perf(semantic): V8-style walk-up reference resolution

[Boshen]

Summary

• Replace bubble-up reference resolution with V8-style walk-up algorithm
• References are collected in a flat Vec<(Ident, ReferenceId)> during traversal instead of per-scope hashmaps
• After full AST visit, all references are resolved in a single pass by walking up the scope chain from each reference's scope
• leave_scope() no longer does any reference resolution work — only direct_eval flag propagation remains
• Early resolution via checkpoint mechanism preserved for function parameters and catch parameters (prevents binding to body variables that share the same scope)
• Deduplicated walk-up logic into a single #[inline(always)] helper shared by both normal and early resolution paths
• Uses to_vec+truncate instead of split_off to preserve Vec capacity and avoid reallocation
• Removes smallvec and oxc_data_structures dependencies from oxc_semantic

Why this is faster

Operation	Bubble-up (old)	Walk-up (new)
Per reference creation	Hashmap insert	Vec push (O(1) amortized)
Per scope exit	Drain + merge hashmap	Nothing
Per reference resolution	Implicit via merging	Walk scope chain, O(depth) lookups
Total hashmap mutations	O(refs × depth) inserts+drains	0 during traversal

Hashmap lookups (read-only) are cheaper than hashmap drain+insert (mutating, allocating). The flat Vec also eliminates creating/recycling per-depth hashmaps entirely.

Snapshot changes

3 jest linter rule snapshots have diagnostic reordering (same diagnostics, different emission order) due to references being processed in insertion order rather than scope-by-scope.

🤖 Generated with Claude Code

[codspeed-hq]

Merging this PR will improve performance by 6.01%

⚡ 4 improved benchmarks
✅ 49 untouched benchmarks
⏩ 3 skipped benchmarks¹

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
⚡	Simulation	semantic[binder.ts]	3 ms	2.9 ms	+5.64%
⚡	Simulation	semantic[react.development.js]	1.2 ms	1.1 ms	+6.01%
⚡	Simulation	semantic[RadixUIAdoptionSection.jsx]	62.1 µs	59.3 µs	+4.73%
⚡	Simulation	pipeline[binder.ts]	15 ms	14.4 ms	+3.82%

---

_{Comparing refactor/walk-up-reference-resolution (ddfa853) with main (5c97b14)²}

Footnotes

1. 3 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩

2. No successful run was found on main (68fb0d0) during the generation of this report, so 5c97b14 was used instead as the comparison base. There might be some changes unrelated to this pull request in this report. ↩

[Boshen]

POC - please redo this PR if necessary.

[Copilot]

Pull request overview

This PR refactors oxc_semantic reference resolution from a per-scope “bubble-up” hashmap merge strategy to a V8-style “walk-up” strategy, aiming to reduce allocation and mutation overhead during traversal while preserving early-resolution correctness for parameters/catch bindings.

Changes:

• Collect unresolved references into a flat Vec<(Ident, ReferenceId)> during AST traversal and resolve them in a single post-traversal pass by walking up the scope chain.
• Preserve early resolution via a checkpoint mechanism for function parameters and catch parameters.
• Remove smallvec and oxc_data_structures dependencies from oxc_semantic and update associated snapshots (coverage, memory tracking, linter).

Reviewed changes

Copilot reviewed 9 out of 12 changed files in this pull request and generated 4 comments.

Show a summary per file

File	Description
tasks/track_memory_allocations/allocs_semantic.snap	Updates allocation tracking snapshot after semantic perf changes.
tasks/track_memory_allocations/allocs_minifier.snap	Updates allocation tracking snapshot after semantic perf changes.
tasks/coverage/snapshots/semantic_typescript.snap	Updates TS semantic coverage snapshot reflecting new reference processing order.
tasks/coverage/snapshots/semantic_misc.snap	Updates misc semantic coverage snapshot reflecting new reference processing order.
crates/oxc_semantic/src/unresolved_stack.rs	Replaces per-scope unresolved stack with a flat unresolved reference list + checkpoints.
crates/oxc_semantic/src/scoping.rs	Adjusts root unresolved reference recording API to match new resolution flow.
crates/oxc_semantic/src/builder.rs	Implements walk-up batch resolution and checkpoint-based early resolution.
crates/oxc_semantic/Cargo.toml	Drops smallvec and oxc_data_structures deps from oxc_semantic.
crates/oxc_linter/src/snapshots/jest_valid_describe_callback.snap	Snapshot reorder due to changed diagnostic emission order.
crates/oxc_linter/src/snapshots/jest_prefer_lowercase_title@jest.snap	Snapshot reorder due to changed diagnostic emission order.
crates/oxc_linter/src/snapshots/jest_no_confusing_set_timeout.snap	Snapshot reorder due to changed diagnostic emission order.
Cargo.lock	Lockfile updates reflecting removed dependencies.

---

You can also share your feedback on Copilot code review. Take the survey.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: e022db3f66

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[Dunqing]

5604bcb change improved 3% performance in each benchmark 😅

[Dunqing]

Pretty cool! Less dependency, more performant, and much cleaner code now!

[Dunqing]

Merge activity

• Mar 13, 6:20 AM UTC: The merge label '0-merge' was detected. This PR will be added to the Graphite merge queue once it meets the requirements.
• Mar 13, 6:20 AM UTC: Dunqing added this pull request to the Graphite merge queue.
• Mar 13, 6:31 AM UTC: Merged by the Graphite merge queue.