Python Asyncio
Asyncio是用来实现异步网络操作,并发,协程
Asyncio关键字
- event_loop 事件循环 程序开启一个事件队列,把一些函数注册到事件队列中,当满足事件发生的条件,调用相应的协程函数
- coroutine 协程 协程对象,指一个使用
async
关键字定义的函数,当调用函数时,不会立即执行函数,而是返回一个协程对象。协程对象需要注册到事件队列中,由事件队列调用 - task 任务 一个协程对象就是一个原生可以挂起的函数,任务则是对协程进一步封装,其中包含了任务的各种状态
- future 代表将来执行或没有执行的的任务结果,它与task没有本质上的区别
- async/await 关键字,python3.5+用于定义协程的关键字,async定义一个协程,await用于挂起阻塞异步调用接口
定义一个协程
使用async
关键字定义协程
from time import time
import asyncio
async def f():
print("i'm coroutine")
start = time()
coroutine = f()
# 获取一个事件loop
loop = asyncio.get_event_loop()
# 将协程加入到事件队列中
loop.run_until_complete(coroutine)
end = time()
print('Time:{}'.format(end-start))
async
关键字定义一个协程,使用get_event_loop()
方法获取当前主线程的事件loop;再使用run_until_complete(coroutine)
将协程注册到事件队列中,并启用事件轮循。
创建task
协程对象不能直接运行,当注册到事件队列中,其实是run_until_complete(coroutine)
将协程封装成一个task任务,用于保存协程的状态,以及未来获取协程结果
from time import time
import asyncio
async def f():
print("i'm coroutine")
return "return i'm coroutine"
start = time()
coroutine = f()
# 获取当前主线程的事件轮循队列
loop = asyncio.get_event_loop()
# 创建任务
task = loop.create_task(coroutine)
# 执行任务
loop.run_until_complete(task)
print(task)
print(task.result())
end = time()
print('it takes %d seconds' % (end-start))
创建task后,在task加入事件循环之前为pending状态,当完成后,状态为finished
关于上面通过loop.create_task(coroutine)创建task,同样的可以通过 asyncio.ensure_future(coroutine)创建task
绑定回调
绑定回调,在task执行完成的时候可以获取执行的结果,回调的最后一个参数是future对象,通过该对象可以获取协程返回值
from time import time
import asyncio
async def f():
print("i'm coroutine")
return "return i'm coroutine"
def callback(future):
print('result is {}'.format(future.result()))
start = time()
coroutine = f()
# 获取当前主线程的事件轮循队列
loop = asyncio.get_event_loop()
# 创建任务
task = loop.create_task(coroutine)
# 绑定回调函数
task.add_done_callback(callback)
# 执行任务
loop.run_until_complete(task)
print(task)
end = time()
print('it takes %d seconds' % (end-start))
通过add_done_callback方法给task任务添加回调函数,当task(也可以说是coroutine)执行完成的时候,就会调用回调函数。并通过参数future获取协程执行的结果。这里我们创建 的task和回调里的future对象实际上是同一个对象
阻塞和await
使用async
可以定义一个协程,使用await可以针对耗时操作进行挂起,就像生成器的yield
,函数让出控制权。协程遇到await
,事件轮循会挂起该协程,执行别的协程,直到协程也挂起或者执行完成,再进行下一个协程的执行
耗时的一般是IO操作,例如网络请求,文件读取。模拟IO操作,使用async.sleep()
模拟IO操作,协程的目的也是让IO操作异步化
from time import time
import asyncio
async def f():
print("i'm coroutine")
await asyncio.sleep(2)
return "return i'm coroutine"
# def callback(future):
# print('result is {}'.format(future.result()))
start = time()
coroutine = f()
# 获取当前主线程的事件轮循队列
loop = asyncio.get_event_loop()
# 创建任务
task = loop.create_task(coroutine)
# 执行任务
loop.run_until_complete(task)
print(task.result())
end = time()
print('it takes %d seconds' % (end-start))
并发和并行
并发是指具有多个活动的系统
并行是指用并发使一个系统运行得更快,并行可以在操作系统的多个抽象层次进行运用
所以并行通常是指有多个任务同时进行,并行则是在同一时刻有多个任务同时进行
from time import time
import asyncio
async def f(x):
print("i'm coroutine{}".format(x))
await asyncio.sleep(2)
return "return i'm coroutine{}".format(x)
# def callback(future):
# print('result is {}'.format(future.result()))
start = time()
coroutine = f(1)
coroutine1 = f(2)
coroutine2 = f(3)
# 获取当前主线程的事件轮循队列
loop = asyncio.get_event_loop()
# 创建任务列表
tasks = [asyncio.ensure_future(coroutine), asyncio.ensure_future(
coroutine1), asyncio.ensure_future(coroutine2)]
# 执行任务
loop.run_until_complete(asyncio.wait(tasks))
for task in tasks:
print('result is {}'.format(task.result()))
end = time()
print('it takes %d seconds' % (end-start))
运行结果
i'm coroutine1
i'm coroutine2
i'm coroutine3
result is return i'm coroutine1
result is return i'm coroutine2
result is return i'm coroutine3
it takes 2 seconds
总共时间为2s。2s阻塞时间,所有协程执行完毕,如果是同步执行,至少需要7s。此时运用asyncio
模块实现了并发,asyncio.wait(tasks)
也可以使用async.gather(*tasks)
,前者接受一个列表,后者接收一堆task
协程嵌套
使用async可以定义协程,协程用于耗时的io操作,我们也可以封装更多的io操作过程,这样就实现了嵌套的协程,即一个协程中await了另外一个协程,如此连接起来
单任务
from time import time
import asyncio
async def f(x):
print("i'm coroutine{}".format(x))
await asyncio.sleep(2)
return "return i'm coroutine{}".format(x)
# def callback(future):
# print('result is {}'.format(future.result()))
async def main():
coroutine = f(1)
result = await coroutine
print(result)
start = time()
loop = asyncio.get_event_loop()
loop.run_until_complete(main())
end = time()
print('time is {}'.format(end-start))
运行结果
i'm coroutine1
return i'm coroutine1
time is 2.015313148498535
当在main这个协成里,等待协程f()
的return
结果,可以使用await f(1)
等待执行的结果,并将其赋值给result
多任务
from time import time
import asyncio
async def f(x):
print("i'm coroutine{}".format(x))
await asyncio.sleep(2)
return "return i'm coroutine{}".format(x)
async def main():
coroutine = f(1)
coroutine1 = f(2)
coroutine2 = f(3)
tasks = [asyncio.ensure_future(coroutine), asyncio.ensure_future(
coroutine1), asyncio.ensure_future(coroutine2)]
dones, pendings = await asyncio.wait(tasks)
for result in dones:
print(result.result())
start = time()
loop = asyncio.get_event_loop()
loop.run_until_complete(main())
end = time()
print('time is {}'.format(end-start))
多任务中,要使用asyncio.wait()
并发执行任务,当协程完成之后,使用dones
,pendings
,dones
,pendings
类型是集合,若是一个参数,则返回的是一个元组,元组中没有result()
方法
运行结果
i'm coroutine1
i'm coroutine2
i'm coroutine3
return i'm coroutine1
return i'm coroutine2
return i'm coroutine3
time is 2.012080430984497
使用asyncio.gather(*tasks)
会将结果保存至一个List
中
from time import time
import asyncio
async def f(x):
print("i'm coroutine{}".format(x))
await asyncio.sleep(2)
return "return i'm coroutine{}".format(x)
async def main():
coroutine = f(1)
coroutine1 = f(2)
coroutine2 = f(3)
tasks = [asyncio.ensure_future(coroutine), asyncio.ensure_future(
coroutine1), asyncio.ensure_future(coroutine2)]
dones, pendings = await asyncio.wait(tasks)
for result in dones:
print(result.result())
start = time()
loop = asyncio.get_event_loop()
loop.run_until_complete(main())
end = time()
print('time is {}'.format(end-start))
运行结果
i'm coroutine1
i'm coroutine2
i'm coroutine3
<class 'list'> ["return i'm coroutine1", "return i'm coroutine2", "return i'm coroutine3"]
return i'm coroutine1
return i'm coroutine2
return i'm coroutine3
time is 2.0033509731292725
不在main协程函数里处理结果,直接返回await的内容,那么最外层的run_until_complete将会返回main协程的结果。 将上述的代码更改为:
from time import time
import asyncio
async def f(x):
print("i'm coroutine{}".format(x))
await asyncio.sleep(2)
return "return i'm coroutine{}".format(x)
async def main():
coroutine = f(1)
coroutine1 = f(2)
coroutine2 = f(3)
tasks = [asyncio.ensure_future(coroutine), asyncio.ensure_future(
coroutine1), asyncio.ensure_future(coroutine2)]
return await asyncio.gather(*tasks)
start = time()
loop = asyncio.get_event_loop()
results = loop.run_until_complete(main())
for result in results:
print(result)
end = time()
print('time is {}'.format(end-start))
运行结果
i'm coroutine1
i'm coroutine2
i'm coroutine3
return i'm coroutine1
return i'm coroutine2
return i'm coroutine3
time is 2.0101969242095947
或者使用asyncio.wait(tasks)
将任务挂起,再从已完成的协程中读取result()
值
from time import time
import asyncio
async def f(x):
print("i'm coroutine{}".format(x))
await asyncio.sleep(2)
return "return i'm coroutine{}".format(x)
# def callback(future):
# print('result is {}'.format(future.result()))
async def main():
coroutine = f(1)
coroutine1 = f(2)
coroutine2 = f(3)
tasks = [asyncio.ensure_future(coroutine), asyncio.ensure_future(
coroutine1), asyncio.ensure_future(coroutine2)]
return await asyncio.wait(tasks)
start = time()
loop = asyncio.get_event_loop()
dones, pendings = loop.run_until_complete(main())
for result in dones:
print(result.result())
end = time()
print('time is {}'.format(end-start))
也可以使用asyncio.as_computed
from time import time
import asyncio
async def f(x):
print("i'm coroutine{}".format(x))
await asyncio.sleep(2)
return "return i'm coroutine{}".format(x)
# def callback(future):
# print('result is {}'.format(future.result()))
async def main():
coroutine = f(1)
coroutine1 = f(2)
coroutine2 = f(3)
tasks = [asyncio.ensure_future(coroutine), asyncio.ensure_future(
coroutine1), asyncio.ensure_future(coroutine2)]
for task in asyncio.as_completed(tasks):
print(await task)
start = time()
loop = asyncio.get_event_loop()
loop.run_until_complete(main())
end = time()
print('time is {}'.format(end-start))
协程的停止
future对象有几个状态:
- Pending
- Running
- Done
- Cacelled
from time import time
import asyncio
async def f(x):
print("i'm coroutine{}".format(x))
await asyncio.sleep(x)
return "return i'm coroutine{}".format(x)
# def callback(future):
# print('result is {}'.format(future.result()))
async def main():
coroutine = f(1)
coroutine1 = f(2)
coroutine2 = f(3)
tasks = [asyncio.ensure_future(coroutine), asyncio.ensure_future(
coroutine1), asyncio.ensure_future(coroutine2)]
return await asyncio.gather(*tasks)
try:
start = time()
loop = asyncio.get_event_loop()
result = loop.run_until_complete(main())
for item in result:
print(item)
end = time()
print('time is {}'.format(end-start))
except KeyboardInterrupt:
print(asyncio.Task.all_tasks())
for item in asyncio.Task.all_tasks():
print(item.cancel())
loop.stop()
loop.run_forever()
finally:
loop.close()
启动事件循环之后,马上ctrl+c,会触发run_until_complete的执行异常 KeyBorardInterrupt。然后通过循环asyncio.Task取消future。可以看到输出如下:
运行结果
i'm coroutine1
i'm coroutine2
i'm coroutine3
{<Task pending coro=<main() running at app.py:29> wait_for=<_GatheringFuture pending cb=[<TaskWakeupMethWrapper object at 0x0000014E4532CD98>()]>>, <Task finished coro=<f() done, defined at app.py:13> result="return i'm coroutine1">, <Task pending coro=<f() running at app.py:15> wait_for=<Future pending cb=[<TaskWakeupMethWrapper object at 0x0000014E4533F4F8>()]> cb=[gather.<locals>._done_callback() at C:\Python3\lib\asyncio\tasks.py:691]>, <Task pending coro=<f() running at app.py:15> wait_for=<Future pending cb=[<TaskWakeupMethWrapper object at 0x0000014E4556E708>()]> cb=[gather.<locals>._done_callback() at C:\Python3\lib\asyncio\tasks.py:691]>}
True
False
True
True
True表示cannel成功,loop stop之后还需要再次开启事件循环,最后在close,不然还会抛出异常
循环task,逐个cancel是一种方案,可是正如上面我们把task的列表封装在main函数中,main函数外进行事件循环的调用。这个时候,main相当于最外出的一个task,那么处理包装的main函数即可。
不同线程的事件轮循
很多时候,我们的事件循环用于注册协程,而有的协程需要动态的添加到事件循环中。一个简单的方式就是使用多线程。当前线程创建一个事件循环,然后在新建一个线程,在新线程中启动事件循环。当前线程不会被block。
import asyncio
from threading import Thread
import time
def now(): return time.time()
def start_loop(loop):
asyncio.set_event_loop(loop)
loop.run_forever()
def more_work(x):
print('More work {}'.format(x))
time.sleep(x)
print('Finished more work {}'.format(x))
start = now()
new_loop = asyncio.new_event_loop()
t = Thread(target=start_loop, args=(new_loop,))
t.start()
print('TIME: {}'.format(time.time() - start))
new_loop.call_soon_threadsafe(more_work, 6)
new_loop.call_soon_threadsafe(more_work, 3)
启动上述代码之后,当前线程不会被block,新线程中会按照顺序执行call_soon_threadsafe方法注册的more_work方法, 后者因为time.sleep操作是同步阻塞的,因此运行完毕more_work需要大致6 + 3
新线程协程
import asyncio
import time
from threading import Thread
now = lambda :time.time()
def start_loop(loop):
asyncio.set_event_loop(loop)
loop.run_forever()
async def do_some_work(x):
print('Waiting {}'.format(x))
await asyncio.sleep(x)
print('Done after {}s'.format(x))
start = now()
new_loop = asyncio.new_event_loop()
t = Thread(target=start_loop, args=(new_loop,))
t.start()
print('TIME: {}'.format(time.time() - start))
asyncio.run_coroutine_threadsafe(do_some_work(6), new_loop)
asyncio.run_coroutine_threadsafe(do_some_work(4), new_loop)
上述的例子,主线程中创建一个new_loop,然后在另外的子线程中开启一个无限事件循环。 主线程通过run_coroutine_threadsafe新注册协程对象。这样就能在子线程中进行事件循环的并发操作,同时主线程又不会被block。一共执行的时间大概在6s左右