fix(analyze & lsp): code action and text mutation

[Sec-ant]

Summary

This PR fixes the unexpected code deletion and repetition when quickfix.biome is enabled and some import-related rules are applied.

This PR includes 3 fixes:

1. Reverse the order of code actions, so the previous code action won't affect its following action.
2. In code actions of rules noUnusedImports and noRedundantUseStrict, remove the trivia of the import statements or "use strict" directives instead of transferring them to their siblings.
3. Calculate TextEdits from CommitChanges directly, instead of .commit().to_string() and diffing the entire code string.

To make this PR code review experience friendly, I grouped the changes into several commits:

1. fix(code_actions): reverse order: 0bf3f09
   Reverse the order of code actions, so actions will be applied from back to front.
   This will make sure following code actions won't be invalidated by their previous code actions.

2. fix(rules): remove leading trivia instead of transferring them: 4f1e9cd
   This helps to get the correct result when applying code actions one by one.
   This can also avoid potential problems where comment directives like @ts-expect-error being transferred to an unexpected place.

3. fix(text_range_and_text_edit): refactor and reimplementation: 22780e8
   The core logic is in file crates/biome_rowan/src/ast/batch.rs. This commit reimplements how we calculate the text range and text edit from batch mutations:

   • SyntaxSlot::Empty now contains an index position of the empty slot. This is to aid the calculation of text edit from commit changes.
   • TextEdit has a new public method with_unicode_words_diff. This is to aid the calculation of text edit from commit changes.
   • The key of CommitChange that will affect its order in a sorted list is changed. This is to ensure changes will be addressed from back to front.
   • Add an is_from_action field to CommitChange. This is to distinguish whether a commit change is from the action or is propagated from the children. This is to aid the calculation of text edit from commit changes.
   • Rename as_text_edits to as_text_range_and_edit for clarification.
   • The core logic of commit is moved into commit_with_text_range_and_edit. That's because when calculating the text range and text edit, we need to commit the changes. So they should be addressed together.
   • When the new tree, text range and text edit are all required, we should use commit_with_text_range_and_edit instead of commit and as_text_range_and_edit separately to avoid unncessary clone. That's because when calculating the text range and text edit, we also need to commit the changes. This should be done in one pass.

4. chore: update test snapshots: 41bc14c
   Most of the updates are the result of the reimplementation of how we calculate the text edits.
   Previously, the text edit is acquired from a naive text diff of the code before and after commit.
   We now populate the text edit struct manually when applying changes.

   Some of the updates are the reuslt of discarding leading trivia when removing a node.

5. chore: regenerate web docs：40cde73
   Regenerate the web docs to reflect the changes.

6. chore: update changelog: 434627f
   Update the changelog to reflect the fix.

7. chore: update changelog: db8b4cf
   Unlink some issues in the changelog.

Test Plan

CI should pass. There'll be many snapshot updates (see above for the reasons). But the final result should be correct.

Code actions are tested locally with VSCode:

1. noUnusedImports

   Video Demos

   before	after
   no-unused-imports-before.mp4	no-unused-imports-after.mp4
   no-unused-imports-before.mp4	no-unused-imports-after.mp4

2. useImportType

   Video Demos

   before	after
   use-import-type-before.mp4	use-import-type-after.mp4

I cannot reproduce others from the linked issues. Please provide the code and settings (include biome and vscode) if someone has a stable minimal reproduction.

[netlify]

✅ Deploy Preview for biomejs ready!

Name	Link
🔨 Latest commit	b4679d3
🔍 Latest deploy log	https://app.netlify.com/sites/biomejs/deploys/660bc3ffe30a8800081ef175
😎 Deploy Preview	https://deploy-preview-2237--biomejs.netlify.app
📱 Preview on mobile	Toggle QR Code... Use your smartphone camera to open QR code link.
Lighthouse	1 paths audited Performance: 100 (no change from production) Accessibility: 97 (no change from production) Best Practices: 100 (no change from production) SEO: 93 (no change from production) PWA: - View the detailed breakdown and full score reports

---

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

[codspeed-hq]

CodSpeed Performance Report

Merging #2237 will improve performances by 7.51%

_{⚠️ No base runs were found}

_{Falling back to comparing Sec-ant:fix/code-action-mutation (b4679d3) with main (2950aeb)}

Summary

⚡ 1 improvements
✅ 92 untouched benchmarks

Benchmarks breakdown

	Benchmark	main	Sec-ant:fix/code-action-mutation	Change
⚡	big5-added.json[cached]	2.4 ms	2.2 ms	+7.51%

[Sec-ant]

!bench_cli

[Sec-ant]

I unlinked #1550 and #1570 because I'm not sure if this PR will fix them. I cannot reproduce them either.

[Sec-ant]

The only concern I have is the regression in readability of the diffs. I would like to hear if you have any ideas on how to address this in the future.

I wouldn't say it is a regression because these kind of diffs already exist in our existing code base. This PR merely changes some of them into another form. I also don't think the solution to address them has anything to do with this PR, but I can take a look into the problem in the future 😉.

Also, have you tested your change on a large repository to see if there is a perf regression or, conversely, a perf gain?

I did attempt to use the command in #2237 (comment) to get a benchmark. I was expecting there would be a report commented by a robot, but I got nothing 😆. What's the correct way to benchmark things?

[Conaclos]

I did attempt to use the command in #2237 (comment) to get a benchmark. I was expecting there would be a report commented by a robot, but I got nothing 😆. What's the correct way to benchmark things?

If I remember correctly, these magic coments were removed when we added automated benchmark using Codspeed.

[Sec-ant]

The release build of this PR has a small perf regression (~%4) when running biome lint compared to the release build on main. I tested the CLI in the TypeScript repo (tests files ignored):

But no regression in biome ci:

[ematipico]

Out of curiosity, how did you compile the binary? We have some settings that we push during the CI

[Sec-ant]

Out of curiosity, how did you compile the binary? We have some settings that we push during the CI

Isn't it cargo build --release 🤯?

[ematipico]

If both benchmarks run on the same built artifact, then it should be fine.

During the CI we pass some environment variables: https://github.com/biomejs/biome/blob/main/.github%2Fworkflows%2Frelease_cli.yml#L128-L132

[Sec-ant]

Oh yes I didn't set the RUSTFLAGS but I used the same command to build these two binaries, one from the main branch, the other from this PR branch (rebased on main). So they may not reflect the real performance of a published version.

[Sec-ant]

When debug symbols are stripped, the performance regression is reduced to less than %1.

[Sec-ant]

@ematipico Do you have other suggestions or can I merge this?

[ematipico]

The changes look good to me; I left only a few nits. That was some hard work you poured into this PR @Sec-ant, fantastic! 💪

[Sec-ant]

Should be good to go. I'm gonna merge this now!