aioresult 1.0

Capture the result of a Trio or anyio task

Welcome to aioresult!

This is a very small library to capture the result of an asynchronous operation, either an async function (with the ResultCapture class) or more generally (with the Future class). It works with Trio nurseries and anyio task groups. It is not needed for Python 3.11 asyncio task groups because those already return an object representing the task, allowing the result to be retrieved.

• Code is hosted on github: https://github.com/arthur-tacca/aioresult

• Documentation is on ReadTheDocs:

  • Overview (this page): https://aioresult.readthedocs.io/en/v1.0/overview.html

  • Capturing a result: https://aioresult.readthedocs.io/en/v1.0/result_capture.html

  • Future objects: https://aioresult.readthedocs.io/en/v1.0/future.html

  • Utility functions for waiting: https://aioresult.readthedocs.io/en/v1.0/wait.html

• The package is on PyPI: https://pypi.org/project/aioresult/

Quick Overview

The ResultCapture class runs an async function in a nursery and stores its return value (or raised exception) for later:

async with trio.open_nursery() as n:
    result1 = ResultCapture.start_soon(n, foo, 1)
    result2 = ResultCapture.start_soon(n, foo, 2)
# At this point the tasks have completed, and results are stashed in ResultCapture objects
print("results", result1.result(), result2.result())

When stored in a list, the effect is very similar to the asyncio gather() function:

async with trio.open_nursery() as n:
    results = [ResultCapture.start_soon(n, foo, i) for i in range(10)]
print("results:", *[r.result() for r in results])

Note

A key design decision about the ResultCapture class is that exceptions are allowed to propagate out of the task into their enclosing nursery. This is unlike some similar libraries, which consume the exception in its original context and rethrow it later. In practice, aioresult’s behaviour is simpler and less error prone.

There is also a simple Future class that shares a lot of its code with ResultCapture. The result is retrieved the same way, but it is set explicitly rather than captured from a task. It is most often used when an API wants to return a value that will be demultiplexed from a shared connection:

# When making a request, create a future, store it for later and return to caller
f = aioresult.Future()

# The result is set, usually inside a networking API
f.set_result(result)

# The calling code can wait for the result then retrieve it
await f.wait_done()
print("result:", f.result())

The interface in Future and ResultCapture to wait for a result and retrieve it is shared in a base class ResultBase.

There are also a few simple utility functions to help waiting for results: wait_any() and wait_all() to wait for one or all of a collection of tasks to complete, and results_to_channel() to allow using the results as they become available.

Installation and Usage

Install into a suitable virtual environment with pip:

pip install aioresult

aioresult can be used with Trio nurseries:

import trio
from aioresult import ResultCapture

async def wait_and_return(i):
    await trio.sleep(i)
    return i

async def use_aioresult():
    async with trio.open_nursery() as n:
        results = [ResultCapture.start_soon(n, wait_and_return, i) for i in range(5)]
    print("results:", *[r.result() for r in results])

if __name__ == "__main__":
    trio.run(use_aioresult)

It can also be used with anyio task groups:

import asyncio
import anyio
from aioresult import ResultCapture

async def wait_and_return(i):
    await anyio.sleep(i)
    return i

async def use_aioresult():
     async with anyio.create_task_group() as tg:
         results = [ResultCapture.start_soon(tg, wait_and_return, i) for i in range(5)]
     print("results:", *[r.result() for r in results])

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

Contributing

This library is deliberately small and limited in scope, so it is essentially “done”. An exception to this is that the typing annotations are not exhaustive and have not been tested with any type checker, so contributions to improve this would be welcome. I could perhaps also be persuaded to add support for optionally including a per-task cancel scope (see issue #2).

To test any changes, install the test requirements (see the pyproject.toml file) and run pytest in the root of the repository:

python -m pytest

To also get coverage information, run it with the coverage command:

coverage run -m pytest

You can then use coverage html to get a nice HTML output of exactly what code has been tested and what has been missed.

License

Copyright Arthur Tacca 2022 - 2024

Distributed under the Boost Software License, Version 1.0. See accompanying file LICENSE or the copy at https://www.boost.org/LICENSE_1_0.txt

This is similar to other liberal licenses like MIT and BSD: you can use this library without the need to share your program’s source code, so long as you provide attribution of aioresult.

The Boost license has the additional provision that you do not even need to provide attribution if you are distributing your software in binary form only, e.g. if you have compiled to an executable with Nuitka. (Bundlers like pyinstaller and py2exe don’t count for this because they still include the source code internally.)