[core] Synchronizers

[davidkpiano]

Added support for synchronizers in XState, allowing state persistence and synchronization across different storage mechanisms.

• Introduced Synchronizer interface for implementing custom synchronization logic
• Added sync option to createActor for attaching synchronizers to actors

import { createActor } from 'xstate';
import { someMachine } from './someMachine';
import { createLocalStorageSync } from './localStorageSynchronizer';

const actor = createActor(someMachine, {
  sync: createLocalStorageSync('someKey')
});

[changeset-bot]

🦋 Changeset detected

Latest commit: 03854c3

The changes in this PR will be included in the next version bump.

This PR includes changesets to release 6 packages

Name	Type
xstate	Minor
@xstate/graph	Major
@xstate/react	Major
@xstate/solid	Major
@xstate/svelte	Major
@xstate/vue	Major

Not sure what this means? Click here to learn what changesets are.

Click here if you're a maintainer who wants to add another changeset to this PR

[Andarist]

We don't use @ for other builtin events like init, done.invoke and more

Suggested change

	this.update(restoredSnapshot, { type: '@xstate.sync' });
	this.update(restoredSnapshot, { type: 'xstate.sync' });

[Andarist]

what snapshot should the synchronizer be aware of when somebody does smth like this:

const actorRef = createActor(m, { sync: someSynchronizer })

// ... do stuff

actorRef.stop()
actorRef.getSnapshot() // .status === 'stopped'

If I'm not mistaken the known snapshot after this operation has a stopped status but the synchronizer won't be aware of it because we have unsubscribed earlier.

[Andarist]

why only for the active status? done and error should likely be passed to it too (and likely stopped too)

[Andarist]

I think it should get enqueued and not handled immediately because with this model we allow for this.update to be potentially reentered (and that's a bad thing™). And if we are in the middle of processing 5 events mailbox and we receive a rawSnapshot here then likely we should append to the mailbox instead of processing it right away

[Andarist]

this could likely be reverted - no tests fails without this if 😉