tower VS bitvec

This question is partially inspired by this PR which is kinda trying to do the same thing.

dd-trace-layer is a middleware for sending Datadog's trace. It's based on Tower and OpenTelemetry Rust.

I'm pretty sure that GCP's APIs (unlike AWS, which uses Smithy for very genuinely, very good reason) are defined using Protobuf and can be communicated with over gRPC, which means that you don't need to bind via cxx to GCP's C++ APIs. Take a look at this example using Tonic. If you're to use Tonic, you'll also be able to use Tower's middleware (main crate, http-specific) to implement retries, timeouts, tracing, and all the other things you need to be production-ready.

#[derive(Clone)] struct MyMiddleware { inner: S, } impl Service> for MyMiddleware where S: Service, Response = Response> + Clone + Send + 'static, S::Future: Send + 'static, { type Response = S::Response; type Error = S::Error; type Future = BoxFuture<'static, Result>; fn poll_ready(&mut self, cx: &mut Context<'_>) -> Poll> { self.inner.poll_ready(cx) } fn call(&mut self, mut req: Request) -> Self::Future { println!("before"); // best practice is to clone the inner service like this // see https://github.com/tower-rs/tower/issues/547 for details let clone = self.inner.clone(); let mut inner = std::mem::replace(&mut self.inner, clone); Box::pin(async move { let res: Response = inner.call(req).await?; println!("after"); Ok(res) }) } } fn main() { let app = Router::new() .route("/", get(|| async { /* ... */ })) .layer(layer_fn(|inner| MyMiddleware { inner })); }

For this tutorial, we're going to use apalis to run cron jobs in an async context. We will also look at how to decorate our jobs with tower middleware allowing us to unlock features like retries, prometheus, sentry etc

So I have now had a look at Axum and think I will give it a try. In the readme in the repository it says something about tower or tower::Service and tonic, what exactly is that? I do not understand that yet.

Better compatibility with tower ecosystem.

ratpack is idealized in the simplicity of the sinatra (ruby) framework in its goal, and attempts to be an alternative to other async HTTP frameworks such as tower, warp, axum, and tide.

i would browse the standard library, tower, nom, or my own bitvec to see layout and trait/record separation. in particular, std::io and std::net may be of use: io::Read and io::Write are pervasive examples of implementing unixy file-descriptor-like behavior in the type system

We're not there yet. I keep an eye on Tower which looks promising to build on top of. And I keep an eye on MoonZoon (full stack framework, unashamedly opinionated!).

I was also under the false impression that bitwise encoding was slow. When I first implemented bitcode with bitvec I got performance 20x worse than bincode. After writing my own implementation I was able to get much better performance.

interesting; i'll have to compare this to my rust counterpart. your numbers indicate some clever implementations i'd love to read

Not to mention having a lackluster std which causes you to use nonstardard not so well documented crates and a 40K LoC library to do "bit-twiddling" (the lib, https://github.com/bitvecto-rs/bitvec the blog that says "twiddle bits" https://blog.adamchalmers.com/making-a-dns-client/ and for crying out loud the blogger also used the language the author mentioned and I quote "ergonomics AND speed AND correctness")

https://github.com/bitvecto-rs/bitvec/issues/135 is a very funny read about how to perform inttoptr with provenance retention

Technically #135 gives me license to yank affected crates, but since the only exploit is "Miri crashes exactly one test out of the suite" it's not really worth it to be a stickler. Call it a truce

My friend Nika wrote a macro that packs a sequence of 1, 0, … tokens into a correctly structured bit-buffer, adaptable over any register type or bit-ordering, at compile time. It's now basically this whole file

if you are interested in a user manual, you can use mdbook as well. for an example, my bitvec project uses mdbook (book.toml) and a github action (.github/workflows/gh-pages.yml) to compile the guide and host it as a github pages website. it's slightly more complicated, and i'd like docs.rs to follow hexdoc.pm's example of hosting both api docs and prose, but until then this is a pretty reasonable solution.

the first one

i would browse the standard library, tower, nom, or my own bitvec to see layout and trait/record separation. in particular, std::io and std::net may be of use: io::Read and io::Write are pervasive examples of implementing unixy file-descriptor-like behavior in the type system