axum-idioms

Axum Idioms and Patterns

Core Philosophy

Axum (0.8+) rewards composability via Tower, type-safe extractors, and zero-cost abstractions. Idiomatic Axum = thin handlers, tower middleware, typed errors.

Version note: This skill targets Axum 0.8+ (released 2025). Key changes from 0.7: path parameter syntax changed from :name to {name}, State extractor is now in axum::extract, and axum::serve replaces axum::Server. If you encounter an existing codebase on 0.7, check the Axum 0.8 changelog before applying these patterns.

Scope: Axum-specific patterns. For Rust fundamentals: @.agents/skills/rust-idioms/SKILL.md. For project structure: @.agents/skills/rust-idioms/references/project-structure.md.

Router and Route Organization

1. Build routers with Router::new() and method routing:

   // ✅ Group by resource, nest for versioning
   fn task_routes() -> Router<AppState> {
       Router::new()
           .route("/tasks", get(list_tasks).post(create_task))
           .route("/tasks/{id}", get(get_task).put(update_task).delete(delete_task))
   }
   
   fn app(state: Arc<AppState>) -> Router {
       let api = Router::new()
           .merge(task_routes())   // .merge() combines peer routers
           .merge(user_routes());
   
       Router::new()
           .nest("/api/v1", api)   // .nest() adds prefix to sub-router
           .fallback(handle_404)   // typed error, not Axum's default plain-text 404
           .with_state(state)
   }
   

2. Path parameters use {name} syntax (not :name).

Extractors

1. Built-in extractors — Path, Query, Json, State, HeaderMap:

   // ✅ Typed and validated at compile time
   async fn get_task(
       State(state): State<Arc<AppState>>,
       Path(id): Path<Uuid>,
   ) -> Result<Json<TaskResponse>, AppError> {
       let task = state.task_service.find(id).await?;
       Ok(Json(task.into()))
   }
   

2. Extractor ordering — body-consuming extractors MUST be last:

   // ✅ Path before Json (Json consumes the body)
   async fn update_task(
       State(state): State<Arc<AppState>>,
       Path(id): Path<Uuid>,
       Json(body): Json<UpdateTaskRequest>,
   ) -> Result<Json<TaskResponse>, AppError> { ... }
   
   // ❌ Json before Path — won't compile or will fail at runtime
   async fn update_task(Json(body): Json<UpdateTaskRequest>, Path(id): Path<Uuid>) { ... }
   

3. Custom extractors via FromRequestParts (non-body) or FromRequest (body):

   // ✅ Custom extractor for authenticated user
   impl<S: Send + Sync> FromRequestParts<S> for AuthUser {
       type Rejection = AppError;
       async fn from_request_parts(parts: &mut Parts, _state: &S) -> Result<Self, Self::Rejection> {
           let token = parts.headers.get(AUTHORIZATION)
               .and_then(|v| v.to_str().ok())
               .and_then(|v| v.strip_prefix("Bearer "))
               .ok_or(AppError::Unauthorized)?;
           decode_jwt(token).map_err(|_| AppError::Unauthorized)
       }
   }
   

4. Handle rejections with custom types — never let Axum's default messages leak to clients.

Application State

1. Wrap in Arc, pass via State:

   pub struct AppState {
       pub db: sqlx::PgPool,
       pub task_service: TaskService,
       pub config: AppConfig,
   }
   
   let state = Arc::new(AppState { db: pool, task_service, config });
   let app = Router::new().route("/tasks", get(list_tasks)).with_state(state);
   

2. Compose state for feature isolation — each feature defines its own state struct, combined at app level.

3. Never clone AppState directly — wrap in Arc, clone the Arc.

Middleware (Tower)

1. ServiceBuilder for layer composition:

   // ✅ Full middleware stack — order matters (outermost runs first)
   let app = Router::new()
       .nest("/api/v1", api_routes())
       .layer(
           ServiceBuilder::new()
               .layer(TraceLayer::new_for_http())
               .layer(CompressionLayer::new())
               .layer(CorsLayer::permissive()) // development ONLY — see below
               .layer(TimeoutLayer::new(Duration::from_secs(30)))
       )
       .with_state(state);
   

   Never ship CorsLayer::permissive() to production. It allows any origin, any method, any header, and sends no Access-Control-Allow-Credentials. Use an explicit, allow-listed layer instead. Load allowed origins from config, not literals:

    use tower_http::cors::CorsLayer;
   use http::HeaderValue;
   
   // ✅ Production-safe — explicit allow-list
   fn cors_layer(allowed_origins: &[String]) -> CorsLayer {
       let origins: Vec<HeaderValue> = allowed_origins
           .iter()
           .filter_map(|o| HeaderValue::try_from(o).ok())
           .collect();
       CorsLayer::new()
           .allow_origin(origins)                                      // explicit list, NEVER Any in prod
           .allow_methods([                                            // methods you actually serve
               axum::http::Method::GET,
               axum::http::Method::POST,
               axum::http::Method::PUT,
               axum::http::Method::DELETE,
           ])
           .allow_headers([
               axum::http::header::AUTHORIZATION,
               axum::http::header::CONTENT_TYPE,
               axum::http::header::ACCEPT,
           ])
           .allow_credentials(true)                                    // required for cookies / auth
           .max_age(Duration::from_secs(3600))                         // cache preflight 1h
   }
   // .layer(cors_layer(&config.cors.allowed_origins))
   

   Rule: allow_credentials(true) is incompatible with allow_origin(Any) — browsers reject it. If you need credentials, you MUST enumerate origins. See security-principles.md §CORS.

2. Custom middleware with from_fn:

   async fn auth_middleware(
       State(state): State<Arc<AppState>>,
       request: Request,
       next: Next,
   ) -> Result<Response, AppError> {
       let token = request.headers().get(AUTHORIZATION)
           .and_then(|v| v.to_str().ok())
           .and_then(|v| v.strip_prefix("Bearer "))
           .ok_or(AppError::Unauthorized)?;
       state.auth_service.validate(token).await?;
       Ok(next.run(request).await)
   }
   
   // Apply to specific routes with route_layer
   let protected = Router::new()
       .route("/tasks", get(list_tasks))
       .route_layer(middleware::from_fn_with_state(state.clone(), auth_middleware));
   

3. Layer ordering: ServiceBuilder applies bottom-to-top — last .layer() wraps closest to handler.

4. Propagate correlation IDs with TraceLayer:

   use tower_http::trace::TraceLayer;
   use tracing::Span;
   
   // ✅ Inject x-request-id into every span for log correlation
   let trace_layer = TraceLayer::new_for_http()
       .make_span_with(|request: &Request<_>| {
            let request_id = request
                .headers()
                .get("x-request-id")
                .and_then(|v| v.to_str().ok())
                .map(String::from)
                .unwrap_or_else(|| uuid::Uuid::new_v4().to_string());
           tracing::info_span!(
               "request",
               method = %request.method(),
               uri = %request.uri(),
               request_id = %request_id,
           )
       });
   
   let app = Router::new()
       .nest("/api/v1", api_routes())
       .layer(trace_layer)
       .with_state(state);
   

   All tracing::info!, warn!, error! calls inside a handler will automatically inherit the span fields above (method, uri, request_id). This satisfies the correlationId requirement from the Logging Mandate. See logging-implementation/SKILL.md §Rust for the full init_tracing() setup.

5. Request body size limits — prevent denial-of-service via large payloads:

   use axum::extract::DefaultBodyLimit;
   
   let app = Router::new()
       .nest("/api/v1", api_routes())
       .layer(DefaultBodyLimit::max(1024 * 1024))  // 1 MB global limit
       .with_state(state);
   
   // Per-route override for file uploads:
   let upload_routes = Router::new()
       .route("/upload", post(upload_file))
       .layer(DefaultBodyLimit::max(50 * 1024 * 1024));  // 50 MB for uploads
   

Error Handling

1. Unified AppError enum with IntoResponse:

   #[derive(Debug, thiserror::Error)]
   pub enum AppError {
       #[error("not found: {0}")]       NotFound(String),
       #[error("validation failed: {0}")] Validation(String),
       #[error("unauthorized")]          Unauthorized,
       #[error("forbidden")]             Forbidden,
       #[error(transparent)]             Internal(#[from] anyhow::Error),
   }
   
   impl IntoResponse for AppError {
       fn into_response(self) -> Response {
           let (status, msg) = match &self {
               Self::NotFound(m) => (StatusCode::NOT_FOUND, m.clone()),
               Self::Validation(m) => (StatusCode::UNPROCESSABLE_ENTITY, m.clone()),
               Self::Unauthorized => (StatusCode::UNAUTHORIZED, "unauthorized".into()),
               Self::Forbidden => (StatusCode::FORBIDDEN, "forbidden".into()),
               Self::Internal(e) => {
                   tracing::error!(error = %e, "internal server error");
                   (StatusCode::INTERNAL_SERVER_ERROR, "internal server error".into())
               }
           };
           (status, Json(serde_json::json!({ "error": msg }))).into_response()
       }
   }
   

2. Handler return type: always Result<impl IntoResponse, AppError>.

3. Convert rejections into AppError for consistent JSON error shape.

Response Types

1. Json<T> for standard responses, (StatusCode, Json<T>) tuple for non-200:

   async fn list_tasks(...) -> Result<Json<Vec<TaskResponse>>, AppError> { ... }
   async fn create_task(...) -> Result<(StatusCode, Json<TaskResponse>), AppError> {
       Ok((StatusCode::CREATED, Json(task.into())))
   }
   

2. Response builder for headers, streaming, or non-JSON (CSV, files).

Validation

1. validator crate with #[derive(Validate)]:

   #[derive(Debug, Deserialize, Validate)]
   pub struct CreateTaskRequest {
       #[validate(length(min = 1, max = 255))]
       pub title: String,
       #[validate(range(min = 1, max = 5))]
       pub priority: u8,
   }
   

2. Custom ValidatedJson<T> extractor — implement FromRequest<S> that deserializes via Json<T> then calls value.validate(), converting failures to AppError::Validation:

   use axum::extract::{FromRequest, Request};
   use axum::Json;
   use serde::de::DeserializeOwned;
   use validator::Validate;
   
   pub struct ValidatedJson<T>(pub T);
   
   impl<S, T> FromRequest<S> for ValidatedJson<T>
   where
       S: Send + Sync,
       T: DeserializeOwned + Validate,
   {
       type Rejection = AppError;
   
       async fn from_request(req: Request, state: &S) -> Result<Self, Self::Rejection> {
           let Json(value) = Json::<T>::from_request(req, state)
                .await
                .map_err(|e| AppError::Validation(e.to_string()))?;
           value.validate().map_err(|e| AppError::Validation(e.to_string()))?;
           Ok(ValidatedJson(value))
       }
   }
   
   // Usage — replaces bare Json<T> in handler signatures
   async fn create_task(
       State(state): State<Arc<AppState>>,
       ValidatedJson(body): ValidatedJson<CreateTaskRequest>,
   ) -> Result<(StatusCode, Json<TaskResponse>), AppError> {
       let task = state.task_service.create(body).await?;
       Ok((StatusCode::CREATED, Json(task.into())))
   }
   

Response Types and Domain Conversion

1. Separate request and response types — never expose domain models directly to the API:

   // --- Request type (deserialize + validate) ---
   #[derive(Debug, Deserialize, Validate)]
   #[serde(rename_all = "camelCase")]
   pub struct CreateTaskRequest {
       #[validate(length(min = 1, max = 255))]
       pub title: String,
       #[serde(default)]
       pub description: Option<String>,
       #[validate(range(min = 1, max = 5))]
       #[serde(default = "default_priority")]
       pub priority: u8,
   }
   
   fn default_priority() -> u8 { 3 }
   
   // --- Response type (serialize) ---
   #[derive(Debug, Serialize)]
   #[serde(rename_all = "camelCase")]
   pub struct TaskResponse {
       pub id: Uuid,
       pub title: String,
       #[serde(skip_serializing_if = "Option::is_none")]
       pub description: Option<String>,
       pub priority: u8,
       pub created_at: DateTime<Utc>,
   }
   
   // --- Domain → Response conversion ---
   impl From<Task> for TaskResponse {
       fn from(task: Task) -> Self {
           Self {
               id: task.id,
               title: task.title,
               description: task.description,
               priority: task.priority,
               created_at: task.created_at,
           }
       }
   }
   

2. Use .into() in handlers for clean conversion:

   async fn get_task(
       State(state): State<Arc<AppState>>,
       Path(id): Path<Uuid>,
   ) -> Result<Json<TaskResponse>, AppError> {
       let task = state.task_service.find(id).await?;
       Ok(Json(task.into()))  // From<Task> for TaskResponse
   }
   

3. For serde attribute patterns (rename_all, deny_unknown_fields, skip_serializing_if), see rust-idioms/references/serde-patterns.md.

Testing

For universal testing principles, see .agents/rules/testing-strategy.md. Below: Axum-specific patterns only.

1. tower::ServiceExt::oneshot — test handlers without spawning a server:

   #[tokio::test]
   async fn test_create_task_returns_201() {
       let app = app(Arc::new(test_app_state().await));
       let response = app.oneshot(
           Request::builder().method("POST").uri("/api/v1/tasks")
               .header("content-type", "application/json")
               .body(Body::from(r#"{"title":"Test","priority":3}"#)).unwrap(),
       ).await.unwrap();
       assert_eq!(response.status(), StatusCode::CREATED);
   }
   

2. Mock state helpers — inject trait-based test doubles for isolation:

   /// Build test AppState with mock services (no real DB needed)
   fn test_app_state() -> Arc<AppState> {
       let mock_task_service = MockTaskService::new();  // implements TaskService trait
       Arc::new(AppState {
           task_service: Box::new(mock_task_service),
           config: test_config(),
       })
   }
   

   This follows the trait-based DI pattern from @.agents/rules/architectural-pattern.md — swap real I/O implementations for test doubles at the AppState level.

3. tower dev-dependency required — add tower = { version = "0.5", features = ["util"] } to [dev-dependencies] to use ServiceExt::oneshot.

4. Integration tests use a real database via sqlx::test or Testcontainers.

Graceful Shutdown

Container orchestrators (Kubernetes, Docker stop) send SIGTERM, not SIGINT — handle BOTH. Use tokio_util::sync::CancellationToken (the recommended cancellation primitive per @.agents/skills/rust-idioms/SKILL.md §Async and Concurrency) so in-flight handlers and background tasks can observe the shutdown and unwind cooperatively. Axum's with_graceful_shutdown then drains active connections before exiting.

use tokio::signal;
use tokio_util::sync::CancellationToken;

// Wire the token into AppState so background tokio tasks can observe shutdown.
let shutdown = CancellationToken::new();
let state = Arc::new(AppState { /* ... */ shutdown: shutdown.clone() });

let listener = tokio::net::TcpListener::bind("0.0.0.0:3000").await?;
axum::serve(listener, app)
    .with_graceful_shutdown(shutdown_signal(shutdown.clone()))
    .await?;

/// ✅ Unified signal handler — fires on SIGINT (ctrl-c) OR SIGTERM (container stop).
async fn shutdown_signal(token: CancellationToken) {
    let ctrl_c = async { signal::ctrl_c().await.expect("install ctrl-c handler") };

    #[cfg(unix)]
    let terminate = async {
        signal::unix::signal(signal::unix::SignalKind::terminate())
            .expect("install SIGTERM handler")
            .recv().await;
    };
    #[cfg(not(unix))]
    let terminate = std::future::pending::<()>();

    tokio::select! {
        _ = ctrl_c    => tracing::info!("SIGINT received, shutting down"),
        _ = terminate => tracing::info!("SIGTERM received, shutting down"),
    }
    token.cancel(); // notify background tasks + long-lived handlers to unwind
}

Why: Axum stops accepting new connections and drains in-flight requests before exiting. The shared CancellationToken lets your background workers (queue consumers, scheduled jobs, long-poll handlers) observe the same shutdown and exit cooperatively instead of being killed mid-write. Add tokio-util (with the rt feature) to your dependencies. See rust-idioms §Async and Concurrency for the cancellation-safety policy.

Anti-Patterns

• ❌ Business logic in handlers — extract to a service/logic layer; handlers only parse, delegate, respond
• ❌ Extension instead of State — Extension is untyped and pre-0.6; use State<T> always
• ❌ Cloning entire state — wrap in Arc, clone the Arc
• ❌ Blocking in async handlers — use tokio::task::spawn_blocking for CPU-bound or blocking I/O
• ❌ Wrong extractor ordering — body-consuming extractors (Json, Form) must be the last parameter
• ❌ Returning string errors — use typed AppError with IntoResponse for consistent error shape
• ❌ Leaking internal error details — log with tracing::error!, return generic message to client

Formatting and Static Analysis

Same tooling as Rust. See @.agents/skills/rust-idioms/SKILL.md#clippy-and-formatting.

Related

• Code Idioms and Conventions @.agents/rules/code-idioms-and-conventions.md
• Rust Idioms @.agents/skills/rust-idioms/SKILL.md
• API Design Principles @.agents/rules/api-design-principles.md
• Security Principles @.agents/rules/security-principles.md
• Error Handling Principles @.agents/rules/error-handling-principles.md
• Architectural Patterns @.agents/rules/architectural-pattern.md
• Testing Strategy @.agents/rules/testing-strategy.md
• Logging and Observability Mandate @.agents/rules/logging-and-observability-mandate.md
• Logging Implementation @.agents/skills/logging-implementation/SKILL.md
• Serde Patterns @.agents/skills/rust-idioms/references/serde-patterns.md
• SQLx Patterns @.agents/skills/rust-idioms/references/sqlx-patterns.md