如何将 asyncio 与现有的阻塞库一起使用？

Question

我几乎没有阻塞函数 foo，bar 并且我无法更改它们（一些内部库我不控制。与一个或多个网络服务对话）。我如何将它用作异步？例如。我不想做以下事情。

results = []
for inp in inps:
    val = foo(inp)
    result = bar(val)
    results.append(result)

这将是低效的，因为我可以在等待第一个输入时为第二个输入调用 foo，bar 也是如此。我如何包装它们以便它们可以与 asyncio 一起使用（即新的 async、await 语法）？

让我们假设函数是可重入的。也就是说，当之前的 foo 已经在处理时，再次调用 foo 是可以的。

更新

使用可重复使用的装饰器扩展答案。例如点击。

def run_in_executor(f):
    @functools.wraps(f)
    def inner(*args, **kwargs):
        loop = asyncio.get_running_loop()
        return loop.run_in_executor(None, functools.partial(f, *args, **kwargs))

    return inner

Answer 1

这里有（某种）两个问题：首先，如何运行异步阻塞代码，其次，如何运行并发异步代码（asyncio 是单线程的，因此GIL 仍然适用，所以它不是真正并行，但我离题了）。

可以使用 asyncio.ensure_future 创建并发任务，如文档所述 here。

要运行同步代码，您将需要run the blocking code in an executor。示例：

import concurrent.futures
import asyncio
import time

def blocking(delay):
    time.sleep(delay)
    print('Completed.')

async def non_blocking(loop, executor):
    # Run three of the blocking tasks concurrently. asyncio.wait will
    # automatically wrap these in Tasks. If you want explicit access
    # to the tasks themselves, use asyncio.ensure_future, or add a
    # "done, pending = asyncio.wait..." assignment
    await asyncio.wait(
        fs={
            # Returns after delay=12 seconds
            loop.run_in_executor(executor, blocking, 12),
            
            # Returns after delay=14 seconds
            loop.run_in_executor(executor, blocking, 14),
            
            # Returns after delay=16 seconds
            loop.run_in_executor(executor, blocking, 16)
        },
        return_when=asyncio.ALL_COMPLETED
    )

loop = asyncio.get_event_loop()
executor = concurrent.futures.ThreadPoolExecutor(max_workers=5)
loop.run_until_complete(non_blocking(loop, executor))

如果您想使用 for 循环安排这些任务（如您的示例所示），您有几种不同的策略，但基本方法是使用安排任务for 循环（或列表理解等），使用 asyncio.wait 等待它们，然后然后检索结果。示例：

done, pending = await asyncio.wait(
    fs=[loop.run_in_executor(executor, blocking_foo, *args) for args in inps],
    return_when=asyncio.ALL_COMPLETED
)

# Note that any errors raise during the above will be raised here; to
# handle errors you will need to call task.exception() and check if it
# is not None before calling task.result()
results = [task.result() for task in done]

Answer 2

扩展已接受的答案以实际解决问题。

注意：需要 python 3.7+

import functools

from urllib.request import urlopen
import asyncio


def legacy_blocking_function():  # You cannot change this function
    r = urlopen("https://example.com")
    return r.read().decode()


def run_in_executor(f):
    @functools.wraps(f)
    def inner(*args, **kwargs):
        loop = asyncio.get_running_loop()
        return loop.run_in_executor(None, lambda: f(*args, **kwargs))

    return inner


@run_in_executor
def foo(arg):  # Your wrapper for async use
    resp = legacy_blocking_function()
    return f"{arg}{len(resp)}"


@run_in_executor
def bar(arg):  # Another wrapper
    resp = legacy_blocking_function()
    return f"{len(resp)}{arg}"


async def process_input(inp):  # Modern async function (coroutine)
    res = await foo(inp)
    res = f"XXX{res}XXX"
    return await bar(res)


async def main():
    inputs = ["one", "two", "three"]
    input_tasks = [asyncio.create_task(process_input(inp)) for inp in inputs]
    print([await t for t in asyncio.as_completed(input_tasks)])
    # This doesn't work as expected :(
    # print([await t for t in asyncio.as_completed([process_input(inp) for inp in input_tasks])])


if __name__ == '__main__':
asyncio.run(main())

单击 here 获取此示例的最新版本并发送拉取请求。

Answer 3

import asyncio
from time import sleep
import logging

logging.basicConfig(
    level=logging.DEBUG, format="%(asctime)s %(thread)s %(funcName)s %(message)s")


def long_task(t):
    """Simulate long IO bound task."""
    logging.info("2. t: %s", t)
    sleep(t)
    logging.info("4. t: %s", t)
    return t ** 2


async def main():
    loop = asyncio.get_running_loop()
    inputs = range(1, 5)
    logging.info("1.")
    futures = [loop.run_in_executor(None, long_task, i) for i in inputs]
    logging.info("3.")
    results = await asyncio.gather(*futures)
    logging.info("5.")
    for (i, result) in zip(inputs, results):
        logging.info("6. Result: %s, %s", i, result)


if __name__ == "__main__":
    asyncio.run(main())

输出：

2020-03-18 17:13:07,523 23964 main 1.
2020-03-18 17:13:07,524 5008 long_task 2. t: 1
2020-03-18 17:13:07,525 21232 long_task 2. t: 2
2020-03-18 17:13:07,525 22048 long_task 2. t: 3
2020-03-18 17:13:07,526 25588 long_task 2. t: 4
2020-03-18 17:13:07,526 23964 main 3.
2020-03-18 17:13:08,526 5008 long_task 4. t: 1
2020-03-18 17:13:09,526 21232 long_task 4. t: 2
2020-03-18 17:13:10,527 22048 long_task 4. t: 3
2020-03-18 17:13:11,527 25588 long_task 4. t: 4
2020-03-18 17:13:11,527 23964 main 5.
2020-03-18 17:13:11,528 23964 main 6. Result: 1, 1
2020-03-18 17:13:11,528 23964 main 6. Result: 2, 4
2020-03-18 17:13:11,529 23964 main 6. Result: 3, 9
2020-03-18 17:13:11,529 23964 main 6. Result: 4, 16

如何将 asyncio 与现有的阻塞库一起使用？

How to use asyncio with existing blocking library?

python

python-3.x

async-await

python-asyncio

python-3.5