@esbuild/freebsd-x64

• Disallow \ in local development server HTTP requests (GHSA-g7r4-m6w7-qqqr)

  This release fixes a security issue where HTTP requests to esbuild's local development server could traverse outside of the serve directory on Windows using a \ backslash character. It happened due to the use of Go's path.Clean() function, which only handles Unix-style / characters. HTTP requests with paths containing \ are no longer allowed.

  Thanks to @dellalibera for reporting this issue.

• Add integrity checks to the Deno API (GHSA-gv7w-rqvm-qjhr)

  The previous release of esbuild added integrity checks to esbuild's npm install script. This release also adds integrity checks to esbuild's Deno install script. Now esbuild's Deno API will also fail with an error if the downloaded esbuild binary contains something other than the expected content.

  Note that esbuild's Deno API installs from registry.npmjs.org by default, but allows the NPM_CONFIG_REGISTRY environment variable to override this with a custom package registry. This change means that the esbuild executable served by NPM_CONFIG_REGISTRY must now match the expected content.

  Thanks to @sondt99 for reporting this issue.

• Avoid inlining using and await using declarations (#4482)

  Previously esbuild's minifier sometimes incorrectly inlined using and await using declarations into subsequent uses of that declaration, which then fails to dispose of the resource correctly. This bug happened because inlining was done for let and const declarations by avoiding doing it for var declarations, which no longer worked when more declaration types were added. Here's an example:

  // Original code
  {
    using x = new Resource()
    x.activate()
  }
  
  // Old output (with --minify)
  new Resource().activate();
  
  // New output (with --minify)
  {using e=new Resource;e.activate()}

• Fix module evaluation when an error is thrown (#4461, #4467)

  If an error is thrown during module evaluation, esbuild previously didn't preserve the state of the module for subsequent module references. This was observable if import() or require() is used to import a module multiple times. The thrown error is supposed to be thrown by every call to import() or require(), not just the first. With this release, esbuild will now throw the same error every time you call import() or require() on a module that throws during its evaluation.

• Fix some edge cases around the new operator (#4477)

  Previously esbuild incorrectly printed certain edge cases involving complex expressions inside the target of a new expression (specifically an optional chain and/or a tagged template literal). The generated code for the new target was not correctly wrapped with parentheses, and either contained a syntax error or had different semantics. These edge cases have been fixed so that they now correctly wrap the new target in parentheses. Here is an example of some affected code:

  // Original code
  new (foo()`bar`)()
  new (foo()?.bar)()
  
  // Old output
  new foo()`bar`();
  new (foo())?.bar();
  
  // New output
  new (foo())`bar`();
  new (foo()?.bar)();

• Fix renaming of nested var declarations (#4471)

  This release fixes a bug where var declarations in nested scopes that are hoisted up to module scope were not correctly being renamed during bundling. That could previously lead to name collisions when minification was disabled, which could potentially cause a behavior change. The bug has been fixed so that these hoisted declarations are now considered to be module-level symbols during the name collision avoidance pass.

• Emit var instead of const for certain TypeScript-only constructs for ES5 (#4448)

  While esbuild doesn't generally support converting const to var for ES5 due to nested scoping rules (which is currently a build-time error), esbuild previously incorrectly converted TypeScript-only import assignment constructs into a const declaration even when targeting ES5. With this release, esbuild will now use var for this case instead:

  // Original code
  import x = require('y')
  
  // Old output (with --target=es5)
  const x = require("y");
  
  // New output (with --target=es5)
  var x = require("y");

• Add support for with { type: 'text' } imports (#4435)

  The import text proposal has reached stage 3 in the TC39 process, which means that it's recommended for implementation. It has also already been implemented by Deno and Bun. So with this release, esbuild also adds support for it. This behaves exactly the same as esbuild's existing text loader. Here's an example:

  import string from './example.txt' with { type: 'text' }
  console.log(string)

• Add integrity checks to fallback download path (#4343)

  Installing esbuild via npm is somewhat complicated with several different edge cases (see esbuild's documentation for details). If the regular installation of esbuild's platform-specific package fails, esbuild's install script attempts to download the platform-specific package itself (first with the npm command, and then with a HTTP request to registry.npmjs.org as a last resort).

  This last resort path previously didn't have any integrity checks. With this release, esbuild will now verify that the hash of the downloaded binary matches the expected hash for the current release. This means the hashes for all of esbuild's platform-specific binary packages will now be embedded in the top-level esbuild package. Hopefully this should work without any problems. But just in case, this change is being done as a breaking change release.

• Update the Go compiler from 1.25.7 to 1.26.1

  This upgrade should not affect anything. However, there have been some significant internal changes to the Go compiler, so esbuild could potentially behave differently in certain edge cases:

  • It now uses the new garbage collector that comes with Go 1.26.
  • The Go compiler is now more aggressive with allocating memory on the stack.
  • The executable format that the Go linker uses has undergone several changes.
  • The WebAssembly build now unconditionally makes use of the sign extension and non-trapping floating-point to integer conversion instructions.

  You can read the Go 1.26 release notes for more information.

• Fix lowering of define semantics for TypeScript parameter properties (#4421)

  The previous release incorrectly generated class fields for TypeScript parameter properties even when the configured target environment does not support class fields. With this release, the generated class fields will now be correctly lowered in this case:

  // Original code
  class Foo {
    constructor(public x = 1) {}
    y = 2
  }
  
  // Old output (with --loader=ts --target=es2021)
  class Foo {
    constructor(x = 1) {
      this.x = x;
      __publicField(this, "y", 2);
    }
    x;
  }
  
  // New output (with --loader=ts --target=es2021)
  class Foo {
    constructor(x = 1) {
      __publicField(this, "x", x);
      __publicField(this, "y", 2);
    }
  }

• Fix for an async generator edge case (#4401, #4417)

  Support for transforming async generators into the equivalent state machine was added in version 0.19.0. However, the generated state machine didn't work correctly when polling async generators concurrently, such as in the following code:

  async function* inner() { yield 1; yield 2 }
  async function* outer() { yield* inner() }
  let gen = outer()
  for await (let x of [gen.next(), gen.next()]) console.log(x)

  Previously esbuild's output of the above code behaved incorrectly when async generators were transformed (such as with --supported:async-generator=false). The transformation should be fixed starting with this release.

  This fix was contributed by @2767mr.

• Fix a regression when metafile is enabled (#4420, #4418)

  This release fixes a regression introduced by the previous release. When metafile: true was enabled in esbuild's JavaScript API, builds with build errors were incorrectly throwing an error about an empty JSON string instead of an object containing the build errors.

• Use define semantics for TypeScript parameter properties (#4421)

  Parameter properties are a TypeScript-specific code generation feature that converts constructor parameters into class fields when they are prefixed by certain keywords. When "useDefineForClassFields": true is present in tsconfig.json, the TypeScript compiler automatically generates class field declarations for parameter properties. Previously esbuild didn't do this, but esbuild will now do this starting with this release:

  // Original code
  class Foo {
    constructor(public x: number) {}
  }
  
  // Old output (with --loader=ts)
  class Foo {
    constructor(x) {
      this.x = x;
    }
  }
  
  // New output (with --loader=ts)
  class Foo {
    constructor(x) {
      this.x = x;
    }
    x;
  }

• Allow es2025 as a target in tsconfig.json (#4432)

  TypeScript recently added es2025 as a compilation target, so esbuild now supports this in the target field of tsconfig.json files, such as in the following configuration file:

  {
    "compilerOptions": {
      "target": "ES2025"
    }
  }

  As a reminder, the only thing that esbuild uses this field for is determining whether or not to use legacy TypeScript behavior for class fields. You can read more in the documentation.

• Fix a regression with CSS media queries (#4395, #4405, #4406)

  Version 0.25.11 of esbuild introduced support for parsing media queries. This unintentionally introduced a regression with printing media queries that use the <media-type> and <media-condition-without-or> grammar. Specifically, esbuild was failing to wrap an or clause with parentheses when inside <media-condition-without-or>. This release fixes the regression.

  Here is an example:

  /* Original code */
  @media only screen and ((min-width: 10px) or (min-height: 10px)) {
    a { color: red }
  }
  
  /* Old output (incorrect) */
  @media only screen and (min-width: 10px) or (min-height: 10px) {
    a {
      color: red;
    }
  }
  
  /* New output (correct) */
  @media only screen and ((min-width: 10px) or (min-height: 10px)) {
    a {
      color: red;
    }
  }

• Fix an edge case with the inject feature (#4407)

  This release fixes an edge case where esbuild's inject feature could not be used with arbitrary module namespace names exported using an export {} from statement with bundling disabled and a target environment where arbitrary module namespace names is unsupported.

  With the fix, the following inject file:

  import jquery from 'jquery';
  export { jquery as 'window.jQuery' };

  Can now always be rewritten as this without esbuild sometimes incorrectly generating an error:

  export { default as 'window.jQuery' } from 'jquery';

• Attempt to improve API handling of huge metafiles (#4329, #4415)

  This release contains a few changes that attempt to improve the behavior of esbuild's JavaScript API with huge metafiles (esbuild's name for the build metadata, formatted as a JSON object). The JavaScript API is designed to return the metafile JSON as a JavaScript object in memory, which makes it easy to access from within a JavaScript-based plugin. Multiple people have encountered issues where this API breaks down with a pathologically-large metafile.

  The primary issue is that V8 has an implementation-specific maximum string length, so using the JSON.parse API with large enough strings is impossible. This release will now attempt to use a fallback JavaScript-based JSON parser that operates directly on the UTF8-encoded JSON bytes instead of using JSON.parse when the JSON metafile is too big to fit in a JavaScript string. The new fallback path has not yet been heavily-tested. The metafile will also now be generated with whitespace removed if the bundle is significantly large, which will reduce the size of the metafile JSON slightly.

  However, hitting this case is potentially a sign that something else is wrong. Ideally you wouldn't be building something so enormous that the build metadata can't even fit inside a JavaScript string. You may want to consider optimizing your project, or breaking up your project into multiple parts that are built independently. Another option could potentially be to use esbuild's command-line API instead of its JavaScript API, which is more efficient (although of course then you can't use JavaScript plugins, so it may not be an option).

• Preserve URL fragments in data URLs (#4370)

  Consider the following HTML, CSS, and SVG:

  • index.html:

    <!DOCTYPE html>
    <html>
      <head><link rel="stylesheet" href="icons.css"></head>
      <body><div class="triangle"></div></body>
    </html>

  • icons.css:

    .triangle {
      width: 10px;
      height: 10px;
      background: currentColor;
      clip-path: url(./triangle.svg#x);
    }

  • triangle.svg:

    <svg xmlns="http://www.w3.org/2000/svg">
      <defs>
        <clipPath id="x">
          <path d="M0 0H10V10Z"/>
        </clipPath>
      </defs>
    </svg>

  The CSS uses a URL fragment (the #x) to reference the clipPath element in the SVG file. Previously esbuild's CSS bundler didn't preserve the URL fragment when bundling the SVG using the dataurl loader, which broke the bundled CSS. With this release, esbuild will now preserve the URL fragment in the bundled CSS:

  /* icons.css */
  .triangle {
    width: 10px;
    height: 10px;
    background: currentColor;
    clip-path: url('data:image/svg+xml,<svg xmlns="http://www.w3.org/2000/svg"><defs><clipPath id="x"><path d="M0 0H10V10Z"/></clipPath></defs></svg>#x');
  }

• Parse and print CSS @scope rules (#4322)

  This release includes dedicated support for parsing @scope rules in CSS. These rules include optional "start" and "end" selector lists. One important consequence of this is that the local/global status of names in selector lists is now respected, which improves the correctness of esbuild's support for CSS modules. Minification of selectors inside @scope rules has also improved slightly.

  Here's an example:

  /* Original code */
  @scope (:global(.foo)) to (:local(.bar)) {
    .bar {
      color: red;
    }
  }
  
  /* Old output (with --loader=local-css --minify) */
  @scope (:global(.foo)) to (:local(.bar)){.o{color:red}}
  
  /* New output (with --loader=local-css --minify) */
  @scope(.foo)to (.o){.o{color:red}}

• Fix a minification bug with lowering of for await (#4378, #4385)

  This release fixes a bug where the minifier would incorrectly strip the variable in the automatically-generated catch clause of lowered for await loops. The code that generated the loop previously failed to mark the internal variable references as used.

• Update the Go compiler from v1.25.5 to v1.25.7 (#4383, #4388)

  This PR was contributed by @MikeWillCook.

• Allow import path specifiers starting with #/ (#4361)

  Previously the specification for package.json disallowed import path specifiers starting with #/, but this restriction has recently been relaxed and support for it is being added across the JavaScript ecosystem. One use case is using it for a wildcard pattern such as mapping #/* to ./src/* (previously you had to use another character such as #_* instead, which was more confusing). There is some more context in nodejs/node#49182.

  This change was contributed by @hybrist.

• Automatically add the -webkit-mask prefix (#4357, #4358)

  This release automatically adds the -webkit- vendor prefix for the mask CSS shorthand property:

  /* Original code */
  main {
    mask: url(x.png) center/5rem no-repeat
  }
  
  /* Old output (with --target=chrome110) */
  main {
    mask: url(x.png) center/5rem no-repeat;
  }
  
  /* New output (with --target=chrome110) */
  main {
    -webkit-mask: url(x.png) center/5rem no-repeat;
    mask: url(x.png) center/5rem no-repeat;
  }

  This change was contributed by @BPJEnnova.

• Additional minification of switch statements (#4176, #4359)

  This release contains additional minification patterns for reducing switch statements. Here is an example:

  // Original code
  switch (x) {
    case 0:
      foo()
      break
    case 1:
    default:
      bar()
  }
  
  // Old output (with --minify)
  switch(x){case 0:foo();break;case 1:default:bar()}
  
  // New output (with --minify)
  x===0?foo():bar();

• Forbid using declarations inside switch clauses (#4323)

  This is a rare change to remove something that was previously possible. The Explicit Resource Management proposal introduced using declarations. These were previously allowed inside case and default clauses in switch statements. This had well-defined semantics and was already widely implemented (by V8, SpiderMonkey, TypeScript, esbuild, and others). However, it was considered to be too confusing because of how scope works in switch statements, so it has been removed from the specification. This edge case will now be a syntax error. See tc39/proposal-explicit-resource-management#215 and rbuckton/ecma262#14 for details.

  Here is an example of code that is no longer allowed:

  switch (mode) {
    case 'read':
      using readLock = db.read()
      return readAll(readLock)
  
    case 'write':
      using writeLock = db.write()
      return writeAll(writeLock)
  }

  That code will now have to be modified to look like this instead (note the additional { and } block statements around each case body):

  switch (mode) {
    case 'read': {
      using readLock = db.read()
      return readAll(readLock)
    }
    case 'write': {
      using writeLock = db.write()
      return writeAll(writeLock)
    }
  }

  This is not being released in one of esbuild's breaking change releases since this feature hasn't been finalized yet, and esbuild always tracks the current state of the specification (so esbuild's previous behavior was arguably incorrect).

• Fix bundler bug with var nested inside if (#4348)

  This release fixes a bug with the bundler that happens when importing an ES module using require (which causes it to be wrapped) and there's a top-level var inside an if statement without being wrapped in a { ... } block (and a few other conditions). The bundling transform needed to hoist these var declarations outside of the lazy ES module wrapper for correctness. See the issue for details.

• Fix minifier bug with for inside try inside label (#4351)

  This fixes an old regression from version v0.21.4. Some code was introduced to move the label inside the try statement to address a problem with transforming labeled for await loops to avoid the await (the transformation involves converting the for await loop into a for loop and wrapping it in a try statement). However, it introduces problems for cross-compiled JVM code that uses all three of these features heavily. This release restricts this transform to only apply to for loops that esbuild itself generates internally as part of the for await transform. Here is an example of some affected code:

  // Original code
  d: {
    e: {
      try {
        while (1) { break d }
      } catch { break e; }
    }
  }
  
  // Old output (with --minify)
  a:try{e:for(;;)break a}catch{break e}
  
  // New output (with --minify)
  a:e:try{for(;;)break a}catch{break e}

• Inline IIFEs containing a single expression (#4354)

  Previously inlining of IIFEs (immediately-invoked function expressions) only worked if the body contained a single return statement. Now it should also work if the body contains a single expression statement instead:

  // Original code
  const foo = () => {
    const cb = () => {
      console.log(x())
    }
    return cb()
  }
  
  // Old output (with --minify)
  const foo=()=>(()=>{console.log(x())})();
  
  // New output (with --minify)
  const foo=()=>{console.log(x())};

• The minifier now strips empty finally clauses (#4353)

  This improvement means that finally clauses containing dead code can potentially cause the associated try statement to be removed from the output entirely in minified builds:

  // Original code
  function foo(callback) {
    if (DEBUG) stack.push(callback.name);
    try {
      callback();
    } finally {
      if (DEBUG) stack.pop();
    }
  }
  
  // Old output (with --minify --define:DEBUG=false)
  function foo(a){try{a()}finally{}}
  
  // New output (with --minify --define:DEBUG=false)
  function foo(a){a()}

• Allow tree-shaking of the Symbol constructor

  With this release, calling Symbol is now considered to be side-effect free when the argument is known to be a primitive value. This means esbuild can now tree-shake module-level symbol variables:

  // Original code
  const a = Symbol('foo')
  const b = Symbol(bar)
  
  // Old output (with --tree-shaking=true)
  const a = Symbol("foo");
  const b = Symbol(bar);
  
  // New output (with --tree-shaking=true)
  const b = Symbol(bar);

This release deliberately contains backwards-incompatible changes. To avoid automatically picking up releases like this, you should either be pinning the exact version of esbuild in your package.json file (recommended) or be using a version range syntax that only accepts patch upgrades such as ^0.26.0 or ~0.26.0. See npm's documentation about semver for more information.

• Use Uint8Array.fromBase64 if available (#4286)

  With this release, esbuild's binary loader will now use the new Uint8Array.fromBase64 function unless it's unavailable in the configured target environment. If it's unavailable, esbuild's previous code for this will be used as a fallback. Note that this means you may now need to specify target when using this feature with Node (for example --target=node22) unless you're using Node v25+.

• Update the Go compiler from v1.23.12 to v1.25.4 (#4208, #4311)

  This raises the operating system requirements for running esbuild:

  • Linux: now requires a kernel version of 3.2 or later
  • macOS: now requires macOS 12 (Monterey) or later

• Enable trusted publishing (#4281)

  GitHub and npm are recommending that maintainers for packages such as esbuild switch to trusted publishing. With this release, a VM on GitHub will now build and publish all of esbuild's packages to npm instead of me. In theory.

  Unfortunately there isn't really a way to test that this works other than to do it live. So this release is that live test. Hopefully this release is uneventful and is exactly the same as the previous one (well, except for the green provenance attestation checkmark on npm that happens with trusted publishing).

• Fix a minification regression with CSS media queries (#4315)

  The previous release introduced support for parsing media queries which unintentionally introduced a regression with the removal of duplicate media rules during minification. Specifically the grammar for @media <media-type> and <media-condition-without-or> { ... } was missing an equality check for the <media-condition-without-or> part, so rules with different suffix clauses in this position would incorrectly compare equal and be deduplicated. This release fixes the regression.

• Update the list of known JavaScript globals (#4310)

  This release updates esbuild's internal list of known JavaScript globals. These are globals that are known to not have side-effects when the property is accessed. For example, accessing the global Array property is considered to be side-effect free but accessing the global scrollY property can trigger a layout, which is a side-effect. This is used by esbuild's tree-shaking to safely remove unused code that is known to be side-effect free. This update adds the following global properties:

  From ES2017:

  • Atomics
  • SharedArrayBuffer

  From ES2020:

  • BigInt64Array
  • BigUint64Array

  From ES2021:

  • FinalizationRegistry
  • WeakRef

  From ES2025:

  • Float16Array
  • Iterator

  Note that this does not indicate that constructing any of these objects is side-effect free, just that accessing the identifier is side-effect free. For example, this now allows esbuild to tree-shake classes that extend from Iterator:

  // This can now be tree-shaken by esbuild:
  class ExampleIterator extends Iterator {}
  

• Add support for the new @view-transition CSS rule (#4313)

  With this release, esbuild now has improved support for pretty-printing and minifying the new @view-transition rule (which esbuild was previously unaware of):

  /* Original code */
  @view-transition {
    navigation: auto;
    types: check;
  }
  
  /* Old output */
  @view-transition { navigation: auto; types: check; }
  
  /* New output */
  @view-transition {
    navigation: auto;
    types: check;
  }
  

  The new view transition feature provides a mechanism for creating animated transitions between documents in a multi-page app. You can read more about view transition rules here.

  This change was contributed by @yisibl.

• Trim CSS rules that will never match

  The CSS minifier will now remove rules whose selectors contain :is() and :where() as those selectors will never match. These selectors can currently be automatically generated by esbuild when you give esbuild nonsensical input such as the following:

  /* Original code */
  div:before {
    color: green;
    &.foo {
      color: red;
    }
  }
  
  /* Old output (with --supported:nesting=false --minify) */
  div:before{color:green}:is().foo{color:red}
  
  /* New output (with --supported:nesting=false --minify) */
  div:before{color:green}
  

  This input is nonsensical because CSS nesting is (unfortunately) not supported inside of pseudo-elements such as :before. Currently esbuild generates a rule containing :is() in this case when you tell esbuild to transform nested CSS into non-nested CSS. I think it's reasonable to do that as it sort of helps explain what's going on (or at least indicates that something is wrong in the output). It shouldn't be present in minified code, however, so this release now strips it out.

• Add support for with { type: 'bytes' } imports (#4292)

  The import bytes proposal has reached stage 2.7 in the TC39 process, which means that although it isn't quite recommended for implementation, it's generally approved and ready for validation. Furthermore it has already been implemented by Deno and Webpack. So with this release, esbuild will also add support for this. It behaves exactly the same as esbuild's existing binary loader. Here's an example:

  import data from './image.png' with { type: 'bytes' }
  const view = new DataView(data.buffer, 0, 24)
  const width = view.getInt32(16)
  const height = view.getInt32(20)
  console.log('size:', width + '\xD7' + height)
  

• Lower CSS media query range syntax (#3748, #4293)

  With this release, esbuild will now transform CSS media query range syntax into equivalent syntax using min-/max- prefixes for older browsers. For example, the following CSS:

  @media (640px <= width <= 960px) {
    main {
      display: flex;
    }
  }
  

  will be transformed like this with a target such as --target=chrome100 (or more specifically with --supported:media-range=false if desired):

  @media (min-width: 640px) and (max-width: 960px) {
    main {
      display: flex;
    }
  }
  

• Fix a panic in a minification edge case (#4287)

  This release fixes a panic due to a null pointer that could happen when esbuild inlines a doubly-nested identity function and the final result is empty. It was fixed by emitting the value undefined in this case, which avoids the panic. This case must be rare since it hasn't come up until now. Here is an example of code that previously triggered the panic (which only happened when minifying):

  function identity(x) { return x }
  identity({ y: identity(123) })
  

• Fix @supports nested inside pseudo-element (#4265)

  When transforming nested CSS to non-nested CSS, esbuild is supposed to filter out pseudo-elements such as ::placeholder for correctness. The CSS nesting specification says the following:

  The nesting selector cannot represent pseudo-elements (identical to the behavior of the ':is()' pseudo-class). We’d like to relax this restriction, but need to do so simultaneously for both ':is()' and '&', since they’re intentionally built on the same underlying mechanisms.

  However, it seems like this behavior is different for nested at-rules such as @supports, which do work with pseudo-elements. So this release modifies esbuild's behavior to now take that into account:

  /* Original code */
  ::placeholder {
    color: red;
    body & { color: green }
    @supports (color: blue) { color: blue }
  }
  
  /* Old output (with --supported:nesting=false) */
  ::placeholder {
    color: red;
  }
  body :is() {
    color: green;
  }
  @supports (color: blue) {
     {
      color: blue;
    }
  }
  
  /* New output (with --supported:nesting=false) */
  ::placeholder {
    color: red;
  }
  body :is() {
    color: green;
  }
  @supports (color: blue) {
    ::placeholder {
      color: blue;
    }
  }
  

• Better support building projects that use Yarn on Windows (#3131, #3663)

  With this release, you can now use esbuild to bundle projects that use Yarn Plug'n'Play on Windows on drives other than the C: drive. The problem was as follows:

  1. Yarn in Plug'n'Play mode on Windows stores its global module cache on the C: drive
  2. Some developers put their projects on the D: drive
  3. Yarn generates relative paths that use ../.. to get from the project directory to the cache directory
  4. Windows-style paths don't support directory traversal between drives via .. (so D:\.. is just D:)
  5. I didn't have access to a Windows machine for testing this edge case

  Yarn works around this edge case by pretending Windows-style paths beginning with C:\ are actually Unix-style paths beginning with /C:/, so the ../.. path segments are able to navigate across drives inside Yarn's implementation. This was broken for a long time in esbuild but I finally got access to a Windows machine and was able to debug and fix this edge case. So you should now be able to bundle these projects with esbuild.

• Preserve parentheses around function expressions (#4252)

  The V8 JavaScript VM uses parentheses around function expressions as an optimization hint to immediately compile the function. Otherwise the function would be lazily-compiled, which has additional overhead if that function is always called immediately as lazy compilation involves parsing the function twice. You can read V8's blog post about this for more details.

  Previously esbuild did not represent parentheses around functions in the AST so they were lost during compilation. With this change, esbuild will now preserve parentheses around function expressions when they are present in the original source code. This means these optimization hints will not be lost when bundling with esbuild. In addition, esbuild will now automatically add this optimization hint to immediately-invoked function expressions. Here's an example:

  // Original code
  const fn0 = () => 0
  const fn1 = (() => 1)
  console.log(fn0, function() { return fn1() }())
  
  // Old output
  const fn0 = () => 0;
  const fn1 = () => 1;
  console.log(fn0, function() {
    return fn1();
  }());
  
  // New output
  const fn0 = () => 0;
  const fn1 = (() => 1);
  console.log(fn0, (function() {
    return fn1();
  })());
  

  Note that you do not want to wrap all function expressions in parentheses. This optimization hint should only be used for functions that are called on initial load. Using this hint for functions that are not called on initial load will unnecessarily delay the initial load. Again, see V8's blog post linked above for details.

• Update Go from 1.23.10 to 1.23.12 (#4257, #4258)

  This should have no effect on existing code as this version change does not change Go's operating system support. It may remove certain false positive reports (specifically CVE-2025-4674 and CVE-2025-47907) from vulnerability scanners that only detect which version of the Go compiler esbuild uses.

• Fix another TypeScript parsing edge case (#4248)

  This fixes a regression with a change in the previous release that tries to more accurately parse TypeScript arrow functions inside the ?: operator. The regression specifically involves parsing an arrow function containing a #private identifier inside the middle of a ?: ternary operator inside a class body. This was fixed by propagating private identifier state into the parser clone used to speculatively parse the arrow function body. Here is an example of some affected code:

  class CachedDict {
    #has = (a: string) => dict.has(a);
    has = window
      ? (word: string): boolean => this.#has(word)
      : this.#has;
  }
  

• Fix a regression with the parsing of source phase imports

  The change in the previous release to parse source phase imports failed to properly handle the following cases:

  import source from 'bar'
  import source from from 'bar'
  import source type foo from 'bar'
  

  Parsing for these cases should now be fixed. The first case was incorrectly treated as a syntax error because esbuild was expecting the second case. And the last case was previously allowed but is now forbidden. TypeScript hasn't added this feature yet so it remains to be seen whether the last case will be allowed, but it's safer to disallow it for now. At least Babel doesn't allow the last case when parsing TypeScript, and Babel was involved with the source phase import specification.

• Parse and print JavaScript imports with an explicit phase (#4238)

  This release adds basic syntax support for the defer and source import phases in JavaScript:

  • defer

    This is a stage 3 proposal for an upcoming JavaScript feature that will provide one way to eagerly load but lazily initialize imported modules. The imported module is automatically initialized on first use. Support for this syntax will also be part of the upcoming release of TypeScript 5.9. The syntax looks like this:

    import defer * as foo from "<specifier>";
    const bar = await import.defer("<specifier>");
    

    Note that this feature deliberately cannot be used with the syntax import defer foo from "<specifier>" or import defer { foo } from "<specifier>".

  • source

    This is a stage 3 proposal for an upcoming JavaScript feature that will provide another way to eagerly load but lazily initialize imported modules. The imported module is returned in an uninitialized state. Support for this syntax may or may not be a part of TypeScript 5.9 (see this issue for details). The syntax looks like this:

    import source foo from "<specifier>";
    const bar = await import.source("<specifier>");
    

    Note that this feature deliberately cannot be used with the syntax import defer * as foo from "<specifier>" or import defer { foo } from "<specifier>".

  This change only adds support for this syntax. These imports cannot currently be bundled by esbuild. To use these new features with esbuild's bundler, the imported paths must be external to the bundle and the output format must be set to esm.

• Support optionally emitting absolute paths instead of relative paths (#338, #2082, #3023)

  This release introduces the --abs-paths= feature which takes a comma-separated list of situations where esbuild should use absolute paths instead of relative paths. There are currently three supported situations: code (comments and string literals), log (log message text and location info), and metafile (the JSON build metadata).

  Using absolute paths instead of relative paths is not the default behavior because it means that the build results are no longer machine-independent (which means builds are no longer reproducible). Absolute paths can be useful when used with certain terminal emulators that allow you to click on absolute paths in the terminal text and/or when esbuild is being automatically invoked from several different directories within the same script.

• Fix a TypeScript parsing edge case (#4241)

  This release fixes an edge case with parsing an arrow function in TypeScript with a return type that's in the middle of a ?: ternary operator. For example:

  x = a ? (b) : c => d;
  y = a ? (b) : c => d : e;
  

  The : token in the value assigned to x pairs with the ? token, so it's not the start of a return type annotation. However, the first : token in the value assigned to y is the start of a return type annotation because after parsing the arrow function body, it turns out there's another : token that can be used to pair with the ? token. This case is notable as it's the first TypeScript edge case that esbuild has needed a backtracking parser to parse. It has been addressed by a quick hack (cloning the whole parser) as it's a rare edge case and esbuild doesn't otherwise need a backtracking parser. Hopefully this is sufficient and doesn't cause any issues.

• Inline small constant strings when minifying

  Previously esbuild's minifier didn't inline string constants because strings can be arbitrarily long, and this isn't necessarily a size win if the string is used more than once. Starting with this release, esbuild will now inline string constants when the length of the string is three code units or less. For example:

  // Original code
  const foo = 'foo'
  console.log({ [foo]: true })
  
  // Old output (with --minify --bundle --format=esm)
  var o="foo";console.log({[o]:!0});
  
  // New output (with --minify --bundle --format=esm)
  console.log({foo:!0});
  

  Note that esbuild's constant inlining only happens in very restrictive scenarios to avoid issues with TDZ handling. This change doesn't change when esbuild's constant inlining happens. It only expands the scope of it to include certain string literals in addition to numeric and boolean literals.

• Fix a memory leak when cancel() is used on a build context (#4231)

  Calling rebuild() followed by cancel() in rapid succession could previously leak memory. The bundler uses a producer/consumer model internally, and the resource leak was caused by the consumer being termianted while there were still remaining unreceived results from a producer. To avoid the leak, the consumer now waits for all producers to finish before terminating.

• Support empty :is() and :where() syntax in CSS (#4232)

  Previously using these selectors with esbuild would generate a warning. That warning has been removed in this release for these cases.

• Improve tree-shaking of try statements in dead code (#4224)

  With this release, esbuild will now remove certain try statements if esbuild considers them to be within dead code (i.e. code that is known to not ever be evaluated). For example:

  // Original code
  return 'foo'
  try { return 'bar' } catch {}
  
  // Old output (with --minify)
  return"foo";try{return"bar"}catch{}
  
  // New output (with --minify)
  return"foo";
  

• Consider negated bigints to have no side effects

  While esbuild currently considers 1, -1, and 1n to all have no side effects, it didn't previously consider -1n to have no side effects. This is because esbuild does constant folding with numbers but not bigints. However, it meant that unused negative bigint constants were not tree-shaken. With this release, esbuild will now consider these expressions to also be side-effect free:

  // Original code
  let a = 1, b = -1, c = 1n, d = -1n
  
  // Old output (with --bundle --minify)
  (()=>{var n=-1n;})();
  
  // New output (with --bundle --minify)
  (()=>{})();
  

• Support a configurable delay in watch mode before rebuilding (#3476, #4178)

  The watch() API now takes a delay option that lets you add a delay (in milliseconds) before rebuilding when a change is detected in watch mode. If you use a tool that regenerates multiple source files very slowly, this should make it more likely that esbuild's watch mode won't generate a broken intermediate build before the successful final build. This option is also available via the CLI using the --watch-delay= flag.

  This should also help avoid confusion about the watch() API's options argument. It was previously empty to allow for future API expansion, which caused some people to think that the documentation was missing. It's no longer empty now that the watch() API has an option.

• Allow mixed array for entryPoints API option (#4223)

  The TypeScript type definitions now allow you to pass a mixed array of both string literals and object literals to the entryPoints API option, such as ['foo.js', { out: 'lib', in: 'bar.js' }]. This was always possible to do in JavaScript but the TypeScript type definitions were previously too restrictive.

• Update Go from 1.23.8 to 1.23.10 (#4204, #4207)

  This should have no effect on existing code as this version change does not change Go's operating system support. It may remove certain false positive reports (specifically CVE-2025-4673 and CVE-2025-22874) from vulnerability scanners that only detect which version of the Go compiler esbuild uses.

• Experimental support for esbuild on OpenHarmony (#4212)

  With this release, esbuild now publishes the @esbuild/openharmony-arm64 npm package for OpenHarmony. It contains a WebAssembly binary instead of a native binary because Go doesn't currently support OpenHarmony. Node does support it, however, so in theory esbuild should now work on OpenHarmony through WebAssembly.

  This change was contributed by @hqzing.

• Fix a regression with browser in package.json (#4187)

  The fix to #4144 in version 0.25.3 introduced a regression that caused browser overrides specified in package.json to fail to override relative path names that end in a trailing slash. That behavior change affected the axios@0.30.0 package. This regression has been fixed, and now has test coverage.

• Add support for certain keywords as TypeScript tuple labels (#4192)

  Previously esbuild could incorrectly fail to parse certain keywords as TypeScript tuple labels that are parsed by the official TypeScript compiler if they were followed by a ? modifier. These labels included function, import, infer, new, readonly, and typeof. With this release, these keywords will now be parsed correctly. Here's an example of some affected code:

  type Foo = [
    value: any,
    readonly?: boolean, // This is now parsed correctly
  ]
  

• Add CSS prefixes for the stretch sizing value (#4184)

  This release adds support for prefixing CSS declarations such as div { width: stretch }. That CSS is now transformed into this depending on what the --target= setting includes:

  div {
    width: -webkit-fill-available;
    width: -moz-available;
    width: stretch;
  }
  

• Add simple support for CORS to esbuild's development server (#4125)

  Starting with version 0.25.0, esbuild's development server is no longer configured to serve cross-origin requests. This was a deliberate change to prevent any website you visit from accessing your running esbuild development server. However, this change prevented (by design) certain use cases such as "debugging in production" by having your production website load code from localhost where the esbuild development server is running.

  To enable this use case, esbuild is adding a feature to allow Cross-Origin Resource Sharing (a.k.a. CORS) for simple requests. Specifically, passing your origin to the new cors option will now set the Access-Control-Allow-Origin response header when the request has a matching Origin header. Note that this currently only works for requests that don't send a preflight OPTIONS request, as esbuild's development server doesn't currently support OPTIONS requests.

  Some examples:

  • CLI:

    esbuild --servedir=. --cors-origin=https://example.com
    

  • JS:

    const ctx = await esbuild.context({})
    await ctx.serve({
      servedir: '.',
      cors: {
        origin: 'https://example.com',
      },
    })
    

  • Go:

    ctx, _ := api.Context(api.BuildOptions{})
    ctx.Serve(api.ServeOptions{
      Servedir: ".",
      CORS: api.CORSOptions{
        Origin: []string{"https://example.com"},
      },
    })
    

  The special origin * can be used to allow any origin to access esbuild's development server. Note that this means any website you visit will be able to read everything served by esbuild.

• Pass through invalid URLs in source maps unmodified (#4169)

  This fixes a regression in version 0.25.0 where sources in source maps that form invalid URLs were not being passed through to the output. Version 0.25.0 changed the interpretation of sources from file paths to URLs, which means that URL parsing can now fail. Previously URLs that couldn't be parsed were replaced with the empty string. With this release, invalid URLs in sources should now be passed through unmodified.

• Handle exports named __proto__ in ES modules (#4162, #4163)

  In JavaScript, the special property name __proto__ sets the prototype when used inside an object literal. Previously esbuild's ESM-to-CommonJS conversion didn't special-case the property name of exports named __proto__ so the exported getter accidentally became the prototype of the object literal. It's unclear what this affects, if anything, but it's better practice to avoid this by using a computed property name in this case.

  This fix was contributed by @magic-akari.

• Fix lowered async arrow functions before super() (#4141, #4142)

  This change makes it possible to call an async arrow function in a constructor before calling super() when targeting environments without async support, as long as the function body doesn't reference this. Here's an example (notice the change from this to null):

  // Original code
  class Foo extends Object {
    constructor() {
      (async () => await foo())()
      super()
    }
  }
  
  // Old output (with --target=es2016)
  class Foo extends Object {
    constructor() {
      (() => __async(this, null, function* () {
        return yield foo();
      }))();
      super();
    }
  }
  
  // New output (with --target=es2016)
  class Foo extends Object {
    constructor() {
      (() => __async(null, null, function* () {
        return yield foo();
      }))();
      super();
    }
  }
  

  Some background: Arrow functions with the async keyword are transformed into generator functions for older language targets such as --target=es2016. Since arrow functions capture this, the generated code forwards this into the body of the generator function. However, JavaScript class syntax forbids using this in a constructor before calling super(), and this forwarding was problematic since previously happened even when the function body doesn't use this. Starting with this release, esbuild will now only forward this if it's used within the function body.

  This fix was contributed by @magic-akari.

• Fix memory leak with --watch=true (#4131, #4132)

  This release fixes a memory leak with esbuild when --watch=true is used instead of --watch. Previously using --watch=true caused esbuild to continue to use more and more memory for every rebuild, but --watch=true should now behave like --watch and not leak memory.

  This bug happened because esbuild disables the garbage collector when it's not run as a long-lived process for extra speed, but esbuild's checks for which arguments cause esbuild to be a long-lived process weren't updated for the new --watch=true style of boolean command-line flags. This has been an issue since this boolean flag syntax was added in version 0.14.24 in 2022. These checks are unfortunately separate from the regular argument parser because of how esbuild's internals are organized (the command-line interface is exposed as a separate Go API so you can build your own custom esbuild CLI).

  This fix was contributed by @mxschmitt.

• More concise output for repeated legal comments (#4139)

  Some libraries have many files and also use the same legal comment text in all files. Previously esbuild would copy each legal comment to the output file. Starting with this release, legal comments duplicated across separate files will now be grouped in the output file by unique comment content.

• Allow a custom host with the development server (#4110)

  With this release, you can now use a custom non-IP host with esbuild's local development server (either with --serve= for the CLI or with the serve() call for the API). This was previously possible, but was intentionally broken in version 0.25.0 to fix a security issue. This change adds the functionality back except that it's now opt-in and only for a single domain name that you provide.

  For example, if you add a mapping in your /etc/hosts file from local.example.com to 127.0.0.1 and then use esbuild --serve=local.example.com:8000, you will now be able to visit http://local.example.com:8000/ in your browser and successfully connect to esbuild's development server (doing that would previously have been blocked by the browser). This should also work with HTTPS if it's enabled (see esbuild's documentation for how to do that).

• Add a limit to CSS nesting expansion (#4114)

  With this release, esbuild will now fail with an error if there is too much CSS nesting expansion. This can happen when nested CSS is converted to CSS without nesting for older browsers as expanding CSS nesting is inherently exponential due to the resulting combinatorial explosion. The expansion limit is currently hard-coded and cannot be changed, but is extremely unlikely to trigger for real code. It exists to prevent esbuild from using too much time and/or memory. Here's an example:

  a,b{a,b{a,b{a,b{a,b{a,b{a,b{a,b{a,b{a,b{a,b{a,b{a,b{a,b{a,b{a,b{a,b{a,b{a,b{a,b{color:red}}}}}}}}}}}}}}}}}}}}
  

  Previously, transforming this file with --target=safari1 took 5 seconds and generated 40mb of CSS. Trying to do that will now generate the following error instead:

  ✘ [ERROR] CSS nesting is causing too much expansion
  
      example.css:1:60:
        1 │ a,b{a,b{a,b{a,b{a,b{a,b{a,b{a,b{a,b{a,b{a,b{a,b{a,b{a,b{a,b{a,b{a,b{a,b{a,b{a,b{color:red}}}}}}}}}}}}}}}}}}}}
          ╵                                                             ^
  
    CSS nesting expansion was terminated because a rule was generated with 65536 selectors. This limit
    exists to prevent esbuild from using too much time and/or memory. Please change your CSS to use
    fewer levels of nesting.
  

• Fix path resolution edge case (#4144)

  This fixes an edge case where esbuild's path resolution algorithm could deviate from node's path resolution algorithm. It involves a confusing situation where a directory shares the same file name as a file (but without the file extension). See the linked issue for specific details. This appears to be a case where esbuild is correctly following node's published resolution algorithm but where node itself is doing something different. Specifically the step LOAD_AS_FILE appears to be skipped when the input ends with ... This release changes esbuild's behavior for this edge case to match node's behavior.

• Update Go from 1.23.7 to 1.23.8 (#4133, #4134)

  This should have no effect on existing code as this version change does not change Go's operating system support. It may remove certain reports from vulnerability scanners that detect which version of the Go compiler esbuild uses, such as for CVE-2025-22871.

  As a reminder, esbuild's development server is intended for development, not for production, so I do not consider most networking-related vulnerabilities in Go to be vulnerabilities in esbuild. Please do not use esbuild's development server in production.

• Support flags in regular expressions for the API (#4121)

  The JavaScript plugin API for esbuild takes JavaScript regular expression objects for the filter option. Internally these are translated into Go regular expressions. However, this translation previously ignored the flags property of the regular expression. With this release, esbuild will now translate JavaScript regular expression flags into Go regular expression flags. Specifically the JavaScript regular expression /\.[jt]sx?$/i is turned into the Go regular expression `(?i)\.[jt]sx?$` internally inside of esbuild's API. This should make it possible to use JavaScript regular expressions with the i flag. Note that JavaScript and Go don't support all of the same regular expression features, so this mapping is only approximate.

• Fix node-specific annotations for string literal export names (#4100)

  When node instantiates a CommonJS module, it scans the AST to look for names to expose via ESM named exports. This is a heuristic that looks for certain patterns such as exports.NAME = ... or module.exports = { ... }. This behavior is used by esbuild to "annotate" CommonJS code that was converted from ESM with the original ESM export names. For example, when converting the file export let foo, bar from ESM to CommonJS, esbuild appends this to the end of the file:

  // Annotate the CommonJS export names for ESM import in node:
  0 && (module.exports = {
    bar,
    foo
  });
  

  However, this feature previously didn't work correctly for export names that are not valid identifiers, which can be constructed using string literal export names. The generated code contained a syntax error. That problem is fixed in this release:

  // Original code
  let foo
  export { foo as "foo!" }
  
  // Old output (with --format=cjs --platform=node)
  ...
  0 && (module.exports = {
    "foo!"
  });
  
  // New output (with --format=cjs --platform=node)
  ...
  0 && (module.exports = {
    "foo!": null
  });
  

• Basic support for index source maps (#3439, #4109)

  The source map specification has an optional mode called index source maps that makes it easier for tools to create an aggregate JavaScript file by concatenating many smaller JavaScript files with source maps, and then generate an aggregate source map by simply providing the original source maps along with some offset information. My understanding is that this is rarely used in practice. I'm only aware of two uses of it in the wild: ClojureScript and Turbopack.

  This release provides basic support for indexed source maps. However, the implementation has not been tested on a real app (just on very simple test input). If you are using index source maps in a real app, please try this out and report back if anything isn't working for you.

  Note that this is also not a complete implementation. For example, index source maps technically allows nesting source maps to an arbitrary depth, while esbuild's implementation in this release only supports a single level of nesting. It's unclear whether supporting more than one level of nesting is important or not given the lack of available test cases.

  This feature was contributed by @clyfish.

• Fix incorrect paths in inline source maps (#4070, #4075, #4105)

  This fixes a regression from version 0.25.0 where esbuild didn't correctly resolve relative paths contained within source maps in inline sourceMappingURL data URLs. The paths were incorrectly being passed through as-is instead of being resolved relative to the source file containing the sourceMappingURL comment, which was due to the data URL not being a file URL. This regression has been fixed, and this case now has test coverage.

• Fix invalid generated source maps (#4080, #4082, #4104, #4107)

  This release fixes a regression from version 0.24.1 that could cause esbuild to generate invalid source maps. Specifically under certain conditions, esbuild could generate a mapping with an out-of-bounds source index. It was introduced by code that attempted to improve esbuild's handling of "null" entries in source maps (i.e. mappings with a generated position but no original position). This regression has been fixed.

  This fix was contributed by @jridgewell.

• Fix a regression with non-file source map paths (#4078)

  The format of paths in source maps that aren't in the file namespace was unintentionally changed in version 0.25.0. Path namespaces is an esbuild-specific concept that is optionally available for plugins to use to distinguish paths from file paths and from paths meant for other plugins. Previously the namespace was prepended to the path joined with a : character, but version 0.25.0 unintentionally failed to prepend the namespace. The previous behavior has been restored.

• Fix a crash with switch optimization (#4088)

  The new code in the previous release to optimize dead code in switch statements accidentally introduced a crash in the edge case where one or more switch case values include a function expression. This is because esbuild now visits the case values first to determine whether any cases are dead code, and then visits the case bodies once the dead code status is known. That triggered some internal asserts that guard against traversing the AST in an unexpected order. This crash has been fixed by changing esbuild to expect the new traversal ordering. Here's an example of affected code:

  switch (x) {
    case '':
      return y.map(z => z.value)
    case y.map(z => z.key).join(','):
      return []
  }
  

• Update Go from 1.23.5 to 1.23.7 (#4076, #4077)

  This should have no effect on existing code as this version change does not change Go's operating system support. It may remove certain reports from vulnerability scanners that detect which version of the Go compiler esbuild uses.

  This PR was contributed by @MikeWillCook.

This release deliberately contains backwards-incompatible changes. To avoid automatically picking up releases like this, you should either be pinning the exact version of esbuild in your package.json file (recommended) or be using a version range syntax that only accepts patch upgrades such as ^0.24.0 or ~0.24.0. See npm's documentation about semver for more information.

• Restrict access to esbuild's development server (GHSA-67mh-4wv8-2f99)

  This change addresses esbuild's first security vulnerability report. Previously esbuild set the Access-Control-Allow-Origin header to * to allow esbuild's development server to be flexible in how it's used for development. However, this allows the websites you visit to make HTTP requests to esbuild's local development server, which gives read-only access to your source code if the website were to fetch your source code's specific URL. You can read more information in the report.

  Starting with this release, CORS will now be disabled, and requests will now be denied if the host does not match the one provided to --serve=. The default host is 0.0.0.0, which refers to all of the IP addresses that represent the local machine (e.g. both 127.0.0.1 and 192.168.0.1). If you want to customize anything about esbuild's development server, you can put a proxy in front of esbuild and modify the incoming and/or outgoing requests.

  In addition, the serve() API call has been changed to return an array of hosts instead of a single host string. This makes it possible to determine all of the hosts that esbuild's development server will accept.

  Thanks to @sapphi-red for reporting this issue.

• Delete output files when a build fails in watch mode (#3643)

  It has been requested for esbuild to delete files when a build fails in watch mode. Previously esbuild left the old files in place, which could cause people to not immediately realize that the most recent build failed. With this release, esbuild will now delete all output files if a rebuild fails. Fixing the build error and triggering another rebuild will restore all output files again.

• Fix correctness issues with the CSS nesting transform (#3620, #3877, #3933, #3997, #4005, #4037, #4038)

  This release fixes the following problems:

  • Naive expansion of CSS nesting can result in an exponential blow-up of generated CSS if each nesting level has multiple selectors. Previously esbuild sometimes collapsed individual nesting levels using :is() to limit expansion. However, this collapsing wasn't correct in some cases, so it has been removed to fix correctness issues.

    /* Original code */
    .parent {
      > .a,
      > .b1 > .b2 {
        color: red;
      }
    }
    
    /* Old output (with --supported:nesting=false) */
    .parent > :is(.a, .b1 > .b2) {
      color: red;
    }
    
    /* New output (with --supported:nesting=false) */
    .parent > .a,
    .parent > .b1 > .b2 {
      color: red;
    }
    

    Thanks to @tim-we for working on a fix.

  • The & CSS nesting selector can be repeated multiple times to increase CSS specificity. Previously esbuild ignored this possibility and incorrectly considered && to have the same specificity as &. With this release, this should now work correctly:

    /* Original code (color should be red) */
    div {
      && { color: red }
      & { color: blue }
    }
    
    /* Old output (with --supported:nesting=false) */
    div {
      color: red;
    }
    div {
      color: blue;
    }
    
    /* New output (with --supported:nesting=false) */
    div:is(div) {
      color: red;
    }
    div {
      color: blue;
    }
    

    Thanks to @CPunisher for working on a fix.

  • Previously transforming nested CSS incorrectly removed leading combinators from within pseudoclass selectors such as :where(). This edge case has been fixed and how has test coverage.

    /* Original code */
    a b:has(> span) {
      a & {
        color: green;
      }
    }
    
    /* Old output (with --supported:nesting=false) */
    a :is(a b:has(span)) {
      color: green;
    }
    
    /* New output (with --supported:nesting=false) */
    a :is(a b:has(> span)) {
      color: green;
    }
    

    This fix was contributed by @NoremacNergfol.

  • The CSS minifier contains logic to remove the & selector when it can be implied, which happens when there is only one and it's the leading token. However, this logic was incorrectly also applied to selector lists inside of pseudo-class selectors such as :where(). With this release, the minifier will now avoid applying this logic in this edge case:

    /* Original code */
    .a {
      & .b { color: red }
      :where(& .b) { color: blue }
    }
    
    /* Old output (with --minify) */
    .a{.b{color:red}:where(.b){color:#00f}}
    
    /* New output (with --minify) */
    .a{.b{color:red}:where(& .b){color:#00f}}
    

• Fix some correctness issues with source maps (#1745, #3183, #3613, #3982)

  Previously esbuild incorrectly treated source map path references as file paths instead of as URLs. With this release, esbuild will now treat source map path references as URLs. This fixes the following problems with source maps:

  • File names in sourceMappingURL that contained a space previously did not encode the space as %20, which resulted in JavaScript tools (including esbuild) failing to read that path back in when consuming the generated output file. This should now be fixed.

  • Absolute URLs in sourceMappingURL that use the file:// scheme previously attempted to read from a folder called file:. These URLs should now be recognized and parsed correctly.

  • Entries in the sources array in the source map are now treated as URLs instead of file paths. The correct behavior for this is much more clear now that source maps has a formal specification. Many thanks to those who worked on the specification.

• Fix incorrect package for @esbuild/netbsd-arm64 (#4018)

  Due to a copy+paste typo, the binary published to @esbuild/netbsd-arm64 was not actually for arm64, and didn't run in that environment. This release should fix running esbuild in that environment (NetBSD on 64-bit ARM). Sorry about the mistake.

• Fix a minification bug with bitwise operators and bigints (#4065)

  This change removes an incorrect assumption in esbuild that all bitwise operators result in a numeric integer. That assumption was correct up until the introduction of bigints in ES2020, but is no longer correct because almost all bitwise operators now operate on both numbers and bigints. Here's an example of the incorrect minification:

  // Original code
  if ((a & b) !== 0) found = true
  
  // Old output (with --minify)
  a&b&&(found=!0);
  
  // New output (with --minify)
  (a&b)!==0&&(found=!0);
  

• Fix esbuild incorrectly rejecting valid TypeScript edge case (#4027)

  The following TypeScript code is valid:

  export function open(async?: boolean): void {
    console.log(async as boolean)
  }
  

  Before this version, esbuild would fail to parse this with a syntax error as it expected the token sequence async as ... to be the start of an async arrow function expression async as => .... This edge case should be parsed correctly by esbuild starting with this release.

• Transform BigInt values into constructor calls when unsupported (#4049)

  Previously esbuild would refuse to compile the BigInt literals (such as 123n) if they are unsupported in the configured target environment (such as with --target=es6). The rationale was that they cannot be polyfilled effectively because they change the behavior of JavaScript's arithmetic operators and JavaScript doesn't have operator overloading.

  However, this prevents using esbuild with certain libraries that would otherwise work if BigInt literals were ignored, such as with old versions of the buffer library before the library fixed support for running in environments without BigInt support. So with this release, esbuild will now turn BigInt literals into BigInt constructor calls (so 123n becomes BigInt(123)) and generate a warning in this case. You can turn off the warning with --log-override:bigint=silent or restore the warning to an error with --log-override:bigint=error if needed.

• Change how console API dropping works (#4020)

  Previously the --drop:console feature replaced all method calls off of the console global with undefined regardless of how long the property access chain was (so it applied to console.log() and console.log.call(console) and console.log.not.a.method()). However, it was pointed out that this breaks uses of console.log.bind(console). That's also incompatible with Terser's implementation of the feature, which is where this feature originally came from (it does support bind). So with this release, using this feature with esbuild will now only replace one level of method call (unless extended by call or apply) and will replace the method being called with an empty function in complex cases:

  // Original code
  const x = console.log('x')
  const y = console.log.call(console, 'y')
  const z = console.log.bind(console)('z')
  
  // Old output (with --drop-console)
  const x = void 0;
  const y = void 0;
  const z = (void 0)("z");
  
  // New output (with --drop-console)
  const x = void 0;
  const y = void 0;
  const z = (() => {
  }).bind(console)("z");
  

  This should more closely match Terser's existing behavior.

• Allow BigInt literals as define values

  With this release, you can now use BigInt literals as define values, such as with --define:FOO=123n. Previously trying to do this resulted in a syntax error.

• Fix a bug with resolve extensions in node_modules (#4053)

  The --resolve-extensions= option lets you specify the order in which to try resolving implicit file extensions. For complicated reasons, esbuild reorders TypeScript file extensions after JavaScript ones inside of node_modules so that JavaScript source code is always preferred to TypeScript source code inside of dependencies. However, this reordering had a bug that could accidentally change the relative order of TypeScript file extensions if one of them was a prefix of the other. That bug has been fixed in this release. You can see the issue for details.

• Better minification of statically-determined switch cases (#4028)

  With this release, esbuild will now try to trim unused code within switch statements when the test expression and case expressions are primitive literals. This can arise when the test expression is an identifier that is substituted for a primitive literal at compile time. For example:

  // Original code
  switch (MODE) {
    case 'dev':
      installDevToolsConsole()
      break
    case 'prod':
      return
    default:
      throw new Error
  }
  
  // Old output (with --minify '--define:MODE="prod"')
  switch("prod"){case"dev":installDevToolsConsole();break;case"prod":return;default:throw new Error}
  
  // New output (with --minify '--define:MODE="prod"')
  return;
  

• Emit /* @__KEY__ */ for string literals derived from property names (#4034)

  Property name mangling is an advanced feature that shortens certain property names for better minification (I say "advanced feature" because it's very easy to break your code with it). Sometimes you need to store a property name in a string, such as obj.get('foo') instead of obj.foo. JavaScript minifiers such as esbuild and Terser have a convention where a /* @__KEY__ */ comment before the string makes it behave like a property name. So obj.get(/* @__KEY__ */ 'foo') allows the contents of the string 'foo' to be shortened.

  However, esbuild sometimes itself generates string literals containing property names when transforming code, such as when lowering class fields to ES6 or when transforming TypeScript decorators. Previously esbuild didn't generate its own /* @__KEY__ */ comments in this case, which means that minifying your code by running esbuild again on its own output wouldn't work correctly (this does not affect people that both minify and transform their code in a single step).

  With this release, esbuild will now generate /* @__KEY__ */ comments for property names in generated string literals. To avoid lots of unnecessary output for people that don't use this advanced feature, the generated comments will only be present when the feature is active. If you want to generate the comments but not actually mangle any property names, you can use a flag that has no effect such as --reserve-props=., which tells esbuild to not mangle any property names (but still activates this feature).

• The text loader now strips the UTF-8 BOM if present (#3935)

  Some software (such as Notepad on Windows) can create text files that start with the three bytes 0xEF 0xBB 0xBF, which is referred to as the "byte order mark". This prefix is intended to be removed before using the text. Previously esbuild's text loader included this byte sequence in the string, which turns into a prefix of \uFEFF in a JavaScript string when decoded from UTF-8. With this release, esbuild's text loader will now remove these bytes when they occur at the start of the file.

• Omit legal comment output files when empty (#3670)

  Previously configuring esbuild with --legal-comment=external or --legal-comment=linked would always generate a .LEGAL.txt output file even if it was empty. Starting with this release, esbuild will now only do this if the file will be non-empty. This should result in a more organized output directory in some cases.

• Update Go from 1.23.1 to 1.23.5 (#4056, #4057)

  This should have no effect on existing code as this version change does not change Go's operating system support. It may remove certain reports from vulnerability scanners that detect which version of the Go compiler esbuild uses.

  This PR was contributed by @MikeWillCook.

• Allow passing a port of 0 to the development server (#3692)

  Unix sockets interpret a port of 0 to mean "pick a random unused port in the ephemeral port range". However, esbuild's default behavior when the port is not specified is to pick the first unused port starting from 8000 and upward. This is more convenient because port 8000 is typically free, so you can for example restart the development server and reload your app in the browser without needing to change the port in the URL. Since esbuild is written in Go (which does not have optional fields like JavaScript), not specifying the port in Go means it defaults to 0, so previously passing a port of 0 to esbuild caused port 8000 to be picked.

  Starting with this release, passing a port of 0 to esbuild when using the CLI or the JS API will now pass port 0 to the OS, which will pick a random ephemeral port. To make this possible, the Port option in the Go API has been changed from uint16 to int (to allow for additional sentinel values) and passing a port of -1 in Go now picks a random port. Both the CLI and JS APIs now remap an explicitly-provided port of 0 into -1 for the internal Go API.

  Another option would have been to change Port in Go from uint16 to *uint16 (Go's closest equivalent of number | undefined). However, that would make the common case of providing an explicit port in Go very awkward as Go doesn't support taking the address of integer constants. This tradeoff isn't worth it as picking a random ephemeral port is a rare use case. So the CLI and JS APIs should now match standard Unix behavior when the port is 0, but you need to use -1 instead with Go API.

• Minification now avoids inlining constants with direct eval (#4055)

  Direct eval can be used to introduce a new variable like this:

  const variable = false
  ;(function () {
    eval("var variable = true")
    console.log(variable)
  })()
  

  Previously esbuild inlined variable here (which became false), which changed the behavior of the code. This inlining is now avoided, but please keep in mind that direct eval breaks many assumptions that JavaScript tools hold about normal code (especially when bundling) and I do not recommend using it. There are usually better alternatives that have a more localized impact on your code. You can read more about this here: https://esbuild.github.io/link/direct-eval/

• Fix regression with --define and import.meta (#4010, #4012, #4013)

  The previous change in version 0.24.1 to use a more expression-like parser for define values to allow quoted property names introduced a regression that removed the ability to use --define:import.meta=.... Even though import is normally a keyword that can't be used as an identifier, ES modules special-case the import.meta expression to behave like an identifier anyway. This change fixes the regression.

  This fix was contributed by @sapphi-red.

• Allow es2024 as a target in tsconfig.json (#4004)

  TypeScript recently added es2024 as a compilation target, so esbuild now supports this in the target field of tsconfig.json files, such as in the following configuration file:

  {
    "compilerOptions": {
      "target": "ES2024"
    }
  }
  

  As a reminder, the only thing that esbuild uses this field for is determining whether or not to use legacy TypeScript behavior for class fields. You can read more in the documentation.

  This fix was contributed by @billyjanitsch.

• Allow automatic semicolon insertion after get/set

  This change fixes a grammar bug in the parser that incorrectly treated the following code as a syntax error:

  class Foo {
    get
    *x() {}
    set
    *y() {}
  }
  

  The above code will be considered valid starting with this release. This change to esbuild follows a similar change to TypeScript which will allow this syntax starting with TypeScript 5.7.

• Allow quoted property names in --define and --pure (#4008)

  The define and pure API options now accept identifier expressions containing quoted property names. Previously all identifiers in the identifier expression had to be bare identifiers. This change now makes --define and --pure consistent with --global-name, which already supported quoted property names. For example, the following is now possible:

  // The following code now transforms to "return true;\n"
  console.log(esbuild.transformSync(
    `return process.env['SOME-TEST-VAR']`,
    { define: { 'process.env["SOME-TEST-VAR"]': 'true' } },
  ))
  

  Note that if you're passing values like this on the command line using esbuild's --define flag, then you'll need to know how to escape quote characters for your shell. You may find esbuild's JavaScript API more ergonomic and portable than writing shell code.

• Minify empty try/catch/finally blocks (#4003)

  With this release, esbuild will now attempt to minify empty try blocks:

  // Original code
  try {} catch { foo() } finally { bar() }
  
  // Old output (with --minify)
  try{}catch{foo()}finally{bar()}
  
  // New output (with --minify)
  bar();
  

  This can sometimes expose additional minification opportunities.

• Include entryPoint metadata for the copy loader (#3985)

  Almost all entry points already include a entryPoint field in the outputs map in esbuild's build metadata. However, this wasn't the case for the copy loader as that loader is a special-case that doesn't behave like other loaders. This release adds the entryPoint field in this case.

• Source mappings may now contain null entries (#3310, #3878)

  With this change, sources that result in an empty source map may now emit a null source mapping (i.e. one with a generated position but without a source index or original position). This change improves source map accuracy by fixing a problem where minified code from a source without any source mappings could potentially still be associated with a mapping from another source file earlier in the generated output on the same minified line. It manifests as nonsensical files in source mapped stack traces. Now the null mapping "resets" the source map so that any lookups into the minified code without any mappings resolves to null (which appears as the output file in stack traces) instead of the incorrect source file.

  This change shouldn't affect anything in most situations. I'm only mentioning it in the release notes in case it introduces a bug with source mapping. It's part of a work-in-progress future feature that will let you omit certain unimportant files from the generated source map to reduce source map size.

• Avoid using the parent directory name for determinism (#3998)

  To make generated code more readable, esbuild includes the name of the source file when generating certain variable names within the file. Specifically bundling a CommonJS file generates a variable to store the lazily-evaluated module initializer. However, if a file is named index.js (or with a different extension), esbuild will use the name of the parent directory instead for a better name (since many packages have files all named index.js but have unique directory names).

  This is problematic when the bundle entry point is named index.js and the parent directory name is non-deterministic (e.g. a temporary directory created by a build script). To avoid non-determinism in esbuild's output, esbuild will now use index instead of the parent directory in this case. Specifically this will happen if the parent directory is equal to esbuild's outbase API option, which defaults to the lowest common ancestor of all user-specified entry point paths.

• Experimental support for esbuild on NetBSD (#3974)

  With this release, esbuild now has a published binary executable for NetBSD in the @esbuild/netbsd-arm64 npm package, and esbuild's installer has been modified to attempt to use it when on NetBSD. Hopefully this makes installing esbuild via npm work on NetBSD. This change was contributed by @bsiegert.

  ⚠️ Note: NetBSD is not one of Node's supported platforms, so installing esbuild may or may not work on NetBSD depending on how Node has been patched. This is not a problem with esbuild. ⚠️

This release deliberately contains backwards-incompatible changes. To avoid automatically picking up releases like this, you should either be pinning the exact version of esbuild in your package.json file (recommended) or be using a version range syntax that only accepts patch upgrades such as ^0.23.0 or ~0.23.0. See npm's documentation about semver for more information.

• Drop support for older platforms (#3902)

  This release drops support for the following operating system:

  • macOS 10.15 Catalina

  This is because the Go programming language dropped support for this operating system version in Go 1.23, and this release updates esbuild from Go 1.22 to Go 1.23. Go 1.23 now requires macOS 11 Big Sur or later.

  Note that this only affects the binary esbuild executables that are published to the esbuild npm package. It's still possible to compile esbuild's source code for these older operating systems. If you need to, you can compile esbuild for yourself using an older version of the Go compiler (before Go version 1.23). That might look something like this:

  git clone https://github.com/evanw/esbuild.git
  cd esbuild
  go build ./cmd/esbuild
  ./esbuild --version
  

• Fix class field decorators in TypeScript if useDefineForClassFields is false (#3913)

  Setting the useDefineForClassFields flag to false in tsconfig.json means class fields use the legacy TypeScript behavior instead of the standard JavaScript behavior. Specifically they use assign semantics instead of define semantics (e.g. setters are triggered) and fields without an initializer are not initialized at all. However, when this legacy behavior is combined with standard JavaScript decorators, TypeScript switches to always initializing all fields, even those without initializers. Previously esbuild incorrectly continued to omit field initializers for this edge case. These field initializers in this case should now be emitted starting with this release.

• Avoid incorrect cycle warning with tsconfig.json multiple inheritance (#3898)

  TypeScript 5.0 introduced multiple inheritance for tsconfig.json files where extends can be an array of file paths. Previously esbuild would incorrectly treat files encountered more than once when processing separate subtrees of the multiple inheritance hierarchy as an inheritance cycle. With this release, tsconfig.json files containing this edge case should work correctly without generating a warning.

• Handle Yarn Plug'n'Play stack overflow with tsconfig.json (#3915)

  Previously a tsconfig.json file that extends another file in a package with an exports map could cause a stack overflow when Yarn's Plug'n'Play resolution was active. This edge case should work now starting with this release.

• Work around more issues with Deno 1.31+ (#3917)

  This version of Deno broke the stdin and stdout properties on command objects for inherited streams, which matters when you run esbuild's Deno module as the entry point (i.e. when import.meta.main is true). Previously esbuild would crash in Deno 1.31+ if you ran esbuild like that. This should be fixed starting with this release.

  This fix was contributed by @Joshix-1.

• Allow using the node: import prefix with es* targets (#3821)

  The node: prefix on imports is an alternate way to import built-in node modules. For example, import fs from "fs" can also be written import fs from "node:fs". This only works with certain newer versions of node, so esbuild removes it when you target older versions of node such as with --target=node14 so that your code still works. With the way esbuild's platform-specific feature compatibility table works, this was added by saying that only newer versions of node support this feature. However, that means that a target such as --target=node18,es2022 removes the node: prefix because none of the es* targets are known to support this feature. This release adds the support for the node: flag to esbuild's internal compatibility table for es* to allow you to use compound targets like this:

  // Original code
  import fs from 'node:fs'
  fs.open
  
  // Old output (with --bundle --format=esm --platform=node --target=node18,es2022)
  import fs from "fs";
  fs.open;
  
  // New output (with --bundle --format=esm --platform=node --target=node18,es2022)
  import fs from "node:fs";
  fs.open;
  

• Fix a panic when using the CLI with invalid build flags if --analyze is present (#3834)

  Previously esbuild's CLI could crash if it was invoked with flags that aren't valid for a "build" API call and the --analyze flag is present. This was caused by esbuild's internals attempting to add a Go plugin (which is how --analyze is implemented) to a null build object. The panic has been fixed in this release.

• Fix incorrect location of certain error messages (#3845)

  This release fixes a regression that caused certain errors relating to variable declarations to be reported at an incorrect location. The regression was introduced in version 0.18.7 of esbuild.

• Print comments before case clauses in switch statements (#3838)

  With this release, esbuild will attempt to print comments that come before case clauses in switch statements. This is similar to what esbuild already does for comments inside of certain types of expressions. Note that these types of comments are not printed if minification is enabled (specifically whitespace minification).

• Fix a memory leak with pluginData (#3825)

  With this release, the build context's internal pluginData cache will now be cleared when starting a new build. This should fix a leak of memory from plugins that return pluginData objects from onResolve and/or onLoad callbacks.

This release deliberately contains backwards-incompatible changes. To avoid automatically picking up releases like this, you should either be pinning the exact version of esbuild in your package.json file (recommended) or be using a version range syntax that only accepts patch upgrades such as ^0.22.0 or ~0.22.0. See npm's documentation about semver for more information.

• Revert the recent change to avoid bundling dependencies for node (#3819)

  This release reverts the recent change in version 0.22.0 that made --packages=external the default behavior with --platform=node. The default is now back to --packages=bundle.

  I've just been made aware that Amazon doesn't pin their dependencies in their "AWS CDK" product, which means that whenever esbuild publishes a new release, many people (potentially everyone?) using their SDK around the world instantly starts using it without Amazon checking that it works first. This change in version 0.22.0 happened to break their SDK. I'm amazed that things haven't broken before this point. This revert attempts to avoid these problems for Amazon's customers. Hopefully Amazon will pin their dependencies in the future.

  In addition, this is probably a sign that esbuild is used widely enough that it now needs to switch to a more complicated release model. I may have esbuild use a beta channel model for further development.

• Fix preserving collapsed JSX whitespace (#3818)

  When transformed, certain whitespace inside JSX elements is ignored completely if it collapses to an empty string. However, the whitespace should only be ignored if the JSX is being transformed, not if it's being preserved. This release fixes a bug where esbuild was previously incorrectly ignoring collapsed whitespace with --jsx=preserve. Here is an example:

  // Original code
  <Foo>
    <Bar />
  </Foo>
  
  // Old output (with --jsx=preserve)
  <Foo><Bar /></Foo>;
  
  // New output (with --jsx=preserve)
  <Foo>
    <Bar />
  </Foo>;
  

This release deliberately contains backwards-incompatible changes. To avoid automatically picking up releases like this, you should either be pinning the exact version of esbuild in your package.json file (recommended) or be using a version range syntax that only accepts patch upgrades such as ^0.21.0 or ~0.21.0. See npm's documentation about semver for more information.

• Omit packages from bundles by default when targeting node (#1874, #2830, #2846, #2915, #3145, #3294, #3323, #3582, #3809, #3815)

  This breaking change is an experiment. People are commonly confused when using esbuild to bundle code for node (i.e. for --platform=node) because some packages may not be intended for bundlers, and may use node-specific features that don't work with a bundler. Even though esbuild's "getting started" instructions say to use --packages=external to work around this problem, many people don't read the documentation and don't do this, and are then confused when it doesn't work. So arguably this is a bad default behavior for esbuild to have if people keep tripping over this.

  With this release, esbuild will now omit packages from the bundle by default when the platform is node (i.e. the previous behavior of --packages=external is now the default in this case). Note that your dependencies must now be present on the file system when your bundle is run. If you don't want this behavior, you can do --packages=bundle to allow packages to be included in the bundle (i.e. the previous default behavior). Note that --packages=bundle doesn't mean all packages are bundled, just that packages are allowed to be bundled. You can still exclude individual packages from the bundle using --external: even when --packages=bundle is present.

  The --packages= setting considers all import paths that "look like" package imports in the original source code to be package imports. Specifically import paths that don't start with a path segment of / or . or .. are considered to be package imports. The only two exceptions to this rule are subpath imports (which start with a # character) and TypeScript path remappings via paths and/or baseUrl in tsconfig.json (which are applied first).

• Drop support for older platforms (#3802)

  This release drops support for the following operating systems:

  • Windows 7
  • Windows 8
  • Windows Server 2008
  • Windows Server 2012

  This is because the Go programming language dropped support for these operating system versions in Go 1.21, and this release updates esbuild from Go 1.20 to Go 1.22.

  Note that this only affects the binary esbuild executables that are published to the esbuild npm package. It's still possible to compile esbuild's source code for these older operating systems. If you need to, you can compile esbuild for yourself using an older version of the Go compiler (before Go version 1.21). That might look something like this:

  git clone https://github.com/evanw/esbuild.git
  cd esbuild
  go build ./cmd/esbuild
  ./esbuild.exe --version
  

  In addition, this release increases the minimum required node version for esbuild's JavaScript API from node 12 to node 18. Node 18 is the oldest version of node that is still being supported (see node's release schedule for more information). This increase is because of an incompatibility between the JavaScript that the Go compiler generates for the esbuild-wasm package and versions of node before node 17.4 (specifically the crypto.getRandomValues function).

• Update await using behavior to match TypeScript

  TypeScript 5.5 subtly changes the way await using behaves. This release updates esbuild to match these changes in TypeScript. You can read more about these changes in microsoft/TypeScript#58624.

• Allow es2024 as a target environment

  The ECMAScript 2024 specification was just approved, so it has been added to esbuild as a possible compilation target. You can read more about the features that it adds here: https://2ality.com/2024/06/ecmascript-2024.html. The only addition that's relevant for esbuild is the regular expression /v flag. With --target=es2024, regular expressions that use the /v flag will now be passed through untransformed instead of being transformed into a call to new RegExp.

• Publish binaries for OpenBSD on 64-bit ARM (#3665, #3674)

  With this release, you should now be able to install the esbuild npm package in OpenBSD on 64-bit ARM, such as on an Apple device with an M1 chip.

  This was contributed by @ikmckenz.

• Publish binaries for WASI (WebAssembly System Interface) preview 1 (#3300, #3779)

  The upcoming WASI (WebAssembly System Interface) standard is going to be a way to run WebAssembly outside of a JavaScript host environment. In this scenario you only need a .wasm file without any supporting JavaScript code. Instead of JavaScript providing the APIs for the host environment, the WASI standard specifies a "system interface" that WebAssembly code can access directly (e.g. for file system access).

  Development versions of the WASI specification are being released using preview numbers. The people behind WASI are currently working on preview 2 but the Go compiler has released support for preview 1, which from what I understand is now considered an unsupported legacy release. However, some people have requested that esbuild publish binary executables that support WASI preview 1 so they can experiment with them.

  This release publishes esbuild precompiled for WASI preview 1 to the @esbuild/wasi-preview1 package on npm (specifically the file @esbuild/wasi-preview1/esbuild.wasm). This binary executable has not been tested and won't be officially supported, as it's for an old preview release of a specification that has since moved in another direction. If it works for you, great! If not, then you'll likely have to wait for the ecosystem to evolve before using esbuild with WASI. For example, it sounds like perhaps WASI preview 1 doesn't include support for opening network sockets so esbuild's local development server is unlikely to work with WASI preview 1.

• Warn about onResolve plugins not setting a path (#3790)

  Plugins that return values from onResolve without resolving the path (i.e. without setting either path or external: true) will now cause a warning. This is because esbuild only uses return values from onResolve if it successfully resolves the path, and it's not good for invalid input to be silently ignored.

• Add a new Go API for running the CLI with plugins (#3539)

  With esbuild's Go API, you can now call cli.RunWithPlugins(args, plugins) to pass an array of esbuild plugins to be used during the build process. This allows you to create a CLI that behaves similarly to esbuild's CLI but with additional Go plugins enabled.

  This was contributed by @edewit.

• Fix Symbol.metadata on classes without a class decorator (#3781)

  This release fixes a bug with esbuild's support for the decorator metadata proposal. Previously esbuild only added the Symbol.metadata property to decorated classes if there was a decorator on the class element itself. However, the proposal says that the Symbol.metadata property should be present on all classes that have any decorators at all, not just those with a decorator on the class element itself.

• Allow unknown import attributes to be used with the copy loader (#3792)

  Import attributes (the with keyword on import statements) are allowed to alter how that path is loaded. For example, esbuild cannot assume that it knows how to load ./bagel.js as type bagel:

  // This is an error with "--bundle" without also using "--external:./bagel.js"
  import tasty from "./bagel.js" with { type: "bagel" }
  

  Because of that, bundling this code with esbuild is an error unless the file ./bagel.js is external to the bundle (such as with --bundle --external:./bagel.js).

  However, there is an additional case where it's ok for esbuild to allow this: if the file is loaded using the copy loader. That's because the copy loader behaves similarly to --external in that the file is left external to the bundle. The difference is that the copy loader copies the file into the output folder and rewrites the import path while --external doesn't. That means the following will now work with the copy loader (such as with --bundle --loader:.bagel=copy):

  // This is no longer an error with "--bundle" and "--loader:.bagel=copy"
  import tasty from "./tasty.bagel" with { type: "bagel" }
  

• Support import attributes with glob-style imports (#3797)

  This release adds support for import attributes (the with option) to glob-style imports (dynamic imports with certain string literal patterns as paths). These imports previously didn't support import attributes due to an oversight. So code like this will now work correctly:

  async function loadLocale(locale: string): Locale {
    const data = await import(`./locales/${locale}.data`, { with: { type: 'json' } })
    return unpackLocale(locale, data)
  }
  

  Previously this didn't work even though esbuild normally supports forcing the JSON loader using an import attribute. Attempting to do this used to result in the following error:

  ✘ [ERROR] No loader is configured for ".data" files: locales/en-US.data
  
      example.ts:2:28:
        2 │   const data = await import(`./locales/${locale}.data`, { with: { type: 'json' } })
          ╵                             ~~~~~~~~~~~~~~~~~~~~~~~~~~
  

  In addition, this change means plugins can now access the contents of with for glob-style imports.

• Support ${configDir} in tsconfig.json files (#3782)

  This adds support for a new feature from the upcoming TypeScript 5.5 release. The character sequence ${configDir} is now respected at the start of baseUrl and paths values, which are used by esbuild during bundling to correctly map import paths to file system paths. This feature lets base tsconfig.json files specified via extends refer to the directory of the top-level tsconfig.json file. Here is an example:

  {
    "compilerOptions": {
      "paths": {
        "js/*": ["${configDir}/dist/js/*"]
      }
    }
  }
  

  You can read more in TypeScript's blog post about their upcoming 5.5 release. Note that this feature does not make use of template literals (you need to use "${configDir}/dist/js/*" not `${configDir}/dist/js/*`). The syntax for tsconfig.json is still just JSON with comments, and JSON syntax does not allow template literals. This feature only recognizes ${configDir} in strings for certain path-like properties, and only at the beginning of the string.

• Fix internal error with --supported:object-accessors=false (#3794)

  This release fixes a regression in 0.21.0 where some code that was added to esbuild's internal runtime library of helper functions for JavaScript decorators fails to parse when you configure esbuild with --supported:object-accessors=false. The reason is that esbuild introduced code that does { get [name]() {} } which uses both the object-extensions feature for the [name] and the object-accessors feature for the get, but esbuild was incorrectly only checking for object-extensions and not for object-accessors. Additional tests have been added to avoid this type of issue in the future. A workaround for this issue in earlier releases is to also add --supported:object-extensions=false.

• Update support for import assertions and import attributes in node (#3778)

  Import assertions (the assert keyword) have been removed from node starting in v22.0.0. So esbuild will now strip them and generate a warning with --target=node22 or above:

  ▲ [WARNING] The "assert" keyword is not supported in the configured target environment ("node22") [assert-to-with]
  
      example.mjs:1:40:
        1 │ import json from "esbuild/package.json" assert { type: "json" }
          │                                         ~~~~~~
          ╵                                         with
  
    Did you mean to use "with" instead of "assert"?
  

  Import attributes (the with keyword) have been backported to node 18 starting in v18.20.0. So esbuild will no longer strip them with --target=node18.N if N is 20 or greater.

• Fix for await transform when a label is present

  This release fixes a bug where the for await transform, which wraps the loop in a try statement, previously failed to also move the loop's label into the try statement. This bug only affects code that uses both of these features in combination. Here's an example of some affected code:

  // Original code
  async function test() {
    outer: for await (const x of [Promise.resolve([0, 1])]) {
      for (const y of x) if (y) break outer
      throw 'fail'
    }
  }
  
  // Old output (with --target=es6)
  function test() {
    return __async(this, null, function* () {
      outer: try {
        for (var iter = __forAwait([Promise.resolve([0, 1])]), more, temp, error; more = !(temp = yield iter.next()).done; more = false) {
          const x = temp.value;
          for (const y of x) if (y) break outer;
          throw "fail";
        }
      } catch (temp) {
        error = [temp];
      } finally {
        try {
          more && (temp = iter.return) && (yield temp.call(iter));
        } finally {
          if (error)
            throw error[0];
        }
      }
    });
  }
  
  // New output (with --target=es6)
  function test() {
    return __async(this, null, function* () {
      try {
        outer: for (var iter = __forAwait([Promise.resolve([0, 1])]), more, temp, error; more = !(temp = yield iter.next()).done; more = false) {
          const x = temp.value;
          for (const y of x) if (y) break outer;
          throw "fail";
        }
      } catch (temp) {
        error = [temp];
      } finally {
        try {
          more && (temp = iter.return) && (yield temp.call(iter));
        } finally {
          if (error)
            throw error[0];
        }
      }
    });
  }
  

• Do additional constant folding after cross-module enum inlining (#3416, #3425)

  This release adds a few more cases where esbuild does constant folding after cross-module enum inlining.

  // Original code: enum.ts
  export enum Platform {
    WINDOWS = 'windows',
    MACOS = 'macos',
    LINUX = 'linux',
  }
  
  // Original code: main.ts
  import { Platform } from './enum';
  declare const PLATFORM: string;
  export function logPlatform() {
    if (PLATFORM == Platform.WINDOWS) console.log('Windows');
    else if (PLATFORM == Platform.MACOS) console.log('macOS');
    else if (PLATFORM == Platform.LINUX) console.log('Linux');
    else console.log('Other');
  }
  
  // Old output (with --bundle '--define:PLATFORM="macos"' --minify --format=esm)
  function n(){"windows"=="macos"?console.log("Windows"):"macos"=="macos"?console.log("macOS"):"linux"=="macos"?console.log("Linux"):console.log("Other")}export{n as logPlatform};
  
  // New output (with --bundle '--define:PLATFORM="macos"' --minify --format=esm)
  function n(){console.log("macOS")}export{n as logPlatform};
  

• Pass import attributes to on-resolve plugins (#3384, #3639, #3646)

  With this release, on-resolve plugins will now have access to the import attributes on the import via the with property of the arguments object. This mirrors the with property of the arguments object that's already passed to on-load plugins. In addition, you can now pass with to the resolve() API call which will then forward that value on to all relevant plugins. Here's an example of a plugin that can now be written:

  const examplePlugin = {
    name: 'Example plugin',
    setup(build) {
      build.onResolve({ filter: /.*/ }, args => {
        if (args.with.type === 'external')
          return { external: true }
      })
    }
  }
  
  require('esbuild').build({
    stdin: {
      contents: `
        import foo from "./foo" with { type: "external" }
        foo()
      `,
    },
    bundle: true,
    format: 'esm',
    write: false,
    plugins: [examplePlugin],
  }).then(result => {
    console.log(result.outputFiles[0].text)
  })
  

• Formatting support for the @position-try rule (#3773)

  Chrome shipped this new CSS at-rule in version 125 as part of the CSS anchor positioning API. With this release, esbuild now knows to expect a declaration list inside of the @position-try body block and will format it appropriately.

• Always allow internal string import and export aliases (#3343)

  Import and export names can be string literals in ES2022+. Previously esbuild forbid any usage of these aliases when the target was below ES2022. Starting with this release, esbuild will only forbid such usage when the alias would otherwise end up in output as a string literal. String literal aliases that are only used internally in the bundle and are "compiled away" are no longer errors. This makes it possible to use string literal aliases with esbuild's inject feature even when the target is earlier than ES2022.

• Implement the decorator metadata proposal (#3760)

  This release implements the decorator metadata proposal, which is a sub-proposal of the decorators proposal. Microsoft shipped the decorators proposal in TypeScript 5.0 and the decorator metadata proposal in TypeScript 5.2, so it's important that esbuild also supports both of these features. Here's a quick example:

  // Shim the "Symbol.metadata" symbol
  Symbol.metadata ??= Symbol('Symbol.metadata')
  
  const track = (_, context) => {
    (context.metadata.names ||= []).push(context.name)
  }
  
  class Foo {
    @track foo = 1
    @track bar = 2
  }
  
  // Prints ["foo", "bar"]
  console.log(Foo[Symbol.metadata].names)
  

  ⚠️ WARNING ⚠️

  This proposal has been marked as "stage 3" which means "recommended for implementation". However, it's still a work in progress and isn't a part of JavaScript yet, so keep in mind that any code that uses JavaScript decorator metadata may need to be updated as the feature continues to evolve. If/when that happens, I will update esbuild's implementation to match the specification. I will not be supporting old versions of the specification.

• Fix bundled decorators in derived classes (#3768)

  In certain cases, bundling code that uses decorators in a derived class with a class body that references its own class name could previously generate code that crashes at run-time due to an incorrect variable name. This problem has been fixed. Here is an example of code that was compiled incorrectly before this fix:

  class Foo extends Object {
    @(x => x) foo() {
      return Foo
    }
  }
  console.log(new Foo().foo())
  

• Fix tsconfig.json files inside symlinked directories (#3767)

  This release fixes an issue with a scenario involving a tsconfig.json file that extends another file from within a symlinked directory that uses the paths feature. In that case, the implicit baseURL value should be based on the real path (i.e. after expanding all symbolic links) instead of the original path. This was already done for other files that esbuild resolves but was not yet done for tsconfig.json because it's special-cased (the regular path resolver can't be used because the information inside tsconfig.json is involved in path resolution). Note that this fix no longer applies if the --preserve-symlinks setting is enabled.

• Correct this in field and accessor decorators (#3761)

  This release changes the value of this in initializers for class field and accessor decorators from the module-level this value to the appropriate this value for the decorated element (either the class or the instance). It was previously incorrect due to lack of test coverage. Here's an example of a decorator that doesn't work without this change:

  const dec = () => function() { this.bar = true }
  class Foo { @dec static foo }
  console.log(Foo.bar) // Should be "true"
  

• Allow es2023 as a target environment (#3762)

  TypeScript recently added es2023 as a compilation target, so esbuild now supports this too. There is no difference between a target of es2022 and es2023 as far as esbuild is concerned since the 2023 edition of JavaScript doesn't introduce any new syntax features.

• Fix a regression with --keep-names (#3756)

  The previous release introduced a regression with the --keep-names setting and object literals with get/set accessor methods, in which case the generated code contained syntax errors. This release fixes the regression:

  // Original code
  x = { get y() {} }
  
  // Output from version 0.21.0 (with --keep-names)
  x = { get y: /* @__PURE__ */ __name(function() {
  }, "y") };
  
  // Output from this version (with --keep-names)
  x = { get y() {
  } };
  

This release doesn't contain any deliberately-breaking changes. However, it contains a very complex new feature and while all of esbuild's tests pass, I would not be surprised if an important edge case turns out to be broken. So I'm releasing this as a breaking change release to avoid causing any trouble. As usual, make sure to test your code when you upgrade.

• Implement the JavaScript decorators proposal (#104)

  With this release, esbuild now contains an implementation of the upcoming JavaScript decorators proposal. This is the same feature that shipped in TypeScript 5.0 and has been highly-requested on esbuild's issue tracker. You can read more about them in that blog post and in this other (now slightly outdated) extensive blog post here: https://2ality.com/2022/10/javascript-decorators.html. Here's a quick example:

  const log = (fn, context) => function() {
    console.log(`before ${context.name}`)
    const it = fn.apply(this, arguments)
    console.log(`after ${context.name}`)
    return it
  }
  
  class Foo {
    @log static foo() {
      console.log('in foo')
    }
  }
  
  // Logs "before foo", "in foo", "after foo"
  Foo.foo()
  

  Note that this feature is different than the existing "TypeScript experimental decorators" feature that esbuild already implements. It uses similar syntax but behaves very differently, and the two are not compatible (although it's sometimes possible to write decorators that work with both). TypeScript experimental decorators will still be supported by esbuild going forward as they have been around for a long time, are very widely used, and let you do certain things that are not possible with JavaScript decorators (such as decorating function parameters). By default esbuild will parse and transform JavaScript decorators, but you can tell esbuild to parse and transform TypeScript experimental decorators instead by setting "experimentalDecorators": true in your tsconfig.json file.

  Probably at least half of the work for this feature went into creating a test suite that exercises many of the proposal's edge cases: https://github.com/evanw/decorator-tests. It has given me a reasonable level of confidence that esbuild's initial implementation is acceptable. However, I don't have access to a significant sample of real code that uses JavaScript decorators. If you're currently using JavaScript decorators in a real code base, please try out esbuild's implementation and let me know if anything seems off.

  ⚠️ WARNING ⚠️

  This proposal has been in the works for a very long time (work began around 10 years ago in 2014) and it is finally getting close to becoming part of the JavaScript language. However, it's still a work in progress and isn't a part of JavaScript yet, so keep in mind that any code that uses JavaScript decorators may need to be updated as the feature continues to evolve. The decorators proposal is pretty close to its final form but it can and likely will undergo some small behavioral adjustments before it ends up becoming a part of the standard. If/when that happens, I will update esbuild's implementation to match the specification. I will not be supporting old versions of the specification.

• Optimize the generated code for private methods

  Previously when lowering private methods for old browsers, esbuild would generate one WeakSet for each private method. This mirrors similar logic for generating one WeakSet for each private field. Using a separate WeakMap for private fields is necessary as their assignment can be observable:

  let it
  class Bar {
    constructor() {
      it = this
    }
  }
  class Foo extends Bar {
    #x = 1
    #y = null.foo
    static check() {
      console.log(#x in it, #y in it)
    }
  }
  try { new Foo } catch {}
  Foo.check()
  

  This prints true false because this partially-initialized instance has #x but not #y. In other words, it's not true that all class instances will always have all of their private fields. However, the assignment of private methods to a class instance is not observable. In other words, it's true that all class instances will always have all of their private methods. This means esbuild can lower private methods into code where all methods share a single WeakSet, which is smaller, faster, and uses less memory. Other JavaScript processing tools such as the TypeScript compiler already make this optimization. Here's what this change looks like:

  // Original code
  class Foo {
    #x() { return this.#x() }
    #y() { return this.#y() }
    #z() { return this.#z() }
  }
  
  // Old output (--supported:class-private-method=false)
  var _x, x_fn, _y, y_fn, _z, z_fn;
  class Foo {
    constructor() {
      __privateAdd(this, _x);
      __privateAdd(this, _y);
      __privateAdd(this, _z);
    }
  }
  _x = new WeakSet();
  x_fn = function() {
    return __privateMethod(this, _x, x_fn).call(this);
  };
  _y = new WeakSet();
  y_fn = function() {
    return __privateMethod(this, _y, y_fn).call(this);
  };
  _z = new WeakSet();
  z_fn = function() {
    return __privateMethod(this, _z, z_fn).call(this);
  };
  
  // New output (--supported:class-private-method=false)
  var _Foo_instances, x_fn, y_fn, z_fn;
  class Foo {
    constructor() {
      __privateAdd(this, _Foo_instances);
    }
  }
  _Foo_instances = new WeakSet();
  x_fn = function() {
    return __privateMethod(this, _Foo_instances, x_fn).call(this);
  };
  y_fn = function() {
    return __privateMethod(this, _Foo_instances, y_fn).call(this);
  };
  z_fn = function() {
    return __privateMethod(this, _Foo_instances, z_fn).call(this);
  };
  

• Fix an obscure bug with lowering class members with computed property keys

  When class members that use newer syntax features are transformed for older target environments, they sometimes need to be relocated. However, care must be taken to not reorder any side effects caused by computed property keys. For example, the following code must evaluate a() then b() then c():

  class Foo {
    [a()]() {}
    [b()];
    static { c() }
  }
  

  Previously esbuild did this by shifting the computed property key forward to the next spot in the evaluation order. Classes evaluate all computed keys first and then all static class elements, so if the last computed key needs to be shifted, esbuild previously inserted a static block at start of the class body, ensuring it came before all other static class elements:

  var _a;
  class Foo {
    constructor() {
      __publicField(this, _a);
    }
    static {
      _a = b();
    }
    [a()]() {
    }
    static {
      c();
    }
  }
  

  However, this could cause esbuild to accidentally generate a syntax error if the computed property key contains code that isn't allowed in a static block, such as an await expression. With this release, esbuild fixes this problem by shifting the computed property key backward to the previous spot in the evaluation order instead, which may push it into the extends clause or even before the class itself:

  // Original code
  class Foo {
    [a()]() {}
    [await b()];
    static { c() }
  }
  
  // Old output (with --supported:class-field=false)
  var _a;
  class Foo {
    constructor() {
      __publicField(this, _a);
    }
    static {
      _a = await b();
    }
    [a()]() {
    }
    static {
      c();
    }
  }
  
  // New output (with --supported:class-field=false)
  var _a, _b;
  class Foo {
    constructor() {
      __publicField(this, _a);
    }
    [(_b = a(), _a = await b(), _b)]() {
    }
    static {
      c();
    }
  }
  

• Fix some --keep-names edge cases

  The NamedEvaluation syntax-directed operation in the JavaScript specification gives certain anonymous expressions a name property depending on where they are in the syntax tree. For example, the following initializers convey a name value:

  var foo = function() {}
  var bar = class {}
  console.log(foo.name, bar.name)
  

  When you enable esbuild's --keep-names setting, esbuild generates additional code to represent this NamedEvaluation operation so that the value of the name property persists even when the identifiers are renamed (e.g. due to minification).

  However, I recently learned that esbuild's implementation of NamedEvaluation is missing a few cases. Specifically esbuild was missing property definitions, class initializers, logical-assignment operators. These cases should now all be handled:

  var obj = { foo: function() {} }
  class Foo0 { foo = function() {} }
  class Foo1 { static foo = function() {} }
  class Foo2 { accessor foo = function() {} }
  class Foo3 { static accessor foo = function() {} }
  foo ||= function() {}
  foo &&= function() {}
  foo ??= function() {}
  

• Support TypeScript experimental decorators on abstract class fields (#3684)

  With this release, you can now use TypeScript experimental decorators on abstract class fields. This was silently compiled incorrectly in esbuild 0.19.7 and below, and was an error from esbuild 0.19.8 to esbuild 0.20.1. Code such as the following should now work correctly:

  // Original code
  const log = (x: any, y: string) => console.log(y)
  abstract class Foo { @log abstract foo: string }
  new class extends Foo { foo = '' }
  
  // Old output (with --loader=ts --tsconfig-raw={\"compilerOptions\":{\"experimentalDecorators\":true}})
  const log = (x, y) => console.log(y);
  class Foo {
  }
  new class extends Foo {
    foo = "";
  }();
  
  // New output (with --loader=ts --tsconfig-raw={\"compilerOptions\":{\"experimentalDecorators\":true}})
  const log = (x, y) => console.log(y);
  class Foo {
  }
  __decorateClass([
    log
  ], Foo.prototype, "foo", 2);
  new class extends Foo {
    foo = "";
  }();
  

• JSON loader now preserves __proto__ properties (#3700)

  Copying JSON source code into a JavaScript file will change its meaning if a JSON object contains the __proto__ key. A literal __proto__ property in a JavaScript object literal sets the prototype of the object instead of adding a property named __proto__, while a literal __proto__ property in a JSON object literal just adds a property named __proto__. With this release, esbuild will now work around this problem by converting JSON to JavaScript with a computed property key in this case:

  // Original code
  import data from 'data:application/json,{"__proto__":{"fail":true}}'
  if (Object.getPrototypeOf(data)?.fail) throw 'fail'
  
  // Old output (with --bundle)
  (() => {
    // <data:application/json,{"__proto__":{"fail":true}}>
    var json_proto_fail_true_default = { __proto__: { fail: true } };
  
    // entry.js
    if (Object.getPrototypeOf(json_proto_fail_true_default)?.fail)
      throw "fail";
  })();
  
  // New output (with --bundle)
  (() => {
    // <data:application/json,{"__proto__":{"fail":true}}>
    var json_proto_fail_true_default = { ["__proto__"]: { fail: true } };
  
    // example.mjs
    if (Object.getPrototypeOf(json_proto_fail_true_default)?.fail)
      throw "fail";
  })();
  

• Improve dead code removal of switch statements (#3659)

  With this release, esbuild will now remove switch statements in branches when minifying if they are known to never be evaluated:

  // Original code
  if (true) foo(); else switch (bar) { case 1: baz(); break }
  
  // Old output (with --minify)
  if(1)foo();else switch(bar){case 1:}
  
  // New output (with --minify)
  foo();
  

• Empty enums should behave like an object literal (#3657)

  TypeScript allows you to create an empty enum and add properties to it at run time. While people usually use an empty object literal for this instead of a TypeScript enum, esbuild's enum transform didn't anticipate this use case and generated undefined instead of {} for an empty enum. With this release, you can now use an empty enum to generate an empty object literal.

  // Original code
  enum Foo {}
  
  // Old output (with --loader=ts)
  var Foo = /* @__PURE__ */ ((Foo2) => {
  })(Foo || {});
  
  // New output (with --loader=ts)
  var Foo = /* @__PURE__ */ ((Foo2) => {
    return Foo2;
  })(Foo || {});
  

• Handle Yarn Plug'n'Play edge case with tsconfig.json (#3698)

  Previously a tsconfig.json file that extends another file in a package with an exports map failed to work when Yarn's Plug'n'Play resolution was active. This edge case should work now starting with this release.

• Work around issues with Deno 1.31+ (#3682)

  Version 0.20.0 of esbuild changed how the esbuild child process is run in esbuild's API for Deno. Previously it used Deno.run but that API is being removed in favor of Deno.Command. As part of this change, esbuild is now calling the new unref function on esbuild's long-lived child process, which is supposed to allow Deno to exit when your code has finished running even though the child process is still around (previously you had to explicitly call esbuild's stop() function to terminate the child process for Deno to be able to exit).

  However, this introduced a problem for Deno's testing API which now fails some tests that use esbuild with error: Promise resolution is still pending but the event loop has already resolved. It's unclear to me why this is happening. The call to unref was recommended by someone on the Deno core team, and calling Node's equivalent unref API has been working fine for esbuild in Node for a long time. It could be that I'm using it incorrectly, or that there's some reference counting and/or garbage collection bug in Deno's internals, or that Deno's unref just works differently than Node's unref. In any case, it's not good for Deno tests that use esbuild to be failing.

  In this release, I am removing the call to unref to fix this issue. This means that you will now have to call esbuild's stop() function to allow Deno to exit, just like you did before esbuild version 0.20.0 when this regression was introduced.

  Note: This regression wasn't caught earlier because Deno doesn't seem to fail tests that have outstanding setTimeout calls, which esbuild's test harness was using to enforce a maximum test runtime. Adding a setTimeout was allowing esbuild's Deno tests to succeed. So this regression doesn't necessarily apply to all people using tests in Deno.

• Fix a bug with the CSS nesting transform (#3648)

  This release fixes a bug with the CSS nesting transform for older browsers where the generated CSS could be incorrect if a selector list contained a pseudo element followed by another selector. The bug was caused by incorrectly mutating the parent rule's selector list when filtering out pseudo elements for the child rules:

  /* Original code */
  .foo {
    &:after,
    & .bar {
      color: red;
    }
  }
  
  /* Old output (with --supported:nesting=false) */
  .foo .bar,
  .foo .bar {
    color: red;
  }
  
  /* New output (with --supported:nesting=false) */
  .foo:after,
  .foo .bar {
    color: red;
  }
  

• Constant folding for JavaScript inequality operators (#3645)

  This release introduces constant folding for the < > <= >= operators. The minifier will now replace these operators with true or false when both sides are compile-time numeric or string constants:

  // Original code
  console.log(1 < 2, '🍕' > '🧀')
  
  // Old output (with --minify)
  console.log(1<2,"🍕">"🧀");
  
  // New output (with --minify)
  console.log(!0,!1);
  

• Better handling of __proto__ edge cases (#3651)

  JavaScript object literal syntax contains a special case where a non-computed property with a key of __proto__ sets the prototype of the object. This does not apply to computed properties or to properties that use the shorthand property syntax introduced in ES6. Previously esbuild didn't correctly preserve the "sets the prototype" status of properties inside an object literal, meaning a property that sets the prototype could accidentally be transformed into one that doesn't and vice versa. This has now been fixed:

  // Original code
  function foo(__proto__) {
    return { __proto__: __proto__ } // Note: sets the prototype
  }
  function bar(__proto__, proto) {
    {
      let __proto__ = proto
      return { __proto__ } // Note: doesn't set the prototype
    }
  }
  
  // Old output
  function foo(__proto__) {
    return { __proto__ }; // Note: no longer sets the prototype (WRONG)
  }
  function bar(__proto__, proto) {
    {
      let __proto__2 = proto;
      return { __proto__: __proto__2 }; // Note: now sets the prototype (WRONG)
    }
  }
  
  // New output
  function foo(__proto__) {
    return { __proto__: __proto__ }; // Note: sets the prototype (correct)
  }
  function bar(__proto__, proto) {
    {
      let __proto__2 = proto;
      return { ["__proto__"]: __proto__2 }; // Note: doesn't set the prototype (correct)
    }
  }
  

• Fix cross-platform non-determinism with CSS color space transformations (#3650)

  The Go compiler takes advantage of "fused multiply and add" (FMA) instructions on certain processors which do the operation x*y + z without intermediate rounding. This causes esbuild's CSS color space math to differ on different processors (currently ppc64le and s390x), which breaks esbuild's guarantee of deterministic output. To avoid this, esbuild's color space math now inserts a float64() cast around every single math operation. This tells the Go compiler not to use the FMA optimization.

• Fix a crash when resolving a path from a directory that doesn't exist (#3634)

  This release fixes a regression where esbuild could crash when resolving an absolute path if the source directory for the path resolution operation doesn't exist. While this situation doesn't normally come up, it could come up when running esbuild concurrently with another operation that mutates the file system as esbuild is doing a build (such as using git to switch branches). The underlying problem was a regression that was introduced in version 0.18.0.

This release deliberately contains backwards-incompatible changes. To avoid automatically picking up releases like this, you should either be pinning the exact version of esbuild in your package.json file (recommended) or be using a version range syntax that only accepts patch upgrades such as ^0.19.0 or ~0.19.0. See npm's documentation about semver for more information.

This time there is only one breaking change, and it only matters for people using Deno. Deno tests that use esbuild will now fail unless you make the change described below.

• Work around API deprecations in Deno 1.40.x (#3609, #3611)

  Deno 1.40.0 was just released and introduced run-time warnings about certain APIs that esbuild uses. With this release, esbuild will work around these run-time warnings by using newer APIs if they are present and falling back to the original APIs otherwise. This should avoid the warnings without breaking compatibility with older versions of Deno.

  Unfortunately, doing this introduces a breaking change. The newer child process APIs lack a way to synchronously terminate esbuild's child process, so calling esbuild.stop() from within a Deno test is no longer sufficient to prevent Deno from failing a test that uses esbuild's API (Deno fails tests that create a child process without killing it before the test ends). To work around this, esbuild's stop() function has been changed to return a promise, and you now have to change esbuild.stop() to await esbuild.stop() in all of your Deno tests.

• Reorder implicit file extensions within node_modules (#3341, #3608)

  In version 0.18.0, esbuild changed the behavior of implicit file extensions within node_modules directories (i.e. in published packages) to prefer .js over .ts even when the --resolve-extensions= order prefers .ts over .js (which it does by default). However, doing that also accidentally made esbuild prefer .css over .ts, which caused problems for people that published packages containing both TypeScript and CSS in files with the same name.

  With this release, esbuild will reorder TypeScript file extensions immediately after the last JavaScript file extensions in the implicit file extension order instead of putting them at the end of the order. Specifically the default implicit file extension order is .tsx,.ts,.jsx,.js,.css,.json which used to become .jsx,.js,.css,.json,.tsx,.ts in node_modules directories. With this release it will now become .jsx,.js,.tsx,.ts,.css,.json instead.

  Why even rewrite the implicit file extension order at all? One reason is because the .js file is more likely to behave correctly than the .ts file. The behavior of the .ts file may depend on tsconfig.json and the tsconfig.json file may not even be published, or may use extends to refer to a base tsconfig.json file that wasn't published. People can get into this situation when they forget to add all .ts files to their .npmignore file before publishing to npm. Picking .js over .ts helps make it more likely that resulting bundle will behave correctly.

• The "preserve" JSX mode now preserves JSX text verbatim (#3605)

  The JSX specification deliberately doesn't specify how JSX text is supposed to be interpreted and there is no canonical way to interpret JSX text. Two most popular interpretations are Babel and TypeScript. Yes they are different (esbuild deliberately follows TypeScript by the way).

  Previously esbuild normalized text to the TypeScript interpretation when the "preserve" JSX mode is active. However, "preserve" should arguably reproduce the original JSX text verbatim so that whatever JSX transform runs after esbuild is free to interpret it however it wants. So with this release, esbuild will now pass JSX text through unmodified:

  // Original code
  let el =
    <a href={'/'} title='&apos;&quot;'> some text
      {foo}
        more text </a>
  
  // Old output (with --loader=jsx --jsx=preserve)
  let el = <a href="/" title={`'"`}>
    {" some text"}
    {foo}
    {"more text "}
  </a>;
  
  // New output (with --loader=jsx --jsx=preserve)
  let el = <a href={"/"} title='&apos;&quot;'> some text
      {foo}
        more text </a>;
  

• Allow JSX elements as JSX attribute values

  JSX has an obscure feature where you can use JSX elements in attribute position without surrounding them with {...}. It looks like this:

  let el = <div data-ab=<><a/><b/></>/>;
  

  I think I originally didn't implement it even though it's part of the JSX specification because it previously didn't work in TypeScript (and potentially also in Babel?). However, support for it was silently added in TypeScript 4.8 without me noticing and Babel has also since fixed their bugs regarding this feature. So I'm adding it to esbuild too now that I know it's widely supported.

  Keep in mind that there is some ongoing discussion about removing this feature from JSX. I agree that the syntax seems out of place (it does away with the elegance of "JSX is basically just XML with {...} escapes" for something arguably harder to read, which doesn't seem like a good trade-off), but it's in the specification and TypeScript and Babel both implement it so I'm going to have esbuild implement it too. However, I reserve the right to remove it from esbuild if it's ever removed from the specification in the future. So use it with caution.

• Fix a bug with TypeScript type parsing (#3574)

  This release fixes a bug with esbuild's TypeScript parser where a conditional type containing a union type that ends with an infer type that ends with a constraint could fail to parse. This was caused by the "don't parse a conditional type" flag not getting passed through the union type parser. Here's an example of valid TypeScript code that previously failed to parse correctly:

  type InferUnion<T> = T extends { a: infer U extends number } | infer U extends number ? U : never
  

• Fix TypeScript-specific class transform edge case (#3559)

  The previous release introduced an optimization that avoided transforming super() in the class constructor for TypeScript code compiled with useDefineForClassFields set to false if all class instance fields have no initializers. The rationale was that in this case, all class instance fields are omitted in the output so no changes to the constructor are needed. However, if all of this is the case and there are #private instance fields with initializers, those private instance field initializers were still being moved into the constructor. This was problematic because they were being inserted before the call to super() (since super() is now no longer transformed in that case). This release introduces an additional optimization that avoids moving the private instance field initializers into the constructor in this edge case, which generates smaller code, matches the TypeScript compiler's output more closely, and avoids this bug:

  // Original code
  class Foo extends Bar {
    #private = 1;
    public: any;
    constructor() {
      super();
    }
  }
  
  // Old output (with esbuild v0.19.9)
  class Foo extends Bar {
    constructor() {
      super();
      this.#private = 1;
    }
    #private;
  }
  
  // Old output (with esbuild v0.19.10)
  class Foo extends Bar {
    constructor() {
      this.#private = 1;
      super();
    }
    #private;
  }
  
  // New output
  class Foo extends Bar {
    #private = 1;
    constructor() {
      super();
    }
  }
  

• Minifier: allow reording a primitive past a side-effect (#3568)

  The minifier previously allowed reordering a side-effect past a primitive, but didn't handle the case of reordering a primitive past a side-effect. This additional case is now handled:

  // Original code
  function f() {
    let x = false;
    let y = x;
    const boolean = y;
    let frag = $.template(`<p contenteditable="${boolean}">hello world</p>`);
    return frag;
  }
  
  // Old output (with --minify)
  function f(){const e=!1;return $.template(`<p contenteditable="${e}">hello world</p>`)}
  
  // New output (with --minify)
  function f(){return $.template('<p contenteditable="false">hello world</p>')}
  

• Minifier: consider properties named using known Symbol instances to be side-effect free (#3561)

  Many things in JavaScript can have side effects including property accesses and ToString operations, so using a symbol such as Symbol.iterator as a computed property name is not obviously side-effect free. This release adds a special case for known Symbol instances so that they are considered side-effect free when used as property names. For example, this class declaration will now be considered side-effect free:

  class Foo {
    *[Symbol.iterator]() {
    }
  }
  

• Provide the stop() API in node to exit esbuild's child process (#3558)

  You can now call stop() in esbuild's node API to exit esbuild's child process to reclaim the resources used. It only makes sense to do this for a long-lived node process when you know you will no longer be making any more esbuild API calls. It is not necessary to call this to allow node to exit, and it's advantageous to not call this in between calls to esbuild's API as sharing a single long-lived esbuild child process is more efficient than re-creating a new esbuild child process for every API call. This API call used to exist but was removed in version 0.9.0. This release adds it back due to a user request.

• Fix glob imports in TypeScript files (#3319)

  This release fixes a problem where bundling a TypeScript file containing a glob import could emit a call to a helper function that doesn't exist. The problem happened because esbuild's TypeScript transformation removes unused imports (which is required for correctness, as they may be type-only imports) and esbuild's glob import transformation wasn't correctly marking the imported helper function as used. This wasn't caught earlier because most of esbuild's glob import tests were written in JavaScript, not in TypeScript.

• Fix require() glob imports with bundling disabled (#3546)

  Previously require() calls containing glob imports were incorrectly transformed when bundling was disabled. All glob imports should only be transformed when bundling is enabled. This bug has been fixed.

• Fix a panic when transforming optional chaining with define (#3551, #3554)

  This release fixes a case where esbuild could crash with a panic, which was triggered by using define to replace an expression containing an optional chain. Here is an example:

  // Original code
  console.log(process?.env.SHELL)
  
  // Old output (with --define:process.env={})
  /* panic: Internal error (while parsing "<stdin>") */
  
  // New output (with --define:process.env={})
  var define_process_env_default = {};
  console.log(define_process_env_default.SHELL);
  

  This fix was contributed by @hi-ogawa.

• Work around a bug in node's CommonJS export name detector (#3544)

  The export names of a CommonJS module are dynamically-determined at run time because CommonJS exports are properties on a mutable object. But the export names of an ES module are statically-determined at module instantiation time by using import and export syntax and cannot be changed at run time.

  When you import a CommonJS module into an ES module in node, node scans over the source code to attempt to detect the set of export names that the CommonJS module will end up using. That statically-determined set of names is used as the set of names that the ES module is allowed to import at module instantiation time. However, this scan appears to have bugs (or at least, can cause false positives) because it doesn't appear to do any scope analysis. Node will incorrectly consider the module to export something even if the assignment is done to a local variable instead of to the module-level exports object. For example:

  // confuseNode.js
  exports.confuseNode = function(exports) {
    // If this local is called "exports", node incorrectly
    // thinks this file has an export called "notAnExport".
    exports.notAnExport = function() {
    };
  };
  

  You can see that node incorrectly thinks the file confuseNode.js has an export called notAnExport when that file is loaded in an ES module context:

  $ node -e 'import("./confuseNode.js").then(console.log)'
  [Module: null prototype] {
    confuseNode: [Function (anonymous)],
    default: { confuseNode: [Function (anonymous)] },
    notAnExport: undefined
  }
  

  To avoid this, esbuild will now rename local variables that use the names exports and module when generating CommonJS output for the node platform.

• Fix the return value of esbuild's super() shim (#3538)

  Some people write constructor methods that use the return value of super() instead of using this. This isn't too common because TypeScript doesn't let you do that but it can come up when writing JavaScript. Previously esbuild's class lowering transform incorrectly transformed the return value of super() into undefined. With this release, the return value of super() will now be this instead:

  // Original code
  class Foo extends Object {
    field
    constructor() {
      console.log(typeof super())
    }
  }
  new Foo
  
  // Old output (with --target=es6)
  class Foo extends Object {
    constructor() {
      var __super = (...args) => {
        super(...args);
        __publicField(this, "field");
      };
      console.log(typeof __super());
    }
  }
  new Foo();
  
  // New output (with --target=es6)
  class Foo extends Object {
    constructor() {
      var __super = (...args) => {
        super(...args);
        __publicField(this, "field");
        return this;
      };
      console.log(typeof __super());
    }
  }
  new Foo();
  

• Terminate the Go GC when esbuild's stop() API is called (#3552)

  If you use esbuild with WebAssembly and pass the worker: false flag to esbuild.initialize(), then esbuild will run the WebAssembly module on the main thread. If you do this within a Deno test and that test calls esbuild.stop() to clean up esbuild's resources, Deno may complain that a setTimeout() call lasted past the end of the test. This happens when the Go is in the middle of a garbage collection pass and has scheduled additional ongoing garbage collection work. Normally calling esbuild.stop() will terminate the web worker that the WebAssembly module runs in, which will terminate the Go GC, but that doesn't happen if you disable the web worker with worker: false.

  With this release, esbuild will now attempt to terminate the Go GC in this edge case by calling clearTimeout() on these pending timeouts.

• Apply /* @__NO_SIDE_EFFECTS__ */ on tagged template literals (#3511)

  Tagged template literals that reference functions annotated with a @__NO_SIDE_EFFECTS__ comment are now able to be removed via tree-shaking if the result is unused. This is a convention from Rollup. Here is an example:

  // Original code
  const html = /* @__NO_SIDE_EFFECTS__ */ (a, ...b) => ({ a, b })
  html`<a>remove</a>`
  x = html`<b>keep</b>`
  
  // Old output (with --tree-shaking=true)
  const html = /* @__NO_SIDE_EFFECTS__ */ (a, ...b) => ({ a, b });
  html`<a>remove</a>`;
  x = html`<b>keep</b>`;
  
  // New output (with --tree-shaking=true)
  const html = /* @__NO_SIDE_EFFECTS__ */ (a, ...b) => ({ a, b });
  x = html`<b>keep</b>`;
  

  Note that this feature currently only works within a single file, so it's not especially useful. This feature does not yet work across separate files. I still recommend using @__PURE__ annotations instead of this feature, as they have wider tooling support. The drawback of course is that @__PURE__ annotations need to be added at each call site, not at the declaration, and for non-call expressions such as template literals you need to wrap the expression in an IIFE (immediately-invoked function expression) to create a call expression to apply the @__PURE__ annotation to.

• Publish builds for IBM AIX PowerPC 64-bit (#3549)

  This release publishes a binary executable to npm for IBM AIX PowerPC 64-bit, which means that in theory esbuild can now be installed in that environment with npm install esbuild. This hasn't actually been tested yet. If you have access to such a system, it would be helpful to confirm whether or not doing this actually works.

• Add support for transforming new CSS gradient syntax for older browsers

  The specification called CSS Images Module Level 4 introduces new CSS gradient syntax for customizing how the browser interpolates colors in between color stops. You can now control the color space that the interpolation happens in as well as (for "polar" color spaces) control whether hue angle interpolation happens clockwise or counterclockwise. You can read more about this in Mozilla's blog post about new CSS gradient features.

  With this release, esbuild will now automatically transform this syntax for older browsers in the target list. For example, here's a gradient that should appear as a rainbow in a browser that supports this new syntax:

  /* Original code */
  .rainbow-gradient {
    width: 100px;
    height: 100px;
    background: linear-gradient(in hsl longer hue, #7ff, #77f);
  }
  
  /* New output (with --target=chrome99) */
  .rainbow-gradient {
    width: 100px;
    height: 100px;
    background:
      linear-gradient(
        #77ffff,
        #77ffaa 12.5%,
        #77ff80 18.75%,
        #84ff77 21.88%,
        #99ff77 25%,
        #eeff77 37.5%,
        #fffb77 40.62%,
        #ffe577 43.75%,
        #ffbb77 50%,
        #ff9077 56.25%,
        #ff7b77 59.38%,
        #ff7788 62.5%,
        #ff77dd 75%,
        #ff77f2 78.12%,
        #f777ff 81.25%,
        #cc77ff 87.5%,
        #7777ff);
  }
  

  You can now use this syntax in your CSS source code and esbuild will automatically convert it to an equivalent gradient for older browsers. In addition, esbuild will now also transform "double position" and "transition hint" syntax for older browsers as appropriate:

  /* Original code */
  .stripes {
    width: 100px;
    height: 100px;
    background: linear-gradient(#e65 33%, #ff2 33% 67%, #99e 67%);
  }
  .glow {
    width: 100px;
    height: 100px;
    background: radial-gradient(white 10%, 20%, black);
  }
  
  /* New output (with --target=chrome33) */
  .stripes {
    width: 100px;
    height: 100px;
    background:
      linear-gradient(
        #e65 33%,
        #ff2 33%,
        #ff2 67%,
        #99e 67%);
  }
  .glow {
    width: 100px;
    height: 100px;
    background:
      radial-gradient(
        #ffffff 10%,
        #aaaaaa 12.81%,
        #959595 15.62%,
        #7b7b7b 21.25%,
        #5a5a5a 32.5%,
        #444444 43.75%,
        #323232 55%,
        #161616 77.5%,
        #000000);
  }
  

  You can see visual examples of these new syntax features by looking at esbuild's gradient transformation tests.

  If necessary, esbuild will construct a new gradient that approximates the original gradient by recursively splitting the interval in between color stops until the approximation error is within a small threshold. That is why the above output CSS contains many more color stops than the input CSS.

  Note that esbuild deliberately replaces the original gradient with the approximation instead of inserting the approximation before the original gradient as a fallback. The latest version of Firefox has multiple gradient rendering bugs (including incorrect interpolation of partially-transparent colors and interpolating non-sRGB colors using the incorrect color space). If esbuild didn't replace the original gradient, then Firefox would use the original gradient instead of the fallback the appearance would be incorrect in Firefox. In other words, the latest version of Firefox supports modern gradient syntax but interprets it incorrectly.

• Add support for color(), lab(), lch(), oklab(), oklch(), and hwb() in CSS

  CSS has recently added lots of new ways of specifying colors. You can read more about this in Chrome's blog post about CSS color spaces.

  This release adds support for minifying colors that use the color(), lab(), lch(), oklab(), oklch(), or hwb() syntax and/or transforming these colors for browsers that don't support it yet:

  /* Original code */
  div {
    color: hwb(90deg 20% 40%);
    background: color(display-p3 1 0 0);
  }
  
  /* New output (with --target=chrome99) */
  div {
    color: #669933;
    background: #ff0f0e;
    background: color(display-p3 1 0 0);
  }
  

  As you can see, colors outside of the sRGB color space such as color(display-p3 1 0 0) are mapped back into the sRGB gamut and inserted as a fallback for browsers that don't support the new color syntax.

• Allow empty type parameter lists in certain cases (#3512)

  TypeScript allows interface declarations and type aliases to have empty type parameter lists. Previously esbuild didn't handle this edge case but with this release, esbuild will now parse this syntax:

  interface Foo<> {}
  type Bar<> = {}
  

  This fix was contributed by @magic-akari.

• Add a treemap chart to esbuild's bundle analyzer (#2848)

  The bundler analyzer on esbuild's website (https://esbuild.github.io/analyze/) now has a treemap chart type in addition to the two existing chart types (sunburst and flame). This should be more familiar for people coming from other similar tools, as well as make better use of large screens.

• Allow decorators after the export keyword (#104)

  Previously esbuild's decorator parser followed the original behavior of TypeScript's experimental decorators feature, which only allowed decorators to come before the export keyword. However, the upcoming JavaScript decorators feature also allows decorators to come after the export keyword. And with TypeScript 5.0, TypeScript now also allows experimental decorators to come after the export keyword too. So esbuild now allows this as well:

  // This old syntax has always been permitted:
  @decorator export class Foo {}
  @decorator export default class Foo {}
  
  // This new syntax is now permitted too:
  export @decorator class Foo {}
  export default @decorator class Foo {}
  

  In addition, esbuild's decorator parser has been rewritten to fix several subtle and likely unimportant edge cases with esbuild's parsing of exports and decorators in TypeScript (e.g. TypeScript apparently does automatic semicolon insertion after interface and export interface but not after export default interface).

• Pretty-print decorators using the same whitespace as the original

  When printing code containing decorators, esbuild will now try to respect whether the original code contained newlines after the decorator or not. This can make generated code containing many decorators much more compact to read:

  // Original code
  class Foo {
    @a @b @c abc
    @x @y @z xyz
  }
  
  // Old output
  class Foo {
    @a
    @b
    @c
    abc;
    @x
    @y
    @z
    xyz;
  }
  
  // New output
  class Foo {
    @a @b @c abc;
    @x @y @z xyz;
  }
  

• Add support for bundling code that uses import attributes (#3384)

  JavaScript is gaining new syntax for associating a map of string key-value pairs with individual ESM imports. The proposal is still a work in progress and is still undergoing significant changes before being finalized. However, the first iteration has already been shipping in Chromium-based browsers for a while, and the second iteration has landed in V8 and is now shipping in node, so it makes sense for esbuild to support it. Here are the two major iterations of this proposal (so far):

  1. Import assertions (deprecated, will not be standardized)

     • Uses the assert keyword
     • Does not affect module resolution
     • Causes an error if the assertion fails
     • Shipping in Chrome 91+ (and in esbuild 0.11.22+)

  2. Import attributes (currently set to become standardized)

     • Uses the with keyword
     • Affects module resolution
     • Unknown attributes cause an error
     • Shipping in node 21+

  You can already use esbuild to bundle code that uses import assertions (the first iteration). However, this feature is mostly useless for bundlers because import assertions are not allowed to affect module resolution. It's basically only useful as an annotation on external imports, which esbuild will then preserve in the output for use in a browser (which would otherwise refuse to load certain imports).

  With this release, esbuild now supports bundling code that uses import attributes (the second iteration). This is much more useful for bundlers because they are allowed to affect module resolution, which means the key-value pairs can be provided to plugins. Here's an example, which uses esbuild's built-in support for the upcoming JSON module standard:

  // On static imports
  import foo from './package.json' with { type: 'json' }
  console.log(foo)
  
  // On dynamic imports
  const bar = await import('./package.json', { with: { type: 'json' } })
  console.log(bar)
  

  One important consequence of the change in semantics between import assertions and import attributes is that two imports with identical paths but different import attributes are now considered to be different modules. This is because the import attributes are provided to the loader, which might then use those attributes during loading. For example, you could imagine an image loader that produces an image of a different size depending on the import attributes.

  Import attributes are now reported in the metafile and are now provided to on-load plugins as a map in the with property. For example, here's an esbuild plugin that turns all imports with a type import attribute equal to 'cheese' into a module that exports the cheese emoji:

  const cheesePlugin = {
    name: 'cheese',
    setup(build) {
      build.onLoad({ filter: /.*/ }, args => {
        if (args.with.type === 'cheese') return {
          contents: `export default "🧀"`,
        }
      })
    }
  }
  
  require('esbuild').build({
    bundle: true,
    write: false,
    stdin: {
      contents: `
        import foo from 'data:text/javascript,' with { type: 'cheese' }
        console.log(foo)
      `,
    },
    plugins: [cheesePlugin],
  }).then(result => {
    const code = new Function(result.outputFiles[0].text)
    code()
  })
  

  Warning: It's possible that the second iteration of this feature may change significantly again even though it's already shipping in real JavaScript VMs (since it has already happened once before). In that case, esbuild may end up adjusting its implementation to match the eventual standard behavior. So keep in mind that by using this, you are using an unstable upcoming JavaScript feature that may undergo breaking changes in the future.

• Adjust TypeScript experimental decorator behavior (#3230, #3326, #3394)

  With this release, esbuild will now allow TypeScript experimental decorators to access both static class properties and #private class names. For example:

  const check =
    <T,>(a: T, b: T): PropertyDecorator =>
      () => console.log(a === b)
  
  async function test() {
    class Foo {
      static #foo = 1
      static bar = 1 + Foo.#foo
      @check(Foo.#foo, 1) a: any
      @check(Foo.bar, await Promise.resolve(2)) b: any
    }
  }
  
  test().then(() => console.log('pass'))
  

  This will now print true true pass when compiled by esbuild. Previously esbuild evaluated TypeScript decorators outside of the class body, so it didn't allow decorators to access Foo or #foo. Now esbuild does something different, although it's hard to concisely explain exactly what esbuild is doing now (see the background section below for more information).

  Note that TypeScript's experimental decorator support is currently buggy: TypeScript's compiler passes this test if only the first @check is present or if only the second @check is present, but TypeScript's compiler fails this test if both checks are present together. I haven't changed esbuild to match TypeScript's behavior exactly here because I'm waiting for TypeScript to fix these bugs instead.

  Some background: TypeScript experimental decorators don't have consistent semantics regarding the context that the decorators are evaluated in. For example, TypeScript will let you use await within a decorator, which implies that the decorator runs outside the class body (since await isn't supported inside a class body), but TypeScript will also let you use #private names, which implies that the decorator runs inside the class body (since #private names are only supported inside a class body). The value of this in a decorator is also buggy (the run-time value of this changes if any decorator in the class uses a #private name but the type of this doesn't change, leading to the type checker no longer matching reality). These inconsistent semantics make it hard for esbuild to implement this feature as decorator evaluation happens in some superposition of both inside and outside the class body that is particular to the internal implementation details of the TypeScript compiler.

• Forbid --keep-names when targeting old browsers (#3477)

  The --keep-names setting needs to be able to assign to the name property on functions and classes. However, before ES6 this property was non-configurable, and attempting to assign to it would throw an error. So with this release, esbuild will no longer allow you to enable this setting while also targeting a really old browser.

• Fix a constant folding bug with bigint equality

  This release fixes a bug where esbuild incorrectly checked for bigint equality by checking the equality of the bigint literal text. This is correct if the bigint doesn't have a radix because bigint literals without a radix are always in canonical form (since leading zeros are not allowed). However, this is incorrect if the bigint has a radix (e.g. 0x123n) because the canonical form is not enforced when a radix is present.

  // Original code
  console.log(!!0n, !!1n, 123n === 123n)
  console.log(!!0x0n, !!0x1n, 123n === 0x7Bn)
  
  // Old output
  console.log(false, true, true);
  console.log(true, true, false);
  
  // New output
  console.log(false, true, true);
  console.log(!!0x0n, !!0x1n, 123n === 0x7Bn);
  

• Add some improvements to the JavaScript minifier

  This release adds more cases to the JavaScript minifier, including support for inlining String.fromCharCode and String.prototype.charCodeAt when possible:

  // Original code
  document.onkeydown = e => e.keyCode === 'A'.charCodeAt(0) && console.log(String.fromCharCode(55358, 56768))
  
  // Old output (with --minify)
  document.onkeydown=o=>o.keyCode==="A".charCodeAt(0)&&console.log(String.fromCharCode(55358,56768));
  
  // New output (with --minify)
  document.onkeydown=o=>o.keyCode===65&&console.log("🧀");
  

  In addition, immediately-invoked function expressions (IIFEs) that return a single expression are now inlined when minifying. This makes it possible to use IIFEs in combination with @__PURE__ annotations to annotate arbitrary expressions as side-effect free without the IIFE wrapper impacting code size. For example:

  // Original code
  const sideEffectFreeOffset = /* @__PURE__ */ (() => computeSomething())()
  use(sideEffectFreeOffset)
  
  // Old output (with --minify)
  const e=(()=>computeSomething())();use(e);
  
  // New output (with --minify)
  const e=computeSomething();use(e);
  

• Automatically prefix the mask-composite CSS property for WebKit (#3493)

  The mask-composite property will now be prefixed as -webkit-mask-composite for older WebKit-based browsers. In addition to prefixing the property name, handling older browsers also requires rewriting the values since WebKit uses non-standard names for the mask composite modes:

  /* Original code */
  div {
    mask-composite: add, subtract, intersect, exclude;
  }
  
  /* New output (with --target=chrome100) */
  div {
    -webkit-mask-composite:
      source-over,
      source-out,
      source-in,
      xor;
    mask-composite:
      add,
      subtract,
      intersect,
      exclude;
  }
  

• Avoid referencing this from JSX elements in derived class constructors (#3454)

  When you enable --jsx=automatic and --jsx-dev, the JSX transform is supposed to insert this as the last argument to the jsxDEV function. I'm not sure exactly why this is and I can't find any specification for it, but in any case this causes the generated code to crash when you use a JSX element in a derived class constructor before the call to super() as this is not allowed to be accessed at that point. For example

  // Original code
  class ChildComponent extends ParentComponent {
    constructor() {
      super(<div />)
    }
  }
  
  // Problematic output (with --loader=jsx --jsx=automatic --jsx-dev)
  import { jsxDEV } from "react/jsx-dev-runtime";
  class ChildComponent extends ParentComponent {
    constructor() {
      super(/* @__PURE__ */ jsxDEV("div", {}, void 0, false, {
        fileName: "<stdin>",
        lineNumber: 3,
        columnNumber: 15
      }, this)); // The reference to "this" crashes here
    }
  }
  

  The TypeScript compiler doesn't handle this at all while the Babel compiler just omits this for the entire constructor (even after the call to super()). There seems to be no specification so I can't be sure that this change doesn't break anything important. But given that Babel is pretty loose with this and TypeScript doesn't handle this at all, I'm guessing this value isn't too important. React's blog post seems to indicate that this value was intended to be used for a React-specific migration warning at some point, so it could even be that this value is irrelevant now. Anyway the crash in this case should now be fixed.

• Allow package subpath imports to map to node built-ins (#3485)

  You are now able to use a subpath import in your package to resolve to a node built-in module. For example, with a package.json file like this:

  {
    "type": "module",
    "imports": {
      "#stream": {
        "node": "stream",
        "default": "./stub.js"
      }
    }
  }
  

  You can now import from node's stream module like this:

  import * as stream from '#stream';
  console.log(Object.keys(stream));
  

  This will import from node's stream module when the platform is node and from ./stub.js otherwise.

• No longer throw an error when a Symbol is missing (#3453)

  Certain JavaScript syntax features use special properties on the global Symbol object. For example, the asynchronous iteration syntax uses Symbol.asyncIterator. Previously esbuild's generated code for older browsers required this symbol to be polyfilled. However, starting with this release esbuild will use Symbol.for() to construct these symbols if they are missing instead of throwing an error about a missing polyfill. This means your code no longer needs to include a polyfill for missing symbols as long as your code also uses Symbol.for() for missing symbols.

• Parse upcoming changes to TypeScript syntax (#3490, #3491)

  With this release, you can now use from as the name of a default type-only import in TypeScript code, as well as of as the name of an await using loop iteration variable:

  import type from from 'from'
  for (await using of of of) ;
  

  This matches similar changes in the TypeScript compiler (#56376 and #55555) which will start allowing this syntax in an upcoming version of TypeScript. Please never actually write code like this.

  The type-only import syntax change was contributed by @magic-akari.

• Fix a regression in 0.19.0 regarding paths in tsconfig.json (#3354)

  The fix in esbuild version 0.19.0 to process tsconfig.json aliases before the --packages=external setting unintentionally broke an edge case in esbuild's handling of certain tsconfig.json aliases where there are multiple files with the same name in different directories. This release adjusts esbuild's behavior for this edge case so that it passes while still processing aliases before --packages=external. Please read the linked issue for more details.

• Fix a CSS font property minification bug (#3452)

  This release fixes a bug where esbuild's CSS minifier didn't insert a space between the font size and the font family in the font CSS shorthand property in the edge case where the original source code didn't already have a space and the leading string token was shortened to an identifier:

  /* Original code */
  .foo { font: 16px"Menlo"; }
  
  /* Old output (with --minify) */
  .foo{font:16pxMenlo}
  
  /* New output (with --minify) */
  .foo{font:16px Menlo}
  

• Fix bundling CSS with asset names containing spaces (#3410)

  Assets referenced via CSS url() tokens may cause esbuild to generate invalid output when bundling if the file name contains spaces (e.g. url(image 2.png)). With this release, esbuild will now quote all bundled asset references in url() tokens to avoid this problem. This only affects assets loaded using the file and copy loaders.

• Fix invalid CSS url() tokens in @import rules (#3426)

  In the future, CSS url() tokens may contain additional stuff after the URL. This is irrelevant today as no CSS specification does this. But esbuild previously had a bug where using these tokens in an @import rule resulted in malformed output. This bug has been fixed.

• Fix browser + false + type: module in package.json (#3367)

  The browser field in package.json allows you to map a file to false to have it be treated as an empty file when bundling for the browser. However, if package.json contains "type": "module" then all .js files will be considered ESM, not CommonJS. Importing a named import from an empty CommonJS file gives you undefined, but importing a named export from an empty ESM file is a build error. This release changes esbuild's interpretation of these files mapped to false in this situation from ESM to CommonJS to avoid generating build errors for named imports.

• Fix a bug in top-level await error reporting (#3400)

  Using require() on a file that contains top-level await is not allowed because require() must return synchronously and top-level await makes that impossible. You will get a build error if you try to bundle code that does this with esbuild. This release fixes a bug in esbuild's error reporting code for complex cases of this situation involving multiple levels of imports to get to the module containing the top-level await.

• Update to Unicode 15.1.0

  The character tables that determine which characters form valid JavaScript identifiers have been updated from Unicode version 15.0.0 to the newly-released Unicode version 15.1.0. I'm not putting an example in the release notes because all of the new characters will likely just show up as little squares since fonts haven't been updated yet. But you can read https://www.unicode.org/versions/Unicode15.1.0/#Summary for more information about the changes.

  This upgrade was contributed by @JLHwung.

• Fix printing of JavaScript decorators in tricky cases (#3396)

  This release fixes some bugs where esbuild's pretty-printing of JavaScript decorators could incorrectly produced code with a syntax error. The problem happened because esbuild sometimes substitutes identifiers for other expressions in the pretty-printer itself, but the decision about whether to wrap the expression or not didn't account for this. Here are some examples:

  // Original code
  import { constant } from './constants.js'
  import { imported } from 'external'
  import { undef } from './empty.js'
  class Foo {
    @constant()
    @imported()
    @undef()
    foo
  }
  
  // Old output (with --bundle --format=cjs --packages=external --minify-syntax)
  var import_external = require("external");
  var Foo = class {
    @123()
    @(0, import_external.imported)()
    @(void 0)()
    foo;
  };
  
  // New output (with --bundle --format=cjs --packages=external --minify-syntax)
  var import_external = require("external");
  var Foo = class {
    @(123())
    @((0, import_external.imported)())
    @((void 0)())
    foo;
  };
  

• Allow pre-release versions to be passed to target (#3388)

  People want to be able to pass version numbers for unreleased versions of node (which have extra stuff after the version numbers) to esbuild's target setting and have esbuild do something reasonable with them. These version strings are of course not present in esbuild's internal feature compatibility table because an unreleased version has not been released yet (by definition). With this release, esbuild will now attempt to accept these version strings passed to target and do something reasonable with them.

• Fix list-style-type with the local-css loader (#3325)

  The local-css loader incorrectly treated all identifiers provided to list-style-type as a custom local identifier. That included identifiers such as none which have special meaning in CSS, and which should not be treated as custom local identifiers. This release fixes this bug:

  /* Original code */
  ul { list-style-type: none }
  
  /* Old output (with --loader=local-css) */
  ul {
    list-style-type: stdin_none;
  }
  
  /* New output (with --loader=local-css) */
  ul {
    list-style-type: none;
  }
  

  Note that this bug only affected code using the local-css loader. It did not affect code using the css loader.

• Avoid inserting temporary variables before use strict (#3322)

  This release fixes a bug where esbuild could incorrectly insert automatically-generated temporary variables before use strict directives:

  // Original code
  function foo() {
    'use strict'
    a.b?.c()
  }
  
  // Old output (with --target=es6)
  function foo() {
    var _a;
    "use strict";
    (_a = a.b) == null ? void 0 : _a.c();
  }
  
  // New output (with --target=es6)
  function foo() {
    "use strict";
    var _a;
    (_a = a.b) == null ? void 0 : _a.c();
  }
  

• Adjust TypeScript enum output to better approximate tsc (#3329)

  TypeScript enum values can be either number literals or string literals. Numbers create a bidirectional mapping between the name and the value but strings only create a unidirectional mapping from the name to the value. When the enum value is neither a number literal nor a string literal, TypeScript and esbuild both default to treating it as a number:

  // Original TypeScript code
  declare const foo: any
  enum Foo {
    NUMBER = 1,
    STRING = 'a',
    OTHER = foo,
  }
  
  // Compiled JavaScript code (from "tsc")
  var Foo;
  (function (Foo) {
    Foo[Foo["NUMBER"] = 1] = "NUMBER";
    Foo["STRING"] = "a";
    Foo[Foo["OTHER"] = foo] = "OTHER";
  })(Foo || (Foo = {}));
  

  However, TypeScript does constant folding slightly differently than esbuild. For example, it may consider template literals to be string literals in some cases:

  // Original TypeScript code
  declare const foo = 'foo'
  enum Foo {
    PRESENT = `${foo}`,
    MISSING = `${bar}`,
  }
  
  // Compiled JavaScript code (from "tsc")
  var Foo;
  (function (Foo) {
    Foo["PRESENT"] = "foo";
    Foo[Foo["MISSING"] = `${bar}`] = "MISSING";
  })(Foo || (Foo = {}));
  

  The template literal initializer for PRESENT is treated as a string while the template literal initializer for MISSING is treated as a number. Previously esbuild treated both of these cases as a number but starting with this release, esbuild will now treat both of these cases as a string. This doesn't exactly match the behavior of tsc but in the case where the behavior diverges tsc reports a compile error, so this seems like acceptible behavior for esbuild. Note that handling these cases completely correctly would require esbuild to parse type declarations (see the declare keyword), which esbuild deliberately doesn't do.

• Ignore case in CSS in more places (#3316)

  This release makes esbuild's CSS support more case-agnostic, which better matches how browsers work. For example:

  /* Original code */
  @KeyFrames Foo { From { OpaCity: 0 } To { OpaCity: 1 } }
  body { CoLoR: YeLLoW }
  
  /* Old output (with --minify) */
  @KeyFrames Foo{From {OpaCity: 0} To {OpaCity: 1}}body{CoLoR:YeLLoW}
  
  /* New output (with --minify) */
  @KeyFrames Foo{0%{OpaCity:0}To{OpaCity:1}}body{CoLoR:#ff0}
  

  Please never actually write code like this.

• Improve the error message for null entries in exports (#3377)

  Package authors can disable package export paths with the exports map in package.json. With this release, esbuild now has a clearer error message that points to the null token in package.json itself instead of to the surrounding context. Here is an example of the new error message:

  ✘ [ERROR] Could not resolve "msw/browser"
  
      lib/msw-config.ts:2:28:
        2 │ import { setupWorker } from 'msw/browser';
          ╵                             ~~~~~~~~~~~~~
  
    The path "./browser" cannot be imported from package "msw" because it was explicitly disabled by
    the package author here:
  
      node_modules/msw/package.json:17:14:
        17 │       "node": null,
           ╵               ~~~~
  
    You can mark the path "msw/browser" as external to exclude it from the bundle, which will remove
    this error and leave the unresolved path in the bundle.
  

• Parse and print the with keyword in import statements

  JavaScript was going to have a feature called "import assertions" that adds an assert keyword to import statements. It looked like this:

  import stuff from './stuff.json' assert { type: 'json' }
  

  The feature provided a way to assert that the imported file is of a certain type (but was not allowed to affect how the import is interpreted, even though that's how everyone expected it to behave). The feature was fully specified and then actually implemented and shipped in Chrome before the people behind the feature realized that they should allow it to affect how the import is interpreted after all. So import assertions are no longer going to be added to the language.

  Instead, the current proposal is to add a feature called "import attributes" instead that adds a with keyword to import statements. It looks like this:

  import stuff from './stuff.json' with { type: 'json' }
  

  This feature provides a way to affect how the import is interpreted. With this release, esbuild now has preliminary support for parsing and printing this new with keyword. The with keyword is not yet interpreted by esbuild, however, so bundling code with it will generate a build error. All this release does is allow you to use esbuild to process code containing it (such as removing types from TypeScript code). Note that this syntax is not yet a part of JavaScript and may be removed or altered in the future if the specification changes (which it already has once, as described above). If that happens, esbuild reserves the right to remove or alter its support for this syntax too.

• Update how CSS nesting is parsed again

  CSS nesting syntax has been changed again, and esbuild has been updated to match. Type selectors may now be used with CSS nesting:

  .foo {
    div {
      color: red;
    }
  }
  

  Previously this was disallowed in the CSS specification because it's ambiguous whether an identifier is a declaration or a nested rule starting with a type selector without requiring unbounded lookahead in the parser. It has now been allowed because the CSS working group has decided that requiring unbounded lookahead is acceptable after all.

  Note that this change means esbuild no longer considers any existing browser to support CSS nesting since none of the existing browsers support this new syntax. CSS nesting will now always be transformed when targeting a browser. This situation will change in the future as browsers add support for this new syntax.

• Fix a scope-related bug with --drop-labels= (#3311)

  The recently-released --drop-labels= feature previously had a bug where esbuild's internal scope stack wasn't being restored properly when a statement with a label was dropped. This could manifest as a tree-shaking issue, although it's possible that this could have also been causing other subtle problems too. The bug has been fixed in this release.

• Make renamed CSS names unique across entry points (#3295)

  Previously esbuild's generated names for local names in CSS were only unique within a given entry point (or across all entry points when code splitting was enabled). That meant that building multiple entry points with esbuild could result in local names being renamed to the same identifier even when those entry points were built simultaneously within a single esbuild API call. This problem was especially likely to happen with minification enabled. With this release, esbuild will now avoid renaming local names from two separate entry points to the same name if those entry points were built with a single esbuild API call, even when code splitting is disabled.

• Fix CSS ordering bug with @layer before @import

  CSS lets you put @layer rules before @import rules to define the order of layers in a stylesheet. Previously esbuild's CSS bundler incorrectly ordered these after the imported files because before the introduction of cascade layers to CSS, imported files could be bundled by removing the @import rules and then joining files together in the right order. But with @layer, CSS files may now need to be split apart into multiple pieces in the bundle. For example:

  /* Original code */
  @layer start;
  @import "data:text/css,@layer inner.start;";
  @import "data:text/css,@layer inner.end;";
  @layer end;
  
  /* Old output (with --bundle) */
  @layer inner.start;
  @layer inner.end;
  @layer start;
  @layer end;
  
  /* New output (with --bundle) */
  @layer start;
  @layer inner.start;
  @layer inner.end;
  @layer end;
  

• Unwrap nested duplicate @media rules (#3226)

  With this release, esbuild's CSS minifier will now automatically unwrap duplicate nested @media rules:

  /* Original code */
  @media (min-width: 1024px) {
    .foo { color: red }
    @media (min-width: 1024px) {
      .bar { color: blue }
    }
  }
  
  /* Old output (with --minify) */
  @media (min-width: 1024px){.foo{color:red}@media (min-width: 1024px){.bar{color:#00f}}}
  
  /* New output (with --minify) */
  @media (min-width: 1024px){.foo{color:red}.bar{color:#00f}}
  

  These rules are unlikely to be authored manually but may result from using frameworks such as Tailwind to generate CSS.

• Fix a regression with baseURL in tsconfig.json (#3307)

  The previous release moved tsconfig.json path resolution before --packages=external checks to allow the paths field in tsconfig.json to avoid a package being marked as external. However, that reordering accidentally broke the behavior of the baseURL field from tsconfig.json. This release moves these path resolution rules around again in an attempt to allow both of these cases to work.

• Parse TypeScript type arguments for JavaScript decorators (#3308)

  When parsing JavaScript decorators in TypeScript (i.e. with experimentalDecorators disabled), esbuild previously didn't parse type arguments. Type arguments will now be parsed starting with this release. For example:

  @foo<number>
  @bar<number, string>()
  class Foo {}
  

• Fix glob patterns matching extra stuff at the end (#3306)

  Previously glob patterns such as ./*.js would incorrectly behave like ./*.js* during path matching (also matching .js.map files, for example). This was never intentional behavior, and has now been fixed.

• Change the permissions of esbuild's generated output files (#3285)

  This release changes the permissions of the output files that esbuild generates to align with the default behavior of node's fs.writeFileSync function. Since most tools written in JavaScript use fs.writeFileSync, this should make esbuild more consistent with how other JavaScript build tools behave.

  The full Unix-y details: Unix permissions use three-digit octal notation where the three digits mean "user, group, other" in that order. Within a digit, 4 means "read" and 2 means "write" and 1 means "execute". So 6 == 4 + 2 == read + write. Previously esbuild uses 0644 permissions (the leading 0 means octal notation) but the permissions for fs.writeFileSync defaults to 0666, so esbuild will now use 0666 permissions. This does not necessarily mean that the files esbuild generates will end up having 0666 permissions, however, as there is another Unix feature called "umask" where the operating system masks out some of these bits. If your umask is set to 0022 then the generated files will have 0644 permissions, and if your umask is set to 0002 then the generated files will have 0664 permissions.

• Fix a subtle CSS ordering issue with @import and @layer

  With this release, esbuild may now introduce additional @layer rules when bundling CSS to better preserve the layer ordering of the input code. Here's an example of an edge case where this matters:

  /* entry.css */
  @import "a.css";
  @import "b.css";
  @import "a.css";
  

  /* a.css */
  @layer a {
    body {
      background: red;
    }
  }
  

  /* b.css */
  @layer b {
    body {
      background: green;
    }
  }
  

  This CSS should set the body background to green, which is what happens in the browser. Previously esbuild generated the following output which incorrectly sets the body background to red:

  /* b.css */
  @layer b {
    body {
      background: green;
    }
  }
  
  /* a.css */
  @layer a {
    body {
      background: red;
    }
  }
  

  This difference in behavior is because the browser evaluates a.css + b.css + a.css (in CSS, each @import is replaced with a copy of the imported file) while esbuild was only writing out b.css + a.css. The first copy of a.css wasn't being written out by esbuild for two reasons: 1) bundlers care about code size and try to avoid emitting duplicate CSS and 2) when there are multiple copies of a CSS file, normally only the last copy matters since the last declaration with equal specificity wins in CSS.

  However, @layer was recently added to CSS and for @layer the first copy matters because layers are ordered using their first location in source code order. This introduction of @layer means esbuild needs to change its bundling algorithm. An easy solution would be for esbuild to write out a.css twice, but that would be inefficient. So what I'm going to try to have esbuild do with this release is to write out an abbreviated form of the first copy of a CSS file that only includes the @layer information, and then still only write out the full CSS file once for the last copy. So esbuild's output for this edge case now looks like this:

  /* a.css */
  @layer a;
  
  /* b.css */
  @layer b {
    body {
      background: green;
    }
  }
  
  /* a.css */
  @layer a {
    body {
      background: red;
    }
  }
  

  The behavior of the bundled CSS now matches the behavior of the unbundled CSS. You may be wondering why esbuild doesn't just write out a.css first followed by b.css. That would work in this case but it doesn't work in general because for any rules outside of a @layer rule, the last copy should still win instead of the first copy.

• Fix a bug with esbuild's TypeScript type definitions (#3299)

  This release fixes a copy/paste error with the TypeScript type definitions for esbuild's JS API:

   export interface TsconfigRaw {
     compilerOptions?: {
  -    baseUrl?: boolean
  +    baseUrl?: string
       ...
     }
   }
  

  This fix was contributed by @privatenumber.

This release deliberately contains backwards-incompatible changes. To avoid automatically picking up releases like this, you should either be pinning the exact version of esbuild in your package.json file (recommended) or be using a version range syntax that only accepts patch upgrades such as ^0.18.0 or ~0.18.0. See npm's documentation about semver for more information.

• Handle import paths containing wildcards (#56, #700, #875, #976, #2221, #2515)

  This release introduces wildcards in import paths in two places:

  • Entry points

    You can now pass a string containing glob-style wildcards such as ./src/*.ts as an entry point and esbuild will search the file system for files that match the pattern. This can be used to easily pass esbuild all files with a certain extension on the command line in a cross-platform way. Previously you had to rely on the shell to perform glob expansion, but that is obviously shell-dependent and didn't work at all on Windows. Note that to use this feature on the command line you will have to quote the pattern so it's passed verbatim to esbuild without any expansion by the shell. Here's an example:

    esbuild --minify "./src/*.ts" --outdir=out
    

    Specifically the * character will match any character except for the / character, and the /**/ character sequence will match a path separator followed by zero or more path elements. Other wildcard operators found in glob patterns such as ? and [...] are not supported.

  • Run-time import paths

    Import paths that are evaluated at run-time can now be bundled in certain limited situations. The import path expression must be a form of string concatenation and must start with either ./ or ../. Each non-string expression in the string concatenation chain becomes a wildcard. The * wildcard is chosen unless the previous character is a /, in which case the /**/* character sequence is used. Some examples:

    // These two forms are equivalent
    const json1 = await import('./data/' + kind + '.json')
    const json2 = await import(`./data/${kind}.json`)
    

    This feature works with require(...) and import(...) because these can all accept run-time expressions. It does not work with import and export statements because these cannot accept run-time expressions. If you want to prevent esbuild from trying to bundle these imports, you should move the string concatenation expression outside of the require(...) or import(...). For example:

    // This will be bundled
    const json1 = await import('./data/' + kind + '.json')
    
    // This will not be bundled
    const path = './data/' + kind + '.json'
    const json2 = await import(path)
    

    Note that using this feature means esbuild will potentially do a lot of file system I/O to find all possible files that might match the pattern. This is by design, and is not a bug. If this is a concern, I recommend either avoiding the /**/ pattern (e.g. by not putting a / before a wildcard) or using this feature only in directory subtrees which do not have many files that don't match the pattern (e.g. making a subdirectory for your JSON files and explicitly including that subdirectory in the pattern).

• Path aliases in tsconfig.json no longer count as packages (#2792, #3003, #3160, #3238)

  Setting --packages=external tells esbuild to make all import paths external when they look like a package path. For example, an import of ./foo/bar is not a package path and won't be external while an import of foo/bar is a package path and will be external. However, the paths field in tsconfig.json allows you to create import paths that look like package paths but that do not resolve to packages. People do not want these paths to count as package paths. So with this release, the behavior of --packages=external has been changed to happen after the tsconfig.json path remapping step.

• Use the local-css loader for .module.css files by default (#20)

  With this release the css loader is still used for .css files except that .module.css files now use the local-css loader. This is a common convention in the web development community. If you need .module.css files to use the css loader instead, then you can override this behavior with --loader:.module.css=css.

• Support advanced CSS @import rules (#953, #3137)

  CSS @import statements have been extended to allow additional trailing tokens after the import path. These tokens sort of make the imported file behave as if it were wrapped in a @layer, @supports, and/or @media rule. Here are some examples:

  @import url(foo.css);
  @import url(foo.css) layer;
  @import url(foo.css) layer(bar);
  @import url(foo.css) layer(bar) supports(display: flex);
  @import url(foo.css) layer(bar) supports(display: flex) print;
  @import url(foo.css) layer(bar) print;
  @import url(foo.css) supports(display: flex);
  @import url(foo.css) supports(display: flex) print;
  @import url(foo.css) print;
  

  You can read more about this advanced syntax here. With this release, esbuild will now bundle @import rules with these trailing tokens and will wrap the imported files in the corresponding rules. Note that this now means a given imported file can potentially appear in multiple places in the bundle. However, esbuild will still only load it once (e.g. on-load plugins will only run once per file, not once per import).

• Implement composes from CSS modules (#20)

  This release implements the composes annotation from the CSS modules specification. It provides a way for class selectors to reference other class selectors (assuming you are using the local-css loader). And with the from syntax, this can even work with local names across CSS files. For example:

  // app.js
  import { submit } from './style.css'
  const div = document.createElement('div')
  div.className = submit
  document.body.appendChild(div)
  

  /* style.css */
  .button {
    composes: pulse from "anim.css";
    display: inline-block;
  }
  .submit {
    composes: button;
    font-weight: bold;
  }
  

  /* anim.css */
  @keyframes pulse {
    from, to { opacity: 1 }
    50% { opacity: 0.5 }
  }
  .pulse {
    animation: 2s ease-in-out infinite pulse;
  }
  

  Bundling this with esbuild using --bundle --outdir=dist --loader:.css=local-css now gives the following:

  (() => {
    // style.css
    var submit = "anim_pulse style_button style_submit";
  
    // app.js
    var div = document.createElement("div");
    div.className = submit;
    document.body.appendChild(div);
  })();
  

  /* anim.css */
  @keyframes anim_pulse {
    from, to {
      opacity: 1;
    }
    50% {
      opacity: 0.5;
    }
  }
  .anim_pulse {
    animation: 2s ease-in-out infinite anim_pulse;
  }
  
  /* style.css */
  .style_button {
    display: inline-block;
  }
  .style_submit {
    font-weight: bold;
  }
  

  Import paths in the composes: ... from syntax are resolved using the new composes-from import kind, which can be intercepted by plugins during import path resolution when bundling is enabled.

  Note that the order in which composed CSS classes from separate files appear in the bundled output file is deliberately undefined by design (see the specification for details). You are not supposed to declare the same CSS property in two separate class selectors and then compose them together. You are only supposed to compose CSS class selectors that declare non-overlapping CSS properties.

  Issue #20 (the issue tracking CSS modules) is esbuild's most-upvoted issue! With this change, I now consider esbuild's implementation of CSS modules to be complete. There are still improvements to make and there may also be bugs with the current implementation, but these can be tracked in separate issues.

• Fix non-determinism with tsconfig.json and symlinks (#3284)

  This release fixes an issue that could cause esbuild to sometimes emit incorrect build output in cases where a file under the effect of tsconfig.json is inconsistently referenced through a symlink. It can happen when using npm link to create a symlink within node_modules to an unpublished package. The build result was non-deterministic because esbuild runs module resolution in parallel and the result of the tsconfig.json lookup depended on whether the import through the symlink or not through the symlink was resolved first. This problem was fixed by moving the realpath operation before the tsconfig.json lookup.

• Add a hash property to output files (#3084, #3293)

  As a convenience, every output file in esbuild's API now includes a hash property that is a hash of the contents field. This is the hash that's used internally by esbuild to detect changes between builds for esbuild's live-reload feature. You may also use it to detect changes between your own builds if its properties are sufficient for your use case.

  This feature has been added directly to output file objects since it's just a hash of the contents field, so it makes conceptual sense to store it in the same location. Another benefit of putting it there instead of including it as a part of the watch mode API is that it can be used without watch mode enabled. You can use it to compare the output of two independent builds that were done at different times.

  The hash algorithm (currently XXH64) is implementation-dependent and may be changed at any time in between esbuild versions. If you don't like esbuild's choice of hash algorithm then you are welcome to hash the contents yourself instead. As with any hash algorithm, note that while two different hashes mean that the contents are different, two equal hashes do not necessarily mean that the contents are equal. You may still want to compare the contents in addition to the hashes to detect with certainty when output files have been changed.

• Avoid generating duplicate prefixed declarations in CSS (#3292)

  There was a request for esbuild's CSS prefixer to avoid generating a prefixed declaration if a declaration by that name is already present in the same rule block. So with this release, esbuild will now avoid doing this:

  /* Original code */
  body {
    backdrop-filter: blur(30px);
    -webkit-backdrop-filter: blur(45px);
  }
  
  /* Old output (with --target=safari12) */
  body {
    -webkit-backdrop-filter: blur(30px);
    backdrop-filter: blur(30px);
    -webkit-backdrop-filter: blur(45px);
  }
  
  /* New output (with --target=safari12) */
  body {
    backdrop-filter: blur(30px);
    -webkit-backdrop-filter: blur(45px);
  }
  

  This can result in a visual difference in certain cases (for example if the browser understands blur(30px) but not blur(45px), it will be able to fall back to blur(30px)). But this change means esbuild now matches the behavior of Autoprefixer which is probably a good representation of how people expect this feature to work.

• Fix asset references with the --line-limit flag (#3286)

  The recently-released --line-limit flag tells esbuild to terminate long lines after they pass this length limit. This includes automatically wrapping long strings across multiple lines using escaped newline syntax. However, using this could cause esbuild to generate incorrect code for references from generated output files to assets in the bundle (i.e. files loaded with the file or copy loaders). This is because esbuild implements asset references internally using find-and-replace with a randomly-generated string, but the find operation fails if the string is split by an escaped newline due to line wrapping. This release fixes the problem by not wrapping these strings. This issue affected asset references in both JS and CSS files.

• Support local names in CSS for @keyframe, @counter-style, and @container (#20)

  This release extends support for local names in CSS files loaded with the local-css loader to cover the @keyframe, @counter-style, and @container rules (and also animation, list-style, and container declarations). Here's an example:

  @keyframes pulse {
    from, to { opacity: 1 }
    50% { opacity: 0.5 }
  }
  @counter-style moon {
    system: cyclic;
    symbols: 🌕 🌖 🌗 🌘 🌑 🌒 🌓 🌔;
  }
  @container squish {
    li { float: left }
  }
  ul {
    animation: 2s ease-in-out infinite pulse;
    list-style: inside moon;
    container: squish / size;
  }
  

  With the local-css loader enabled, that CSS will be turned into something like this (with the local name mapping exposed to JS):

  @keyframes stdin_pulse {
    from, to {
      opacity: 1;
    }
    50% {
      opacity: 0.5;
    }
  }
  @counter-style stdin_moon {
    system: cyclic;
    symbols: 🌕 🌖 🌗 🌘 🌑 🌒 🌓 🌔;
  }
  @container stdin_squish {
    li {
      float: left;
    }
  }
  ul {
    animation: 2s ease-in-out infinite stdin_pulse;
    list-style: inside stdin_moon;
    container: stdin_squish / size;
  }
  

  If you want to use a global name within a file loaded with the local-css loader, you can use a :global selector to do that:

  div {
    /* All symbols are global inside this scope (i.e.
     * "pulse", "moon", and "squish" are global below) */
    :global {
      animation: 2s ease-in-out infinite pulse;
      list-style: inside moon;
      container: squish / size;
    }
  }
  

  If you want to use @keyframes, @counter-style, or @container with a global name, make sure it's in a file that uses the css or global-css loader instead of the local-css loader. For example, you can configure --loader:.module.css=local-css so that the local-css loader only applies to *.module.css files.

• Support strings as keyframe animation names in CSS (#2555)

  With this release, esbuild will now parse animation names that are specified as strings and will convert them to identifiers. The CSS specification allows animation names to be specified using either identifiers or strings but Chrome only understands identifiers, so esbuild will now always convert string names to identifier names for Chrome compatibility:

  /* Original code */
  @keyframes "hide menu" {
    from { opacity: 1 }
    to { opacity: 0 }
  }
  menu.hide {
    animation: 0.5s ease-in-out "hide menu";
  }
  
  /* Old output */
  @keyframes "hide menu" { from { opacity: 1 } to { opacity: 0 } }
  menu.hide {
    animation: 0.5s ease-in-out "hide menu";
  }
  
  /* New output */
  @keyframes hide\ menu {
    from {
      opacity: 1;
    }
    to {
      opacity: 0;
    }
  }
  menu.hide {
    animation: 0.5s ease-in-out hide\ menu;
  }
  

• Support An+B syntax and :nth-*() pseudo-classes in CSS

  This adds support for the :nth-child(), :nth-last-child(), :nth-of-type(), and :nth-last-of-type() pseudo-classes to esbuild, which has the following consequences:

  • The An+B syntax is now parsed, so parse errors are now reported
  • An+B values inside these pseudo-classes are now pretty-printed (e.g. a leading + will be stripped because it's not in the AST)
  • When minification is enabled, An+B values are reduced to equivalent but shorter forms (e.g. 2n+0 => 2n, 2n+1 => odd)
  • Local CSS names in an of clause are now detected (e.g. in :nth-child(2n of :local(.foo)) the name foo is now renamed)

  /* Original code */
  .foo:nth-child(+2n+1 of :local(.bar)) {
    color: red;
  }
  
  /* Old output (with --loader=local-css) */
  .stdin_foo:nth-child(+2n + 1 of :local(.bar)) {
    color: red;
  }
  
  /* New output (with --loader=local-css) */
  .stdin_foo:nth-child(2n+1 of .stdin_bar) {
    color: red;
  }
  

• Adjust CSS nesting parser for IE7 hacks (#3272)

  This fixes a regression with esbuild's treatment of IE7 hacks in CSS. CSS nesting allows selectors to be used where declarations are expected. There's an IE7 hack where prefixing a declaration with a * causes that declaration to only be applied in IE7 due to a bug in IE7's CSS parser. However, it's valid for nested CSS selectors to start with *. So esbuild was incorrectly parsing these declarations and anything following it up until the next { as a selector for a nested CSS rule. This release changes esbuild's parser to terminate the parsing of selectors for nested CSS rules when a ; is encountered to fix this edge case:

  /* Original code */
  .item {
    *width: 100%;
    height: 1px;
  }
  
  /* Old output */
  .item {
    *width: 100%; height: 1px; {
    }
  }
  
  /* New output */
  .item {
    *width: 100%;
    height: 1px;
  }
  

  Note that the syntax for CSS nesting is about to change again, so esbuild's CSS parser may still not be completely accurate with how browsers do and/or will interpret CSS nesting syntax. Expect additional updates to esbuild's CSS parser in the future to deal with upcoming CSS specification changes.

• Adjust esbuild's warning about undefined imports for TypeScript import equals declarations (#3271)

  In JavaScript, accessing a missing property on an import namespace object is supposed to result in a value of undefined at run-time instead of an error at compile-time. This is something that esbuild warns you about by default because doing this can indicate a bug with your code. For example:

  // app.js
  import * as styles from './styles'
  console.log(styles.buton)
  

  // styles.js
  export let button = {}
  

  If you bundle app.js with esbuild you will get this:

  ▲ [WARNING] Import "buton" will always be undefined because there is no matching export in "styles.js" [import-is-undefined]
  
      app.js:2:19:
        2 │ console.log(styles.buton)
          │                    ~~~~~
          ╵                    button
  
    Did you mean to import "button" instead?
  
      styles.js:1:11:
        1 │ export let button = {}
          ╵            ~~~~~~
  

  However, there is TypeScript-only syntax for import equals declarations that can represent either a type import (which esbuild should ignore) or a value import (which esbuild should respect). Since esbuild doesn't have a type system, it tries to only respect import equals declarations that are actually used as values. Previously esbuild always generated this warning for unused imports referenced within import equals declarations even when the reference could be a type instead of a value. Starting with this release, esbuild will now only warn in this case if the import is actually used. Here is an example of some code that no longer causes an incorrect warning:

  // app.ts
  import * as styles from './styles'
  import ButtonType = styles.Button
  

  // styles.ts
  export interface Button {}
  

• Fix a regression with whitespace inside :is() (#3265)

  The change to parse the contents of :is() in version 0.18.14 introduced a regression that incorrectly flagged the contents as a syntax error if the contents started with a whitespace token (for example div:is( .foo ) {}). This regression has been fixed.

• Add the --serve-fallback= option (#2904)

  The web server built into esbuild serves the latest in-memory results of the configured build. If the requested path doesn't match any in-memory build result, esbuild also provides the --servedir= option to tell esbuild to serve the requested path from that directory instead. And if the requested path doesn't match either of those things, esbuild will either automatically generate a directory listing (for directories) or return a 404 error.

  Starting with this release, that last step can now be replaced with telling esbuild to serve a specific HTML file using the --serve-fallback= option. This can be used to provide a "not found" page for missing URLs. It can also be used to implement a single-page app that mutates the current URL and therefore requires the single app entry point to be served when the page is loaded regardless of whatever the current URL is.

• Use the tsconfig field in package.json during extends resolution (#3247)

  This release adds a feature from TypeScript 3.2 where if a tsconfig.json file specifies a package name in the extends field and that package's package.json file has a tsconfig field, the contents of that field are used in the search for the base tsconfig.json file.

• Implement CSS nesting without :is() when possible (#1945)

  Previously esbuild would always produce a warning when transforming nested CSS for a browser that doesn't support the :is() pseudo-class. This was because the nesting transform needs to generate an :is() in some complex cases which means the transformed CSS would then not work in that browser. However, the CSS nesting transform can often be done without generating an :is(). So with this release, esbuild will no longer warn when targeting browsers that don't support :is() in the cases where an :is() isn't needed to represent the nested CSS.

  In addition, esbuild's nested CSS transform has been updated to avoid generating an :is() in cases where an :is() is preferable but there's a longer alternative that is also equivalent. This update means esbuild can now generate a combinatorial explosion of CSS for complex CSS nesting syntax when targeting browsers that don't support :is(). This combinatorial explosion is necessary to accurately represent the original semantics. For example:

  /* Original code */
  .first,
  .second,
  .third {
    & > & {
      color: red;
    }
  }
  
  /* Old output (with --target=chrome80) */
  :is(.first, .second, .third) > :is(.first, .second, .third) {
    color: red;
  }
  
  /* New output (with --target=chrome80) */
  .first > .first,
  .first > .second,
  .first > .third,
  .second > .first,
  .second > .second,
  .second > .third,
  .third > .first,
  .third > .second,
  .third > .third {
    color: red;
  }
  

  This change means you can now use CSS nesting with esbuild when targeting an older browser that doesn't support :is(). You'll now only get a warning from esbuild if you use complex CSS nesting syntax that esbuild can't represent in that older browser without using :is(). There are two such cases:

  /* Case 1 */
  a b {
    .foo & {
      color: red;
    }
  }
  
  /* Case 2 */
  a {
    > b& {
      color: red;
    }
  }
  

  These two cases still need to use :is(), both for different reasons, and cannot be used when targeting an older browser that doesn't support :is():

  /* Case 1 */
  .foo :is(a b) {
    color: red;
  }
  
  /* Case 2 */
  a > a:is(b) {
    color: red;
  }
  

• Automatically lower inset in CSS for older browsers

  With this release, esbuild will now automatically expand the inset property to the top, right, bottom, and left properties when esbuild's target is set to a browser that doesn't support inset:

  /* Original code */
  .app {
    position: absolute;
    inset: 10px 20px;
  }
  
  /* Old output (with --target=chrome80) */
  .app {
    position: absolute;
    inset: 10px 20px;
  }
  
  /* New output (with --target=chrome80) */
  .app {
    position: absolute;
    top: 10px;
    right: 20px;
    bottom: 10px;
    left: 20px;
  }
  

• Add support for the new @starting-style CSS rule (#3249)

  This at rule allow authors to start CSS transitions on first style update. That is, you can now make the transition take effect when the display property changes from none to block.

  /* Original code */
  @starting-style {
    h1 {
      background-color: transparent;
    }
  }
  
  /* Output */
  @starting-style{h1{background-color:transparent}}
  

  This was contributed by @yisibl.

• Implement local CSS names (#20)

  This release introduces two new loaders called global-css and local-css and two new pseudo-class selectors :local() and :global(). This is a partial implementation of the popular CSS modules approach for avoiding unintentional name collisions in CSS. I'm not calling this feature "CSS modules" because although some people in the community call it that, other people in the community have started using "CSS modules" to refer to something completely different and now CSS modules is an overloaded term.

  Here's how this new local CSS name feature works with esbuild:

  • Identifiers that look like .className and #idName are global with the global-css loader and local with the local-css loader. Global identifiers are the same across all files (the way CSS normally works) but local identifiers are different between different files. If two separate CSS files use the same local identifier .button, esbuild will automatically rename one of them so that they don't collide. This is analogous to how esbuild automatically renames JS local variables with the same name in separate JS files to avoid name collisions.

  • It only makes sense to use local CSS names with esbuild when you are also using esbuild's bundler to bundle JS files that import CSS files. When you do that, esbuild will generate one export for each local name in the CSS file. The JS code can import these names and use them when constructing HTML DOM. For example:

    // app.js
    import { outerShell } from './app.css'
    const div = document.createElement('div')
    div.className = outerShell
    document.body.appendChild(div)
    

    /* app.css */
    .outerShell {
      position: absolute;
      inset: 0;
    }
    

    When you bundle this with esbuild app.js --bundle --loader:.css=local-css --outdir=out you'll now get this (notice how the local CSS name outerShell has been renamed):

    // out/app.js
    (() => {
      // app.css
      var outerShell = "app_outerShell";
    
      // app.js
      var div = document.createElement("div");
      div.className = outerShell;
      document.body.appendChild(div);
    })();
    

    /* out/app.css */
    .app_outerShell {
      position: absolute;
      inset: 0;
    }
    

    This feature only makes sense to use when bundling is enabled both because your code needs to import the renamed local names so that it can use them, and because esbuild needs to be able to process all CSS files containing local names in a single bundling operation so that it can successfully rename conflicting local names to avoid collisions.

  • If you are in a global CSS file (with the global-css loader) you can create a local name using :local(), and if you are in a local CSS file (with the local-css loader) you can create a global name with :global(). So the choice of the global-css loader vs. the local-css loader just sets the default behavior for identifiers, but you can override it on a case-by-case basis as necessary. For example:

    :local(.button) {
      color: red;
    }
    :global(.button) {
      color: blue;
    }
    

    Processing this CSS file with esbuild with either the global-css or local-css loader will result in something like this:

    .stdin_button {
      color: red;
    }
    .button {
      color: blue;
    }
    

  • The names that esbuild generates for local CSS names are an implementation detail and are not intended to be hard-coded anywhere. The only way you should be referencing the local CSS names in your JS or HTML is with an import statement in JS that is bundled with esbuild, as demonstrated above. For example, when --minify is enabled esbuild will use a different name generation algorithm which generates names that are as short as possible (analogous to how esbuild minifies local identifiers in JS).

  • You can easily use both global CSS files and local CSS files simultaneously if you give them different file extensions. For example, you could pass --loader:.css=global-css and --loader:.module.css=local-css to esbuild so that .css files still use global names by default but .module.css files use local names by default.

  • Keep in mind that the css loader is different than the global-css loader. The :local and :global annotations are not enabled with the css loader and will be passed through unchanged. This allows you to have the option of using esbuild to process CSS containing while preserving these annotations. It also means that local CSS names are disabled by default for now (since the css loader is currently the default for CSS files). The :local and :global syntax may be enabled by default in a future release.

  Note that esbuild's implementation does not currently have feature parity with other implementations of modular CSS in similar tools. This is only a preliminary release with a partial implementation that includes some basic behavior to get the process started. Additional behavior may be added in future releases. In particular, this release does not implement:

  • The composes pragma
  • Tree shaking for unused local CSS
  • Local names for keyframe animations, grid lines, @container, @counter-style, etc.

  Issue #20 (the issue for this feature) is esbuild's most-upvoted issue! While this release still leaves that issue open, it's an important first step in that direction.

• Parse :is, :has, :not, and :where in CSS

  With this release, esbuild will now parse the contents of these pseudo-class selectors as a selector list. This means you will now get syntax warnings within these selectors for invalid selector syntax. It also means that esbuild's CSS nesting transform behaves slightly differently than before because esbuild is now operating on an AST instead of a token stream. For example:

  /* Original code */
  div {
    :where(.foo&) {
      color: red;
    }
  }
  
  /* Old output (with --target=chrome90) */
  :where(.foo:is(div)) {
    color: red;
  }
  
  /* New output (with --target=chrome90) */
  :where(div.foo) {
    color: red;
  }