feat: loose parser mode

[dummdidumm]

This implements a new option for the parser, loose, which means the parser is forgiving with the syntax and tries to continue parsing even if it encounters broken code. It does not parse no matter what, but handles the most common "in the middle of typing" scenarios gracefully.

closes #4818

Our language tools can then make use of this to provide better intellisense in more situations.

Before submitting the PR, please make sure you do the following

• It's really useful if your PR references an issue where it is discussed ahead of time. In many cases, features are absent for a reason. For large changes, please create an RFC: https://github.com/sveltejs/rfcs
• Prefix your PR title with feat:, fix:, chore:, or docs:.
• This message body should clearly illustrate what problems it solves.
• Ideally, include a test that fails without this PR but passes with it.
• If this PR changes code within packages/svelte/src, add a changeset (npx changeset).

Tests and linting

• Run the tests with pnpm test and lint the project with pnpm lint

[changeset-bot]

🦋 Changeset detected

Latest commit: e079744

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

This PR includes changesets to release 1 package

Name	Type
svelte	Minor

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

[Rich-Harris]

preview: https://svelte-dev-git-preview-svelte-14691-svelte.vercel.app/

this is an automated message

[github-actions]

Playground

pnpm add https://pkg.pr.new/svelte@14691

[Rich-Harris]

Should we just always parse in loose mode, and push errors to an array as we go? Then, if loose !== true, at the end of parsing we throw errors[0], otherwise the errors are accessible just as warnings are today, which feels necessary. Might that simplify things?

[dummdidumm]

For language tools I'm not interested in where the first error happened. Logic-wise, it's a few if blocks here and there now, and trading those for not throwing directly in certain cases would not give us much (i.e. we trade less code in one place for almost the same additional code in another). So I vote for keeping it as it's now.

[Rich-Harris]

Wouldn't you want to show red squigglies for all the errors? And surely if you don't capture the errors during loose parse you need to do a separate strict parse to show just the first error, so we're not even saving any work?

[dummdidumm]

• I would only want the very first error. All other ones might be wrong since the parser might be in a different state than it should be
• I have to do a separate parse anyway. I have to run the the compiler, not the parser, in order to get all warnings etc. - and again, the parser could parse things wrong and so the compiler gives wrong warnings. It might also be that the compiler isn't equipped to handle the somewhat broken AST and throws. So even if there is at some point a way for the compiler to receive the AST to save some work, I would only want to go that step when I'm sure that the parser succeeded.

Bottom line, in the future it might make sense to have an error property which returns the very first error in case of loose parsing, but that's about it.