Thruster

A fast and intuitive rust web framework

Don't have time to read the docs? Check out

Why you should use Thruster
Why you shouldn't use Thruster (definitely read this one)

✅ Runs in stable ✅ Runs fast ✅ Doesn't use unsafe

Features

built with async/await in mind
hyper compatible
ssl ready
testable
static file serving
socketio
gRPC, and more experimental non-tonic based gRPC
dependency injection

Motivation

Thruster is a web framework that aims for developers to be productive and consistent across projects and teams. Its goals are to be:

Performant
Simple
Intuitive

Thruster also

Does not use unsafe
Works in stable rust

Fast

Thruster can be run with different server backends and represents a nicely packaged layer over them. This means that it can keep up with the latest and greatest changes from the likes of Hyper, Actix, or even ThrusterServer, a home-grown http engine.

Intuitive

Based on frameworks like Koa, and Express, Thruster aims to be a pleasure to develop with.

Example

To run the example cargo run --example <example-name>. For example, cargo run --example hello_world and open http://localhost:4321/

Middleware Based

The core parts that make the new async await code work is designating middleware functions with the #[middleware_fn] attribute (which marks the middleware so that it's compatible with the stable futures version that Thruster is built on,) and then the async_middleware! macro in the actual routes.

A simple example for using async await is:

use std::boxed::Box;
use std::future::Future;
use std::pin::Pin;
use std::time::Instant;

use thruster::{App, BasicContext as Ctx, Request};
use thruster::{m, middleware_fn, MiddlewareNext, MiddlewareResult, Server, ThrusterServer};

#[middleware_fn]
async fn profile(context: Ctx, next: MiddlewareNext<Ctx>) -> MiddlewareResult<Ctx> {
    let start_time = Instant::now();

    context = next(context).await;

    let elapsed_time = start_time.elapsed();
    println!(
        "[{}μs] {} -- {}",
        elapsed_time.as_micros(),
        context.request.method(),
        context.request.path()
    );

    Ok(context)
}

#[middleware_fn]
async fn plaintext(mut context: Ctx, _next: MiddlewareNext<Ctx>) -> MiddlewareResult<Ctx> {
    let val = "Hello, World!";
    context.body(val);
    Ok(context)
}

#[middleware_fn]
async fn four_oh_four(mut context: Ctx, _next: MiddlewareNext<Ctx>) -> MiddlewareResult<Ctx> {
    context.status(404);
    context.body("Whoops! That route doesn't exist!");
    Ok(context)
}

#[tokio::main]
fn main() {
    println!("Starting server...");

    let mut app = App::<Request, Ctx, ()>::new_basic();

    app.get("/plaintext", m![profile, plaintext]);
    app.set404(m![four_oh_four]);

    let server = Server::new(app);
    server.build("0.0.0.0", 4321).await;
}

Error handling

It's recommended to use the map_try! macro from the main package. This has the same function as try!, but with the ability to properly map the error in a way that the compiler knows that execution ends (so there's no movement issues with context.)

This ends up looking like:

use thruster::errors::ThrusterError as Error;
use thruster::proc::{async_middleware, middleware_fn};
use thruster::{map_try, App, BasicContext as Ctx, Request};
use thruster::{MiddlewareNext, MiddlewareResult, MiddlewareReturnValue, Server, ThrusterServer};

#[middleware_fn]
async fn plaintext(mut context: Ctx, _next: MiddlewareNext<Ctx>) -> MiddlewareResult<Ctx> {
    let val = "Hello, World!";
    context.body(val);
    Ok(context)
}

#[middleware_fn]
async fn error(mut context: Ctx, _next: MiddlewareNext<Ctx>) -> MiddlewareResult<Ctx> {
    let res = "Hello, world".parse::<u32>();
    let non_existent_param = map_try!(res, Err(_) => {
        Error {
          context,
          message: "Parsing failure!".to_string(),
          status: 400
        }
      }
    );

    context.body(&format!("{}", non_existent_param));

    Ok(context)
}

#[middleware_fn]
async fn json_error_handler(context: Ctx, next: MiddlewareNext<Ctx>) -> MiddlewareResult<Ctx> {
    let res = next(context).await;

    let ctx = match res {
        Ok(val) => val,
        Err(e) => {
            let mut context = e.context;
            context.body(&format!(
                "{{\"message\": \"{}\",\"success\":false}}",
                e.message
            ));
            context.status(e.status);
            context
        }
    };

    Ok(ctx)
}

#[tokio::main]
fn main() {
    println!("Starting server...");

    let mut app = App::<Request, Ctx, ()>::new_basic();

    app.use_middleware("/", m![json_error_handler]);

    app.get("/plaintext", m![plaintext]);
    app.get("/error", m![error]);

    let server = Server::new(app);
    server.build("0.0.0.0", 4321).await;
}

Quick setup without a DB

The easiest way to get started is to just clone the starter kit

> git clone git@github.com:trezm/thruster-starter-kit.git
> cd thruster-starter-kit
> cargo run

The example provides a simple plaintext route, a route with JSON serialization, and the preferred way to organize sub routes using sub apps.

Quick setup with Postgres

The easiest way to get started with Postgres is to install thruster-cli,

> cargo install thruster-cli

And then to run

> thruster-cli init MyAwesomeProject
> thruster-cli component Users
> thruster-cli migrate

Which will generate everything you need to get started! Note that this requires a running Postgres connection and assumes the following connection string is valid:

postgres://postgres@localhost/<Your Project Name>

This is all configurable and none of it is hidden from the developer. It's like seeing the magic trick and learning how it's done! Check out the docs for thruster-cli here.

Testing

Thruster provides an easy test suite to test your endpoints, simply include the testing module as below:

let mut app = App::<Request, Ctx, ()>::new_basic();

...

app.get("/plaintext", m![plaintext]);

...

let result = testing::get(app, "/plaintext");

assert!(result.body == "Hello, World!");

Make your own middleware modules

Middleware is super easy to make! Simply create a function and export it at a module level. Below, you'll see a piece of middleware that allows profiling of requests:

#[middleware_fn]
async fn profiling<C: 'static + Context + Send>(
    mut context: C,
    next: MiddlewareNext<C>,
) -> MiddlewareResult<C> {
    let start_time = Instant::now();

    context = next(context).await?;

    let elapsed_time = start_time.elapsed();
    info!("[{}μs] {}", elapsed_time.as_micros(), context.route());

    Ok(context)
}

You might find that you want to allow for more specific data stored on the context, for example, perhaps you want to be able to hydrate query parameters into a hashmap for later use by other middlewares. In order to do this, you can create an additional trait for the context that middlewares downstream must adhere to. Check out the provided query_params middleware for an example.

Other, or Custom Backends

Thruster is capable of just providing the routing layer on top of a server of some sort, for example, in the Hyper snippet above. This can be applied broadly to any backend, as long as the server implements ThrusterServer.

use async_trait::async_trait;

#[async_trait]
pub trait ThrusterServer {
    type Context: Context + Send;
    type Response: Send;
    type Request: RequestWithParams + Send;

    fn new(App<Self::Request, Self::Context>) -> Self;
    async fn build(self, host: &str, port: u16);
}

There needs to be:

An easy way to create a server.
A function to build the server into a future that could be loaded into an async runtime.

Within the build function, the server implementation should:

Start up some sort of listener for connections
Call let matched = app.resolve_from_method_and_path(<some method>, <some path>); (This is providing the actual routing.)
Call app.resolve(<incoming request>, matched) (This runs the chained middleware.)

Using cargo generate

Note: This hasn't yet been updated for the latest version of Thruster

If you have cargo generate installed, you can simply run the cargo generator

cargo generate --git https://github.com/ami44/thruster-basic-template.git --name myproject

Why you should use Thruster

Change your backends at will. Out of the box, Thruster now can be used over: actix-web, hyper, or a custom backend
Thruster supports testing from the framework level
@trezm gets lonely when no one makes PRs or opens issues.
Thruster is more succinct for more middleware-centric concepts -- like a route guard. Take this example in actix to restrict IPs:

fn ip_guard(head: &RequestHead) -> bool {
    // Check for the cloudflare IP header
    let ip = if let Some(val) = head.headers().get(CF_IP_HEADER) {
        val.to_str().unwrap_or("").to_owned()
    } else if let Some(val) = head.peer_addr {
        val.to_string()
    } else {
        return false;
    };

    "1.2.3.4".contains(&ip)
}

#[actix_web::post("/ping")]
async fn ping() -> Result<HttpResponse, UserPersonalError> {
    Ok(HttpResponse::Ok().body("pong"))
}

...
        web::scope("/*")
            // This is confusing, but we catch all routes that _aren't_
            // ip guarded and return an error.
            .guard(guard::Not(ip_guard))
            .route("/*", web::to(HttpResponse::Forbidden)),
    )
    .service(ping);
...

Here is Thruster:

#[middleware_fn]
async fn ip_guard(mut context: Ctx, next: MiddlewareNext<Ctx>) -> MiddlewareResult<Ctx> {
    if "1.2.3.4".contains(&context.headers().get("Auth-Token").unwrap_or("")) {
        context = next(context).await?;

        Ok(context)
    } else {
        Err(Error::unauthorized_error(context))
    }

}

#[middleware_fn]
async fn ping(mut context: Ctx, _next: MiddlewareNext<Ctx>) -> MiddlewareResult<Ctx> {
    context.body("pong");
    Ok(context)
}

...
    app.get("/ping", m![ip_guard, plaintext]);
...

A bit more direct is nice!

Why you shouldn't use Thruster

It's got few maintainers (pretty much just one.)
There are other projects that have been far more battle tested. Thruster is in use in production, but nowhere that you'd know or that matters.
It hasn't been optimized by wicked smarties. @trezm tries his best, but keeps getting distracted by his dog.
Serously, this framework could be great, but it definitely hasn't been poked and proded like others. Your help could go a long way to making it more secure and robust, but we might now be there just yet.

If you got this far, thanks for reading! Always feel free to reach out.

ynuwenhof/Thruster