perf(parser): try hybrid parsing for jsx children and closing element/fragments#18789
Conversation
How to use the Graphite Merge QueueAdd either label to this PR to merge it via 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. |
Merging this PR will not alter performance
Comparing Footnotes
|
b1efb06 to
db4eee4
Compare
There was a problem hiding this comment.
Pull request overview
This PR optimizes JSX parsing performance by 1-2% through a refactoring that eliminates the need to rewind parser state. Previously, when encountering a closing fragment, the parser would rewind and call a separate function to parse the closing tag. Now, children and closing tags are parsed in a single pass using the new parse_jsx_children_and_closing function.
Changes:
- Introduced a unified parsing approach that combines JSX children and closing tag parsing in one function
- Improved error diagnostics to be more similar to TypeScript's error messages
- Removed separate
parse_jsx_closing_fragment,parse_jsx_closing_element, andparse_jsx_childfunctions in favor of integrated parsing
Reviewed changes
Copilot reviewed 2 out of 3 changed files in this pull request and generated 1 comment.
| File | Description |
|---|---|
| crates/oxc_parser/src/jsx/mod.rs | Refactored JSX parsing to use hybrid approach with new JSXClosing enum and parse_jsx_children_and_closing function that eliminates parser state rewinding |
| crates/oxc_parser/src/diagnostics.rs | Added new jsx_fragment_no_match diagnostic function for better error messages when fragment closing tags don't match |
| tasks/coverage/snapshots/parser_typescript.snap | Updated test snapshots to reflect improved error diagnostics that provide clearer context for JSX fragment mismatches |
Comments suppressed due to low confidence (1)
crates/oxc_parser/src/jsx/mod.rs:288
- Consider adding an explicit
continuestatement after line 287 for consistency with the other branches in this match statement (lines 254, 260, 280). While the code is functionally correct without it, adding it would make the control flow more explicit and consistent.
Kind::LCurly => {
let span_start = self.start_span();
self.bump_any(); // bump `{`
// {...expr}
if self.eat(Kind::Dot3) {
children.push(JSXChild::Spread(self.parse_jsx_spread_child(span_start)));
continue;
}
// {expr}
children.push(JSXChild::ExpressionContainer(
self.parse_jsx_expression_container(
span_start, /* in_jsx_child */ true,
),
));
}
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
Merge activity
|
…/fragments (#18789) +1-2% on parsing JSX files. <img width="713" height="363" alt="Screenshot 2026-02-03 at 10 25 55 PM" src="https://github.com/user-attachments/assets/693c725d-460c-4bb0-bae4-4e85c44efd53" /> Previously, we were rewinding the parser state whenever we hit a closing fragment and then calling a different function to continue parsing the applicable type of closing tag. Now, we do all of the children and closing tag parsing in one function. This allows us to never rewind the parser state which should be faster. It also allows us to provide slightly better diagnostics as we know what state we are in. Diagnostic changes are expected, it's a little bit more similar to TypeScript's diagnostics: https://www.typescriptlang.org/play/?#code/CYUwxgNghgTiAEA3W8BKIpgC4C55QDsBPAbgCgyAeAPgAsBLSgemHsWviafgFEYYA9jAoUaIAQDN40rr35CgA
db4eee4 to
f78c525
Compare
### 💥 BREAKING CHANGES - 2bf7293 mangler: [**BREAKING**] Enable `top_level` by default for modules and commonjs (#18278) (sapphi-red) - 48b0542 span: [**BREAKING**] SourceType::ts should set module to unambigious (#18873) (Boshen) ### 🚀 Features - 500d071 minifier: Local traverse ctx and generated minifier traverse (#19106) (Boshen) - 142a1be parser: Detect binary files with TS1490 error (#19047) (Boshen) - e316857 allocator/bitset: Add `Ones` iterator to `BitSet` (#19027) (sapphi-red) - 742ad3f minifier: Default `invalid_import_side_effects` to `false` (#18916) (sapphi-red) - 0eff6be parser: Error JSX-like type assertions and generics in `.mts`/`.cts` (#18910) (Boshen) - 18320c6 span: Store file extension in `SourceType` (#18893) (Boshen) ### 🐛 Bug Fixes - a7514e4 isolated-declarations: Preserve const context in literal type inference (#19178) (camc314) - 312e756 isolated-declarations: Preserve readonly literal initializers (#19177) (camc314) - d0ca8d0 isolated-declarations: Skip parenthesis when inferring type (#19176) (camc314) - 110c300 oxc_ecmascript: `+[false]` and `+[true]` should evaluate to `NaN` (#19174) (copilot-swe-agent) - f32ea19 semantic: Report redeclaration error for import bindings conflicting with value declarations (#19068) (Boshen) - 3aeba7a semantic: Report redeclaration error for `function a() {} var a` in module mode (#19041) (Boshen) - 35e32c6 coverage: Match Babel's options.json inheritance for test fixtures (#19002) (Boshen) - 463d60d semantic: Skip TS2391 for standalone computed-name class methods (#19025) (Boshen) - 56c086b parser: Add modifier ordering validation (TS1029) (#19024) (Boshen) - 6067a49 linter/jsdoc: False positive in `check-tag-names` for `@` in email addresses and npm scopes (#19021) (Boshen) - b13bb70 semantic/jsdoc: Inline tags like `{@link}` break jsdoc parsing (#19019) (Boshen) - e3609e3 regular_expression: Preserve UnicodeEscape CharacterKind in string literals (#18998) (Boshen) - 57917ee parser: Parse decorators on rest parameters (#18938) (Boshen) - 487601b napi: Disable mimalloc on Windows to fix worker_threads crash (#18923) (Boshen) - 1f6b193 parser: Validate TypeScript import type options (#18889) (Boshen) - 1663184 parser: Allow conditional types in function type parameters (#18886) (Boshen) - 5758046 parser: Error on property access after instantiation expression (#18887) (Boshen) - 5eb4a94 parser: Handle `<<` as two `<` tokens in type argument contexts (#18885) (Boshen) ### ⚡ Performance - ed8c054 oxc_str: Add precomputed hash to Ident for fast HashMap lookups (#19143) (Boshen) - d4a0867 transformer_plugins: Switch ReplaceGlobalDefines from Traverse to VisitMut (#19146) (Boshen) - 9eb16b3 syntax: Pack ASCII identifier tables into single bitflag table (#19088) (Boshen) - e7595d1 mangler: Use BitSet for exported symbols set (#19023) (sapphi-red) - 2537924 semantic: Optimize scope resolution with fast paths and inlining (#19029) (Boshen) - 69a8d85 mangler: Use BitSet for keep_names symbols set (#19028) (sapphi-red) - f78c525 parser: Try hybrid parsing for jsx children and closing element/fragments (#18789) (camchenry) Co-authored-by: camc314 <[email protected]>
### 💥 BREAKING CHANGES - 2bf7293 mangler: [**BREAKING**] Enable `top_level` by default for modules and commonjs (oxc-project#18278) (sapphi-red) - 48b0542 span: [**BREAKING**] SourceType::ts should set module to unambigious (oxc-project#18873) (Boshen) ### 🚀 Features - 500d071 minifier: Local traverse ctx and generated minifier traverse (oxc-project#19106) (Boshen) - 142a1be parser: Detect binary files with TS1490 error (oxc-project#19047) (Boshen) - e316857 allocator/bitset: Add `Ones` iterator to `BitSet` (oxc-project#19027) (sapphi-red) - 742ad3f minifier: Default `invalid_import_side_effects` to `false` (oxc-project#18916) (sapphi-red) - 0eff6be parser: Error JSX-like type assertions and generics in `.mts`/`.cts` (oxc-project#18910) (Boshen) - 18320c6 span: Store file extension in `SourceType` (oxc-project#18893) (Boshen) ### 🐛 Bug Fixes - a7514e4 isolated-declarations: Preserve const context in literal type inference (oxc-project#19178) (camc314) - 312e756 isolated-declarations: Preserve readonly literal initializers (oxc-project#19177) (camc314) - d0ca8d0 isolated-declarations: Skip parenthesis when inferring type (oxc-project#19176) (camc314) - 110c300 oxc_ecmascript: `+[false]` and `+[true]` should evaluate to `NaN` (oxc-project#19174) (copilot-swe-agent) - f32ea19 semantic: Report redeclaration error for import bindings conflicting with value declarations (oxc-project#19068) (Boshen) - 3aeba7a semantic: Report redeclaration error for `function a() {} var a` in module mode (oxc-project#19041) (Boshen) - 35e32c6 coverage: Match Babel's options.json inheritance for test fixtures (oxc-project#19002) (Boshen) - 463d60d semantic: Skip TS2391 for standalone computed-name class methods (oxc-project#19025) (Boshen) - 56c086b parser: Add modifier ordering validation (TS1029) (oxc-project#19024) (Boshen) - 6067a49 linter/jsdoc: False positive in `check-tag-names` for `@` in email addresses and npm scopes (oxc-project#19021) (Boshen) - b13bb70 semantic/jsdoc: Inline tags like `{@link}` break jsdoc parsing (oxc-project#19019) (Boshen) - e3609e3 regular_expression: Preserve UnicodeEscape CharacterKind in string literals (oxc-project#18998) (Boshen) - 57917ee parser: Parse decorators on rest parameters (oxc-project#18938) (Boshen) - 487601b napi: Disable mimalloc on Windows to fix worker_threads crash (oxc-project#18923) (Boshen) - 1f6b193 parser: Validate TypeScript import type options (oxc-project#18889) (Boshen) - 1663184 parser: Allow conditional types in function type parameters (oxc-project#18886) (Boshen) - 5758046 parser: Error on property access after instantiation expression (oxc-project#18887) (Boshen) - 5eb4a94 parser: Handle `<<` as two `<` tokens in type argument contexts (oxc-project#18885) (Boshen) ### ⚡ Performance - ed8c054 oxc_str: Add precomputed hash to Ident for fast HashMap lookups (oxc-project#19143) (Boshen) - d4a0867 transformer_plugins: Switch ReplaceGlobalDefines from Traverse to VisitMut (oxc-project#19146) (Boshen) - 9eb16b3 syntax: Pack ASCII identifier tables into single bitflag table (oxc-project#19088) (Boshen) - e7595d1 mangler: Use BitSet for exported symbols set (oxc-project#19023) (sapphi-red) - 2537924 semantic: Optimize scope resolution with fast paths and inlining (oxc-project#19029) (Boshen) - 69a8d85 mangler: Use BitSet for keep_names symbols set (oxc-project#19028) (sapphi-red) - f78c525 parser: Try hybrid parsing for jsx children and closing element/fragments (oxc-project#18789) (camchenry) Co-authored-by: camc314 <[email protected]>
### 💥 BREAKING CHANGES - 2bf7293 mangler: [**BREAKING**] Enable `top_level` by default for modules and commonjs (oxc-project#18278) (sapphi-red) - 48b0542 span: [**BREAKING**] SourceType::ts should set module to unambigious (oxc-project#18873) (Boshen) ### 🚀 Features - 500d071 minifier: Local traverse ctx and generated minifier traverse (oxc-project#19106) (Boshen) - 142a1be parser: Detect binary files with TS1490 error (oxc-project#19047) (Boshen) - e316857 allocator/bitset: Add `Ones` iterator to `BitSet` (oxc-project#19027) (sapphi-red) - 742ad3f minifier: Default `invalid_import_side_effects` to `false` (oxc-project#18916) (sapphi-red) - 0eff6be parser: Error JSX-like type assertions and generics in `.mts`/`.cts` (oxc-project#18910) (Boshen) - 18320c6 span: Store file extension in `SourceType` (oxc-project#18893) (Boshen) ### 🐛 Bug Fixes - a7514e4 isolated-declarations: Preserve const context in literal type inference (oxc-project#19178) (camc314) - 312e756 isolated-declarations: Preserve readonly literal initializers (oxc-project#19177) (camc314) - d0ca8d0 isolated-declarations: Skip parenthesis when inferring type (oxc-project#19176) (camc314) - 110c300 oxc_ecmascript: `+[false]` and `+[true]` should evaluate to `NaN` (oxc-project#19174) (copilot-swe-agent) - f32ea19 semantic: Report redeclaration error for import bindings conflicting with value declarations (oxc-project#19068) (Boshen) - 3aeba7a semantic: Report redeclaration error for `function a() {} var a` in module mode (oxc-project#19041) (Boshen) - 35e32c6 coverage: Match Babel's options.json inheritance for test fixtures (oxc-project#19002) (Boshen) - 463d60d semantic: Skip TS2391 for standalone computed-name class methods (oxc-project#19025) (Boshen) - 56c086b parser: Add modifier ordering validation (TS1029) (oxc-project#19024) (Boshen) - 6067a49 linter/jsdoc: False positive in `check-tag-names` for `@` in email addresses and npm scopes (oxc-project#19021) (Boshen) - b13bb70 semantic/jsdoc: Inline tags like `{@link}` break jsdoc parsing (oxc-project#19019) (Boshen) - e3609e3 regular_expression: Preserve UnicodeEscape CharacterKind in string literals (oxc-project#18998) (Boshen) - 57917ee parser: Parse decorators on rest parameters (oxc-project#18938) (Boshen) - 487601b napi: Disable mimalloc on Windows to fix worker_threads crash (oxc-project#18923) (Boshen) - 1f6b193 parser: Validate TypeScript import type options (oxc-project#18889) (Boshen) - 1663184 parser: Allow conditional types in function type parameters (oxc-project#18886) (Boshen) - 5758046 parser: Error on property access after instantiation expression (oxc-project#18887) (Boshen) - 5eb4a94 parser: Handle `<<` as two `<` tokens in type argument contexts (oxc-project#18885) (Boshen) ### ⚡ Performance - ed8c054 oxc_str: Add precomputed hash to Ident for fast HashMap lookups (oxc-project#19143) (Boshen) - d4a0867 transformer_plugins: Switch ReplaceGlobalDefines from Traverse to VisitMut (oxc-project#19146) (Boshen) - 9eb16b3 syntax: Pack ASCII identifier tables into single bitflag table (oxc-project#19088) (Boshen) - e7595d1 mangler: Use BitSet for exported symbols set (oxc-project#19023) (sapphi-red) - 2537924 semantic: Optimize scope resolution with fast paths and inlining (oxc-project#19029) (Boshen) - 69a8d85 mangler: Use BitSet for keep_names symbols set (oxc-project#19028) (sapphi-red) - f78c525 parser: Try hybrid parsing for jsx children and closing element/fragments (oxc-project#18789) (camchenry) Co-authored-by: camc314 <[email protected]>

+1-2% on parsing JSX files.
Previously, we were rewinding the parser state whenever we hit a closing fragment and then calling a different function to continue parsing the applicable type of closing tag. Now, we do all of the children and closing tag parsing in one function. This allows us to never rewind the parser state which should be faster. It also allows us to provide slightly better diagnostics as we know what state we are in.
Diagnostic changes are expected, it's a little bit more similar to TypeScript's diagnostics: https://www.typescriptlang.org/play/?#code/CYUwxgNghgTiAEA3W8BKIpgC4C55QDsBPAbgCgyAeAPgAsBLSgemHsWviafgFEYYA9jAoUaIAQDN40rr35CgA