flow

A minimalistic functional state machine

This is an attempt to build the simplest functional state machine without any external dependencies. Reviewers who can contribute to make it simpler or more efficient without compromising funcionality are welcome.

Borrowing from functional programming, coroutines, and generator functions; this "coded by convention" loop sets the foundation to compose more complex applications with just three basic constructs:

• Actions - Pure functions with logic to drive internal transitions

• Reducer - Pure function with logic to reduce the state from internal and external transition payloads

• Flow - A closure with the coroutine implementing the loop and holding the state

Conventions

• Anonymous actions yield control back to the caller (User). Think of it as a one-time yielding generator function.

• Actions can optionally return:

  • An Object - action payload
  • An Action - (named or anonymous) to allow composition
  • An Array of the above - to be executed in order

The loop

The flow is initialized with a map of actions, optional parameters, the reducer, and optional callbacks (invoked, shifted) for tracing and debugging actions. The injected action map allows composition without coupling modules.

The returned coroutine must be started with a root action and can be successively invoked with transition payloads. It always returns its internal structure including the current state and scope.

Actions are internally invoked with 3 arguments: (state, scope, { params, actions }).

The flow keeps track of action recursion, indentation, and stack depth levels in the scope object.

Schema

flow = {
  state: {...} // current reduced state
  scope: { // current scope
    name: 'string', // action name
    recur: 'int', // recurrence counter
    parent: {...}, // parent scope
    level: 'int', // indentation level
    depth: 'int' // stack depth
  },
  yielding: 'function' // yielding action
  stack: [] // current stack
  done: 'bool', // true when end of stack reached
}

How to use

const flow = require('@rotorsoft/flow')

function action1(state, scope, { params }) {
  return [
    { ask: `Am I speaking with ${params.name}?` },
    (state, { recur }) => {
      if (state.action1.answer === 'yes') return
      if (recur < 2) return action1
    }
  ]
}

const root = (state, scope, { params, actions }) => {
  return [
    actions.action1,
    function root(state) {
      if (state.action1.answer === 'yes') return { say: `Hello ${params.name}. How are you today?` }
      return { say: "I'm sorry for the inconvenience." }
    }
  ]
}

const next = flow({
  params: { name: 'John Doe' },
  actions: { action1 },
  reducer: (state, scope, payload) => ({ ...state, ...payload })
})

let $ = next(root) // start root action
$ = next({ action1: { answer: 'what?' } }) // update action1 state
$ = next({ action1: { answer: 'yes' } }) // update action1 state
console.log($)

Test

npm test

The provided tests are self explanatory and should log a trace like this:

  simple test
[ 0] root() { // [authenticate(state,scope,{params}), next({authenticate,verifyPhone,canComeToThePhone})]
[ 2]    authenticate() { // [{"ask":"Am I speakin...}, (state,{recur})]
[ 2]       {"ask":"Am I speaking with John Doe?"}
[ 2]       (state,{recur}) ... {"authenticate":{"answer":"whatever"}}
[ 3]       authenticate:1() { // [{"ask":"Am I speakin...}, (state,{recur})]
[ 3]          {"ask":"Am I speaking with John Doe?"}
[ 3]          (state,{recur}) ... {"authenticate":{"answer":"yes"}}
[ 3]       } // authenticate:1
[ 2]    } // authenticate
[ 1]    next() {
[ 1]       {"say":"Hello John Doe. How are you today?","authenticated":true}
[ 1]    } // next
[ 0] } // root

===
{
  state: {
    ask: 'Am I speaking with John Doe?',
    authenticate: { answer: 'yes' },
    say: 'Hello John Doe. How are you today?',
    authenticated: true
  },
  scope: {},
  stack: [],
  done: true
}
===

    √ should authenticate

---

Simplicity is the ultimate sophistication

Leonardo da Vinci

---

Contributing

In lieu of a formal style guide, take care to maintain the existing coding style. Add unit tests for any new or changed functionality. Lint and test your code.

License

MIT