feat: synchronous atomEffect

[dmaskasky]

Related Discussion

#52

Summary

Refactor atomEffect to be synchronous. Synchronous atom effects run in the current task before store subscribers (such as store.sub and atom.onMount.

TODO

• check tests
• update docs

[dmaskasky]

FIXME: replace with official version once available.

[codesandbox-ci]

This pull request is automatically built and testable in CodeSandbox.

To see build info of the built libraries, click here or the icon next to each commit SHA.

[sebinsua]

Is this published anywhere?

I think I ran into this 'problem' in the following situation:

1. I had an observable source that could sometimes eventually emit some data.
2. I had an effect that would subscribe to this and 'resolve' a Promise.withResolvers deferred with this data (or 'resolve' with empty data if a timeout passed), and which also handled cleanly unsubscribing, etc.
3. An async atom used this effect and then awaited the deferred.promise so that our React code would attempt to block until it became available.
4. The deferred promise never resolved as the effect was never called.

I could be wrong and my issue may have had nothing to do with this. But, either way, I'd be interested to find out if I could remove the slightly strange logic I wrote to make things work.

[dmaskasky]

Could you make a simple reproduction of this logic in a codesandbox? I'm wrapping this PR up quickly and can test the new build against what you have there.

I think what you have sounds reasonable. Currently, atomEffects have a microtask delay from when their atom dependency changed (e.g the atom was set), to when the effect runs. This PR removes that delay.

But the effect's get and set are already both synchronous. I think you need the deferred promise because you are waiting for next signal, which is reasonable.

So for next steps, let me see what you currently have in a codesandbox.

[sebinsua]

@dmaskasky Something like this?

https://codesandbox.io/p/sandbox/spring-glade-hvrczc?file=%2Fsrc%2FApp.js%3A31%2C1

Btw, is there any difference between wrapping an atom with withAtomEffect to run effects on mount/unmount of an atom and just using get(someEffect) within the 'read' of an atom? I wonder if I'm just misunderstanding how things are meant to be used sometimes.

[dmaskasky]

@dmaskasky Something like this?

https://codesandbox.io/p/sandbox/spring-glade-hvrczc?file=%2Fsrc%2FApp.js%3A31%2C1

Btw, is there any difference between wrapping an atom with withAtomEffect to run effects on mount/unmount of an atom and just using get(someEffect) within the 'read' of an atom? I wonder if I'm just misunderstanding how things are meant to be used sometimes.

Ah, I see what's going on. Let's move this to a GH discussion. #54