v0.17.1

• Make it possible to cancel a build (#2725)

  The context object introduced in version 0.17.0 has a new cancel() method. You can use it to cancel a long-running build so that you can start a new one without needing to wait for the previous one to finish. When this happens, the previous build should always have at least one error and have no output files (i.e. it will be a failed build).

  Using it might look something like this:

  • JS:

    let ctx = await esbuild.context({
      // ...
    })
    
    let rebuildWithTimeLimit = timeLimit => {
      let timeout = setTimeout(() => ctx.cancel(), timeLimit)
      return ctx.rebuild().finally(() => clearTimeout(timeout))
    }
    
    let build = await rebuildWithTimeLimit(500)

  • Go:

    ctx, err := api.Context(api.BuildOptions{
      // ...
    })
    if err != nil {
      return
    }
    
    rebuildWithTimeLimit := func(timeLimit time.Duration) api.BuildResult {
      t := time.NewTimer(timeLimit)
      go func() {
        <-t.C
        ctx.Cancel()
      }()
      result := ctx.Rebuild()
      t.Stop()
      return result
    }
    
    build := rebuildWithTimeLimit(500 * time.Millisecond)

  This API is a quick implementation and isn't maximally efficient, so the build may continue to do some work for a little bit before stopping. For example, I have added stop points between each top-level phase of the bundler and in the main module graph traversal loop, but I haven't added fine-grained stop points within the internals of the linker. How quickly esbuild stops can be improved in future releases. This means you'll want to wait for cancel() and/or the previous rebuild() to finish (i.e. await the returned promise in JavaScript) before starting a new build, otherwise rebuild() will give you the just-canceled build that still hasn't ended yet. Note that onEnd callbacks will still be run regardless of whether or not the build was canceled.

• Fix server-sent events without servedir (#2827)

  The server-sent events for live reload were incorrectly using servedir to calculate the path to modified output files. This means events couldn't be sent when servedir wasn't specified. This release uses the internal output directory (which is always present) instead of servedir (which might be omitted), so live reload should now work when servedir is not specified.

• Custom entry point output paths now work with the copy loader (#2828)

  Entry points can optionally provide custom output paths to change the path of the generated output file. For example, esbuild foo=abc.js bar=xyz.js --outdir=out generates the files out/foo.js and out/bar.js. However, this previously didn't work when using the copy loader due to an oversight. This bug has been fixed. For example, you can now do esbuild foo=abc.html bar=xyz.html --outdir=out --loader:.html=copy to generate the files out/foo.html and out/bar.html.

• The JS API can now take an array of objects (#2828)

  Previously it was not possible to specify two entry points with the same custom output path using the JS API, although it was possible to do this with the Go API and the CLI. This will not cause a collision if both entry points use different extensions (e.g. if one uses .js and the other uses .css). You can now pass the JS API an array of objects to work around this API limitation:

  // The previous API didn't let you specify duplicate output paths
  let result = await esbuild.build({
    entryPoints: {
      // This object literal contains a duplicate key, so one is ignored
      'dist': 'foo.js',
      'dist': 'bar.css',
    },
  })
  
  // You can now specify duplicate output paths as an array of objects
  let result = await esbuild.build({
    entryPoints: [
      { in: 'foo.js', out: 'dist' },
      { in: 'bar.css', out: 'dist' },
    ],
  })

v0.17.0

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 ^0.16.0 or ~0.16.0. See npm's documentation about semver for more information.

At a high level, the breaking changes in this release fix some long-standing issues with the design of esbuild's incremental, watch, and serve APIs. This release also introduces some exciting new features such as live reloading. In detail:

• Move everything related to incremental builds to a new context API (#1037, #1606, #2280, #2418)

  This change removes the incremental and watch options as well as the serve() method, and introduces a new context() method. The context method takes the same arguments as the build() method but only validates its arguments and does not do an initial build. Instead, builds can be triggered using the rebuild(), watch(), and serve() methods on the returned context object. The new context API looks like this:

  // Create a context for incremental builds
  const context = await esbuild.context({
    entryPoints: ['app.ts'],
    bundle: true,
  })
  
  // Manually do an incremental build
  const result = await context.rebuild()
  
  // Enable watch mode
  await context.watch()
  
  // Enable serve mode
  await context.serve()
  
  // Dispose of the context
  context.dispose()

  The switch to the context API solves a major issue with the previous API which is that if the initial build fails, a promise is thrown in JavaScript which prevents you from accessing the returned result object. That prevented you from setting up long-running operations such as watch mode when the initial build contained errors. It also makes tearing down incremental builds simpler as there is now a single way to do it instead of three separate ways.

  In addition, this release also makes some subtle changes to how incremental builds work. Previously every call to rebuild() started a new build. If you weren't careful, then builds could actually overlap. This doesn't cause any problems with esbuild itself, but could potentially cause problems with plugins (esbuild doesn't even give you a way to identify which overlapping build a given plugin callback is running on). Overlapping builds also arguably aren't useful, or at least aren't useful enough to justify the confusion and complexity that they bring. With this release, there is now only ever a single active build per context. Calling rebuild() before the previous rebuild has finished now "merges" with the existing rebuild instead of starting a new build.

• Allow using watch and serve together (#805, #1650, #2576)

  Previously it was not possible to use watch mode and serve mode together. The rationale was that watch mode is one way of automatically rebuilding your project and serve mode is another (since serve mode automatically rebuilds on every request). However, people want to combine these two features to make "live reloading" where the browser automatically reloads the page when files are changed on the file system.

  This release now allows you to use these two features together. You can only call the watch() and serve() APIs once each per context, but if you call them together on the same context then esbuild will automatically rebuild both when files on the file system are changed and when the server serves a request.

• Support "live reloading" through server-sent events (#802)

  Server-sent events are a simple way to pass one-directional messages asynchronously from the server to the client. Serve mode now provides a /esbuild endpoint with an change event that triggers every time esbuild's output changes. So you can now implement simple "live reloading" (i.e. reloading the page when a file is edited and saved) like this:

  new EventSource('/esbuild').addEventListener('change', () => location.reload())

  The event payload is a JSON object with the following shape:

  interface ChangeEvent {
    added: string[]
    removed: string[]
    updated: string[]
  }

  This JSON should also enable more complex live reloading scenarios. For example, the following code hot-swaps changed CSS <link> tags in place without reloading the page (but still reloads when there are other types of changes):

  new EventSource('/esbuild').addEventListener('change', e => {
    const { added, removed, updated } = JSON.parse(e.data)
    if (!added.length && !removed.length && updated.length === 1) {
      for (const link of document.getElementsByTagName("link")) {
        const url = new URL(link.href)
        if (url.host === location.host && url.pathname === updated[0]) {
          const next = link.cloneNode()
          next.href = updated[0] + '?' + Math.random().toString(36).slice(2)
          next.onload = () => link.remove()
          link.parentNode.insertBefore(next, link.nextSibling)
          return
        }
      }
    }
    location.reload()
  })

  Implementing live reloading like this has a few known caveats:

  • These events only trigger when esbuild's output changes. They do not trigger when files unrelated to the build being watched are changed. If your HTML file references other files that esbuild doesn't know about and those files are changed, you can either manually reload the page or you can implement your own live reloading infrastructure instead of using esbuild's built-in behavior.

  • The EventSource API is supposed to automatically reconnect for you. However, there's a bug in Firefox that breaks this if the server is ever temporarily unreachable: https://bugzilla.mozilla.org/show_bug.cgi?id=1809332. Workarounds are to use any other browser, to manually reload the page if this happens, or to write more complicated code that manually closes and re-creates the EventSource object if there is a connection error. I'm hopeful that this bug will be fixed.

  • Browser vendors have decided to not implement HTTP/2 without TLS. This means that each /esbuild event source will take up one of your precious 6 simultaneous per-domain HTTP/1.1 connections. So if you open more than six HTTP tabs that use this live-reloading technique, you will be unable to use live reloading in some of those tabs (and other things will likely also break). The workaround is to enable HTTPS, which is now possible to do in esbuild itself (see below).

• Add built-in support for HTTPS (#2169)

  You can now tell esbuild's built-in development server to use HTTPS instead of HTTP. This is sometimes necessary because browser vendors have started making modern web features unavailable to HTTP websites. Previously you had to put a proxy in front of esbuild to enable HTTPS since esbuild's development server only supported HTTP. But with this release, you can now enable HTTPS with esbuild without an additional proxy.

  To enable HTTPS with esbuild:

  1. Generate a self-signed certificate. There are many ways to do this. Here's one way, assuming you have openssl installed:

     openssl req -x509 -newkey rsa:4096 -keyout key.pem -out cert.pem -days 9999 -nodes -subj /CN=127.0.0.1
     

  2. Add --keyfile=key.pem and --certfile=cert.pem to your esbuild development server command

  3. Click past the scary warning in your browser when you load your page

  If you have more complex needs than this, you can still put a proxy in front of esbuild and use that for HTTPS instead. Note that if you see the message "Client sent an HTTP request to an HTTPS server" when you load your page, then you are using the incorrect protocol. Replace http:// with https:// in your browser's URL bar.

  Keep in mind that esbuild's HTTPS support has nothing to do with security. The only reason esbuild now supports HTTPS is because browsers have made it impossible to do local development with certain modern web features without jumping through these extra hoops. Please do not use esbuild's development server for anything that needs to be secure. It's only intended for local development and no considerations have been made for production environments whatsoever.

• Better support copying index.html into the output directory (#621, #1771)

  Right now esbuild only supports JavaScript and CSS as first-class content types. Previously this meant that if you were building a website with a HTML file, a JavaScript file, and a CSS file, you could use esbuild to build the JavaScript file and the CSS file into the output directory but not to copy the HTML file into the output directory. You needed a separate cp command for that.

  Or so I thought. It turns out that the copy loader added in version 0.14.44 of esbuild is sufficient to have esbuild copy the HTML file into the output directory as well. You can add something like index.html --loader:.html=copy and esbuild will copy index.html into the output directory for you. The benefits of this are a) you don't need a separate cp command and b) the index.html file wi...

v0.16.17

• Fix additional comment-related regressions (#2814)

  This release fixes more edge cases where the new comment preservation behavior that was added in 0.16.14 could introduce syntax errors. Specifically:

  x = () => (/* comment */ {})
  for ((/* comment */ let).x of y) ;
  function *f() { yield (/* comment */class {}) }

  These cases caused esbuild to generate code with a syntax error in version 0.16.14 or above. These bugs have now been fixed.

v0.16.16

• Fix a regression caused by comment preservation (#2805)

  The new comment preservation behavior that was added in 0.16.14 introduced a regression where comments in certain locations could cause esbuild to omit certain necessary parentheses in the output. The outermost parentheses were incorrectly removed for the following syntax forms, which then introduced syntax errors:

  (/* comment */ { x: 0 }).x;
  (/* comment */ function () { })();
  (/* comment */ class { }).prototype;

  This regression has been fixed.

v0.16.15

• Add format to input files in the JSON metafile data

  When --metafile is enabled, input files may now have an additional format field that indicates the export format used by this file. When present, the value will either be cjs for CommonJS-style exports or esm for ESM-style exports. This can be useful in bundle analysis.

  For example, esbuild's new Bundle Size Analyzer now uses this information to visualize whether ESM or CommonJS was used for each directory and file of source code (click on the CJS/ESM bar at the top).

  This information is helpful when trying to reduce the size of your bundle. Using the ESM variant of a dependency instead of the CommonJS variant always results in a faster and smaller bundle because it omits CommonJS wrappers, and also may result in better tree-shaking as it allows esbuild to perform tree-shaking at the statement level instead of the module level.

• Fix a bundling edge case with dynamic import (#2793)

  This release fixes a bug where esbuild's bundler could produce incorrect output. The problematic edge case involves the entry point importing itself using a dynamic import() expression in an imported file, like this:

  // src/a.js
  export const A = 42;
  
  // src/b.js
  export const B = async () => (await import(".")).A
  
  // src/index.js
  export * from "./a"
  export * from "./b"

• Remove new type syntax from type declarations in the esbuild package (#2798)

  Previously you needed to use TypeScript 4.3 or newer when using the esbuild package from TypeScript code due to the use of a getter in an interface in node_modules/esbuild/lib/main.d.ts. This release removes this newer syntax to allow people with versions of TypeScript as far back as TypeScript 3.5 to use this latest version of the esbuild package. Here is change that was made to esbuild's type declarations:

   export interface OutputFile {
     /** "text" as bytes */
     contents: Uint8Array;
     /** "contents" as text (changes automatically with "contents") */
  -  get text(): string;
  +  readonly text: string;
   }

v0.16.14

• Preserve some comments in expressions (#2721)

  Various tools give semantic meaning to comments embedded inside of expressions. For example, Webpack and Vite have special "magic comments" that can be used to affect code splitting behavior:

  import(/* webpackChunkName: "foo" */ '../foo');
  import(/* @vite-ignore */ dynamicVar);
  new Worker(/* webpackChunkName: "bar" */ new URL("../bar.ts", import.meta.url));
  new Worker(new URL('./path', import.meta.url), /* @vite-ignore */ dynamicOptions);

  Since esbuild can be used as a preprocessor for these tools (e.g. to strip TypeScript types), it can be problematic if esbuild doesn't do additional work to try to retain these comments. Previously esbuild special-cased Webpack comments in these specific locations in the AST. But Vite would now like to use similar comments, and likely other tools as well.

  So with this release, esbuild now will attempt to preserve some comments inside of expressions in more situations than before. This behavior is mainly intended to preserve these special "magic comments" that are meant for other tools to consume, although esbuild will no longer only preserve Webpack-specific comments so it should now be tool-agnostic. There is no guarantee that all such comments will be preserved (especially when --minify-syntax is enabled). So this change does not mean that esbuild is now usable as a code formatter. In particular comment preservation is more likely to happen with leading comments than with trailing comments. You should put comments that you want to be preserved before the relevant expression instead of after it. Also note that this change does not retain any more statement-level comments than before (i.e. comments not embedded inside of expressions). Comment preservation is not enabled when --minify-whitespace is enabled (which is automatically enabled when you use --minify).

v0.16.13

• Publish a new bundle visualization tool

  While esbuild provides bundle metadata via the --metafile flag, previously esbuild left analysis of it completely up to third-party tools (well, outside of the rudimentary --analyze flag). However, the esbuild website now has a built-in bundle visualization tool:

  • https://esbuild.github.io/analyze/

  You can pass --metafile to esbuild to output bundle metadata, then upload that JSON file to this tool to visualize your bundle. This is helpful for answering questions such as:

  • Which packages are included in my bundle?
  • How did a specific file get included?
  • How small did a specific file compress to?
  • Was a specific file tree-shaken or not?

  I'm publishing this tool because I think esbuild should provide some answer to "how do I visualize my bundle" without requiring people to reach for third-party tools. At the moment the tool offers two types of visualizations: a radial "sunburst chart" and a linear "flame chart". They serve slightly different but overlapping use cases (e.g. the sunburst chart is more keyboard-accessible while the flame chart is easier with the mouse). This tool may continue to evolve over time.

• Fix --metafile and --mangle-cache with --watch (#1357)

  The CLI calls the Go API and then also writes out the metafile and/or mangle cache JSON files if those features are enabled. This extra step is necessary because these files are returned by the Go API as in-memory strings. However, this extra step accidentally didn't happen for all builds after the initial build when watch mode was enabled. This behavior used to work but it was broken in version 0.14.18 by the introduction of the mangle cache feature. This release fixes the combination of these features, so the metafile and mangle cache features should now work with watch mode. This behavior was only broken for the CLI, not for the JS or Go APIs.

• Add an original field to the metafile

  The metadata file JSON now has an additional field: each import in an input file now contains the pre-resolved path in the original field in addition to the post-resolved path in the path field. This means it's now possible to run certain additional analysis over your bundle. For example, you should be able to use this to detect when the same package subpath is represented multiple times in the bundle, either because multiple versions of a package were bundled or because a package is experiencing the dual-package hazard.

v0.16.12

• Loader defaults to js for extensionless files (#2776)

  Certain packages contain files without an extension. For example, the yargs package contains the file yargs/yargs which has no extension. Node, Webpack, and Parcel can all understand code that imports yargs/yargs because they assume that the file is JavaScript. However, esbuild was previously unable to understand this code because it relies on the file extension to tell it how to interpret the file. With this release, esbuild will now assume files without an extension are JavaScript files. This can be customized by setting the loader for "" (the empty string, representing files without an extension) to another loader. For example, if you want files without an extension to be treated as CSS instead, you can do that like this:

  • CLI:

    esbuild --bundle --loader:=css
    

  • JS:

    esbuild.build({
      bundle: true,
      loader: { '': 'css' },
    })

  • Go:

    api.Build(api.BuildOptions{
      Bundle: true,
      Loader: map[string]api.Loader{"": api.LoaderCSS},
    })

  In addition, the "type" field in package.json files now only applies to files with an explicit .js, .jsx, .ts, or .tsx extension. Previously it was incorrectly applied by esbuild to all files that had an extension other than .mjs, .mts, .cjs, or .cts including extensionless files. So for example an extensionless file in a "type": "module" package is now treated as CommonJS instead of ESM.

v0.16.11

• Avoid a syntax error in the presence of direct eval (#2761)

  The behavior of nested function declarations in JavaScript depends on whether the code is run in strict mode or not. It would be problematic if esbuild preserved nested function declarations in its output because then the behavior would depend on whether the output was run in strict mode or not instead of respecting the strict mode behavior of the original source code. To avoid this, esbuild transforms nested function declarations to preserve the intended behavior of the original source code regardless of whether the output is run in strict mode or not:

  // Original code
  if (true) {
    function foo() {}
    console.log(!!foo)
    foo = null
    console.log(!!foo)
  }
  console.log(!!foo)
  
  // Transformed code
  if (true) {
    let foo2 = function() {
    };
    var foo = foo2;
    console.log(!!foo2);
    foo2 = null;
    console.log(!!foo2);
  }
  console.log(!!foo);

  In the above example, the original code should print true false true because it's not run in strict mode (it doesn't contain "use strict" and is not an ES module). The code that esbuild generates has been transformed such that it prints true false true regardless of whether it's run in strict mode or not.

  However, this transformation is impossible if the code contains direct eval because direct eval "poisons" all containing scopes by preventing anything in those scopes from being renamed. That prevents esbuild from splitting up accesses to foo into two separate variables with different names. Previously esbuild still did this transformation but with two variables both named foo, which is a syntax error. With this release esbuild will now skip doing this transformation when direct eval is present to avoid generating code with a syntax error. This means that the generated code may no longer behave as intended since the behavior depends on the run-time strict mode setting instead of the strict mode setting present in the original source code. To fix this problem, you will need to remove the use of direct eval.

• Fix a bundling scenario involving multiple symlinks (#2773, #2774)

  This release contains a fix for a bundling scenario involving an import path where multiple path segments are symlinks. Previously esbuild was unable to resolve certain import paths in this scenario, but these import paths should now work starting with this release. This fix was contributed by @onebytegone.

v0.16.10

• Change the default "legal comment" behavior again (#2745)

  The legal comments feature automatically gathers comments containing @license or @preserve and puts the comments somewhere (either in the generated code or in a separate file). This behavior used to be on by default but was disabled by default in version 0.16.0 because automatically inserting comments is potentially confusing and misleading. These comments can appear to be assigning the copyright of your code to another entity. And this behavior can be especially problematic if it happens automatically by default since you may not even be aware of it happening. For example, if you bundle the TypeScript compiler the preserving legal comments means your source code would contain this comment, which appears to be assigning the copyright of all of your code to Microsoft:

  /*! *****************************************************************************
  Copyright (c) Microsoft Corporation. All rights reserved.
  Licensed under the Apache License, Version 2.0 (the "License"); you may not use
  this file except in compliance with the License. You may obtain a copy of the
  License at http://www.apache.org/licenses/LICENSE-2.0
  
  THIS CODE IS PROVIDED ON AN *AS IS* BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
  KIND, EITHER EXPRESS OR IMPLIED, INCLUDING WITHOUT LIMITATION ANY IMPLIED
  WARRANTIES OR CONDITIONS OF TITLE, FITNESS FOR A PARTICULAR PURPOSE,
  MERCHANTABLITY OR NON-INFRINGEMENT.
  
  See the Apache Version 2.0 License for specific language governing permissions
  and limitations under the License.
  ***************************************************************************** */

  However, people have asked for this feature to be re-enabled by default. To resolve the confusion about what these comments are applying to, esbuild's default behavior will now be to attempt to describe which package the comments are coming from. So while this feature has been re-enabled by default, the output will now look something like this instead:

  /*! Bundled license information:
  
  typescript/lib/typescript.js:
    (*! *****************************************************************************
    Copyright (c) Microsoft Corporation. All rights reserved.
    Licensed under the Apache License, Version 2.0 (the "License"); you may not use
    this file except in compliance with the License. You may obtain a copy of the
    License at http://www.apache.org/licenses/LICENSE-2.0
  
    THIS CODE IS PROVIDED ON AN *AS IS* BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
    KIND, EITHER EXPRESS OR IMPLIED, INCLUDING WITHOUT LIMITATION ANY IMPLIED
    WARRANTIES OR CONDITIONS OF TITLE, FITNESS FOR A PARTICULAR PURPOSE,
    MERCHANTABLITY OR NON-INFRINGEMENT.
  
    See the Apache Version 2.0 License for specific language governing permissions
    and limitations under the License.
    ***************************************************************************** *)
  */

  Note that you can still customize this behavior with the --legal-comments= flag. For example, you can use --legal-comments=none to turn this off, or you can use --legal-comments=linked to put these comments in a separate .LEGAL.txt file instead.

• Enable external legal comments with the transform API (#2390)

  Previously esbuild's transform API only supported none, inline, or eof legal comments. With this release, external legal comments are now also supported with the transform API. This only applies to the JS and Go APIs, not to the CLI, and looks like this:

  • JS:

    const { code, legalComments } = await esbuild.transform(input, {
      legalComments: 'external',
    })

  • Go:

    result := api.Transform(input, api.TransformOptions{
      LegalComments: api.LegalCommentsEndOfFile,
    })
    code := result.Code
    legalComments := result.LegalComments

• Fix duplicate function declaration edge cases (#2757)

  The change in the previous release to forbid duplicate function declarations in certain cases accidentally forbid some edge cases that should have been allowed. Specifically duplicate function declarations are forbidden in nested blocks in strict mode and at the top level of modules, but are allowed when they are declared at the top level of function bodies. This release fixes the regression by re-allowing the last case.

• Allow package subpaths with alias (#2715)

  Previously the names passed to the alias feature had to be the name of a package (with or without a package scope). With this release, you can now also use the alias feature with package subpaths. So for example you can now create an alias that substitutes @org/pkg/lib with something else.