feat(utils): Add loadable

[Pinpickle]

Fixes: #672 (or at least tries to)

This implements loadable as discussed in the mentioned issue.

I can't get all edge-cases to work. In particular, recovering from error, when an atom goes through these states:

suspense -> error -> suspense -> resolved

Going from error -> suspense does not trigger a re-render.

I think this might need some changes to core, as I suspect by catching the error/promise, we are preventing some state-keeping logic from happening.

@dai-shi could you have a look? It's the last test in loadable.test.tsx that's failing.

[vercel]

This pull request is being automatically deployed with Vercel (learn more).
To see the status of your deployment, click below or on the icon next to each commit.

🔍 Inspect: https://vercel.com/pmndrs/jotai/7vC8KaT7kiUF2X1EP16WUV92zd1o
✅ Preview: https://jotai-git-fork-pinpickle-add-loadable-pmndrs.vercel.app

[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.

Latest deployment of this branch, based on commit 15fbff3:

Sandbox	Source
React	Configuration
React Typescript	Configuration
React Browserify	Configuration
React Snowpack	Configuration
Next.js	Configuration

[dai-shi]

Nice work.
I have some coding preference to apply, reducing bundle size, fixing refAtom usage, but in general, it looks good.

However:

I can't get all edge-cases to work. In particular, recovering from error

Oh, right... This is not resolved yet, if it's same as #601.
Yeah, I agree it's probably core responsibility.
I think it's going to take some time for this.
Should we fix it first, or could loadable be useful with this limitation for now?

[Pinpickle]

@dai-shi I wasn't aware that it was a bug for all atoms

Given that this is a bug with all atoms, I think it's OK to release as-is. It should probably be documented somewhere that error recovery doesn't work. Should I mark the test as skipped or remove altogether?

[dai-shi]

I'm not 100% sure if it's the same bug as #601. (tbh, I haven't spent much time on it yet.)

Please leave the tests as skipped. Let's tackle it separately, if this is useful as is.
I'll request some changes on code.

[dai-shi]

My final recommendation is to use WeakMap instead of refAtom, but let me shamelessly leave my intermediate comments.

[Pinpickle]

@dai-shi I've addressed your comments, it's a lot cleaner now.

Let me know what you think.

I also looked into #601 and it is a core bug, opened a PR at #705

[dai-shi]

Looks very nice!
Added a small suggestion to refactor.

[Pinpickle]

@dai-shi done! I did the same for the loadable cache as well, it's only one key

[Pinpickle]

As we're using a WeakMap, I added this test case. If a primitive value ever became a key for error, it wouldn't work.

[dai-shi]

I did the same for the loadable cache as well, it's only one key

Interestingly, I didn't notice it. Thanks. (There might be other utils we could do the same.)

[dai-shi]

Looks good! Thanks so much.

[dai-shi]

Would you also work on docs? For now, docs are in a separate repo for now.
https://github.com/pmndrs/website/tree/docs/docs/jotai/api/utils.mdx
(and, we should probably split utils page in to some pages.

[Pinpickle]

@dai-shi yep I can have a look at that, will probably be next week

[TwistedMinda]

EDIT: Moving my comment in the open discussion instead
EDIT2: Not sure it was the best move