多线程
GIL
GIL(Global Interpreter Lock)即全局解释器锁。
- 在Python中一个线程对应于C(Cpython)中的一个线程;
- GIL使得同一个时刻只有一个线程在一个cpu上执行字节码,而且无法将多个线程映射到多个cpu上执行(无法利用多核优势),查看Python字节码:
import dis
def add(a):
a = a+1
return a
print(dis.dis(add))
- 释放:非线程的整个过程完全占有
- 根据执行的字节码行数以及时间片释放GIL;
- 在遇到io的操作时候主动释放。
import threading
total = 0
def add():
global total
for i in range(1000000):
total += 1
def desc():
global total
for i in range(1000000):
total -= 1 # 执行的过程中会释放锁,让给另一个线程
thread1 = threading.Thread(target=add)
thread2 = threading.Thread(target=desc)
thread1.start()
thread2.start()
thread1.join()
thread2.join()
# 每次执行最终结果都不确定,即加和减的次数不定
print(total)
多线程编程
- 对于io操作来说,多线程和多进程性能差别不大(线程调度更轻量);
- 可以通过Thread类实例化或集成Thread来实现多线程。
# 模拟多线程爬虫(并发爬取列表页和详情页)
import time
import threading
# 爬取详情页
def get_detail_html(url):
print("get detail html started")
time.sleep(2)
print("get detail html end")
# 从列表页爬取详情页url
def get_detail_url(url):
print("get detail url started")
time.sleep(4)
print("get detail url end")
class GetDetailHtml(threading.Thread):
def __init__(self, name):
super().__init__(name=name)
def run(self):
print("get detail html started")
time.sleep(2)
print("get detail html end")
class GetDetailUrl(threading.Thread):
def __init__(self, name):
super().__init__(name=name)
def run(self):
print("get detail url started")
time.sleep(4)
print("get detail url end")
if __name__ == "__main__":
thread1 = GetDetailHtml("get_detail_html")
thread2 = GetDetailUrl("get_detail_url")
start_time = time.time()
thread1.start()
thread2.start()
thread1.join() # 等待完成后再继续执行下面的
thread2.join()
# 当主线程退出的时候,子线程才会杀死
print ("last time: {}".format(time.time() - start_time))
线程间通信
共享变量 + 锁
import time
import threading
from threading import Condition
# 生产者当生产10个url以后就就等待,保证detail_url_list中最多只有十个url
# 当url_list为空的时候,消费者就暂停
detail_url_list = [] # list非线程安全,需要加锁
# global引用过多时可以创建一个模块专门存放共享变量
# from chapter11 import variables
# 不可以from chapter11.variables import detail_url_list
def get_detail_html(lock):
# 爬取文章详情页
global detail_url_list
while True:
if len(detail_url_list):
lock.acquire()
if len(detail_url_list):
url = detail_url_list.pop()
lock.release()
print("get detail html started")
time.sleep(2)
print("get detail html end")
else:
lock.release()
time.sleep(1)
def get_detail_url(lock):
global detail_url_list
# 爬取文章列表页(列表页爬速度比详情页快,可以开启多个线程爬去详情页)
while True:
print("get detail url started")
time.sleep(4)
for i in range(20):
lock.acquire()
if len(detail_url_list) >= 10:
lock.release()
time.sleep(1)
else:
detail_url_list.append("http://projectsedu.com/{id}".format(id=i))
lock.release()
print("get detail url end")
if __name__ == "__main__":
lock = RLock()
thread_detail_url = threading.Thread(target=get_detail_url, args=(lock,))
for i in range(10):
html_thread = threading.Thread(target=get_detail_html, args=(lock,))
html_thread.start()
#当主线程退出的时候, 子线程kill掉
print ("last time: {}".format(time.time() - start_time))
队列
import time
import threading
from queue import Queue
def get_detail_html(queue):
# 爬取文章详情页
while True:
url = queue.get() # Queue默认阻塞
# for url in detail_url_list:
print("get detail html started")
time.sleep(2)
print("get detail html end")
def get_detail_url(queue):
# 爬取文章列表页
while True:
print("get detail url started")
time.sleep(4)
for i in range(20):
queue.put("http://projectsedu.com/{id}".format(id=i))
print("get detail url end")
if __name__ == "__main__":
detail_url_queue = Queue(maxsize=1000)
thread_detail_url = threading.Thread(target=get_detail_url, args=(detail_url_queue,))
for i in range(10):
html_thread = threading.Thread(target=get_detail_html, args=(detail_url_queue,))
html_thread.start()
start_time = time.time()
detail_url_queue.task_done() # 主动使Queue退出
detail_url_queue.join()
# 当主线程退出的时候, 子线程kill掉
print ("last time: {}".format(time.time() - start_time))
锁:线程间同步
- 使用锁可以实现线程同步,但会影响性能,也可能导致死锁;
- 重入锁:在同一个线程里,可以连续调用多次acquire, 注意acquire的次数要和release的次数相等;
from threading import Lock, RLock, Condition
import threading
total = 0
lock = RLock() # 重入锁(在同一线程中可多次acquire)
# lock = Lock() # 一般锁,多次申请会造成死锁
def add():
global lock
global total
for i in range(1000000):
lock.acquire() # 申请锁(失败则等待)
lock.acquire()
total += 1
lock.release()
lock.release()
def desc():
global total
global lock
for i in range(1000000):
lock.acquire()
total -= 1
lock.release()
thread1 = threading.Thread(target=add)
thread2 = threading.Thread(target=desc)
thread1.start()
thread2.start()
thread1.join()
thread2.join()
print(total)
"""
死锁:
互斥
不可抢占
请求且占有
循环等待
A(a, b)
acquire (a)
acquire (b)
B(a, b)
acquire (b)
acquire (a)
"""
条件变量
- 只用锁无法确保两个线程交替运行(进度不一致),需要一个“通知-等待”的机制,在本线程工作完成后由下一个线程工作,并等待该线程的通知;
- 使用通知-等待机制要注意线程启动的顺序(先启动需要被notify的,即被动方);
- 在调用with...cond之后(在cond的作用域中)才能调用wait或者notify方法;
- condition有两层锁, 一把底层锁会在线程调用了wait方法的时候释放,上层锁会在每次调用wait时分配一把并放入到cond等待队列中,等到notify方法的唤醒。
import threading
from concurrent import futures
class XiaoAi(threading.Thread):
def __init__(self, cond):
super().__init__(name="小爱")
self.cond = cond
def run(self):
with self.cond:
self.cond.wait()
print("{} : 在 ".format(self.name))
self.cond.notify()
self.cond.wait()
print("{} : 好啊 ".format(self.name))
self.cond.notify()
self.cond.wait()
print("{} : 君住长江尾 ".format(self.name))
self.cond.notify()
self.cond.wait()
print("{} : 共饮长江水 ".format(self.name))
self.cond.notify()
self.cond.wait()
print("{} : 此恨何时已 ".format(self.name))
self.cond.notify()
self.cond.wait()
print("{} : 定不负相思意 ".format(self.name))
self.cond.notify()
class TianMao(threading.Thread):
def __init__(self, cond):
super().__init__(name="天猫精灵")
self.cond = cond
def run(self):
with self.cond:
print("{} : 小爱同学 ".format(self.name))
self.cond.notify()
self.cond.wait()
print("{} : 我们来对古诗吧 ".format(self.name))
self.cond.notify()
self.cond.wait()
print("{} : 我住长江头 ".format(self.name))
self.cond.notify()
self.cond.wait()
print("{} : 日日思君不见君 ".format(self.name))
self.cond.notify()
self.cond.wait()
print("{} : 此水几时休 ".format(self.name))
self.cond.notify()
self.cond.wait()
print("{} : 只愿君心似我心 ".format(self.name))
self.cond.notify()
self.cond.wait()
if __name__ == "__main__":
cond = threading.Condition()
xiaoai = XiaoAi(cond)
tianmao = TianMao(cond)
xiaoai.start()
tianmao.start()
信号量
- 在读写分离的场景,一般只能被一个线程写,但允许多个线程读,通过Semaphore可以控制进入的数量;
- 对于一个semaphore,调用acquire时数值 - 1,直到数值减到0则会加锁;调用release释放,数值 + 1;
- 在需要控制并发程度的场景,信号量也能很好地发挥作用。
import threading
import time
class HtmlSpider(threading.Thread):
def __init__(self, url, sem):
super().__init__()
self.url = url
self.sem = sem
def run(self):
time.sleep(2)
print("got html text success")
self.sem.release()
class UrlProducer(threading.Thread):
def __init__(self, sem):
super().__init__()
self.sem = sem
def run(self):
for i in range(20):
self.sem.acquire() # 启动线程前申请,在线程内部释放
html_thread = HtmlSpider("https://baidu.com/{}".format(i), self.sem)
html_thread.start()
if __name__ == "__main__":
sem = threading.Semaphore(3)
url_producer = UrlProducer(sem)
url_producer.start()
线程池
- 使用线程池实现线程重用、状态与返回值管理(使用done方法当一个线程完成的时候主线程能立即知道)
- futures包中多线程与多进程接口一致,能减少开发难度
- task的返回容器:Future对象(当时未完成,但完成后可以通过对象获取结果)。
from concurrent.futures import ThreadPoolExecutor
import time
def get_html(times):
time.sleep(times)
print("get page {} success".format(times))
return times
executor = ThreadPoolExecutor(max_workers=2)
# 通过submit函数提交执行的函数到线程池中, 立即返回
task1 = executor.submit(get_html, (3))
task2 = executor.submit(get_html, (2))
task1.done() # 获取task1执行状态
task1.result() # 获取task1执行结果
task2.cancel() # 取消task2执行
批量提交线程,获取成功执行的线程
from concurrent.futures import ThreadPoolExecutor, as_completed, wait, FIRST_COMPLETED
executor = ThreadPoolExecutor(max_workers=2)
# 使用as_completed生成器,每有一个线程完成即yield
urls = [3, 2, 4]
all_task = [executor.submit(get_html, (url)) for url in urls]
wait(all_task, return_when=FIRST_COMPLETED) # 等待首个子线程执行完成,主线程再继续执行
print('main')
for future in as_completed(all_task):
data = future.result()
print("get {} page".format(data))
# 通过executor的map获取已经完成的task的值(将每个url传入函数中一一执行)
for data in executor.map(get_html, urls):
print("get {} page".format(data))
多进程
- 对于在Python中存在GIL,消耗CPU的操作无法利用多核优势,使用多线程无法实现并行操作,此时应使用多进程;
- 进程切换代价比较高,对于频繁IO操作使用多线程更好(开销更小、更稳定);
- Windows下多线程多进程编程必须加入
if __name__ == '__main__'
CPU操作:
from concurrent.futures import ThreadPoolExecutor, as_completed
from concurrent.futures import ProcessPoolExecutor
def fib(n):
if n<=2:
return 1
return fib(n-1)+fib(n-2)
# 使用多线程
with ThreadPoolExecutor(3) as executor:
all_task = [executor.submit(fib, (num)) for num in range(25, 40)]
start_time = time.time()
for future in as_completed(all_task):
data = future.result()
print("exe result: {}".format(data))
print("last time is: {}".format(time.time()-start_time))
# 使用多进程
with ProcessPoolExecutor(3) as executor:
all_task = [executor.submit(fib, (num)) for num in range(25, 40)]
start_time = time.time()
for future in as_completed(all_task):
data = future.result()
print("exe result: {}".format(data))
print("last time is: {}".format(time.time()-start_time))
IO操作:
def random_sleep(n):
time.sleep(n)
return n
# 使用多线程
with ThreadPoolExecutor(3) as executor:
all_task = [executor.submit(random_sleep, (num)) for num in [2] * 30]
start_time = time.time()
for future in as_completed(all_task):
data = future.result()
print("exe result: {}".format(data))
print("last time is: {}".format(time.time() - start_time))
# 使用多进程
with ProcessPoolExecutor(3) as executor:
all_task = [executor.submit(fib, (num)) for num in range(25, 40)]
start_time = time.time()
for future in as_completed(all_task):
data = future.result()
print("exe result: {}".format(data))
print("last time is: {}".format(time.time()-start_time))
多进程编程
- 执行fork会马上创建一个子进程,同时主进程继续向下执行;
- 子进程会把主进程的数据复制一份独自执行,与主进程隔离;
import os
# fork只能用于unix/linux中
pid = os.fork()
print("ywh") # 从这句开始,主进程、子进程都会执行
if pid == 0:
print('子进程 {} ,父进程是: {}.' .format(os.getpid(), os.getppid()))
else:
print('我是父进程:{}.'.format(pid))
使用multiprocessing和concurrent.futures包
def get_html(n):
time.sleep(n)
print("sub_progress success")
return n
# 方法1
progress = multiprocessing.Process(target=get_html, args=(2,))
print(progress.pid)
progress.start()
print(progress.pid)
progress.join()
print("main progress end")
# 方法2
pool = multiprocessing.Pool(multiprocessing.cpu_count()) # 默认为系统CPU数
result = pool.apply_async(get_html, args=(3,)) # 异步提交
pool.close() # 必须关闭,不再接收新的任务
pool.join() # 等待任务完成
print(result.get()) # 获取返回结果
# 方法3
for result in pool.imap(get_html, [1,5,3]):
print("{} sleep success".format(result))
# 方法4
for result in pool.imap_unordered(get_html, [1, 5, 3]):
print("{} sleep success".format(result))
进程间通信
- 注意多线程和多进程通信的包不一样,不能重用;
- 多线程中共享全局变量的方法不能用于多进程(数据全部复制到子进程);
- 线程池:multiprocessing中的Queue不能用于进程池,而应使用Manager.Queue;
- 管道性能比Queue高,但只适用于两个进程之间的通信;
- Python内置有很多内存共享的数据结构,在multiprocessing.Manager,需要注意数据同步。
多线程通信
import time
from multiprocessing import Process, Queue, Pool
def producer(queue):
queue.put("a")
time.sleep(2)
def consumer(queue):
time.sleep(2)
data = queue.get()
print(data)
queue = Queue(10)
my_producer = Process(target=producer, args=(queue,))
my_consumer = Process(target=consumer, args=(queue,))
my_producer.start()
my_consumer.start()
my_producer.join()
my_consumer.join()
进程池
from multiprocessing import Process, Manager
def producer(queue):
queue.put("a")
time.sleep(2)
def consumer(queue):
time.sleep(2)
data = queue.get()
print(data)
queue = Manager().Queue(10)
pool = Pool(2)
pool.apply_async(producer, args=(queue,))
pool.apply_async(consumer, args=(queue,))
pool.close()
pool.join()
from queue import Queue # 多线程
from multiprocessing import Queue # 多进程
from multiprocessing import Manager # 进程池
管道
from multiprocessing import Process, Pipe
def producer(pipe):
pipe.send("bobby")
def consumer(pipe):
print(pipe.recv())
if __name__ == "__main__":
recevie_pipe, send_pipe = Pipe()
my_producer = Process(target=producer, args=(send_pipe,))
my_consumer = Process(target=consumer, args=(recevie_pipe,))
my_producer.start()
my_consumer.start()
my_producer.join()
my_consumer.join()