我正在尝试从异步队列中提取任务，并在发生异常时调用给定的错误处理程序。排队的项目以字典（由enqueue_task列为字典）形式给出，其中包含任务，可能的错误处理程序以及错误处理程序可能需要的所有args / kwarg。由于我想在任务完成时处理所有错误，因此我将每个任务映射到出队的字典，并在发生异常时尝试访问它。
async def _check_tasks(self):
try:
while self._check_tasks_task or not self._check_task_queue.empty():
tasks = []
details = {}
try:
while len(tasks) < self._CHECK_TASKS_MAX_COUNT:
detail = self._check_task_queue.get_nowait()
task = detail['task']
tasks.append(task)
details[task] = detail
except asyncio.QueueEmpty:
pass
if tasks:
for task in asyncio.as_completed(tasks):
try:
await task
except Exception as e:
logger.exception('')
detail = details[task]
error_handler = detail.get('error_handler')
error_handler_args = detail.get('error_handler_args', [])
error_handler_kwargs = detail.get('error_handler_kwargs', {})
if error_handler:
logger.info('calling error handler')
if inspect.iscoroutinefunction(error_handler):
self.enqueue_task(
task=error_handler(
e,
*error_handler_args,
**error_handler_kwargs
)
)
else:
error_handler(e, *error_handler_args, **error_handler_kwargs)
else:
logger.exception(f'Exception encountered while handling task: {str(e)}')
else:
await asyncio.sleep(self._QUEUE_EMPTY_SLEEP_TIME)
except:
logger.exception('')
def enqueue_task(self, task, error_handler=None, error_handler_args=[],
error_handler_kwargs={}):
if not asyncio.isfuture(task):
task = asyncio.ensure_future(task)
self._app.gateway._check_task_queue.put_nowait({
'task': task,
'error_handler': error_handler,
'error_handler_args': error_handler_args,
'error_handler_kwargs': error_handler_kwargs,
})
但是，当发生异常时，似乎在details词典中找不到用作键的任务，并且我收到以下错误：
KeyError: <generator object as_completed.<locals>._wait_for_one at 0x7fc2d1cea308>
Exception encountered while handling task: <generator object as_completed.<locals>._wait_for_one at 0x7fc2d1cea308>
Traceback (most recent call last):
File "/app/app/gateway/gateway.py", line 64, in _check_tasks
detail = details[task]
KeyError: <generator object as_completed.<locals>._wait_for_one at 0x7fc2d1cea308>
当task产生asyncio.as_completed时，它似乎是生成器
<generator object as_completed.<locals>._wait_for_one at 0x7fc2d1cea308>
当我希望这是一项任务
<Task pending coro=<GatewayL1Component._save_tick_to_stream() running at /app/app/gateway/l1.py:320> wait_for=<Future pending cb=[<TaskWakeupMethWrapper object at 0x7fc2d4380d98>()]>>
在task产生后，为什么asyncio.as_complete是生成器而不是原始任务？有没有办法访问原始任务？
答案 0 :(得分：1)
为什么
asyncio.as_complete产生任务后，任务是生成器而不是原始任务？
问题在于as_completed不是async iterator（用async for用尽的），而是普通的迭代器。如果异步迭代器的__aiter__可以在等待异步事件时挂起，则普通迭代器的__iter__必须立即提供结果。显然，它不能产生已完成的任务，因为还没有时间来完成任务，因此它会产生一个等待对象，实际上等待任务完成。这是看起来像发电机的对象。
实现的另一个结果是，与原始concurrent.futures.as_completed相比，等待该任务将为您提供原始任务的结果，而不是对任务对象的引用。这使asyncio.as_completed不太直观，更难使用，并且there is a bug report认为as_completed也应可用作异步迭代器，并提供正确的语义。 （这可以通过向后兼容的方式来完成。）
是否可以访问原始任务？
作为一种解决方法，您可以通过将原始任务包装到一个协程中来创建as_completed的异步版本，该协程在任务完成时会完成，并将任务作为结果：
async def as_completed_async(futures):
loop = asyncio.get_event_loop()
wrappers = []
for fut in futures:
assert isinstance(fut, asyncio.Future) # we need Future or Task
# Wrap the future in one that completes when the original does,
# and whose result is the original future object.
wrapper = loop.create_future()
fut.add_done_callback(wrapper.set_result)
wrappers.append(wrapper)
for next_completed in asyncio.as_completed(wrappers):
# awaiting next_completed will dereference the wrapper and get
# the original future (which we know has completed), so we can
# just yield that
yield await next_completed
这应该可以让您获得原始任务-这是一个简单的测试用例：
async def main():
loop = asyncio.get_event_loop()
fut1 = loop.create_task(asyncio.sleep(.2))
fut1.t = .2
fut2 = loop.create_task(asyncio.sleep(.3))
fut2.t = .3
fut3 = loop.create_task(asyncio.sleep(.1))
fut3.t = .1
async for fut in as_completed_async([fut1, fut2, fut3]):
# using the `.t` attribute shows that we've got the original tasks
print('completed', fut.t)
asyncio.get_event_loop().run_until_complete(main())
答案 1 :(得分：0)
使用asyncio.gather代替：
async def _check_tasks(self):
while self._check_tasks_task or not self._check_task_queue.empty():
tasks = []
details = []
try:
while len(tasks) < self._CHECK_TASKS_MAX_COUNT:
detail = self._check_task_queue.get_nowait()
task = detail['task']
tasks.append(task)
details.append(detail)
except asyncio.QueueEmpty:
pass
if tasks:
results = await asyncio.gather(*tasks, return_exceptions=True)
for i, result in enumerate(results):
if isinstance(result, Exception):
detail = details[i]
error_handler = detail.get('error_handler')
error_handler_args = detail.get('error_handler_args', [])
error_handler_kwargs = detail.get('error_handler_kwargs', {})
if error_handler:
logger.info('calling error handler')
if inspect.iscoroutinefunction(error_handler):
self.enqueue_task(
task=error_handler(
result,
*error_handler_args,
**error_handler_kwargs
)
)
else:
error_handler(
result, *error_handler_args, **error_handler_kwargs
)
else:
msg = f'Exception encountered while handling task: {str(result)}'
logger.exception(msg)
else:
await asyncio.sleep(self._QUEUE_EMPTY_SLEEP_TIME)