tower-lsp-server 0.23.0

//! Show how to server an language server over TCP.

use tokio::net::{TcpListener, TcpStream};
use tower_lsp_server::{LspService, Server};

// You can check the language server implementation in `examples/common/lsp.rs`.
// The dummy language server provided is common to all transports examples.
#[path = "common/lsp.rs"]
mod lsp;

#[tokio::main]
async fn main() -> Result<(), Box<dyn std::error::Error>> {
    #[cfg(feature = "runtime-agnostic")]
    use tokio_util::compat::{TokioAsyncReadCompatExt, TokioAsyncWriteCompatExt};

    tracing_subscriber::fmt().init();

    let mut args = std::env::args();
    let stream = match args.nth(1).as_deref() {
        None => {
            // If no argument is supplied (args is just the program name), then
            // we presume that the client has opened the TCP port and is waiting
            // for us to connect. This is the connection pattern used by clients
            // built with vscode-langaugeclient.
            TcpStream::connect("127.0.0.1:9257").await?
        }
        Some("--listen") => {
            // If the `--listen` argument is supplied, then the roles are
            // reversed: we need to start a server and wait for the client to
            // connect.
            let listener = TcpListener::bind("127.0.0.1:9257").await?;
            let (stream, _) = listener.accept().await?;
            stream
        }
        Some(arg) => {
            panic!("Unrecognized argument: {arg}. Use --listen to listen for connections.")
        }
    };

    let (read, write) = tokio::io::split(stream);
    #[cfg(feature = "runtime-agnostic")]
    let (read, write) = (read.compat(), write.compat_write());

    let (service, socket) = LspService::new(lsp::Backend::new);
    Server::new(read, write, socket).serve(service).await;

    Ok(())
}