feat(compiler-sfc): support defineAttrs macro

[rudyxu1102]

RFC

• Infer attrs for implementing HOCs rfcs#477

[github-actions]

Size Report

Bundles

File	Size	Gzip	Brotli
runtime-dom.global.prod.js	85.9 kB (+38 B)	32.7 kB (+4 B)	29.5 kB (+6 B)
vue.global.prod.js	132 kB (+38 B)	49.3 kB (+5 B)	44.3 kB (-38 B)

Usages

Name	Size	Gzip	Brotli
createApp	47.9 kB	18.8 kB	17.2 kB
createSSRApp	50.6 kB	19.9 kB	18.2 kB
defineCustomElement	50.3 kB	19.6 kB	17.9 kB
overall	61.2 kB	23.7 kB	21.6 kB

[sxzz]

Is there any difference between this PR and #7444?

[sadeghbarati]

Is there any difference between this PR and #7444?

If I understand correctly, this PR is for typed/inferring attrs with macro in <script setup> (SFC)

and this PR #7444 is the same but for defineComponent or functional components (TS/TSX)

[sxzz]

#7444 did the same thing (defineAttrs macro) before my review. As my review comment in #7444, I don't think it's necessary to introduce a new macro for attrs. Reusing useAttrs is enough for declaring types.

In you PR, defineAttrs will be replaced with useAttrs. So why not use it directly, without any transformation?

const attrs = defineAttrs<{
	bar?: number
}>()

Let's discuss in #7444 since they're basically doing the same feature.

[rudyxu1102]

#7444 did the same thing (defineAttrs macro) before my review. As my review comment in #7444, I don't think it's necessary to introduce a new macro for attrs. Reusing useAttrs is enough for declaring types.

In you PR, defineAttrs will be replaced with useAttrs. So why not use it directly, without any transformation?

const attrs = defineAttrs<{
	bar?: number
}>()

Let's discuss in #7444 since they're basically doing the same feature.

I also planned to use useAttrs<T> at the beginning, but after I saw the implementation of defineSlots, I realize using useAttrs<T> without using <script setup> would confuse users.

e.g. Vue SFC Playground

Judging from the discussion in RFC#477, most people like defineAttrs macro and don't like the attrs option. I think it may be easier to discuss if it is separated into two PRs.

[sxzz]

Sorry, I was misled and thought these two PRs belonged to two different authors. (P.S hope to have a more detailed PR description in the future :P )

Compared to defineSlots, you're right. However, we need only one generic support, which means maybe adding a macro defineAttrs is enough. For non <script setup>, users can reach attrs by args of setup function, which already has the right type declaration.

defineComponent({
  setup(_, { attrs }){}
})

Sorry again! I should discuss this with you before giving review suggestions. Please hold on, I will discuss it with Evan for a final decision.

[sxzz]

After discussion, we decided to do something similar to defineSlots.
useAttrs will not be changed, but add a defineAttrs<T>() macro.

[sxzz]

I think this PR and #7444 can be combined into one. WDYT?

[rudyxu1102]

I think this PR and #7444 can be combined into one. WDYT?

Of course.Thank you for your positive response.