Crate tokio_fsm 

tokio-fsm

Compile-time validated, zero-overhead async finite state machines for Tokio.

tokio-fsm allows you to define complex asynchronous state machines using a declarative macro. It eliminates the boilerplate of manual event loops, channel management, and state transitions while ensuring that your FSM logic is verified at compile-time.

Problem Statement

Traditional FSM implementations in Rust often fall into two categories:

1. Hand-rolled: Extremely fast and flexible, but error-prone. Managing mpsc channels, tokio::select! loops, and ensuring “invalid state” transitions don’t occur requires significant boilerplate.
2. Runtime Engines: Flexible, but often involve virtual dispatch, heavy allocations, or complex trait hierarchies that add overhead and obfuscate the mental model.

tokio-fsm provides a third way: Write standard Rust impl blocks, and let the proc-macro generate the high-performance, type-safe boilerplate for you.

Core Concepts

States & Events

States and Events are discovered automatically from your #[on] handlers. You don’t need to manually define enums for them—the macro generates a [FsmName]State and [FsmName]Event enum for you.

Context

The Context is the shared, mutable data owned by the FSM. Every handler has access to &mut self.context.

Transitions

Handlers return a Transition<NextState>. This explicitly defines the next state the FSM should move to. The macro validates that NextState is a known state and that the transition is reachable.

Timeouts

tokio-fsm supports state-level timeouts via #[state_timeout]. These are implemented using single, stack-pinned tokio::time::Sleep futures, ensuring zero heap allocations during transitions.

Architecture

1. Validation Layer: At compile-time, the macro builds a directed graph of your FSM. It verifies that all states are reachable and that transitions are logically consistent.
2. Codegen Layer: Generates a tight, state-gated match loop. There is no runtime engine; the generated code is identical to what a senior engineer would write by hand.

Example: Basic Worker

use tokio_fsm::{fsm, Transition};

pub struct WorkerContext { iterations: usize }

#[fsm(initial = Idle)]
impl WorkerFsm {
    type Context = WorkerContext;
    type Error = std::convert::Infallible;

    #[on(state = Idle, event = Start)]
    async fn on_start(&mut self) -> Transition<Running> {
        Transition::to(Running)
    }

    #[on(state = Running, event = Tick)]
    async fn on_tick(&mut self) -> Transition<Running> {
        self.context.iterations += 1;
        Transition::to(Running)
    }
}

FAQ & Troubleshooting

Why is my event ignored?

If an event is sent while the FSM is in a state where no #[on(state = current, event = event)] handler is defined, the event is silently dropped by default. This is intentional to prevent deadlocks in high-throughput systems.

Compile Error: “State X not found”

Ensure that state X is either the initial state or is used as a target in a Transition<X> or a source in an #[on(state = X, ...)] attribute.

Compile Error: “Event Y not found”

Ensure that event Y is defined in at least one #[on(..., event = Y)] attribute.

Enums

ShutdownMode

Shutdown mode for the FSM.

TaskError

Error type returned by the FSM background task.

Transition

Represents a state transition in the FSM.

Attribute Macros

fsm

Generates an asynchronous Finite State Machine (FSM) from an impl block.