
(66, '64:65')
(67, '65:66')
(68, '63:64')
(69, '72:73')
(70, '71:72')
(71, '73:74')
(72, '70:71')
(73, '69:70')
(74, '66:67')
(75, '68:69')
(76, '67:68')
(77, '75:76')
(78, '74:75')
(79, '76:77')
(80, '78:79')
(81, '77:78')
(82, '80:81')
(83, '79:80')
(84, '82:83')
(85, '81:82')
(86, '85:86')
(87, '83:84')
(88, '84:85')
(89, '87:88')
(90, '86:87')
(91, '89:90')
(92, '88:89')
(93, '92:93')
(94, '91:92')
(95, '90:91')
(96, '94:95')
(97, '93:94')
(98, '96:97')
(99, '95:96')
('n=', 97)
这里可以看到线程7和8，29和30，60和61分别获取了同一份n的值。
加锁代码示例，加锁代码段可以理解为串行执行：
import threading
import time
def f(i):
global n
time.sleep(1)
# print("begin += 1: i=%s,n=%s"%(i,n))
lock.acquire()
d[i] , n = "%s:%s"%(n,n +1),n+1
lock.release()
# print("end +=1: i=%s,n=%s"%(i,n))
if __name__ == "__main__":
t_list = []
n = 0
d= {}
lock= threading.Lock()
for i in range(100):
t = threading.Thread(target=f,args=(i,))
t.start()
t_list.append(t)
for i in t_list:
i.join()
time.sleep(2)
for i,v in d.items():
print(i,v)
print("n=",n)
如果time.sleep(1)被加锁，由于加锁代码段串行执行，那么这个sleep(x)也是串行的。
lock.acquire()
time.sleep(0.1)
d[i] , n = "%s:%s"%(n,n +1),n+1
lock.release()
递归锁,存在函数之间调用，而函数中又都有锁时用递归锁：
RLock
import threading
import time
def a(i):
global data
lock.acquire()
data += 2
print("a,%s,%s"%(i,data))
lock.release()
def b(i):
global data
lock.acquire()
data *= 2
print("b,%s,%s\n"%(i,data))
lock.release()
def c(i):
time.sleep(1)
lock.acquire()
print("c,%s,%s"%(i,data))
a(i)
print("----------a,b---------------")
b(i)
lock.release()
if __name__ == "__main__":
lock = threading.RLock()
data = 100
t_list = []
for i in range(5):
t = threading.Thread(target=c,args=(i,))
t.start()
t_list.append(t)
for t in t_list:
t.join()
print("main:",data)
运行结果：
c,1,100 a,1,102 ----------a,b--------------- b,1,204 c,0,204 a,0,206 ----------a,b--------------- b,0,412 c,3,412 a,3,414 ----------a,b--------------- b,3,828 c,4,828 a,4,830 ----------a,b--------------- b,4,1660 c,2,1660 a,2,1662 ----------a,b--------------- b,2,3324 main: 3324
可以看到被锁住的代码串行运行。
