Possible future direction of the project

[matthewp]

About a year ago I experienced with a library built on top of Robot that combined state machines with DOM manipulation. Ultimately I decided to stop pursuing it. The one thing that really resonated, however, was that it defined your state machine in a schema style syntax (similar to Zod) rather than the functional way we do in Robot now.

I want to discuss this as a possible new syntax for Robot. Here's an example of what a machine would look like:

import { r } from 'robot3';

let machine = r
  .model({
    mode: r.string(),
    dark: r.boolean(),
    emoji: r.string(),
  })
  .states(['idle'])
  .events('idle', ['toggle'])
  .transition(
    'idle',
    'toggle',
    'idle',
    r.assign('mode', ({ model }) => (model.mode === 'dark' ? 'light' : 'dark')),
    r.assign('dark', ({ model }) => model.mode === 'dark'),
    r.assign('emoji', ({ model }) => model.dark ? '🌛' : '☀️')
  )

This renames context to model, and you define the properties on your model. Then in assign you get typesafe manipulation of model values.

To be clear, I'm not sold on this idea as a replacement for Robot's current functional style. But I'm curious to hear what people think. There is no decision being made now, and it's not likely to do so for a long time, if ever.

[nolde]

Hmm. It looks nice for small state machines like this, but larger ones might become harder to understand.