perf: bump to sugar_path@2 for SIMD-based search and on-demand memory allocation#8432
Conversation
How to use the Graphite Merge QueueAdd the label graphite: merge-when-ready to this PR to add it to the merge queue. 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. |
✅ Deploy Preview for rolldown-rs canceled.
|
There was a problem hiding this comment.
Pull request overview
This pull request upgrades the sugar_path crate from version 1.2.1 to 2.0.0, which introduces a breaking API change where the .normalize() method now returns Cow<'_, Path> instead of PathBuf. This is a performance optimization that allows the method to return a borrowed reference when normalization doesn't change the path.
Changes:
- Bumped
sugar_pathdependency from 1.2.1 to 2.0.0 in Cargo.toml - Updated all call sites to handle the new
Cow<Path>return type by splitting chained operations and adding.into_owned()where ownedPathBufinstances are required - Cargo.lock updated with transitive dependency updates (windows-sys 0.59.0 → 0.61.2, new memchr dependency for sugar_path)
Reviewed changes
Copilot reviewed 15 out of 16 changed files in this pull request and generated no comments.
Show a summary per file
| File | Description |
|---|---|
| Cargo.toml | Version bump of sugar_path from 1.2.1 to 2.0.0 |
| Cargo.lock | Dependency tree updates including sugar_path version, windows-sys, and new memchr dependency |
| crates/rolldown_resolver/src/resolver.rs | Split chained .join().normalize() into separate statements for compatibility |
| crates/rolldown_plugin_vite_resolve/src/vite_resolve_plugin.rs | Split chained operation and added intermediate variable |
| crates/rolldown_plugin_vite_build_import_analysis/src/ast_visit.rs | Split chained operation to handle Cow return type |
| crates/rolldown_plugin_vite_asset_import_meta_url/src/lib.rs | Split operations in two locations, one with simplified inline chaining |
| crates/rolldown_plugin_utils/src/check_public_file.rs | Added .into_owned() to convert Cow to owned PathBuf |
| crates/rolldown_common/src/inner_bundler_options/types/tsconfig.rs | Added .into_owned() for owned PathBuf storage |
| crates/rolldown_binding/src/options/plugin/config/binding_vite_transform_plugin_config.rs | Added .into_owned() for owned PathBuf storage |
| crates/rolldown_binding/src/options/plugin/config/binding_vite_reporter_plugin_config.rs | Added .into_owned() for owned PathBuf storage |
| crates/rolldown/src/stages/generate_stage/mod.rs | Added .into_owned() in two locations for absolutize_with results |
| crates/rolldown/examples/watch_new.rs | Added .into_owned() for Option storage |
| crates/rolldown/examples/watch.rs | Added .into_owned() for Option storage |
| crates/rolldown/examples/lazy.rs | Added .into_owned() for Option storage |
| crates/rolldown/examples/dev.rs | Added .into_owned() for Option storage |
| crates/rolldown/examples/basic.rs | Added .into_owned() for Option storage |
Merging this PR will improve performance by 9.59%
Performance Changes
Comparing |
0a20da3 to
b2179f3
Compare
Benchmarks Rust |
Merge activity
|
## Summary Bump `sugar_path` from 1.2.1 to 2^ ### What changed in sugar_path 2.0 The 2.0 release is focused on reducing allocations in hot paths. The key optimizations: - **`normalize()` returns `Cow<'_, Path>` instead of `PathBuf`** — a `needs_normalization()` fast-path check (using `memchr`) detects already-clean paths and returns `Cow::Borrowed` with zero allocation ([#32](hyf0/sugar_path#32)) - **`absolutize()` / `absolutize_with()` return `Cow<'_, Path>`** — same idea: already-absolute clean paths are returned borrowed ([#34](hyf0/sugar_path#34)) - **`memchr`-accelerated fast path for `relative()`** — replaces the component-iterator approach with SIMD-accelerated `/` scanning, avoids the `absolutize()` → `current_dir()` syscall when both paths are already absolute, and uses `SmallVec<[&str; 8]>` to stay on the stack ([#27](hyf0/sugar_path#27)) - **Reduced allocations across the board** — reuse buffers, `SmallVec` for component lists, avoid `collect()` into `Vec` ([#26](hyf0/sugar_path#26)) ### Breaking change `normalize()`, `absolutize()`, and `absolutize_with()` now return `Cow<'_, Path>` instead of `PathBuf`. Call sites that need an owned `PathBuf` require `.into_owned()`, and chained operations like `.join().normalize().to_slash_lossy()` need to be split so the intermediate `Cow` lives long enough. ## Test plan - [x] CI passes (same API surface, just `Cow` unwrapping at call sites)
b712bda to
7c62b0e
Compare
sugar_path@2sugar_path@2 for SIMD-based search and on-demand memory allocation
## [1.0.0-rc.6] - 2026-02-26 ### 💥 BREAKING CHANGES - css: remove `css_entry_filenames` , `css_chunk_filenames` and related code (#8402) by @hyf0 - css: drop builtin CSS bundling to explore alternative solutions (#8399) by @hyf0 ### 🚀 Features - rust/data-url: use hash as id for data url modules to prevent long string overhead (#8420) by @hyf0 - validate bundle stays within output dir (#8441) by @sapphi-red - rust: support `PluginOrder::PinPost` (#8417) by @hyf0 - support `ModuleType:Copy` (#8407) by @hyf0 - expose `ESTree` types from `rolldown/utils` (#8400) by @sapphi-red ### 🐛 Bug Fixes - incorrect sourcemap when postBanner/postFooter is used with shebang (#8459) by @Copilot - resolver: disable node_path option to align ESM resolver behavior (#8472) by @sapphi-red - parse `.js` within `"type": "commonjs"` as ESM for now (#8470) by @sapphi-red - case-insensitive filename conflict detection for chunk deduplication (#8458) by @Copilot - prevent inlining CJS exports that are mutated by importers (#8456) by @IWANABETHATGUY - parse `.cjs` / `.cts` / `.js` within `"type": "commonjs"` as CommonJS (#8455) by @sapphi-red - plugin/copy-module: correct hooks' priority (#8423) by @hyf0 - plugin/chunk-import-map: ensure `render_chunk_meta` run after users plugin (#8422) by @hyf0 - rust: correct hooks order of `DataUriPlugin` (#8418) by @hyf0 - `jsx.preserve` should also considering tsconfig json preserve (#8324) by @IWANABETHATGUY - `deferred_scan_data.rs "Should have resolved id: NotFound"` error (#8379) by @sapphi-red - cli: require value for `--dir`/`-d` and `--file`/`-o` (#8378) by @Copilot - dev: avoid mutex deadlock caused by inconsistent lock order (#8370) by @sapphi-red ### 🚜 Refactor - watch: rename TaskStart/TaskEnd to BundleStart/BundleEnd (#8463) by @hyf0 - rust: rename `rolldown_plugin_data_uri` to `rolldown_plugin_data_url` (#8421) by @hyf0 - bindingify-build-hook: extract helper for PluginContextImpl (#8438) by @ShroXd - give source loading a proper name (#8436) by @IWANABETHATGUY - ban holding DashMap refs across awaits (#8362) by @sapphi-red ### 📚 Documentation - add glob pattern usage example to input option (#8469) by @IWANABETHATGUY - remove `https://rolldown.rs` from links in reference docs (#8454) by @sapphi-red - mention execution order issue in `output.codeSplitting` docs (#8452) by @sapphi-red - clarify `output.comments` behavior a bit (#8451) by @sapphi-red - replace npmjs package links with npmx.dev (#8439) by @Boshen - reference: add `Exported from` for values / types exported from subpath exports (#8394) by @sapphi-red - add JSDocs for APIs exposed from subpath exports (#8393) by @sapphi-red - reference: generate reference pages for APIs exposed from subpath exports (#8392) by @sapphi-red - avoid pipe character in codeSplitting example to fix broken rendering (#8391) by @IWANABETHATGUY ### ⚡ Performance - avoid redundant PathBuf allocations in resolve paths (#8435) by @Brooooooklyn - bump to `sugar_path@2` (#8432) by @hyf0 - use flag-based convergence detection in include_statements (#8412) by @Brooooooklyn ### 🧪 Testing - execute `_test.mjs` even if `executeOutput` is false (#8398) by @sapphi-red - add retry to tree-shake/module-side-effects-proxy4 as it is flaky (#8397) by @sapphi-red - avoid `expect.assertions()` as it is not concurrent test friendly (#8383) by @sapphi-red - disable `mockReset` option (#8382) by @sapphi-red - fix flaky failure caused by concurrent resolveId calls (#8381) by @sapphi-red ### ⚙️ Miscellaneous Tasks - deps: update dependency rollup to v4.59.0 [security] (#8471) by @renovate[bot] - ai/design: add design doc about watch mode (#8453) by @hyf0 - deps: update oxc resolver to v11.19.0 (#8461) by @renovate[bot] - ai: introduce progressive spec-driven development pattern (#8446) by @hyf0 - deprecate output.legalComments (#8450) by @sapphi-red - deps: update dependency oxlint-tsgolint to v0.15.0 (#8448) by @renovate[bot] - ai: make CLAUDE.md a symlink of AGENTS.md (#8445) by @hyf0 - deps: update rollup submodule for tests to v4.59.0 (#8433) by @sapphi-red - deps: update test262 submodule for tests (#8434) by @sapphi-red - deps: update oxc to v0.115.0 (#8430) by @renovate[bot] - deps: update oxc apps (#8429) by @renovate[bot] - deps: update npm packages (#8426) by @renovate[bot] - deps: update rust crate owo-colors to v4.3.0 (#8428) by @renovate[bot] - deps: update github-actions (#8424) by @renovate[bot] - deps: update rust crates (#8425) by @renovate[bot] - deps: update oxc resolver to v11.18.0 (#8406) by @renovate[bot] - deps: update dependency oxlint-tsgolint to v0.14.2 (#8405) by @renovate[bot] - ban `expect.assertions` in all fixture tests (#8395) by @sapphi-red - deps: update oxc apps (#8389) by @renovate[bot] - ban `expect.assertions` in fixture tests (#8387) by @sapphi-red - enable lint for `_config.ts` files (#8386) by @sapphi-red - deps: update dependency oxlint-tsgolint to v0.14.1 (#8385) by @renovate[bot] Co-authored-by: shulaoda <[email protected]>

Summary
Bump
sugar_pathfrom 1.2.1 to 2^What changed in sugar_path 2.0
The 2.0 release is focused on reducing allocations in hot paths. The key optimizations:
normalize()returnsCow<'_, Path>instead ofPathBuf— aneeds_normalization()fast-path check (usingmemchr) detects already-clean paths and returnsCow::Borrowedwith zero allocation (#32)absolutize()/absolutize_with()returnCow<'_, Path>— same idea: already-absolute clean paths are returned borrowed (#34)memchr-accelerated fast path forrelative()— replaces the component-iterator approach with SIMD-accelerated/scanning, avoids theabsolutize()→current_dir()syscall when both paths are already absolute, and usesSmallVec<[&str; 8]>to stay on the stack (#27)SmallVecfor component lists, avoidcollect()intoVec(#26)Breaking change
normalize(),absolutize(), andabsolutize_with()now returnCow<'_, Path>instead ofPathBuf. Call sites that need an ownedPathBufrequire.into_owned(), and chained operations like.join().normalize().to_slash_lossy()need to be split so the intermediateCowlives long enough.Test plan
Cowunwrapping at call sites)