runtime 0.3.0-alpha.1

Empowering everyone to build asynchronous software.

About

Runtime is what we imagine async APIs could look like if they were part of stdlib. We want async Rust to be an experience that mirrors the quality of the standard lib. We believe that in order for Rust to succeed it's not only important to make async Rust possible, it's crucial to make async Rust feel seamless.

And the embodiment of these values is Runtime: a library crafted to empower everyone to build asynchronous software.

  • runtime agnostic: Runtime comes with minimal OS bindings out of the box, but switching to a different runtime is a matter of changing a single line.
  • await anywhere: Runtime allows you to write async main functions, async tests, and async benchmarks. Experience what first-class async support in Rust feels like.
  • built for performance: Runtime is the thinnest layer possible on top of the backing implementations. All of the speed, none of the boilerplate.

Examples

UDP Echo Server

#![feature(async_await, await_macro, futures_api)]

use runtime::net::UdpSocket;

#[runtime::main]
async fn main() -> std::io::Result<()> {
    let mut socket = UdpSocket::bind("127.0.0.1:8080")?;
    let mut buf = vec![0u8; 1024];

    println!("Listening on {}", socket.local_addr()?);

    loop {
        let (recv, peer) = await!(socket.recv_from(&mut buf))?;
        let sent = await!(socket.send_to(&buf[..recv], &peer))?;
        println!("Sent {} out of {} bytes to {}", sent, recv, peer);
    }
}

To send messages do:

$ nc -u localhost 8080

More Examples

Attributes

Runtime introduces 3 attributes to enable the use of await anywhere, and swap between different runtimes. Each Runtime is bound locally to the initializing thread. This enables the testing of different runtimes during testing or benchmarking.

#[runtime::main]
async fn main() {}

#[runtime::test]
async fn my_test() {}

#[runtime::bench]
async fn my_bench() {}

Runtimes

Switching runtimes is a one-line change:

/// Use the default Native Runtime
#[runtime::main]
async fn main() {}

/// Use the Tokio Runtime
#[runtime::main(runtime_tokio::Tokio)]
async fn main() {}

The following backing runtimes are available:

  • Runtime Native (default) provides a thread pool, bindings to the OS, and a concurrent scheduler.
  • Runtime Tokio provides a thread pool, bindings to the OS, and a work-stealing scheduler.

Performance

Runtime provides performance that's competitive with most other systems languages, and great ergonomics to match.

Because we don't know what your workload is like, we can't predict which runtime will be able to maximize resource consumption for your use case.

But we can tell from our benchmarks that the difference between using Runtime and not using Runtime doesn't show up for IO-bound applications.

 name          baseline:: ns/iter  native:: ns/iter  diff ns/iter    diff %  speedup
 notify_self   1,350,882           1,237,416             -113,466    -8.40%   x 1.09
 poll_reactor  2,270,428           2,162,264             -108,164    -4.76%   x 1.05

Installation

With cargo-edit do:

$ cargo add runtime --allow-prerelease

FAQ

When is it useful to switch Runtimes?

What might be the best solution now, might not stay the best in the future. As Rust grows, so will the ecosystem. By making runtimes pluggable, your code can be forward compatible with any future changes. And as things evolve, you'll be able to test out the benefit new developments in the ecosystem have on your code by just changing a single line.

How is Runtime versioned?

We're currently in the 0.3-alpha range of releases, mirroring the Futures libraries. Once Futures hits 1.0, we'll follow suit and move over to semver proper.

This doesn't mean that Runtime won't release breaking changes. But if we do, we'll release a new major version, and provide instructions on how to upgrade. We view Runtime to be a foundational piece of technology, and that means we have to be serious about our stability guarantees.

Can I use Runtime in production?

Runtime is a thin layer that sits between your code and the backing runtimes. If you trust the backing runtime in production, then you can probably trust Runtime too.

Why is Runtime Native the default?

We believe Runtime Native provides a balanced implementation that works well for most scenarios. The codebase is small and comprehensive, and the algorithms simple yet performant.

Specific runtimes might introduce different trade-offs, and with Runtime you're able to compare, and pick the best fit for your requirements.

Can Runtime be used on embedded devices?

Runtime is designed to be compatible with micro processors, but not with micro controllers. Out of the box Runtime works on embedded devices such as Raspberry Pis, and with the appropriate backends it should also work on phones.

Micro controllers are very specific in what they provide, and while a Runtime-like library might be possible in the future, it's still early for the ecosystem and APIs would likely also need to be different. We don't know what the future holds, but for now we've chosen not to target micro controllers.

When will Timers and File System support land?

Timers are next up on the list of things we want to target, together with Unix Domain Sockets. Filesystem is a bit further behind because currently the implementations in the backing runtimes are changing, and we're not sure yet how to best abstract that.

Getting things right takes time. But if you'd like to move the state of async forward, we'd love for you to get involved!

Safety

This crate uses #![deny(unsafe_code)] to ensure everything is implemented in 100% Safe Rust.

Contributing

Want to join us? Check out our The "Contributing" section of the guide and take a look at some of these issues:

Conduct

The Runtime project adheres to the Contributor Covenant Code of Conduct. This describes the minimum behavior expected from all contributors.

License

Licensed under either of

at your option.

Contribution

Unless you explicitly state otherwise, any contribution intentionally submitted for inclusion in the work by you, as defined in the Apache-2.0 license, shall be dual licensed as above, without any additional terms or conditions.