Creating and managing tasks

A task is a unit of execution that lets you do many things concurrently that need waiting on. This works so that while you can have any number of tasks, the asynchronous event loop can only run one of them at a time. When the task encounters an await statement that requires the task to sleep until something happens, the event loop is then free to work on another task. When the thing the first task was waiting is complete, the event loop will resume the execution of that task on the first opportunity it gets.

Task handling in AnyIO loosely follows the trio model. Tasks can be created (spawned) using task groups. A task group is an asynchronous context manager that makes sure that all its child tasks are finished one way or another after the context block is exited. If a child task, or the code in the enclosed context block raises an exception, all child tasks are cancelled. Otherwise the context manager just waits until all child tasks have exited before proceeding.

Here’s a demonstration:

from anyio import sleep, create_task_group, run

async def sometask(num):
    print('Task', num, 'running')
    await sleep(1)
    print('Task', num, 'finished')

async def main():
    async with create_task_group() as tg:
        for num in range(5):
            tg.start_soon(sometask, num)

    print('All tasks finished!')


Starting and initializing tasks

Sometimes it is very useful to be able to wait until a task has successfully initialized itself. For example, when starting network services, you can have your task start the listener and then signal the caller that initialization is done. That way, the caller can now start another task that depends on that service being up and running. Also, if the socket bind fails or something else goes wrong during initialization, the exception will be propagated to the caller which can then catch and handle it.

This can be done with TaskGroup.start():

from anyio import TASK_STATUS_IGNORED, create_task_group, connect_tcp, create_tcp_listener, run
from import TaskStatus

async def handler(stream):

async def start_some_service(port: int, *, task_status: TaskStatus[None] = TASK_STATUS_IGNORED):
    async with await create_tcp_listener(local_host='', local_port=port) as listener:
        await listener.serve(handler)

async def main():
    async with create_task_group() as tg:
        await tg.start(start_some_service, 5000)
        async with await connect_tcp('', 5000) as stream:


The target coroutine function must call task_status.started() because the task that is calling with TaskGroup.start() will be blocked until then. If the spawned task never calls it, then the TaskGroup.start() call will raise a RuntimeError.


Unlike start_soon(), start() needs an await.

Handling multiple errors in a task group

It is possible for more than one task to raise an exception in a task group. This can happen when a task reacts to cancellation by entering either an exception handler block or a finally: block and raises an exception there. This raises the question: which exception is propagated from the task group context manager? The answer is “both”. In practice this means that a special exception, ExceptionGroup is raised which contains both exception objects. Unfortunately this complicates any code that wishes to catch a specific exception because it could be wrapped in an ExceptionGroup.

Context propagation

Whenever a new task is spawned, context will be copied to the new task. It is important to note which content will be copied to the newly spawned task. It is not the context of the task group’s host task that will be copied, but the context of the task that calls TaskGroup.start() or TaskGroup.start_soon().


Context propagation does not work on asyncio when using Python 3.6, as asyncio support for this only landed in v3.7.