Struct RouteBuilder 

pub struct RouteBuilder<S> { /* private fields */ }

Builder for configuring routes and settings before creating an App.

RouteBuilder uses a fluent API to register routes, configure CORS, and adjust server settings. Call .seal() to produce the final App<S>.

Example

use mini_serve::{RouteBuilder, handler, body};
use hyper::Response;
use hyper::body::Bytes;

#[tokio::main]
async fn main() -> Result<(), Box<dyn std::error::Error>> {
    let app = RouteBuilder::stateless()
        .get("/health", handler(|_, _| async {
            Ok(Response::new(body(Bytes::from("OK"))))
        }))
        .seal();

    app.bind("127.0.0.1:8080".parse()?).await?;
    Ok(())
}

Implementations

impl<S: Send + Sync + 'static> RouteBuilder<S>

pub fn new(state: S) -> Self

Create a new builder with shared state.

pub fn wrap(self, middleware: Middleware<S>) -> Self

Register a middleware. Applies to every route registered after this call — middlewares wrap in registration order, so the first .wrap() call becomes the outermost layer and runs first.

pub fn with_max_body_size(self, max: usize) -> Self

Set the maximum request body size in bytes (default: 2 MiB).

pub fn with_header_read_timeout(self, d: Duration) -> Self

Set the header read timeout (default: 30 seconds).

pub fn with_response_header( self, name: &str, value: &str, ) -> Result<Self, ServeError>

Log one line per request to stderr, plus handler panics and the internal detail behind 5xx responses.

Off by default: a library writing to its host’s stderr uninvited is a surprise. See RouteBuilder::with_request_logging_to to choose the destination. Send name: value on every response that does not already carry that header.

For the policy headers an API wants applied uniformly — Strict-Transport-Security, Content-Security-Policy, Referrer-Policy.

Apply-if-absent, deliberately: a handler that sets the header itself wins. This is the opposite of mini-static’s rule, and for the opposite reason — that crate computes every header itself, so a fixed value fighting a computed one is a bug, whereas handlers here are arbitrary user code that may legitimately vary a policy per route.

Errors

ServeError if name or value is not a valid HTTP header, or if name is one the connection layer owns (Content-Length, Connection, Transfer-Encoding) — those describe framing this crate does not choose, so a fixed value could only contradict it.

pub fn with_request_logging(self) -> Self

pub fn with_request_logging_to(self, writer: Box<dyn Write + Send>) -> Self

Log one line per request to writer, plus handler panics and 5xx internals.

Each served request writes GET /path 200 0.421ms — method, path exactly as received, status, and handling time. Two failures that are otherwise invisible also come here:

• Handler panics. The task’s result was previously discarded, so a panicking handler dropped the client’s connection and left no trace anywhere.
• 5xx internal detail. The client body is sanitized to {"message":"internal server error"} deliberately; without a sink the real message was discarded with it, leaving an operator nothing to debug from.

The path is logged exactly as received, not decoded: it is attacker-controlled input, and whoever reads the log deserves the bytes that actually arrived. Writes are serialized across connections; write failures are ignored rather than allowed to fail a request.

pub fn with_connect_timeout(self, d: Duration) -> Self

How long a transport has to turn an accepted connection into a usable one (default: 10 seconds).

Bounds the connect step of App::run_with_transport, so a transport that negotiates forever cannot hold a connection slot forever. It has no effect on the identity transport App::run uses, which cannot block — the guarantee exists for transports that do work, which is the only kind worth plugging in.

Replaces the former with_tls_handshake_timeout: the bound belongs to the server’s connection lifecycle rather than to any one transport.

pub fn with_upgrades(self) -> Self

Set the maximum concurrent connections (default: 1024). Allow handlers to take over a connection with OnUpgrade.

Off by default: an application with no upgrade route should not pay for the capability, and enabling it changes how every connection on this server is served. Without it, a 101 response is written and the callback never runs — which is reported to the log sink if one is configured, since the client would otherwise see a dead connection and the author would see nothing.

The server takes the request’s upgrade future before routing, so a handler that calls hyper::upgrade::on itself receives nothing. Use OnUpgrade instead; it exists so the upgraded stream is serviced inside the connection’s own task, keeping it inside the connection ceiling and the shutdown drain.

pub fn with_max_connections(self, max: usize) -> Self

pub fn with_error_handler( self, f: impl Fn(StatusCode, &str) -> Response<ResponseBody> + Send + Sync + 'static, ) -> Self

pub fn with_cors(self, config: CorsConfig) -> Self

pub fn get(self, path: &str, handler: Handler<S>) -> Self

pub fn post(self, path: &str, handler: Handler<S>) -> Self

pub fn put(self, path: &str, handler: Handler<S>) -> Self

pub fn delete(self, path: &str, handler: Handler<S>) -> Self

pub fn patch(self, path: &str, handler: Handler<S>) -> Self

Register a PATCH handler.

Dispatch and the Allow header are generic over Method, so this needs nothing from the router that its siblings do not — it was simply missing.

pub fn seal(self) -> App<S>

impl RouteBuilder<()>

pub fn stateless() -> Self