### 🚀 Features -8dd67a6transform-react: Expose TypeScript transform options (Tomas Särnhammar) -deed7dcparser: More friendly error for duplicate import attribute key (#28262) (sapphi-red) ### 🐛 Bug Fixes -eaa8e90ecmascript: Avoid numeric `+` folding when an operand may be a string (#28544) (camc314) -b03c7b8regular_expression: Allow incomplete oversized legacy quantifiers (#28449) (Boshen) -128c67dsemantic: Check strict eval and arguments writes by reference flags (#28445) (Boshen) -c97c43cparser: Reject TypeScript default exports in JavaScript (#28440) (Boshen) -fdd119fspan: Count Unicode characters when filtering typo suggestions (#28452) (Boshen) -cb64638regular_expression: Reject incomplete Unicode atom escapes (#28448) (Boshen) -fc0eed3semantic: Stop let-binding scans at arrow functions (#28444) (Boshen) -3421516parser: Preserve default import names in reexports (#28442) (Boshen) -a1e7e5fparser: Reject optional class members in JavaScript (#28439) (Boshen) -5b9a7bcparser: Parse sloppy let expressions and labels (#28438) (Boshen) -8e61652regex: Reject templates without cooked values (#28345) (Boshen) -3b36543parser: Preserve in-context after TypeScript type arguments (#28300) (Boshen) -2e3126cregex: Decode constructor patterns and flags consistently (#28298) (Boshen) -8a90be2regex: Track capture groups and disjunction identity (#28297) (Boshen) -3dc2cb0regex: Preserve escape values and boundaries (#28296) (Boshen) -fc8acbcparser: Handle leading-zero numeric exponents (#28295) (Boshen) -91a638dparser: Normalize raw template line endings with invalid escapes (#28285) (Boshen) -2ac9a42parser: Allow parenthesized private-in expressions (#28284) (Boshen) -052bca9ast: Prevent greyed out fields in AST types in IDEs (#28335) (overlookmotel) -ed7df7dast/estree: Include comments in TS Program span (#28326) (camc314) ### ⚡ Performance -9b6b7bfregular_expression: Advance validated Unicode escapes directly (#28514) (Boshen) -82e1982regular_expression: Share Unicode property name prefixes (#28513) (Boshen) -5395a40regular_expression: Parse class set character prefixes once (#28512) (Boshen) -3708934regular_expression: Defer literal range dash allocation (#28511) (Boshen) -d85ee3dsemantic: Skip arguments ancestry checks outside classes (#28510) (Boshen) -cf81d07semantic: Reuse private reference resolution (#28509) (Boshen) -a79eebcsemantic: Avoid duplicate maps for public JavaScript members (#28508) (Boshen) -235ca3bsemantic: Skip escape checks for unescaped strings (#28507) (Boshen) -babf44cparser: Reuse shorthand assignment target identifiers (#28504) (Boshen) -444f24aecmascript: Short-circuit simple parameter checks for rest (#28503) (Boshen) -04b20c0semantic: Compute duplicate parameter policy once per list (#28446) (Boshen) -dbf8482parser: Use token lookahead for simple function types (#28489) (Boshen) -bca1275parser: Avoid unused destructuring rest capacity (#28499) (Boshen) -691be5cparser: Compact cover initialized name storage (#28495) (Boshen) -b3b5444parser: Build simple arrow bindings directly (#28493) (Boshen) -e1c9bafparser: Disambiguate single-identifier arrow heads (#28491) (Boshen) -550e98aregular_expression: Avoid rescanning unique capture names (#28450) (Boshen) -10e9fc7parser: Skip duplicate lookup for singleton import attributes (#28441) (Boshen) -5238839parser: Use a stack buffer for float separators (#28435) (Boshen) -6e35e3cparser: Fast-path leaf expressions before delimiters (#28432) (Boshen) -3d7512cparser: Decode Unicode identifiers once (#28414) (Boshen) -30f59f4parser: Factor object type opening brace (#28430) (Boshen) -028a267parser: Always inline postfix update parsing (#28429) (Boshen) -1b5a171parser: Reuse export kind classification (#28424) (Boshen) -179a432parser: Factor return type identifier parsing (#28416) (Boshen) -553dd1dparser: Inline hot paths and fold function dispatch (#28404) (Boshen) -de75d56parser: Combine parenthesized arrow lookahead (#28412) (Boshen) -a768a0cparser: Reuse comment annotations after rewind (#28408) (Boshen) -563d92aparser: Skip redundant source cursor updates (#28409) (Boshen) -7941ca9parser: Factor tuple identifier parsing (#28398) (Boshen) -426c151parser: Construct completed tokens without intermediate writes (#28291) (Boshen) -df42953minifier: Don't request another pass for block unwrapping (#28315) (Dunqing) -c2d663dparser: Avoid vectors for single parenthesized expressions (#28374) (Boshen) -50bd0a9parser: Size delimited lists after the first element (#28373) (Boshen) -bd869b5parser: Bypass redundant operand parsing (#28372) (Boshen) -455fc09parser: Skip keyword checks for plain identifier expressions (#28371) (Boshen) -61303fbparser: Fast-path identifier operands (#28306) (Boshen) -6cf5b1aparser: Guard arrow and conditional parsing (#28305) (Boshen) -3666939parser: Specialize identifier reference parsing (#28303) (Boshen) -0d6641aparser: Skip inapplicable expression parsing paths (#28294) (Boshen) -3be7997parser: Reduce lexer search batch size (#28292) (Boshen) -ab5655aregex: Remove redundant character-class and quantifier checks (#28288) (Boshen) -ec8e8d4parser: Remove redundant grammar and lexer checks (#28287) (Boshen) -c903f5cparser: Specialize known-token parsing (#28286) (Boshen) -2683225str: `static_ident!` and `static_str!` macros guarantee const evaluation (#28340) (overlookmotel) -df90836ast: Reduce compile time of `#[ast]` macro (#28336) (overlookmotel) -8239dacparser: Replace dragonbox_ecma with zmij_ecma (#28282) (Boshen)
Oxc Minify
See usage instructions.
This is alpha software and may yield incorrect results, feel free to submit a bug report.
Performance and Compression Size
See minification-benchmarks for details.
The current version already outperforms esbuild,
but it still lacks a few key minification techniques
such as constant inlining and dead code removal,
which we plan to implement next.
Caveats
To maximize performance, oxc-minify assumes the input code is semantically correct.
It uses oxc-parser's fast mode to parse the input code,
which does not check for semantic errors related to symbols and scopes.
API
Functions
// Synchronous minification
minifySync(
filename: string,
sourceText: string,
options?: MinifyOptions,
): MinifyResult
// Asynchronous minification
minify(
filename: string,
sourceText: string,
options?: MinifyOptions,
): Promise<MinifyResult>
Use minifySync for synchronous minification. Use minify for asynchronous minification, which can be beneficial in I/O-bound or concurrent scenarios, though it adds async overhead.
Example
import { minifySync } from "oxc-minify";
const filename = "test.js";
const code = "const x = 'a' + 'b'; console.log(x);";
const options = {
compress: {
target: "esnext",
},
mangle: {
toplevel: false,
},
codegen: {
removeWhitespace: true,
},
sourcemap: true,
};
const result = minifySync(filename, code, options);
// Or use async version: const result = await minify(filename, code, options);
console.log(result.code);
console.log(result.map);
Property-name mangling
Property mangling is opt-in and independent from identifier mangling. include is a JavaScript
RegExp and is required when mangleProps is present. Its source and flags are compiled with
Rust's regex engine. Flags i, m, s, and u are
supported. Other JavaScript flags and unsupported syntax are reported in errors.
const result = minifySync("component.js", source, {
mangle: false,
mangleProps: {
include: /^_/,
exclude: /^__public/,
reserved: ["_externalApi"],
quoted: false,
cache: previousResult?.mangleCache,
},
});
saveCache(result.mangleCache);
The returned mangleCache contains the input cache plus newly assigned names when parsing
finishes without errors, sorted by original name. A false cache value keeps that property
unchanged. Feeding the cache back keeps recorded mappings stable for later unminified input, but
the cache does not discover unchanged names that exist only in another input. Callers coordinating
separate files must pin or reserve those names explicitly. Property mappings are
single-application: do not minify already-mangled output with the same cache. Custom target names
may be shared deliberately, but must be valid JavaScript IdentifierName values and cannot be
__proto__, constructor, or prototype. The original name __proto__ is always reserved and
cannot be used as a cache key.
With quoted: false, quoting is handled per occurrence: obj._field is eligible while
obj["_field"] is not. Property mangling assumes matching properties are never accessed through
arbitrary dynamic strings. Use /* @__KEY__ */ "_field" for a string that semantically names a
property, such as a reflective API argument. Names reached through direct eval, the Function
constructor, or with must be reserved or excluded explicitly.
Properties owned by unminified code, imported module namespaces, globals, DOM objects, or other host APIs must also be excluded or reserved.
Assumptions
oxc-minify makes some assumptions about the source code.
See https://github.com/oxc-project/oxc/blob/main/crates/oxc_minifier/README.md#assumptions for details.
Supports WASM
See https://stackblitz.com/edit/oxc-minify for usage example.