⚠️ No Changeset found

Latest commit: 88f10aa

Merging this PR will not cause a version bump for any packages. If these changes should not result in a new version, you're good to go. If these changes should result in a version bump, you need to add a changeset.

Click here to learn what changesets are, and how to add one.

Click here if you're a maintainer who wants to add a changeset to this PR

Binary Sizes

Commit: 0a5c1bb

Merging this PR will not alter performance

✅ 192 untouched benchmarks

_{Comparing kdy1/swc-es-parser-internal-complete (88f10aa) with main (325170f)}

PR Review: feat(es_parser): complete internal parser wiring without ecma runtime dep

Overall this is a solid PR that systematically wires syntax options (decorators_before_export, export_default_from, allow_super_outside_method, dts, disallow_ambiguous_jsx_like) into swc_es_parser and tightens super context handling. The test coverage is good and the approach aligns with parity goals against swc_ecma_parser.

Code Quality

Good:

• The export default from disambiguation logic at parser.rs:489-494 is well-designed — checking for Str after from to distinguish export default from 'mod' from export default from; (expression).
• The ALLOW_DIRECT_SUPER context flag follows the same pattern as swc_ecma_parser's AllowDirectSuper: set in class/object methods, removed in regular function bodies.
• Extracting contains_swc_ecma_parser_import helper in no_ecma_dependency.rs reduces duplication.
• The typescript.rs change to use matches!(ext, "cts" | "mts") instead of file_name.contains(...) is more precise and avoids false positives from directory names.

Nits:

• parity_suite.rs:215-221 — The matches!(path.as_str(), p if p.ends_with(...) || ...) pattern is valid but a bit unusual. Since the pattern p is an irrefutable catch-all that always binds, all filtering happens in the guard. A simple let p = path.as_str(); p.ends_with(...) || ... reads more naturally and avoids reader confusion about what the matches! is actually pattern-matching on.

Potential Issues

1. super inside object shorthand methods — parse_class_method_function sets ALLOW_DIRECT_SUPER and is called for both class methods and object literal shorthand methods (line 3714). This is correct for super property access in shorthand object methods (which is valid ES6+), but worth documenting since the function name says "class_method" while it's shared with object methods. Consider a comment or rename for clarity.

2. parse_class_method_function context before params — The context change (line 3554-3562) now sets IN_FUNCTION, ALLOW_DIRECT_SUPER, IN_ASYNC, IN_GENERATOR before parsing parameters. The original code set context only before the body. This means parameter default expressions are now parsed with ALLOW_DIRECT_SUPER set, which would allow super.foo inside parameter defaults of class methods. Verify this matches swc_ecma_parser behavior — in the ecma parser, AllowDirectSuper is set at the do_inside_of_context call site which wraps the entire method parse including params, so this should be fine, but it's a subtle semantic change worth double-checking.

3. expr_is_super_reference traverses through parens — At line 5166, (super) is treated as a super reference. This correctly matches the test262 fail case 5864c96b99472ffc.js (function f() { (super)() }). Good.

Performance

• The ensure_super_property_or_call_allowed check is called on every member access and call expression (lines 2857, 2877, 2952, 2967). Since it's a simple flag check that short-circuits on non-super expressions via expr_is_super_reference (which is typically a single match arm), the overhead is negligible.

Test Coverage

• Good coverage for all five syntax options with dedicated unit tests.
• Parity suite correctly marks the super span fixtures as expected failures.
• Dependency guard tests properly verify no runtime dependency on swc_ecma_parser.
• The removed skip entries (stale fixture ignores) indicate these cases now pass, which is a good sign of parser maturation.

Minor Suggestions

• The export default, * as ns from error message (around line 532) mentions the AST limitation. This is fine as a diagnostic but consider whether this should be a TODO for future AST model extension.
• The error message "decorators before export are disabled by syntax option" could be slightly more actionable: "decorators must appear after export (set decorators_before_export to allow this)".

LGTM with the minor suggestions above. The behavioral changes around super handling are well-motivated and align with swc_ecma_parser semantics.