refactor!: remove resolvePackageEntry and resolvePackageData APIs

[bluwy]

Description

They are preventing caching optimizations that we can do. IIRC the setResolvedCache API isn't correct too and is causing bugs. (have to find the issue #12957 (comment))

Additional context

After a lot of yak shaving, I finally got to what I wanted to do!

---

Admittedly I didn't plan this ahead since it would be better with a deprecation phase, but I think it's fine as these APIs are rather internal.

GitHub search for resolvePackageEntry and resolvePackageData usage

Unfortunately resolvePackageEntry affects some Vite plugins, and resolvePackageData affects two big projects: qwik and slidev. But their usage are simple enough that the migration guide should work.

---

What is the purpose of this pull request?

• Bug fix
• New Feature
• Documentation update
• Other

[sapphi-red]

IIRC the setResolvedCache API isn't correct too and is causing bugs. (have to find the issue)

I guess it's #12957 (comment).

[patak-dev]

Sounds good to me! You're really pushing hard to clean up things in Vite 😄
It would be good to see what Anthony thinks given that it affects sli.dev though. And I wonder if Vite should export similar utilities in the future (once optimizations are done and you are happy with the new API). It seems useful for plugins to get the exact same handling of package.json as Vite and also take advantage of the cache.

[bluwy]

Yeah once this is unblocked we should be able to refine the API to be public again for sure 👍 (especially the package.json handling). I have also thought about creating new @vitejs/* packages to really generalize it, but maybe that's for another day.

[bluwy]

I think we can hold off merging for now and get qwik and slidev migrated first.

[aleclarson]

Out of curiosity, is there a detailed explanation of the optimizations blocked by this API, the bugs caused by it, and how removing it will fix them?

From a quick look, I have no packages depending on the API.

[bluwy]

Out of curiosity, is there a detailed explanation of the optimizations blocked by this API, the bugs caused by it, and how removing it will fix them?

1. We need to modify the PackageCache to only cache the package.json object, not PackageData.
2. We need two package cache with different value types. Right now we're mixing both for compat. The keys are prefixed like so (one and two) to workaround it.
3. 预览模式中element-ui el-table jsx写法不显示 #12957 (comment) highlights the bug that we can't rely on a single resolve cache for each package. We might need to remove PackageData completely, or cache it in a different way.
4. And IIRC it also blocks us being able to cache for CSS import resolvers

[bluwy]

I've sent the PRs upstream to remove them. Will go ahead and merge this.