Async

Posted recently Updated recently

By Administrator

23~29 min read

协程可以使用更少的资源实现“并发”的效果。

在开发过程中要注意协程只有在遇到await才会被切换，因此在web开发中如果出现会阻塞的代码（消耗时间的同步代码或者cpu密集型代码块）需要将其添加到其他进程或者线程，才能避免阻塞主其他的请求。

ps：与go相比，GIL锁的存在，在单线程中启动python时，相当于只有一套的GMP。

普通的异步调用

import asyncio
import time

async def say_after(delay, what):
    await asyncio.sleep(delay)
    print(what)

async def main():
    print(f"started at {time.strftime('%X')}")

    await say_after(1, 'hello')
    await say_after(2, 'world')

    print(f"finished at {time.strftime('%X')}")

asyncio.run(main())

create_task：异步并发调用

async def main():
    task1 = asyncio.create_task(
        say_after(1, 'hello'))

    task2 = asyncio.create_task(
        say_after(2, 'world'))

    print(f"started at {time.strftime('%X')}")

    # wait until both tasks are completed (should take
    # around 2 seconds.)
    await task1
    await task2

    print(f"finished at {time.strftime('%X')}")

TaskGroup：异步任务组

async def main():
    async with asyncio.TaskGroup() as tg:
        task1 = tg.create_task(
            say_after(1, 'hello'))

        task2 = tg.create_task(
            say_after(2, 'world'))

        print(f"started at {time.strftime('%X')}")

    # The await is implicit when the context manager exits.

    print(f"finished at {time.strftime('%X')}")

批量迭代创建任务：

注意：需要添加到集合，否则会导致任务执行期间被垃圾回收

background_tasks = set()

for i in range(10):
    task = asyncio.create_task(some_coro(param=i))

    # Add task to the set. This creates a strong reference.
    background_tasks.add(task)

    # To prevent keeping references to finished tasks forever,
    # make each task remove its own reference from the set after
    # completion:
    task.add_done_callback(background_tasks.discard)

任务组的使用及终止

通过在任务组中的任务抛出异常可以终止剩余任务的执行

import asyncio
from asyncio import TaskGroup

class TerminateTaskGroup(Exception):
    """Exception raised to terminate a task group."""

async def force_terminate_task_group():
    """Used to force termination of a task group."""
    raise TerminateTaskGroup()

async def job(task_id, sleep_time):
    print(f'Task {task_id}: start')
    await asyncio.sleep(sleep_time)
    print(f'Task {task_id}: done')

async def main():
    try:
        async with TaskGroup() as group:
            # spawn some tasks
            group.create_task(job(1, 0.5))
            group.create_task(job(2, 1.5))
            # sleep for 1 second
            await asyncio.sleep(1)
            # add an exception-raising task to force the group to terminate
            group.create_task(force_terminate_task_group())
    except* TerminateTaskGroup:
        pass

asyncio.run(main())

gather:并发的创建和运行任务

异常不会导致其他任务停止

import asyncio

async def factorial(name, number):
    f = 1
    for i in range(2, number + 1):
        print(f"Task {name}: compute factorial({number}), currently i={i}...")
        await asyncio.sleep(1)
        f *= i
    print(f"Task {name}: factorial({number}) = {f}")
    return f

async def main():
    # schedule three calls *concurrently*:
    L = await asyncio.gather(
        factorial("A", 2),
        factorial("B", 3),
        factorial("C", 4),
    )
    print(L)

asyncio.run(main())

wait_for：控制协程的运行时间

async def main():
    async with asyncio.timeout(10):
        await long_running_task()
        
# 其他
time_at
time_for

async def eternity():
    # Sleep for one hour
    await asyncio.sleep(3600)
    print('yay!')

async def main():
    # Wait for at most 1 second
    try:
        await asyncio.wait_for(eternity(), timeout=1.0)
    except TimeoutError:
        print('timeout!')

asyncio.run(main())

在其他thread中运行同步函数

def blocking_io():
    print(f"start blocking_io at {time.strftime('%X')}")
    # Note that time.sleep() can be replaced with any blocking
    # IO-bound operation, such as file operations.
    time.sleep(1)
    print(f"blocking_io complete at {time.strftime('%X')}")

async def main():
    print(f"started main at {time.strftime('%X')}")

    await asyncio.gather(
        asyncio.to_thread(blocking_io),
        asyncio.sleep(1))

    print(f"finished main at {time.strftime('%X')}")


asyncio.run(main())

与run_coroutine_threadsafe的区别

在非异步代码中需要运行协程，使用run_coroutine_threadsafe

import asyncio
import threading

# 定义一个协程
async def my_coroutine():
 await asyncio.sleep(1)
 return "Done"

# 在另一个线程中运行事件循环
def start_event_loop(loop):
 asyncio.set_event_loop(loop)
 loop.run_forever()

# 主线程中提交协程到事件循环
def submit_coroutine(loop):
 future = asyncio.run_coroutine_threadsafe(my_coroutine(), loop)
 print(future.result())  # 等待并获取结果

# 创建新的事件循环
loop = asyncio.new_event_loop()

# 启动事件循环线程
thread = threading.Thread(target=start_event_loop, args=(loop,))
thread.start()

# 在主线程中提交协程
submit_coroutine(loop)

# 关闭事件循环
loop.call_soon_threadsafe(loop.stop)
thread.join()

其他


# 通过将参数设置为0，可以在长时间的阻塞任务中，使得其他任务可以执行
asyncio.sleep

# 提高协程效率
# 只有在程序阻塞时才会切换，减少loop导致的性能开销
asyncio.create_eager_task_factory(custom_task_constructor)

# 防止协程被取消
asyncio.shield(aw)

# 当执行成功时，可以返回执行成功的时间
done, pending = await asyncio.wait(aws, *, timeout=None, return_when=ALL_COMPLETED)

# 遍历可等待的对象，获得
asyncio.as_completed

在以上函数外，python还提供了一系列的获取协程状态的函数。

问题：

fastapi中，如果在async中调用了一个阻塞的函数，会阻塞其他请求吗

会,可以将阻塞的函书放到其他线程或者进程（CPU密集）中

# 放到进程中
import concurrent.futures
import math

PRIMES = [
    112272535095293,
    112582705942171,
    112272535095293,
    115280095190773,
    115797848077099,
    1099726899285419]

def is_prime(n):
    if n < 2:
        return False
    if n == 2:
        return True
    if n % 2 == 0:
        return False

    sqrt_n = int(math.floor(math.sqrt(n)))
    for i in range(3, sqrt_n + 1, 2):
        if n % i == 0:
            return False
    return True

def main():
    with concurrent.futures.ProcessPoolExecutor() as executor:
        for number, prime in zip(PRIMES, executor.map(is_prime, PRIMES)):
            print('%d is prime: %s' % (number, prime))

if __name__ == '__main__':
    main()

# 放到线程中
import asyncio

def blocking_function():
    # 模拟阻塞操作
    import time
    time.sleep(5)
    return "Done"

async def main():
    result = await asyncio.to_thread(blocking_function)
    print(result)

asyncio.run(main())

License: CC BY 4.0