@xstate/fsm
The @xstate/fsm package contains a minimal, 1kb implementation of XState for finite state machines.
Featuresβ
| @xstate/fsm | XState | |
|---|---|---|
| Finite states | β | β | 
| Initial state | β | β | 
| Transitions (object) | β | β | 
| Transitions (string target) | β | β | 
| Delayed transitions | β | β | 
| Eventless transitions | β | β | 
| Nested states | β | β | 
| Parallel states | β | β | 
| History states | β | β | 
| Final states | β | β | 
| Context | β | β | 
| Entry actions | β | β | 
| Exit actions | β | β | 
| Transition actions | β | β | 
| Parameterized actions | β | β | 
| Transition guards | β | β | 
| Parameterized guards | β | β | 
| Spawned actors | β | β | 
| Invoked actors | β | β | 
- Finite states (non-nested)
- Initial state
- Transitions (object or strings)
- Context
- Entry actions
- Exit actions
- Transition actions
- state.changed
If you want to use statechart features such as nested states, parallel states, history states, activities, invoked services, delayed transitions, transient transitions, etc. please use XState.
Quick startβ
Installationβ
npm i @xstate/fsm
Usage (machine)β
import { createMachine } from '@xstate/fsm';
const toggleMachine = createMachine({
  id: 'toggle',
  initial: 'inactive',
  states: {
    inactive: { on: { TOGGLE: 'active' } },
    active: { on: { TOGGLE: 'inactive' } }
  }
});
const { initialState } = toggleMachine;
const toggledState = toggleMachine.transition(initialState, 'TOGGLE');
toggledState.value;
const untoggledState = toggleMachine.transition(toggledState, 'TOGGLE');
untoggledState.value;
// => 'inactive'
Usage (service)β
import { createMachine, interpret } from '@xstate/fsm';
const toggleMachine = createMachine({});
const toggleService = interpret(toggleMachine).start();
toggleService.subscribe((state) => {
  console.log(state.value);
});
toggleService.send('TOGGLE');
toggleService.send('TOGGLE');
toggleService.stop();
APIβ
createMachine(config, options)β
Creates a new finite state machine from the config.
| Argument | Type | Description | 
|---|---|---|
| config | object (see below) | The config object for creating the machine. | 
| options | object (see below) | The optional options object. | 
Returnsβ
A Machine, which provides:
- machine.initialState: the machineβs resolved initial state
- machine.transition(state, event): a pure transition function that returns the next state given the current- stateand- event
The machine config has this schema:
Machine configβ
- id(string) - an identifier for the type of machine this is. Useful for debugging.
- initial(string) - the key of the initial state.
- states(object) - an object mapping state names (keys) to states
State configβ
- on(object) - an object mapping event types (keys) to transitions
Transition configβ
String syntax:
- (string) - the state name to transition to.- Same as { target: stateName }
 
- Same as 
Object syntax:
- target?(string) - the state name to transition to.
- actions?(Action | Action[]) - the action(s) to execute when this transition is taken.
- cond?(Guard) - the condition (predicate function) to test. If it returns- true, the transition will be taken.
Machine optionsβ
- actions?(object) - a lookup object for your string actions.
Action configβ
Function syntax:
- (function) - the action function to execute. Resolves to { type: actionFn.name, exec: actionFn }and the function takes the following arguments:- context(any) - the machineβs current- context.
- event(object) - the event that caused the action to be executed.
 
Object syntax:
- type(string) - the action type.
- exec?(function) - the action function to execute.
String syntax:
- (string) - the action type.- By default it resolves to { type: actionType, exec: undefined }. It can resolve to resolved function or resolved object action if the action can be looked up in theoptions.actionsobject.
 
- By default it resolves to 
Why use a string or object for defining actions?
Using the string or object syntax is useful for handling actions in a custom way, rather than baking in the implementation details to your machine:
const nextState = machine.transition();
nextState.actions.forEach((action) => {
  if (action.type === 'focus') {
  }
});
machine.initialStateβ
The resolved initial state of the machine.
machine.transition(state, event)β
A pure transition function that returns the next state given the current state and event.
The state can be a string state name, or a State object (the return type of machine.transition(...)).
| Argument | Type | Description | 
|---|---|---|
| state | stringorStateobject | The current state to transition from | 
| event | stringor{ type: string, ... } | The event that transitions the current stateto the next state | 
Returnsβ
A State object, which represents the next state.
Exampleβ
const yellowState = machine.transition('green', 'TIMER');
const redState = machine.transition(yellowState, 'TIMER');
const greenState = machine.transition(yellowState, { type: 'TIMER' });
// => { value: 'green', ... }
Stateβ
An object that represents the state of a machine with the following schema:
- value(string) - the finite state value
- context(object) - the extended state (context)
- actions(array) - an array of action objects representing the side-effects (actions) to be executed
- changed(boolean) - whether this state is changed from the previous state (- trueif the- state.valueand- state.contextare the same, and there are no side-effects)
- matches(value)(boolean) - whether this stateβs value matches (i.e., is equal to) the- value. This is useful for typestate checking.
interpret(machine)β
Creates an instance of an interpreted machine, also known as a service. This is a stateful representation of the running machine, which you can subscribe to, send events to, start, and stop.
Actions will also be executed by the interpreter.
| Argument | Type | Description | 
|---|---|---|
| machine | StateMachine | The machine to be interpreted. | 
Exampleβ
import { createMachine, interpret } from '@xstate/fsm';
const machine = createMachine({});
const service = interpret(machine);
const subscription = service.subscribe((state) => {
  console.log(state);
});
service.start();
service.send('SOME_EVENT');
service.send({ type: 'ANOTHER_EVENT' });
subscription.unsubscribe();
service.stop();
service.subscribe(stateListener)β
A service (created from interpret(machine)) can be subscribed to via the .subscribe(...) method. The subscription will be notified of all state changes (including the initial state) and can be unsubscribed.
| Argument | Type | Description | 
|---|---|---|
| stateListener | (state) => void | The listener that is called with the interpreted machineβs current statewhenever it changes. | 
Returnsβ
A subscription object with an unsubscribe method.
service.send(event)β
Sends an event to the interpreted machine. The event can be a string (e.g., "EVENT") or an object with a type property (e.g., { type: "EVENT" }).
| Argument | Type | Description | 
|---|---|---|
| event | stringor{ type: string, ... } | The event to be sent to the interpreted machine. | 
service.start()β
Starts the interpreted machine.
Events sent to the interpreted machine will not trigger any transitions until the service is started. All listeners (via service.subscribe(listener)) will receive the machine.initialState.
service.stop()β
Stops the interpreted machine.
Events sent to a stopped service will no longer trigger any transitions. All listeners (via service.subscribe(listener)) will be unsubscribed.
TypeScriptβ
A machine can be strictly typed by passing in 3 generic types:
- TContext- the machineβs- context
- TEvent- all events that the machine accepts
- TState- all states that the machine can be in
The TContext type should be an object that represents all possible combined types of state.context.
The TEvent type should be the union of all event objects that the machine can accept, where each event object has a { type: string } property, as well as any other properties that may be present.
The TState type should be the union of all typestates (value and contexts) that the machine can be in, where each typestate has:
- value(string) - the value (name) of the state
- context(object) - an object that extends- TContextand narrows the shape of the context to what it should be in this state.
Exampleβ
interface User {
  name: string;
}
interface UserContext {
  user?: User;
  error?: string;
}
type UserEvent =
  | { type: 'FETCH'; id: string }
  | { type: 'RESOLVE'; user: User }
  | { type: 'REJECT'; error: string };
type UserState =
  | {
      value: 'idle';
      context: UserContext & {
        user: undefined;
        error: undefined;
      };
    }
  | {
      value: 'loading';
      context: UserContext;
    }
  | {
      value: 'success';
      context: UserContext & { user: User; error: undefined };
    }
  | {
      value: 'failure';
      context: UserContext & { user: undefined; error: string };
    };
const userMachine = createMachine<UserContext, UserEvent, UserState>({
  /* ... */
});
const userService = interpret(userMachine);
userService.subscribe((state) => {
  if (state.matches('success')) {
    // from UserState, `user` will be defined
    state.context.user.name;
  }
});
Exampleβ
import { createMachine, assign, interpret } from '@xstate/fsm';
const lightMachine = createMachine({
  id: 'light',
  initial: 'green',
  context: { redLights: 0 },
  states: {
    green: {
      on: {
        TIMER: 'yellow'
      }
    },
    yellow: {
      on: {
        TIMER: {
          target: 'red',
          actions: () => console.log('Going to red!')
        }
      }
    },
    red: {
      entry: assign({ redLights: (ctx) => ctx.redLights + 1 }),
      on: {
        TIMER: 'green'
      }
    }
  }
});
const lightService = interpret(lightMachine);
lightService.subscribe((state) => {
  console.log(state);
});
lightService.start();
lightService.send('TIMER');
lightService.send('TIMER');
// => logs {
//   value: 'red',
//   context: { redLights: 1 },
//   actions: [],
//   changed: true
// }