Advice on getting valid transition targets for node

[mattpocock]

I'm working on xstate-codegen, and I've been trying to think of a way to get all the valid transition targets from a node. I'd like to get them as strings, with the delimiter, so that I can use them for type checking.

Here's what I need to get:

1. The siblings of a node.
2. The children of all siblings, including their children.
3. Any nodes declared with an id property, to allow for '#id' transitions. I also need all of their children/grandchildren.

I am getting there, but I wondered if you have any shortcuts/utils I can use?

[davidkpiano]

Do you mean all the targets of all the transitions from a state node?

function getAllTransitionTargets(stateNode) {
  return flatten(stateNode.transitions.map(t => {
    return t.target || [];
  }));
}

That should do it. You can dedupe using Set() and recursively get all the transition targets of the sibling nodes themselves (Object.values(stateNode.states)).

This is after the machine does the "processing" of the config object.

[Andarist]

Hm, could you clarify this a little bit more? Non-declared transitions are not valid and thus not possible. I know you know that so we must clearly just not understand each other correctly. Could you rephrase?

[mattpocock]

@Andarist Of course, my fault.

A loom for clarity:

https://www.loom.com/share/a3f9eb337338495e8f40c9553acce8c0

[Andarist]

There are no utils for this, as far as I know. Maybe we could include such in @xstate/graph package? although not 100% sure what's it exact purpose, it sounds like a good place for this and if not then a separate package with utils like this could be created

For now, I would just implement this in the codegen's codebase - we can later extract it.

To get all ID targets you could probably do something like (just a pseudo code, not validated with real-life functions etc):

import { getAllStateNodes } from 'xstate/lib/stateUtils'

const explicitIds = getAllStateNodes(machine).filter(node => node.config.id).map(node => node.id)

[davidkpiano]

Yeah, the possible targets would be a combination of the explicit IDs (as stated above) and the "relative" dot-representation of those IDs relative to that state node. All state nodes can share the explicitIds, so for the relative paths, you would filter the state nodes that are either siblings, children, or cousins (children of siblings) and generate the relative paths.

[mattpocock]

Got this working, thanks for your help 👍