infer action creators from createSlice as PayloadActionCreator

[phryneas]

see #157

This changes the createSlice types to use PayloadActionCreator and modifies the behaviour of PayloadActionCreator to account for P = void and P = Something | undefined.

Here are some examples:

const someNumber: number = 5;
const someString: string = "";

const voidActionCreator = createAction<void>('a');
voidActionCreator(); // Action<'a'>
voidActionCreator(undefined); // Action<'a'>

const stringActionCreator = createAction<string>('b');
stringActionCreator('x'); // PayloadAction<'x', 'b'>

const optionalStringActionCreator = createAction<string | undefined>('c');
optionalStringActionCreator(); // Action<'c'>
optionalStringActionCreator(undefined); // Action<'c'>
optionalStringActionCreator('x') // PayloadAction<'x', 'c'>

const mixedActionCreator = createAction<string | number>('d');
mixedActionCreator('x') // PayloadAction<'x', 'd'>
mixedActionCreator(someNumber) // PayloadAction<number, 'd'>
mixedActionCreator(5) // PayloadAction<5, 'd'>
mixedActionCreator(someString) // PayloadAction<string, 'd'>

const anyActionCreator = createAction('e');
anyActionCreator() // Action<'e'>
anyActionCreator(someNumber) // PayloadAction<number, 'e'>
anyActionCreator(5) // PayloadAction<5, 'e'>
anyActionCreator(someString) // PayloadAction<string, 'e'>

so payload types are now interfered much more accurate, which is especially useful for the <any> case.

There is one open question open regarding the typing of the undefined argument: if undefined is not omitted, but explicitly specified, should we return an Action or a PayloadAction? (See https://github.com/reduxjs/redux-starter-kit/pull/158/files#diff-437575fb1d5c39e3edd424ab38899f8eR27 )

PS: I could easily do away with the excessive type interference, but that would hurt the <any> case a bit in my eyes and I think it won't hurt, so I'm looking forward to your opinion on it ;)

[netlify]

Deploy preview for redux-starter-kit-docs ready!

Built with commit a960dc6

https://deploy-preview-158--redux-starter-kit-docs.netlify.com

[markerikson]

I vote that createAction() should always return a PayloadAction, for consistency (and on the grounds that we'll also probably end up throwing meta in there too in the near future too).

[markerikson]

Why are these not "counter/increment"? That's really confusing to me.

[phryneas]

Good catch.
Because I typed them wrongly and those tests only check for types at compile time, not validate that it is the real output.
That said, I'll have to remove that and only type the type as string then - can't do concatenation with types :(

[markerikson]

One other question to throw out there. If we do eventually tack on meta, is PayloadAction really the best term? Should it maybe be FSA instead or something?

[phryneas]

One other question to throw out there. If we do eventually tack on meta, is PayloadAction really the best term? Should it maybe be FSA instead or something?

I'm all for that. 👍

[phryneas]

PS: Generally, I'm for a renaming, but having changed it to FSA everywhere in the code, it looks really more like a generic than a type to me, with it being only capital letters. Not really sure if that would be good for readablility. Do we have other options?

I was thinking along the lines of:

• FSAction
• ExtendedAction
• SpecifiedAction
• DetailedAction

But reading all those above, I have to say that I'd also be fine with keeping PayloadAction and just adding a third parameter for meta... Not 100% happy with any of those.

[markerikson]

Eh, let's keep PayloadAction for now. Less churn.

[markerikson]

Cool. All right, let's get this out. Thank you so much for working on this!

[phryneas]

Happy to help 😎

Just because I noticed, in case you forgot (I don't know your release workflow):

While the 0.6 is available on npm, the commit & tag are still missing on GitHub. Did you forget to push?

[markerikson]

Not quite. I actually did the release from work, but I'm behind a corporate proxy. I use https://github.com/sindresorhus/np to do my releases, but it seems that doesn't work great behind proxies. It was able to hit NPM to publish, but not do the Github work.

I'll get those up tonight after I get home.