async-plus 0.3.0

Async-related stuff you miss in standard library

Async-related stuff you miss in standard library

“The horrors of hanging tasks when using asyncio” by Jacques Callot

Safely run coroutines concurrently

The asyncio.gather() function has an issue: in case of error in one of coroutines the rest coroutines are left running detached. This might cause hard to detect problems. On the contrary, the async_plus.try_gather() insures all tasks are cancelled on error:

result1, result2 = await async_plus.try_gather(
    coroutine_func1(...),
    coroutine_func2(...),
)

Fire-and-forget task

With create_task() it’s your responsibility to retrieve exception. Usually this causes that exception is not seen until process finishes or even is not seen at all if process is killed. With launch_watched() exception is logged immediately when it’s raised.

async_plus.launch_watched(your_coroutine_func(...))

Structuring groups of tasks

async with async_plus.task_scope() as scope:
    scope.launch(coroutine_func1(...))
    scope.launch(coroutine_func2(...))
    await scope.wait()

By default, wait() call returns when all tasks finish or first exception occurs. In all cases all unfinished tasks are cancelled at the end of async with block.

Increase delay between attempts in supervisor

retry_delayer = async_plus.RetryDelayer([0, 10, 60], random_shift=1)
while True:
    try:
        await run_service_x()
    # In Python <3.8 it inherits from Exception
    except asyncio.CancelledError:
        raise
    except Exception:
        logger.exception('Error in service X:')
        await retry_delayer.sleep()

Log long waits

Does your program hang and you don’t know what it’s waiting for? Wrap suspicious coroutines with impatient() to see bottlenecks:

await async_plus.impatient(asyncio.sleep(10), log_after=5)

Or just log the time it took:

await async_plus.impatient(asyncio.sleep(10), log_completion='always')

Change log

See CHANGELOG.