select! & Concurrency

Q: select! vs join!?

select! first completion; join! waits for all. Different use cases.

Q: Loop pattern?

loop { select! { ... } } is standard for Tokio services until shutdown branch breaks.

Q: select with streams?

some_stream.next() => branch supported in tokio::select!.

Q: pin required?

Some branch futures need pin in select - follow macro expansion errors with pin!.

Q: Axum shutdown?

with_graceful_shutdown uses signal future - related but at server level.

Q: testing select?

tokio::time::pause + advance for deterministic timer branches.

Q: priority events?

Put high-priority branch first with biased - document starvation risk.

Q: select across crates?

Use tokio::select! for Tokio futures; futures::select for non-Tokio contexts.

tokio::select! multiplexes multiple async branches, completing when the first one finishes. It is the core pattern for event loops, shutdown handling, and racing I/O with timers.

Recipe

Quick-reference recipe card - copy-paste ready.

use tokio::sync::mpsc;
use tokio::time::{sleep, Duration};
 
#[tokio::main]
async fn main() {
    let (tx, mut rx) = mpsc::channel(4);
 
    tokio::select! {
        Some(msg) = rx.recv() => println!("{msg}"),
        _ = sleep(Duration::from_millis(100)) => println!("timeout"),
        else => println!("channel closed"),
    }
 
    let _ = tx.send("hi").await;
}

When to reach for this: Server main loops, combining channel messages with shutdown signals, and timeout races.

Working Example

use tokio::sync::mpsc;
use tokio::time::{interval, Duration};
 
async fn worker(mut rx: mpsc::Receiver<u32>, mut shutdown: mpsc::Receiver<()>) {
    let mut ticker = interval(Duration::from_millis(200));
 
    loop {
        tokio::select! {
            _ = shutdown.recv() => {
                println!("shutdown");
                break;
            }
            Some(job) = rx.recv() => {
                println!("job {job}");
            }
            _ = ticker.tick() => {
                println!("heartbeat");
            }
        }
    }
}
 
#[tokio::main]
async fn main() {
    let (job_tx, job_rx) = mpsc::channel(8);
    let (stop_tx, stop_rx) = mpsc::channel(1);
 
    let handle = tokio::spawn(worker(job_rx, stop_rx));
    job_tx.send(1).await.unwrap();
    stop_tx.send(()).await.unwrap();
    handle.await.unwrap();
}

What this demonstrates:

Long-running loop with multiple event sources.
Shutdown branch exits cleanly.
interval tick branch for periodic work.

Deep Dive

How It Works

First ready branch wins: Other branches are disabled for this iteration.
Cancellation: Losing branches are dropped at await points.
biased: Optional biased; prefix checks branches in declaration order (faster, can starve).
else: Runs when all branches are disabled (e.g., closed channel).

Cancel Safety

Operation	Cancel-safe in loop?
`mpsc::recv`	yes
`sleep`	yes
`read` partial buffer	often no - use buffer carefully
`async fn` with state	depends - audit docs

Rust Notes

// Random fair selection (pattern for rare cases)
use rand::Rng;
 
let branches = [async { 1 }, async { 2 }];
// Prefer declarative select! for readability in most servers

Document which branches are safe to cancel in loop select!.
Combine with Semaphore to limit concurrent selected work.
select! on multiple JoinHandles races task completion.

Gotchas

Non-cancel-safe read in loop - Partial read lost on cancel. Fix: Buffer per protocol rules or use dedicated reader task.
biased select starvation - First branch always ready blocks others. Fix: Reorder or remove biased for fairness.
select without loop exit - Busy spin if branch immediately ready. Fix: Add sleep or await blocking recv.
Mutex across select - Guard not Send. Fix: Acquire lock inside branch body only.
Forgotten else on closed channels - Spin on disabled branches. Fix: Use else or break on None.

Alternatives

Alternative	Use When	Don't Use When
`join!`	Need all branches	Event loop first-ready
Dedicated reader task + mpsc	Complex protocol parsing	Simple timer + one channel
`FuturesUnordered`	Many dynamic futures	Fixed small branch set
epoll manually	Custom runtime	Tokio application code

FAQs

select! vs join!?

select! first completion; join! waits for all. Different use cases.

What does biased do?

Checks branches in order without randomization - predictable but can starve later branches.

Loop pattern?

loop { select! { ... } } is standard for Tokio services until shutdown branch breaks.

Multiple channels?

One branch per channel recv - classic multiplexed server design.

select with streams?

some_stream.next() => branch supported in tokio::select!.

pin required?

Some branch futures need pin in select - follow macro expansion errors with pin!.

Axum shutdown?

with_graceful_shutdown uses signal future - related but at server level.

testing select?

tokio::time::pause + advance for deterministic timer branches.

priority events?

Put high-priority branch first with biased - document starvation risk.

select across crates?

Use tokio::select! for Tokio futures; futures::select for non-Tokio contexts.

Combining Futures - join and select theory
Channels in Tokio - recv branches
Timers & Intervals - sleep and interval
Graceful Shutdown - shutdown select
Cancellation & Timeouts - cancel safety

Stack versions: This page was written for Rust 1.97.0 (edition 2024), Tokio 1.x, Axum 0.8, serde 1.0, sqlx 0.8, clap 4, and Polars 0.46+.

Recipe

Quick-reference recipe card - copy-paste ready.

use tokio::sync::mpsc;
use tokio::time::{sleep, Duration};
 
#[tokio::main]
async fn main() {
    let (tx, mut rx) = mpsc::channel(4);
 
    tokio::select! {
        Some(msg) = rx.recv() => println!("{msg}"),
        _ = sleep(Duration::from_millis(100)) => println!("timeout"),
        else => println!("channel closed"),
    }
 
    let _ = tx.send("hi").await;
}

When to reach for this: Server main loops, combining channel messages with shutdown signals, and timeout races.

Working Example

use tokio::sync::mpsc;
use tokio::time::{interval, Duration};
 
async fn worker(mut rx: mpsc::Receiver<u32>, mut shutdown: mpsc::Receiver<()>) {
    let mut ticker = interval(Duration::from_millis(200));
 
    loop {
        tokio::select! {
            _ = shutdown.recv() => {
                println!("shutdown");
                break;
            }
            Some(job) = rx.recv() => {
                println!("job {job}");
            }
            _ = ticker.tick() => {
                println!("heartbeat");
            }
        }
    }
}
 
#[tokio::main]
async fn main() {
    let (job_tx, job_rx) = mpsc::channel(8);
    let (stop_tx, stop_rx) = mpsc::channel(1);
 
    let handle = tokio::spawn(worker(job_rx, stop_rx));
    job_tx.send(1).await.unwrap();
    stop_tx.send(()).await.unwrap();
    handle.await.unwrap();
}

What this demonstrates:

Long-running loop with multiple event sources.
Shutdown branch exits cleanly.
interval tick branch for periodic work.

Deep Dive

How It Works

First ready branch wins: Other branches are disabled for this iteration.
Cancellation: Losing branches are dropped at await points.
biased: Optional biased; prefix checks branches in declaration order (faster, can starve).
else: Runs when all branches are disabled (e.g., closed channel).

Cancel Safety

Operation	Cancel-safe in loop?
`mpsc::recv`	yes
`sleep`	yes
`read` partial buffer	often no - use buffer carefully
`async fn` with state	depends - audit docs

Rust Notes

// Random fair selection (pattern for rare cases)
use rand::Rng;
 
let branches = [async { 1 }, async { 2 }];
// Prefer declarative select! for readability in most servers

Document which branches are safe to cancel in loop select!.
Combine with Semaphore to limit concurrent selected work.
select! on multiple JoinHandles races task completion.

Gotchas

Non-cancel-safe read in loop - Partial read lost on cancel. Fix: Buffer per protocol rules or use dedicated reader task.
biased select starvation - First branch always ready blocks others. Fix: Reorder or remove biased for fairness.
select without loop exit - Busy spin if branch immediately ready. Fix: Add sleep or await blocking recv.
Mutex across select - Guard not Send. Fix: Acquire lock inside branch body only.
Forgotten else on closed channels - Spin on disabled branches. Fix: Use else or break on None.

Alternatives

Alternative	Use When	Don't Use When
`join!`	Need all branches	Event loop first-ready
Dedicated reader task + mpsc	Complex protocol parsing	Simple timer + one channel
`FuturesUnordered`	Many dynamic futures	Fixed small branch set
epoll manually	Custom runtime	Tokio application code

FAQs

select! vs join!?

select! first completion; join! waits for all. Different use cases.

What does biased do?

Checks branches in order without randomization - predictable but can starve later branches.

Loop pattern?

loop { select! { ... } } is standard for Tokio services until shutdown branch breaks.

Multiple channels?

One branch per channel recv - classic multiplexed server design.

select with streams?

some_stream.next() => branch supported in tokio::select!.

pin required?

Some branch futures need pin in select - follow macro expansion errors with pin!.

Axum shutdown?

with_graceful_shutdown uses signal future - related but at server level.

testing select?

tokio::time::pause + advance for deterministic timer branches.

priority events?

Put high-priority branch first with biased - document starvation risk.

select across crates?

Use tokio::select! for Tokio futures; futures::select for non-Tokio contexts.

Combining Futures - join and select theory
Channels in Tokio - recv branches
Timers & Intervals - sleep and interval
Graceful Shutdown - shutdown select
Cancellation & Timeouts - cancel safety

Stack versions: This page was written for Rust 1.97.0 (edition 2024), Tokio 1.x, Axum 0.8, serde 1.0, sqlx 0.8, clap 4, and Polars 0.46+.

select! & Concurrency

Search Documentation

select! & Concurrency