1.协程介绍
协程:是单线程下的并发,又称微线程,纤程。英文名Coroutine。一句话说明什么是线程:协程是一种用户态的轻量级线程,即协程是由用户程序自己控制调度的。、
需要强调的是:
#<span style="color: #800080">1</span><span style="color: #000000">. python的线程属于内核级别的,即由操作系统控制调度(如单线程遇到io或执行时间过长就会被迫交出cpu执行权限,<a href="https://www.gaodaima.com/tag/%e5%88%87%e6%8d%a2" title="查看更多关于切换的文章" target="_blank">切换</a>其他线程运行) #</span><span style="color: #800080">2</span>. 单线程内开启协程,一旦遇到io,就会从应用程序级别(而非操作系统)控制切换,以此来提升效率(非io操作的切换与效率无关)
对比操作系统控制线程的切换,用户在单线程内控制协程的切换
优点如下:
#<span style="color: #800080">1</span><span style="color: #000000">. 协程的切换开销更小,属于程序级别的切换,操作系统完全感知不到,因而更加轻量级 #</span><span style="color: #800080">2</span>. 单线程内就可以实现并发的效果,最大限度地利用cpu
缺点如下:
#<span style="color: #800080">1</span>. 协程的本质是单线程下,无法利用多核,可以是一个程序开启多个进程,每个进程内开启多个线程,每个线程内开启协程
总结协程特点:
- 必须在只有一个单线程里实现并发
- 修改共享数据不需加锁
- 用户程序里自己保存多个控制流的上下文栈
2.Greenlet
如果在单个线程内有20个任务,要想实现在多个任务之间切换,使用yield生成器的方式过于麻烦(需要先得到初始化一次的生成器,然后再调用send。。。非常麻烦),而使用greenlet模块可以非常简单地实现这20个任务直接的切换
<span style="color: #000000">#安装 pip3 install greenlet</span>
<span style="color: #000000"> #真正的协程模块就是使用greenlet完成的切换 </span><span style="color: #0000ff">from</span><span style="color: #000000"> greenlet import greenlet def eat(name): print(</span><span style="color: #800000">"</span><span style="color: #800000">%s eat 1</span><span style="color: #800000">"</span> %name) #<span style="color: #800080">2</span><span style="color: #000000"> g2.</span><span style="color: #0000ff">switch</span>(<span style="color: #800000">"</span><span style="color: #800000">yanhui</span><span style="color: #800000">"</span>) #<span style="color: #800080">3</span><span style="color: #000000"> print(</span><span style="color: #800000">"</span><span style="color: #800000">%s eat 2</span><span style="color: #800000">"</span> %name) #<span style="color: #800080">6</span><span style="color: #000000"> g2.</span><span style="color: #0000ff">switch</span>() #<span style="color: #800080">7</span><span style="color: #000000"> def play(name): print(</span><span style="color: #800000">"</span><span style="color: #800000">%s play 1</span><span style="color: #800000">"</span> %name) #<span style="color: #800080">4</span><span style="color: #000000"> g1.</span><span style="color: #0000ff">switch</span>() #<span style="color: #800080">5</span><span style="color: #000000"> print(</span><span style="color: #800000">"</span><span style="color: #800000">%s play 2</span><span style="color: #800000">"</span> %name) #<span style="color: #800080">8</span><span style="color: #000000"> g1</span>=<span style="color: #000000">greenlet(eat) g2</span>=<span style="color: #000000">greenlet(play) g1.</span><span style="color: #0000ff">switch</span>(<span style="color: #800000">"</span><span style="color: #800000">yanhui</span><span style="color: #800000">"</span>)#可以在第一次switch时传入参数,以后都不需要 <span style="color: #800080">1</span>
单纯的切换(在没有io的情况下或者没有重复开辟内存空间的操作),反而会降低程序的执行速度
<span style="color: #000000">#顺序执行 import time def f1(): res</span>=<span style="color: #800080">1</span> <span style="color: #0000ff">for</span> i <span style="color: #0000ff">in</span> range(<span style="color: #800080">100000000</span><span style="color: #000000">): res</span>+=<span style="color: #000000">i def f2(): res</span>=<span style="color: #800080">1</span> <span style="color: #0000ff">for</span> i <span style="color: #0000ff">in</span> range(<span style="color: #800080">100000000</span><span style="color: #000000">): res</span>*=<span style="color: #000000">i start</span>=<span style="color: #000000">time.time() f1() f2() stop</span>=<span style="color: #000000">time.time() print(</span><span style="color: #800000">"</span><span style="color: #800000">run time is %s</span><span style="color: #800000">"</span> %(stop-start)) #<span style="color: #800080">10.985628366470337</span><span style="color: #000000"> #切换 </span><span style="color: #0000ff">from</span><span style="color: #000000"> greenlet import greenlet import time def f1(): res</span>=<span style="color: #800080">1</span> <span style="color: #0000ff">for</span> i <span style="color: #0000ff">in</span> range(<span style="color: #800080">100000000</span><span style="color: #000000">): res</span>+=<span style="color: #000000">i g2.</span><span style="color: #0000ff">switch</span><span style="color: #000000">() def f2(): res</span>=<span style="color: #800080">1</span> <span style="color: #0000ff">for</span> i <span style="color: #0000ff">in</span> range(<span style="color: #800080">100000000</span><span style="color: #000000">): res</span>*=<span style="color: #000000">i g1.</span><span style="color: #0000ff">switch</span><span style="color: #000000">() start</span>=<span style="color: #000000">time.time() g1</span>=<span style="color: #000000">greenlet(f1) g2</span>=<span style="color: #000000">greenlet(f2) g1.</span><span style="color: #0000ff">switch</span><span style="color: #000000">() stop</span>=<span style="color: #000000">time.time() print(</span><span style="color: #800000">"</span><span style="color: #800000">run time is %s</span><span style="color: #800000">"</span> %(stop-start)) # <span style="color: #800080">52.763017892837524</span>
效率对比
greenlet只是提供了一种比generator更加便捷的切换方式,当切到一个任务执行时如果遇到io,那就原地阻塞,仍然是没有解决遇到IO自动切换来提升效率的问题,协程虽然没有规避固有的I/O时间,但是我们使用这个时间来做别的事情了,一般在工作中我们都是进程+线程+协程的方式来实现并发,以达到最好的并发效果,如果是4核的cpu,一般起5个进程,每个进程中20个线程(5倍cpu数量),每个线程可以起500个协程,大规模爬取页面的时候,等待网络延迟的时间的时候,我们就可以用协程去实现并发。 并发数量 = 5 * 20 * 500 = 50000个并发,这是一般一个4cpu的机器最大的并发数。nginx在负载均衡的时候最大承载量就是5w个
单线程里的这20个任务的代码通常会既有计算操作又有阻塞操作,我们完全可以在执行任务1时遇到阻塞,就利用阻塞的时间去执行任务2。。。。如此,才能提高效率,这就用到了Gevent模块。
3.Gevent介绍
<span style="color: #000000">#安装 pip3 install gevent</span>
Gevent 是一个第三方库,可以轻松通过gevent实现并发同步或异步编程,在gevent中用到的主要模式是Greenlet, 它是以C扩展模块形式接入Python的轻量级协程。 Greenlet全部运行在主程序操作系统进程的内部,但它们被协作式地调度
<span style="color: #000000">#用法 g1</span>=gevent.spawn(func,<span style="color: #800080">1</span>,<span style="color: #800080">2</span>,<span style="color: #800080">3</span>,x=<span style="color: #800080">4</span>,y=<span style="color: #800080">5</span><span style="color: #000000">)创建一个协程对象g1,spawn括号内第一个参数是函数名,如eat,后面可以有多个参数,可以是位置实参或关键字实参,都是传给函数eat的,spawn是异步提交任务 g2</span>=<span style="color: #000000">gevent.spawn(func2) g1.join() #等待g1结束 g2.join() #等待g2结束 有人测试的时候会发现,不写第二个join也能执行g2,是的,协程帮你切换执行了,但是你会发现,如果g2里面的任务执行的时间长,但是不写join的话,就不会执行完等到g2剩下的任务了 #或者上述两步合作一步:gevent.joinall([g1,g2]) g1.value#拿到func1的返回值</span>
遇到IO阻塞时会自动切换任务
<span style="color: #000000">import gevent def eat(name): print(</span><span style="color: #800000">"</span><span style="color: #800000">%s eat 1</span><span style="color: #800000">"</span> %<span style="color: #000000">name) gevent.sleep(</span><span style="color: #800080">2</span><span style="color: #000000">) print(</span><span style="color: #800000">"</span><span style="color: #800000">%s eat 2</span><span style="color: #800000">"</span> %<span style="color: #000000">name) def play(name): print(</span><span style="color: #800000">"</span><span style="color: #800000">%s play 1</span><span style="color: #800000">"</span> %<span style="color: #000000">name) gevent.sleep(</span><span style="color: #800080">1</span><span style="color: #000000">) print(</span><span style="color: #800000">"</span><span style="color: #800000">%s play 2</span><span style="color: #800000">"</span> %<span style="color: #000000">name) g1</span>=gevent.spawn(eat,<span style="color: #800000">"</span><span style="color: #800000">egon</span><span style="color: #800000">"</span><span style="color: #000000">) g2</span>=gevent.spawn(play,name=<span style="color: #800000">"</span><span style="color: #800000">egon</span><span style="color: #800000">"</span><span style="color: #000000">) g1.join() g2.join() #或者gevent.joinall([g1,g2]) print(</span><span style="color: #800000">"</span><span style="color: #800000">主</span><span style="color: #800000">"</span><span style="color: #000000">) 遇到I</span>/O切换
遇到I/O切换
上例gevent.sleep(2)模拟的是gevent可以识别的io阻塞,
而time.sleep(2)或其他的阻塞,gevent是不能直接识别的需要用下面一行代码,打补丁,就可以识别了
from gevent import monkey;monkey.patch_all()必须放到被打补丁者的前面,如time,socket模块之前
或者我们干脆记忆成:要用gevent,需要将from gevent import monkey;monkey.patch_all()放到文件的开头
<span style="color: #0000ff">from</span><span style="color: #000000"> gevent import monkey;monkey.patch_all() #必须写在最上面,这句话后面的所有阻塞全部能够识别了 import gevent #直接导入即可 import time def eat(): #print() print(</span><span style="color: #800000">"</span><span style="color: #800000">eat food 1</span><span style="color: #800000">"</span><span style="color: #000000">) time.sleep(</span><span style="color: #800080">2</span><span style="color: #000000">) #加上mokey就能够识别到time模块的sleep了 print(</span><span style="color: #800000">"</span><span style="color: #800000">eat food 2</span><span style="color: #800000">"</span><span style="color: #000000">) def play(): print(</span><span style="color: #800000">"</span><span style="color: #800000">play 1</span><span style="color: #800000">"</span><span style="color: #000000">) time.sleep(</span><span style="color: #800080">1</span>) #来回切换,直到一个I/<span style="color: #000000">O的时间结束,这里都是我们个gevent做得,不再是控制不了的操作系统了。 print(</span><span style="color: #800000">"</span><span style="color: #800000">play 2</span><span style="color: #800000">"</span><span style="color: #000000">) g1</span>=<span style="color: #000000">gevent.spawn(eat) g2</span>=<span style="color: #000000">gevent.spawn(play_phone) gevent.joinall([g1,g2]) print(</span><span style="color: #800000">"</span><span style="color: #800000">主</span><span style="color: #800000">"</span>)
可以用threading.current_thread().getName()来查看每个g1和g2,查看的结果为DummyThread-n,即假线程,虚拟线程,其实都在一个线程里面
进程线程的任务切换是由操作系统自行切换的,你自己不能控制
协程是通过自己的程序(代码)来进行切换的,自己能够控制,只有遇到协程模块能够识别的IO操作的时候,程序才会进行任务切换,实现并发效果,如果所有程序都没有IO操作,那么就基本属于串行执行了
4.Gevent之同步与异步
<span style="color: #0000ff">from</span><span style="color: #000000"> gevent import spawn,joinall,monkey;monkey.patch_all() import time def task(pid): </span><span style="color: #800000">""" </span> Some non-<span style="color: #000000">deterministic task </span><span style="color: #800000">""" </span> time.sleep(<span style="color: #800080">0.5</span><span style="color: #000000">) print(</span><span style="color: #800000">"</span><span style="color: #800000">Task %s done</span><span style="color: #800000">"</span> %<span style="color: #000000"> pid) def synchronous(): </span><span style="color: #0000ff">for</span> i <span style="color: #0000ff">in</span> range(<span style="color: #800080">10</span><span style="color: #000000">): task(i) def asynchronous(): g_l</span>=[spawn(task,i) <span style="color: #0000ff">for</span> i <span style="color: #0000ff">in</span> range(<span style="color: #800080">10</span><span style="color: #000000">)] joinall(g_l) </span><span style="color: #0000ff">if</span> __name__ == <span style="color: #800000">"</span><span style="color: #800000">__main__</span><span style="color: #800000">"</span><span style="color: #000000">: print(</span><span style="color: #800000">"</span><span style="color: #800000">Synchronous:</span><span style="color: #800000">"</span><span style="color: #000000">) synchronous() print(</span><span style="color: #800000">"</span><span style="color: #800000">Asynchronous:</span><span style="color: #800000">"</span><span style="color: #000000">) asynchronous() #上面程序的重要部分是将task函数封装到Greenlet内部线程的gevent.spawn。 初始化的greenlet列表存放在数组threads中,此数组被传给gevent.joinall 函数,后者阻塞当前流程,并执行所有给定的greenlet。执行流程只会在 所有greenlet执行完后才会继续向下走。 协程:同步异步对比</span>
同步异步对比
5.Gevent之应用举例一
<span style="color: #0000ff">from</span><span style="color: #000000"> gevent import monkey;monkey.patch_all() import gevent import requests import time def get_page(url): print(</span><span style="color: #800000">"</span><span style="color: #800000">GET: %s</span><span style="color: #800000">"</span> %<span style="color: #000000">url) response</span>=requests.<span style="color: #0000ff">get</span><span style="color: #000000">(url) </span><span style="color: #0000ff">if</span> response.status_code == <span style="color: #800080">200</span><span style="color: #000000">: print(</span><span style="color: #800000">"</span><span style="color: #800000">%d bytes received from %s</span><span style="color: #800000">"</span> %<span style="color: #000000">(len(response.text),url)) start_time</span>=<span style="color: #000000">time.time() gevent.joinall([ gevent.spawn(get_page,</span><span style="color: #800000">"</span><span style="color: #800000">https://www.python.org/</span><span style="color: #800000">"</span><span style="color: #000000">), gevent.spawn(get_page,</span><span style="color: #800000">"</span><span style="color: #800000">https://www.yahoo.com/</span><span style="color: #800000">"</span><span style="color: #000000">), gevent.spawn(get_page,</span><span style="color: #800000">"</span><span style="color: #800000">https://github.com/</span><span style="color: #800000">"</span><span style="color: #000000">), ]) stop_time</span>=<span style="color: #000000">time.time() print(</span><span style="color: #800000">"</span><span style="color: #800000">run time is %s</span><span style="color: #800000">"</span> %(stop_time-<span style="color: #000000">start_time)) 协程应用:爬虫</span>
协程应用:爬虫
将上面的程序最后加上一段串行的代码看看效率:如果你的程序不需要太高的效率,那就不用什么并发啊协程啊之类的东西
print(<span style="color: #800000">"</span><span style="color: #800000">--------------------------------</span><span style="color: #800000">"</span><span style="color: #000000">) s </span>=<span style="color: #000000"> time.time() requests.</span><span style="color: #0000ff">get</span>(<span style="color: #800000">"</span><span style="color: #800000">https://www.python.org/</span><span style="color: #800000">"</span><span style="color: #000000">) requests.</span><span style="color: #0000ff">get</span>(<span style="color: #800000">"</span><span style="color: #800000">https://www.yahoo.com/</span><span style="color: #800000">"</span><span style="color: #000000">) requests.</span><span style="color: #0000ff">get</span>(<span style="color: #800000">"</span><span style="color: #800000">https://github.com/</span><span style="color: #800000">"</span><span style="color: #000000">) t </span>=<span style="color: #000000"> time.time() print(</span><span style="color: #800000">"</span><span style="color: #800000">串行时间>></span><span style="color: #800000">"</span>,t-s)
View Code
6.Gevent之应用举例二
通过gevent实现单线程下的socket并发(from gevent import monkey;monkey.patch_all()一定要放到导入socket模块之前,否则gevent无法识别socket的阻塞)
一个网络请求里面经过多个时间延迟time.
<span style="color: #0000ff">from</span><span style="color: #000000"> gevent import monkey;monkey.patch_all() </span><span style="color: #0000ff">from</span> socket import *<span style="color: #000000"> import gevent #如果不想用money.patch_all()打补丁,可以用gevent自带的socket # </span><span style="color: #0000ff">from</span><span style="color: #000000"> gevent import socket # s</span>=<span style="color: #000000">socket.socket() def server(server_ip,port): s</span>=<span style="color: #000000">socket(AF_INET,SOCK_STREAM) s.setsockopt(SOL_SOCKET,SO_REUSEADDR,</span><span style="color: #800080">1</span><span style="color: #000000">) s.bind((server_ip,port)) s.listen(</span><span style="color: #800080">5</span><span style="color: #000000">) </span><span style="color: #0000ff">while</span><span style="color: #000000"> True: conn,addr</span>=<span style="color: #000000">s.accept() gevent.spawn(talk,conn,addr) def talk(conn,addr): </span><span style="color: #0000ff">try</span><span style="color: #000000">: </span><span style="color: #0000ff">while</span><span style="color: #000000"> True: res</span>=conn.recv(<span style="color: #800080">1024</span><span style="color: #000000">) print(</span><span style="color: #800000">"</span><span style="color: #800000">client %s:%s msg: %s</span><span style="color: #800000">"</span> %(addr[<span style="color: #800080">0</span>],addr[<span style="color: #800080">1</span><span style="color: #000000">],res)) conn.send(res.upper()) except Exception </span><span style="color: #0000ff">as</span><span style="color: #000000"> e: print(e) </span><span style="color: #0000ff">finally</span><span style="color: #000000">: conn.close() </span><span style="color: #0000ff">if</span> __name__ == <span style="color: #800000">"</span><span style="color: #800000">__main__</span><span style="color: #800000">"</span><span style="color: #000000">: server(</span><span style="color: #800000">"</span><span style="color: #800000">127.0.0.1</span><span style="color: #800000">"</span>,<span style="color: #800080">8080</span>)
服务端
<span style="color: #0000ff">from</span> socket import *<span style="color: #000000"> client</span>=<span style="color: #000000">socket(AF_INET,SOCK_STREAM) client.connect((</span><span style="color: #800000">"</span><span style="color: #800000">127.0.0.1</span><span style="color: #800000">"</span>,<span style="color: #800080">8080</span><span style="color: #000000">)) </span><span style="color: #0000ff">while</span><span style="color: #000000"> True: msg</span>=input(<span style="color: #800000">"</span><span style="color: #800000">>>: </span><span style="color: #800000">"</span><span style="color: #000000">).strip() </span><span style="color: #0000ff">if</span> not msg:<span style="color: #0000ff">continue</span><span style="color: #000000"> client.send(msg.encode(</span><span style="color: #800000">"</span><span style="color: #800000">utf-8</span><span style="color: #800000">"</span><span style="color: #000000">)) msg</span>=client.recv(<span style="color: #800080">1024</span>)
客户端
<span style="color: #0000ff">from</span><span style="color: #000000"> threading import Thread </span><span style="color: #0000ff">from</span> socket import *<span style="color: #000000"> import threading def client(server_ip,port): c</span>=<span style="color: #000000">socket(AF_INET,SOCK_STREAM) #套接字对象一定要加到函数内,即局部名称空间内,放在函数外则被所有线程共享,则大家公用一个套接字对象,那么客户端端口永远一样了 c.connect((server_ip,port)) count</span>=<span style="color: #800080">0</span> <span style="color: #0000ff">while</span><span style="color: #000000"> True: c.send((</span><span style="color: #800000">"</span><span style="color: #800000">%s say hello %s</span><span style="color: #800000">"</span> %(threading.current_thread().getName(),count)).encode(<span style="color: #800000">"</span><span style="color: #800000">utf-8</span><span style="color: #800000">"</span><span style="color: #000000">)) msg</span>=c.recv(<span style="color: #800080">1024</span><span style="color: #000000">) print(msg.decode(</span><span style="color: #800000">"</span><span style="color: #800000">utf-8</span><span style="color: #800000">"</span><span style="color: #000000">)) count</span>+=<span style="color: #800080">1</span> <span style="color: #0000ff">if</span> __name__ == <span style="color: #800000">"</span><span style="color: #800000">__main__</span><span style="color: #800000">"</span><span style="color: #000000">: </span><span style="color: #0000ff">for</span> i <span style="color: #0000ff">in</span> range(<span style="color: #800080">500</span><span style="color: #000000">): t</span>=Thread(target=client,args=(<span style="color: #800000">"</span><span style="color: #800000">127.0.0.1</span><span style="color: #800000">"</span>,<span style="color: #800080">8080</span><span style="color: #000000">)) t.start()</span>
多线程并发多个客户端,去请求上面的服务端是没问题的