#oneshot-channel #oneshot #channel #async #performance #concurrency #future

no-std async-oneshot

A fast, small, full-featured, async-aware oneshot channel

10 releases (4 breaking)

0.5.9 Mar 13, 2021
0.5.0 Mar 13, 2021
0.4.2 Oct 11, 2020
0.3.3 Aug 25, 2020
0.1.0 May 30, 2019

#11 in #oneshot-channel

Download history 2250/week @ 2023-12-18 1580/week @ 2023-12-25 1526/week @ 2024-01-01 1954/week @ 2024-01-08 1824/week @ 2024-01-15 2128/week @ 2024-01-22 2152/week @ 2024-01-29 2172/week @ 2024-02-05 2313/week @ 2024-02-12 2671/week @ 2024-02-19 2733/week @ 2024-02-26 3715/week @ 2024-03-04 4973/week @ 2024-03-11 4837/week @ 2024-03-18 4872/week @ 2024-03-25 5846/week @ 2024-04-01

20,926 downloads per month
Used in 38 crates (17 directly)

MPL-2.0 license

252 lines


License Package Documentation

A fast, small, full-featured, async-aware oneshot channel.


  • Blazing fast! See Performance section below.
  • Tiny code, only one dependency and a lightning quick build.
  • Complete no_std support (with alloc for Arc).
  • Unique feature: sender may wait for a receiver to be waiting.


fn success_one_thread() {
    let (s,r) = oneshot::<bool>();
    assert_eq!((), s.send(true).unwrap());
    assert_eq!(Ok(true), future::block_on(r));


async-oneshot comes with a benchmark suite which you can run with cargo bench.

All benches are single-threaded and take double digit nanoseconds on my machine. async benches use futures_lite::future::block_on as an executor.

Numbers from my machine

Here are benchmark numbers from my primary machine, a Ryzen 9 3900X running alpine linux 3.12 that I attempted to peg at maximum cpu:

create_destroy          time:   [51.596 ns 51.710 ns 51.835 ns]
send/success            time:   [13.080 ns 13.237 ns 13.388 ns]
send/closed             time:   [25.304 ns 25.565 ns 25.839 ns]
try_recv/success        time:   [26.136 ns 26.246 ns 26.335 ns]
try_recv/empty          time:   [10.764 ns 11.161 ns 11.539 ns]
try_recv/closed         time:   [27.048 ns 27.159 ns 27.248 ns]
async.recv/success      time:   [30.532 ns 30.774 ns 31.011 ns]
async.recv/closed       time:   [28.112 ns 28.208 ns 28.287 ns]
async.wait/success      time:   [56.449 ns 56.603 ns 56.737 ns]
async.wait/closed       time:   [34.014 ns 34.154 ns 34.294 ns]

In short, we are very fast. Close to optimal, I think.

Compared to other libraries

The oneshot channel in futures isn't very fast by comparison.

Tokio put up an excellent fight and made us work hard to improve. In general I'd say we're slightly faster overall, but it's incredibly tight.

Note on safety

This crate uses UnsafeCell and manually synchronises with atomic bitwise ops for performance. We believe it is correct, but we would welcome more eyes on it.

See Also

Note on benchmarking

The benchmarks are synthetic and a bit of fun.



Breaking changes:

  • Make Sender.send() only take a mut ref instead of move.



  • Added some tests to cover repeated fix released in last version.
  • Inline more aggressively for some nice benchmark boosts.



  • Remove some overzealous debug_asserts that caused crashes in development in case of repeated waking. Thanks @nazar-pc!


  • Better benchmarks, based on criterion.


Breaking changes:

  • Sender.wait()'s function signature has changed to be a non-async fn returning an impl Future. This reduces binary size, runtime and possibly memory usage too. Thanks @zserik!


  • Race condition where the sender closes in a narrow window during receiver poll and doesn't wake the Receiver. Thanks @zserik!


  • Static assertions. Thanks @zserik!



  • Update futures-micro and improve the tests



  • Segfault when dropping receiver. Caused by a typo, d'oh! Thanks @boardwalk!



  • Remove redundant use of ManuallyDrop with UnsafeCell. Thanks @cynecx!



  • Rewrote, benchmarked and optimised.


  • First real release.

Copyright (c) 2020 James Laver, async-oneshot contributors.

This Source Code Form is subject to the terms of the Mozilla Public License, v. 2.0. If a copy of the MPL was not distributed with this file, You can obtain one at http://mozilla.org/MPL/2.0/.