feat: add new directives plugin

[lxsmnsyc]

This PR intends to replace the current directives plugin (TanStack) with our own.

Features:

• shorter function ids
  • previous plugin used SHA, which produced a very long id. The new one has shorter length (around 8 characters as uid) and has a descriptive name for easier debugging.
• API change
  • previously we only required createServerReference. This limited the server functions for only top-level declaration. The new plugin now allows inner declarations via cloning (this introduces a new API called cloneServerReference), so now the server functions are able to be registered ahead of time while deferring the actual instance in the exact time it is declared by intention.
  • the previous manifest mechanism is deprecated by server preloading. The plugin keeps track which files has the server functions and then preloads those ahead of time. The function lookup is now done through a map so we don't need to do some dynamic imports. Not sure of the repercussions but there shouldn't be any issues.
• backwards compatibility
  • should still retain the required features
• proper treeshaking
  • previous plugin didn't remove some code from a module with a module-level directive. The new one creates an entirely new module from scratch.
• aliasing
  • the new plugin can now keep track if the server function is exported by alias, so it will be able to retain the export.

[netlify]

✅ Deploy Preview for solid-start-landing-page ready!

Name	Link
🔨 Latest commit	de83996
🔍 Latest deploy log	https://app.netlify.com/projects/solid-start-landing-page/deploys/69a6f28dac00d30008751c46
😎 Deploy Preview	https://deploy-preview-2070--solid-start-landing-page.netlify.app
📱 Preview on mobile	Toggle QR Code... Use your smartphone camera to open QR code link.

---

To edit notification comments on pull requests, go to your Netlify project configuration.

[changeset-bot]

⚠️ No Changeset found

Latest commit: de83996

Merging this PR will not cause a version bump for any packages. If these changes should not result in a new version, you're good to go. If these changes should result in a version bump, you need to add a changeset.

Click here to learn what changesets are, and how to add one.

Click here if you're a maintainer who wants to add a changeset to this PR

[pkg-pr-new]

Open in StackBlitz

npm i https://pkg.pr.new/solidjs/solid-start/@solidjs/start@2070

npm i https://pkg.pr.new/solidjs/solid-start/@solidjs/vite-plugin-nitro-2@2070

commit: e78934e

[agoldstein03]

Hey! Looking to support in any way I can to get this ready, and ideally backported to v1 🙏

This issue has been holding us back from upgrading our very large main codebase past 1.0.10, we have way too much "use server" functionality with transitive imports affected by this, would end up leaking server code to the client without this patch.

I ran a couple tests with this locally on my system and it looks great, there are a couple extra tests I would recommend adding which got hairy with previous hacky implementations I had attempted, and would be nice to have for regression testing:

• top-level conditional error throwing
  • can be based on isServer for testing (as long as there's no chance of tree shaking identifying that as a tree-shakable const)
    • though more commonly in real code it's based on whether an env var is set or not
  • worth testing both directly in "use server" file and via transitive imports
• top-level non-function const which calls server code
  • toy example but e.g. import { randomUUID } from "node:crypto"; const GENERATED_ID = randomUUID();
• checking through production built client code to ensure no server-only consts end up in the client bundle

Happy to write these up and submit them as my best guess to match current test patterns just don't want to contribute anything that might slow down the timeline if the tests don't match expected structure

Anything else I can do please let me know!

[lxsmnsyc]

@agoldstein03 Not really sure if it can be backported to 1.0

The two you listed can probably be added, not the third one just because it's tricky to do so