问题

Question

我的函数run_tasks(all_tasks, window_size)，它接受asyncio个任务的生成器，并在以下情况下返回其值：

同时从window_size

窗口

all_tasks

保留返回结果的顺序（all_tasks[i]结果为results[i]）
处理每次运行的异常

我目前的实施：

import asyncio
from itertools import islice


# run all tasks and return their results in the same order
# window is the max number of tasks that will run in parallel
def run_tasks(all_tasks, window_size=4):
    loop = asyncio.get_event_loop()

    while True:
        window_tasks = list(islice(all_tasks, window_size))
        if not window_tasks:
            break

        futures = asyncio.wait(window_tasks, loop=loop)
        finished, unfinished = loop.run_until_complete(futures)

        # sort finished tasks by their launch order.
        # removing this line makes returned tasks unordered
        finished = sorted(finished, key=lambda f: window_tasks.index(f._coro))

        for finished_task in finished:
            try:
                yield finished_task.result()
            except Exception as e:
                yield repr(e)

# Example Usage:

# a coroutine that sometime raises exception
async def sleepy(i):
    print(f'{i} started')
    await asyncio.sleep(10 - i)
    print(f'{i} finished')
    if i == 5:
        raise ValueError('5 is the worst')
    return i

# a generator of tasks
all_tasks = (sleepy(i) for i in range(10))

for result in list(run_tasks(all_tasks)):
    print(result)

问题

我的实现问题是，我无法在不访问f._coro对象的内部属性asyncio.Task的情况下找到对任务进行排序的方法。

# removing this line makes returned tasks unordered
finished = sorted(finished, key=lambda f: window_tasks.index(f._coro))

我可以使用asyncio.gather(*tasks)，但这不会处理错误。

我愿意接受有关如何为run_tasks()实施此三个属性而无法访问f._coro的建议。

Answer 1

如果您指定了关键字参数

void main(){
    gl_PointSize = 100.;
    gl_Position = projectionMatrix * modelViewMatrix * vec4(position,1.);
}

，则

asyncio.gather can会返回错误。为了区分真实异常和异常对象返回作为协程的结果，您可以使用return_exceptions将window_tasks包裹在任务中：

ensure_future

asyncio运行多个任务，同时保留顺序和处理错误

问题

1 个答案: