Lifecycle Hooks

Enable the hook feature to intercept request lifecycle events for observability, tracing, logging, or custom middleware.

Example

#![allow(unused)]
fn main() {
use afast::hook::{Hook, RequestContext, RequestGuard};

struct TimingHook;

impl Hook for TimingHook {
    fn before_request(&self, ctx: &RequestContext) -> Option<Box<dyn RequestGuard>> {
        println!("→ {} ({})", ctx.handler_name, ctx.transport);
        Some(Box::new(std::time::Instant::now()))
    }
}

impl RequestGuard for std::time::Instant {
    fn on_response(&mut self, ctx: &RequestContext, _resp: &[u8]) {
        println!("← {} OK ({:?})", ctx.handler_name, self.elapsed());
    }
    fn on_error(&mut self, ctx: &RequestContext, err: &afast::Error) {
        println!("✗ {} error: {}", ctx.handler_name, err);
    }
}
}

Global and Service Hooks

#![allow(unused)]
fn main() {
let app = AFast::new()
    .hook(TimingHook)                    // Global: all handlers
    .service(
        service!("api" => { h(handler) })
            .hook(ApiSpecificHook)       // Service: only this service's handlers
    );
}

• Global hooks run for every handler.
• Service hooks run only for handlers in that service.
• Both always execute — they never replace each other.
• Execution order: global first, then service (onion model 🧅).
  • before_request: Hook1 → Hook2 → handler
  • on_response/on_error: Guard2 → Guard1 (reversed)

Long Connection Hooks

For WebSocket/TCP long-connection handlers, the full lifecycle is:

on_connect → before_request → handler → on_response/on_error → on_disconnect

RequestContext Fields