genawaiter VS tokio-tungstenite

Yes, Letlang is translated to Rust and the runtime is implemented in Rust, using tokio and genawaiter. The compiler itself is also built in Rust.

(note that genawaiter itself doesn't support no_std environments, but there's a merge request for that.)

> `async` to make fake generators.

Genawaiter[0] is one of them.

[0]: https://github.com/whatisaphone/genawaiter

Are you aware of the genawaiter crate?

A warning though: genawaiter doesn't seem to be maintained. The last commit is 2 years old, and issues are not active (I opened one that I find somewhat critical: https://github.com/whatisaphone/genawaiter/issues/35).

There's e.g. https://github.com/whatisaphone/genawaiter, but you can also use yield directly (https://doc.rust-lang.org/beta/unstable-book/language-features/generators.html).

If you're interested in something that can be used right now, on stable, you should take a look at the genawaiter crate, it reimplements generators with async.

I can't help you debug your code if you do not provide it. Have you looked at the example client?

It passed all test of the autobahn testsuite And web-socket-benchmark show about 3x faster then tokio-tungstenite

There are example files in the tokio-tungstenite crate called `autobahn-client.rs` and `autobahn-server.rs`. Why are they called autobahn? I googled and can't understand what autobahn is all about. Is it a websocket pattern? Or some protocol?

I'm using another crate that requires tls, specifically tokio-tungstenite, I'll try your suggestions later today once I get home

Tokio-tungstenite - It looks like in this example, it's spamming the task thread with wakeup calls from all of the active connections. This design choice makes me doubt that this was well written in general.

tokio-tungstenite is the more popular library. In frameworks, offhand Axum supports websockets (docs, example)

If you only have one task sending data to the sink, you probably don't need forward, as you can just write to the sink directly (you might not even need to split the stream in the first place, but i'm not sure about that). But often you want to write to the sink from different tasks (e.g. this example takes messages sent from one websocket connection, and broadcasts it onto every other connected websocket, so the sink for each websocket needs to be accessed by every other websocket handler task), and you can't do that with only the sink as you can't clone it. Either need to wrap it into a Mutex and clone that around the different tasks (and lock it every time you need to write to it, like OP did originally) or you can use forward to map the rx (receiver) of a channel to the sink, and clone the tx (sender) part of the channel for each task that wants to write to the sink. That way, you only have one task that is accessing the sink directly, so no issues with synchronization.

Has anyone worked with websockets before? Particularly with the tokio-tungstenite crate? I'm getting a Protocol(ResetWithoutClosingHandshake) error in my request. I send in some text, and i'm supposed to receive an audio file back.

tokio-tungstenite

i based my code mostly on the client example from the tokio-tungstenite project: https://github.com/snapview/tokio-tungstenite/blob/master/examples/client.rs