Question: How to aggregate outputs from Parallel States in v5?

[powfan]

I'm using XState v5 and trying to get the aggregated output from a Parallel State when it completes (onDone).

In XState v4 or previous discussions, it was implied that event.output in the onDone transition of a parallel state would contain the merged outputs of its child final states. However, in v5, I am getting undefined for event.output.

Here is a minimal reproduction:

import { setup, fromPromise, createActor } from 'xstate'

const machine = setup({
  actors: {
    worker: fromPromise(async () => ({ data: 'result' })),
  },
}).createMachine({
  initial: 'parallel',
  states: {
    parallel: {
      type: 'parallel',
      onDone: {
        actions: ({ event }) => {
          // This logs 'undefined'
          console.log('Aggregated Output:', event.output)
        },
        target: 'end',
      },
      states: {
        branch1: {
          initial: 'working',
          states: {
            working: {
              invoke: {
                src: 'worker',
                onDone: 'finished',
              },
            },
            finished: {
              type: 'final',
              output: {
                res1: 1,
              },
            },
          },
        },
        branch2: {
          initial: 'working',
          states: {
            working: {
              invoke: {
                src: 'worker',
                onDone: 'finished',
              },
            },
            finished: {
              type: 'final',
              output: {
                res2: 2,
              },
            },
          },
        },
      },
    },
    end: { type: 'final' },
  },
})

const actor = createActor(machine).start()

Question:

1. Does XState v5 support automatic output aggregation for Parallel States?
2. If yes, what am I missing?
3. If no, is using context and assign the only recommended way to collect results from parallel branches?

Thank you!

[FloNeu]

Good question - didn't know that worked in V4, i currently, as you stated, assign the single output-results on the context for later processing.

[davidkpiano]

In v5, it does not, but we are probably going to add support for aggregating the outputs in v6