perf(parser): use SmallVec for duplicate default export detection

[camc314]

Avoid heap allocation for the common case (0-2 default exports) by using
SmallVec with inline capacity of 2 instead of Vec.

🤖 generated with help from Claude Opus 4.5

[camc314]

• perf(parser): use SmallVec for duplicate default export detection #16801 👈 (View in Graphite)
• main

---

How to use the Graphite Merge Queue

Add either label to this PR to merge it via the merge queue:

• 0-merge - adds this PR to the back of the merge queue
• hotfix - for urgent hot fixes, skip the queue and merge this PR next

You must have a Graphite account in order to use the merge queue. Sign up using this link.

An organization admin has enabled the Graphite Merge Queue in this repository.

Please do not merge from GitHub as this will restart CI on PRs being processed by the merge queue.

This stack of pull requests is managed by Graphite. Learn more about stacking.

[codspeed-hq]

CodSpeed Performance Report

Merging #16801 will not alter performance

_{Comparing c/12-13-refactor_parser_use_smallvec_for_duplicate_default_export_detection (7856ab2) with main (3f5d6b7)}

Summary

✅ 42 untouched
⏩ 3 skipped¹

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. ↩

[charliecreates]

Overall the change is small and reasonable, but the current SmallVec usage still eagerly collects all default-export spans. If you want the perf tweak to be maximally effective, consider short-circuiting once duplication is detected (or only collecting all labels when needed).

Additional notes (2)

• Performance | crates/oxc_parser/src/module_record.rs:57-63
SmallVec is introduced for a simple len() > 1 check, but this still collects all default-export spans. If a module has many default exports (pathological, but possible), this will still iterate and store everything (and will heap-allocate once it exceeds inline capacity). Since the only behavior difference is emitting an error with labels, consider an approach that short-circuits after the second hit for the common/valid case while still producing good diagnostics (e.g., collect first span, detect second, then optionally continue collecting only if you want all labels). This keeps the perf win but avoids extra work in invalid cases too.

• Performance | crates/oxc_parser/src/module_record.rs:57-63
  The change to SmallVec is fine, but with_labels(default_exports) will typically consume an IntoIterator of labels. It’s safer and clearer to pass an iterator (or slice) explicitly, rather than relying on SmallVec’s trait impls staying compatible with the diagnostics API. This also avoids accidentally moving default_exports if future code wants to use it after with_labels.

Also, this function only needs to know whether there are 2+ defaults; collecting all spans is unnecessary work in degenerate cases with many exports. You can short-circuit once you’ve found a second default, while still preserving the labels you want to show.

Summary of changes

What changed

• Added smallvec as a dependency for oxc_parser (crates/oxc_parser/Cargo.toml).
• Switched duplicate default export span collection in ModuleRecordBuilder from std::vec::Vec to SmallVec<[Span; 2]> to avoid heap allocation in the common case (crates/oxc_parser/src/module_record.rs).
• Updated imports to include use smallvec::SmallVec;.

[Copilot]

Pull request overview

This PR optimizes memory allocation in the parser's duplicate default export detection by using SmallVec<[Span; 2]> instead of std::vec::Vec<_>. This avoids heap allocation for the common case where modules have 0-2 default exports (the typical scenario being 0 or 1 valid export).

Key Changes:

• Replaced Vec with SmallVec with inline capacity of 2 for collecting default export spans
• Added smallvec dependency to oxc_parser crate

Reviewed changes

Copilot reviewed 2 out of 3 changed files in this pull request and generated no comments.

File	Description
crates/oxc_parser/src/module_record.rs	Changed collection type from Vec to SmallVec<[Span; 2]> for default export spans
crates/oxc_parser/Cargo.toml	Added smallvec workspace dependency
Cargo.lock	Updated lock file to reflect new dependency

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Boshen]

default_exports is not used elsewhere, why not .count() 🤔 ?

[camc314]

default_exports is not used elsewhere, why not .count() 🤔 ?

we use it for the labels in L63

[Boshen]

@CharlieHelps change to use .count()?

[charliecreates]

Expand this to see my work.

• Previously refactored ModuleRecordBuilder::errors in crates/oxc_parser/src/module_record.rs to remove SmallVec, avoid allocations in the 0–1 default export case, and verified via cargo test -p oxc_parser --all-features and cargo ck that the parser and workspace still build and pass tests, then pushed commits 045d54a8a and 66081aa06 to the branch c/12-13-refactor_parser_use_smallvec_for_duplicate_default_export_detection referenced by PR #16801.
• Implemented the reviewer-suggested .count() approach by introducing a default_export_spans iterator factory in module_record.rs and using default_export_spans().take(2).count() > 1 to detect duplicates before calling OxcDiagnostic::error("Duplicated default export").with_labels(default_export_spans()).
• Ran rg over .github/workflows to confirm there’s no cargo fmt gate in CI, ensuring the new default_export_spans closure style won’t trip formatting checks.
• Re-ran cargo ck after the .count() refactor and confirmed from the output that crates/oxc_parser and the rest of the workspace compile cleanly with the new duplicate default export logic.
• Executed cargo test -p oxc_parser --all-features and confirmed all 54 tests, including module_record::module_record_tests::* covering export behavior, pass with the .count()-based implementation.
• Committed and pushed commit 102071a2c (perf(parser): use count() to detect duplicate default exports) updating only crates/oxc_parser/src/module_record.rs on branch c/12-13-refactor_parser_use_smallvec_for_duplicate_default_export_detection, so PR #16801 now reflects the .count() approach Boshen asked about.

[charliecreates]

#16801 (comment)

Updated the PR to use .take(2).count() (vs collecting) to detect duplicate default exports, while still keeping the labels for the error case. Also dropped the smallvec dep since it’s no longer needed.

Changes

• Switched duplicate-default detection to default_export_spans().take(2).count() > 1, and only builds labels when we actually have a duplicate (102071a2c, crates/oxc_parser/src/module_record.rs around L35-L70).
• Removed smallvec from oxc_parser and the corresponding Cargo.lock entry (045d54a8).

Verification

# Cargo check: ok
$ cargo ck

# Tests (oxc_parser): 54 passed
$ cargo test -p oxc_parser --all-features

Skipped cargo test --workspace --all-features here because it fails on this devbox due to an oxc_codegen snapshot that includes the local Node.js version (unrelated to the parser change).

[Boshen]

Merge activity

• Dec 15, 3:38 PM UTC: The merge label '0-merge' was detected. This PR will be added to the Graphite merge queue once it meets the requirements.
• Dec 15, 3:48 PM UTC: Boshen added this pull request to the Graphite merge queue.
• Dec 15, 3:57 PM UTC: Merged by the Graphite merge queue.