fix: workaround for missing zimmerframe types

[sacrosanctic]

Closes #1289

Description

Checklist

• Update snapshots (if applicable)
• Add a changeset (if applicable)
• Allow maintainers to edit this PR
• I care about what I'm doing, no matter the tool I use (Notepad, Sublime, VSCode, AI...)

[pkg-svelte-dev]

Install the latest version of sv from 05bc073:

pnx https://pkg.svelte.dev/sv/c/05bc07388abf567aa09f6e50028dd8cbd50d7ef4 create

Open in pkg.svelte.dev: https://pkg.svelte.dev/repos/cli/pr/1290

Note

This PR is from a fork. A maintainer must approve approve each commit before it can be built and installed.

[changeset-bot]

🦋 Changeset detected

Latest commit: 05bc073

The changes in this PR will be included in the next version bump.

This PR includes changesets to release 2 packages

Name	Type
@sveltejs/sv-utils	Patch
sv	Patch

Not sure what this means? Click here to learn what changesets are.

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

[jycouet]

Right now, it string-replaces generated output and hardcodes index_d_exports, a name rolldown-plugin-dts owns. I don't think that it's the way to go.

---

I think that we have 3 personas:

• cli devs
• community add-on devs
• cli & add-on users

This issue is annoying for community add-on devs. We could instruct them what to install as devDeps to have a better DX?

• we should probably fix sveltejs/svelte#17520 ASAP this is the REAL way to go.
• We could get rid of tsdown (using rolldown-plugin-dts) and move to JSDoc?

[sacrosanctic]

Oops, did not mean to represent this as a through fix. Also made some adjustment to see if this is a passable short term fix if we're blocked upstream.

[jycouet]

I have to say that a few types inferred in this PR are a nice positive thing.

But it's not really fixing anything and I'm not sure how much is this a "problem" today.
You had an issue with zimmerframe and that's why you want to improve it? Because other libs have the same problem... (because of the same root cause) it's ok to not fix them?
I'm a bit worried that we add code that we don't remember that we can remove!

Anyhow, I'm approving because it's removing a lot of as ___ but I would like to not generalize this, so if you think it's a big deal and other maintainers agree it will be merged.

[sacrosanctic]

It does work a little better...but we still have to cast the first param as AstTypes.Node. There isn't a lot of documentation around walking the ast so getting intellisense on the methods is nice.