import WebpackLicense from '@components/WebpackLicense'; import { ApiMeta, Stability } from '../../../components/ApiMeta'; import { Tabs, Tab } from '@theme';

Output

The top-level output key contains a set of options instructing Rspack on how and where it should output your bundles, assets, and anything else you bundle or load with Rspack.

Type: Object

output.assetModuleFilename

Type: string | ((pathData: PathData, assetInfo?: AssetInfo) => string)
Default: '[hash][ext][query]'

The same as output.filename but for Asset Modules.

[name], [file], [query], [fragment], [base], and [path] are set to an empty string for the assets built from data URI replacements.

The name of the file to be output by the Asset module. This value can be overridden by rules[].generator.filename.

:::info Asset module output as a separate file

Module type is 'asset' and asset is set to satisfy rules[].parser.dataUrlCondition
Module type is 'asset/resource'

:::

output.asyncChunks

Type: boolean
Default: true

Controls whether dynamically imported modules are emitted as separate async chunks or bundled into existing chunks.

true: Modules loaded via import() are split into independent async chunks. These chunks are emitted as separate files and are loaded on demand at runtime. This enables code splitting and keeps the initial bundle smaller.
false: Dynamically imported modules are bundled into existing chunks instead of being emitted as separate files. No additional async chunk files are generated.

export default {
  output: {
    asyncChunks: false,
  },
};

output.bundlerInfo

In versions 1.x, use experiments.rspackFuture.bundlerInfo instead.

Type:

type BundlerInfo = {
  version?: string,
  bundler?: string,
  force?: ('version' | 'uniqueId')[] ｜ boolean;
};

Used to inject the currently used Rspack information into the generated asset:

version: Used to specify the Rspack version, defaults to the version field in @rspack/core/package.json.
bundler: Used to specify the name of the packaging tool, defaults to rspack.
force: Whether to force the injection of Rspack information, which will be added to chunk as a runtime module, and defaults to false. An array can be used to select the items to be forced injected.

By default, injection occurs when either __rspack_version__ or __rspack_unique_id__ is detected in the code, and only the corresponding runtime module is injected. Set force to true to always inject both:

__rspack_version__: Inject version information.
__rspack_unique_id__: Inject the unique ID of the bundler.

export default {
  output: {
    bundlerInfo: { force: true },
  },
};

output.chunkFilename

Type: string | (pathData: PathData, assetInfo?: AssetInfo) => string
Default: Determined by output.filename when it is not a function, otherwise '[id].js'.

This option determines the name of non-initial chunk files. See output.filename option for details on the possible values.

Note that these filenames need to be generated at runtime to send the requests for chunks. Because of this, placeholders like [name] and [chunkhash] need to add a mapping from chunk id to placeholder value to the output bundle with the Rspack runtime. This increases the size and may invalidate the bundle when placeholder value for any chunk changes.

By default [id].js is used or a value inferred from output.filename([name] is replaced with [id] or [id]. is prepended).

export default {
  output: {
    chunkFilename: '[id].js',
  },
};

:::tip See Filename placeholders for more information. :::

Usage as a function:

export default {
  output: {
    chunkFilename: (pathData) => {
      return pathData.chunk.name === 'main' ? '[name].js' : '[name]/[name].js';
    },
  },
};

output.chunkFormat

Type: false | 'array-push' | 'commonjs' | 'module' | string
Default: Determined by target and output.module

The format of chunks (formats included by default are 'array-push' (web/webworker), 'commonjs' (node.js), 'module' (ESM), but others might be added by plugins).

:::tip

The default value of this option depends on the target and output.module setting. For more details, search for "chunkFormat" in the Rspack defaults.

:::

export default {
  output: {
    chunkFormat: 'commonjs',
  },
};

output.chunkLoading

Type: false | 'jsonp' | 'import-scripts' | 'require' | 'async-node' | 'import' | string

The method to load chunks (methods included by default are 'jsonp' (web), 'import' (ESM), 'import-scripts' (webworker), 'require' (sync node.js), 'async-node' (async node.js), but others might be added by plugins). The default value will be determined based on the configuration of target and chunkFormat.

:::tip

The default value of this option depends on the target and chunkFormat setting. For more details, search for "chunkLoading" in the Rspack defaults.

:::

export default {
  output: {
    chunkLoading: 'async-node',
  },
};

output.chunkLoadingGlobal

Type: string
Default: Determined by output.uniqueName

The global variable is used by Rspack for loading chunks.

export default {
  output: {
    chunkLoadingGlobal: 'myCustomFunc',
  },
};

output.chunkLoadTimeout

Type: number
Default: 120000

The Number of milliseconds before chunk request timed out.

export default {
  output: {
    chunkLoadTimeout: 30000, // 30 seconds before chunk request timed out.
  },
};

output.clean

Type: boolean | { keep?: string | RegExp | ((path: string) => boolean) }
Default: false

Before generating the products, delete all files in the output directory.

export default {
  output: {
    clean: true, // Clean the output directory before emit.
  },

  // or
  output: {
    clean: {
      keep: 'ignored/dir', // keep these assets under 'dist/ignored/dir'.
    },
  },

  // or
  output: {
    clean: {
      keep: /ignored\/dir/, // keep these assets under 'dist/ignored/dir'.
    },
  },

  // or
  output: {
    clean: {
      keep: (path) => path.includes('ignored/dir'), // keep these assets under 'dist/ignored/dir'.
    },
  },
};

output.compareBeforeEmit

Type: boolean
Default: true

Tells Rspack to check if to be emitted file already exists and has the same content before writing to the output file system.

:::warning Rspack will not write output file when file already exists on disk with the same content. :::

export default {
  output: {
    compareBeforeEmit: false,
  },
};

output.crossOriginLoading

Type: false | 'anonymous' | 'use-credentials'
Default: false

The crossOriginLoading config allows you to set the crossorigin attribute for dynamically loaded chunks.

If target is 'web', Rspack will dynamically create <script> and <link> tags to load asynchronous JavaScript and CSS resources. Rspack will add the crossorigin attribute to the <script> and <link> tags if the URLs of these resources are on other domains and crossOriginLoading is not false.

Optional values

crossOriginLoading has the following optional values:

false: Do not set the crossorigin attribute.
'anonymous': Set crossorigin to 'anonymous' to enable cross-origin without user credentials.
'use-credentials': Set crossorigin to 'use-credentials' enable cross-origin with user credentials.

Example

For example, set output.publicPath to https://example.com/ and output.crossOriginLoading to 'anonymous':

import path from 'node:path';

export default {
  output: {
    publicPath: 'https://example.com/',
    crossOriginLoading: 'anonymous',
  },
};

When Rspack dynamically loads JavaScript resources, it will generate the following HTML:

html

<script src="https://example.com/foo.js" crossorigin="anonymous"></script>

output.cssChunkFilename

Type: string | (pathData: PathData, assetInfo?: AssetInfo) => string
Default: Determined by output.chunkFilename when it is not a function, otherwise '[id].css'.

This option determines the name of non-initial CSS output files on disk. See output.filename option for details on the possible values.

You must not specify an absolute path here. However, feel free to include folders separated by '/'. This specified path combines with the output.path value to pinpoint the location on the disk.

output.cssFilename

Type: string | (pathData: PathData, assetInfo?: AssetInfo) => string
Default: Determined by output.filename

This option determines the name of CSS output files on disk. See output.filename option for details on the possible values.

output.devtoolFallbackModuleFilenameTemplate

Type:

type DevtoolFallbackModuleFilenameTemplate =
  | string
  | ((context: ModuleFilenameTemplateContext) => string);

Default: undefined

A fallback is used when the template string or function above yields duplicates.

See output.devtoolModuleFilenameTemplate.

output.devtoolModuleFilenameTemplate

Type:

type DevtoolModuleFilenameTemplate =
  | string
  | ((context: ModuleFilenameTemplateContext) => string);

Default: webpack://[namespace]/[resource-path]?[loaders]'

This option is only used when devtool uses an option that requires module names.

Customize the names used in each source map's sources array. This can be done by passing a template string or function. For example, when using devtool: 'eval'.

export default {
  output: {
    devtoolModuleFilenameTemplate:
      'webpack://[namespace]/[resource-path]?[loaders]',
  },
};

The following substitutions are available in template strings

Template	Description
[absolute-resource-path]	The absolute filename
[relative-resource-path]	Resource path relative to the source map file’s directory (inline source maps are not supported)
[all-loaders]	Automatic and explicit loaders and params up to the name of the first loader
[hash]	The hash of the module identifier
[id]	The module identifier
[loaders]	Explicit loaders and params up to the name of the first loader
[resource]	The path used to resolve the file and any query params used on the first loader
[resource-path]	The path used to resolve the file without any query params
[namespace]	The modules namespace. This is usually the library name when building as a library, empty otherwise

When using a function, the same options are available camel-cased via the info parameter:

export default {
  output: {
    devtoolModuleFilenameTemplate: (info) => {
      return `webpack:///${info.resourcePath}?${info.loaders}`;
    },
  },
};

If multiple modules would result in the same name, output.devtoolFallbackModuleFilenameTemplate is used instead for these modules.

output.devtoolNamespace

Type: string
Default: undefined

This option determines the module's namespace used with the output.devtoolModuleFilenameTemplate. When not specified, it will default to the value of: output.uniqueName. It's used to prevent source file path collisions in sourcemaps when loading multiple libraries built with Rspack.

For example, if you have 2 libraries, with namespaces library1 and library2, which both have a file ./src/index.js (with potentially different contents), they will expose these files as webpack://library1/./src/index.js and webpack://library2/./src/index.js.

output.enabledChunkLoadingTypes

Type: ('jsonp' | 'import-scripts' | 'require' | 'async-node' | string)[]
Default: Determined by output.chunkLoading, output.workerChunkLoading and Entry's chunkLoading config.

List of chunk loading types enabled for use by entry points. Will be automatically filled by Rspack. Only needed when using a function as entry option and returning chunkLoading option from there.

export default {
  output: {
    enabledChunkLoadingTypes: ['jsonp', 'require'],
  },
};

output.enabledLibraryTypes

Type: string[]
Default: Determined by output.library and Entry

List of library types enabled for use by entry points.

export default {
  output: {
    enabledLibraryTypes: ['module'],
  },
};

output.enabledWasmLoadingTypes

Type: ('fetch-streaming' | 'fetch' | 'async-node' | string | false)[]
Default: Determined by output.wasmLoading and output.workerWasmLoading

List of Wasm loading types enabled for use by entry points.

export default {
  output: {
    enabledWasmLoadingTypes: ['fetch'],
  },
};

output.environment

Type:

type Environment = {
  /** The environment supports arrow functions ('() => { ... }'). */
  arrowFunction?: boolean;
  /** The environment supports async function and await ('async function () { await ... }'). */
  asyncFunction?: boolean;
  /** The environment supports BigInt as literal (123n). */
  bigIntLiteral?: boolean;
  /** The environment supports const and let for variable declarations. */
  const?: boolean;
  /** The environment supports destructuring ('{ a, b } = obj'). */
  destructuring?: boolean;
  /** The environment supports 'document' variable. */
  document?: boolean;
  /** The environment supports an async import() function to import EcmaScript modules. */
  dynamicImport?: boolean;
  /** The environment supports an async import() when creating a worker, only for web targets at the moment. */
  dynamicImportInWorker?: boolean;
  /** The environment supports `import.meta.dirname` and `import.meta.filename`. */
  importMetaDirnameAndFilename?: boolean;
  /** The environment supports 'for of' iteration ('for (const x of array) { ... }'). */
  forOf?: boolean;
  /** The environment supports 'globalThis'. */
  globalThis?: boolean;
  /** The environment supports { fn() {} } */
  methodShorthand?: boolean;
  /** The environment supports ECMAScript Module syntax to import ECMAScript modules (import ... from '...'). */
  module?: boolean;
  /**
   * Determines if the node: prefix is generated for core module imports in environments that support it.
   * This is only applicable to Rspack runtime code.
   * */
  nodePrefixForCoreModules?: boolean;
  /** The environment supports optional chaining ('obj?.a' or 'obj?.()'). */
  optionalChaining?: boolean;
  /** The environment supports template literals. */
  templateLiteral?: boolean;
};

output.environment specifies which ECMAScript language features and host environment capabilities Rspack is allowed to use when generating its runtime code.

This setting affects only code emitted by Rspack itself, such as runtime helpers. It does not downgrade your application source code.

Default behavior

By default, output.environment is automatically inferred from your target configuration.

Rspack evaluates target to determine which features your target environment supports. For example:

If the target browsers support arrow functions, Rspack will use arrow functions in the runtime to produce more compact output.
If target specifies a particular Node.js version, Rspack will infer the supported syntax from that version’s capabilities.

In general, you do not need to configure this option manually. You would override it only when you want to explicitly enforce or restrict specific features. Manual configuration replaces the inferred values and gives you direct control over the syntax used in the generated runtime.

Example

Suppose target is set to ['web', 'es2018']. In this case, Rspack will infer that const is supported and enable it by default. You can disable it by setting output.environment.const: false. Rspack will then emit runtime code that uses var declarations instead.

export default {
  target: ['web', 'es2018'],
  output: {
    environment: {
      const: false,
    },
  },
};

output.filename

Type: string | (pathData: PathData, assetInfo?: AssetInfo) => string
Default: When output.module is true, it is '[name].mjs', otherwise it is '[name].js'.

This option determines the name of each output bundle. The bundle is written to the directory specified by the output.path option.

For a single entry point, this can be a static name.

export default {
  output: {
    filename: 'bundle.js',
  },
};

However, when creating multiple bundles via more than one entry point, code splitting, or various plugins, you should use one of the following substitutions to give each bundle a unique name...

:::info Description of other cases where multiple bundles can be split

Rspack performs code splitting optimizations on user input code, which may include, but are not limited to, code splitting, bundle splitting, or splitting implemented through other plugins. These splitting actions can result in multiple bundles being generated, so the filenames of the bundles need to be generated dynamically.

{ // TODO: add the Glossary link }

:::

Use Entry name:

export default {
  output: {
    filename: '[name].bundle.js',
  },
};

Using internal chunk id:

export default {
  output: {
    filename: '[id].bundle.js',
  },
};

Using hashes generated from the generated content:

export default {
  output: {
    filename: '[contenthash].bundle.js',
  },
};

Combining multiple substitutions:

export default {
  output: {
    filename: '[name].[contenthash].bundle.js',
  },
};

:::tip See Filename placeholders for more information. :::

Using the function to return the filename:

export default {
  output: {
    filename: (pathData) => {
      return pathData.chunk.name === 'main' ? '[name].js' : '[name]/[name].js';
    },
  },
};

Note this option is called filename but you are still allowed to use something like 'js/[name]/bundle.js' to create a folder structure.

Note this option does not affect output files for on-demand-loaded chunks. It only affects output files that are initially loaded. For on-demand-loaded chunk files, the output.chunkFilename option is used. Files created by loaders also aren't affected. In this case, you would have to try the specific loader's available options.

output.globalObject

Type: string
Default: 'self'

When targeting a library, especially when library.type is 'umd', this option indicates what global object will be used to mount the library. To make UMD build available on both browsers and Node.js, set output.globalObject option to 'this'. Defaults to self for Web-like targets.

The return value of your entry point will be assigned to the global object using the value of output.library.name. Depending on the value of the type option, the global object could change respectively, e.g., self, global, or globalThis.

For example:

export default {
  output: {
    library: {
      name: 'myLib',
      type: 'umd',
    },
    filename: 'myLib.js',
    globalObject: 'this',
  },
};

output.hashDigest

Type: string
Default: 'hex'

The encoding to use when generating the hash. Using 'base64' for filenames might be problematic since it has the character / in its alphabet. Likewise 'latin1' could contain any character.

output.hashDigestLength

Type: number
Default: 16

The prefix length of the hash digest to use, see Hash length for more details.

export default {
  output: {
    hashDigestLength: 8,
  },
};

output.hashFunction

Type: 'md4' | 'xxhash64' | 'sha256'
Default: 'xxhash64'

The hashing algorithm to use.

export default {
  output: {
    hashFunction: 'xxhash64',
  },
};

:::tip Rspack uses the faster xxhash64 algorithm by default since v1.1. :::

output.hashSalt

Type: string
Default: undefined

An optional salt to update the hash.

output.hotUpdateChunkFilename

Type: string
Default: "[id].[fullhash].hot-update.js"

Customize the filenames of hot update chunks. See output.filename option for details on the possible values.

The only placeholders allowed here are [id] and [fullhash], the default being:

export default {
  output: {
    hotUpdateChunkFilename: '[id].[fullhash].hot-update.js',
  },
};

:::tip Typically you don't need to change output.hotUpdateChunkFilename. :::

output.hotUpdateGlobal

Type: string
Default: "rspackHotUpdate" + output.uniqueName

Only used when target is set to 'web', which uses JSONP for loading hot updates.

A JSONP function is used to asynchronously load hot-update chunks.

For details see output.chunkLoadingGlobal.

output.hotUpdateMainFilename

Type: string
Default: "[runtime].[fullhash].hot-update.json"

Customize the main hot update filename. [fullhash] and [runtime] are available as placeholder.

:::tip Typically you don't need to change output.hotUpdateMainFilename. :::

output.iife

Type: boolean
Default: true

Tells Rspack to add IIFE wrapper around emitted code.

export default {
  output: {
    iife: true,
  },
};

output.importFunctionName

Type: string
Default: 'import'

The name of the native import() function. Can be used for polyfilling, e.g. with dynamic-import-polyfill.

export default {
  output: {
    importFunctionName: '__import__',
  },
};

output.importMetaName

Type: string
Default: 'import.meta'

The name of the native import.meta object (can be exchanged for a polyfill).

export default {
  output: {
    importMetaName: 'pseudoImport.meta',
  },
};

output.library

Output a library exposing the exports of your entry point.

Type: string | string[] | object

Let's take a look at an example.

export default {
  entry: './src/index.js',
  output: {
    library: 'MyLibrary',
  },
};

Say you have exported a function in your src/index.js entry:

export function hello(name) {
  console.log(`hello ${name}`);
}

Now the variable MyLibrary will be bound with the exports of your entry file, and here's how to consume the Rspack bundled library:

html

<script src="https://example.org/path/to/my-library.js"></script>
<script>
  MyLibrary.hello('rspack');
</script>

In the above example, we're passing a single entry file to entry, however, Rspack can accept many kinds of entry point, e.g., an array, or an object.

If you provide an array as the entry point, only the last one in the array will be exposed.

export default {
  entry: ['./src/a.js', './src/b.js'], // only exports in b.js will be exposed
  output: {
    library: 'MyLibrary',
  },
};

If an object is provided as the entry point, all entries can be exposed using the array syntax of library:

export default {
  entry: {
    a: './src/a.js',
    b: './src/b.js',
  },
  output: {
    filename: '[name].js',
    library: ['MyLibrary', '[name]'], // name is a placeholder here
  },
};

Assuming that both a.js and b.js export a function hello, here's how to consume the libraries:

html

<script src="https://example.org/path/to/a.js"></script>
<script src="https://example.org/path/to/b.js"></script>
<script>
  MyLibrary.a.hello('rspack');
  MyLibrary.b.hello('rspack');
</script>

output.library.amdContainer

Type: string

Use a container(defined in global space) for calling define/require functions in an AMD module.

:::warning Note that the value of amdContainer must be set as a global variable. :::

export default {
  output: {
    library: {
      amdContainer: 'window["clientContainer"]',
      type: 'amd', // or 'amd-require'
    },
  },
};

Which will result in the following bundle:

window['clientContainer'].define(/*define args*/); // or 'amd-require' window['clientContainer'].require(/*require args*/);

output.library.name

Specify a name for the library.

Type: string | string[] | {amd?: string, commonjs?: string, root?: string | string[]}

export default {
  output: {
    library: {
      name: 'MyLibrary',
    },
  },
};

output.library.type

Configure how the library will be exposed.

Type: string

Types included by default are 'var', 'module', 'modern-module', 'system', 'assign', 'assign-properties', 'this', 'window', 'self', 'global', 'commonjs', 'commonjs2', 'commonjs-module', 'commonjs-static', 'amd', 'amd-require', 'umd', 'umd2', 'jsonp', but others might be added by plugins.

For the following examples, we'll use _entry_return_ to indicate the values returned by the entry point.

Expose a variable

These options assign the return value of the entry point (e.g. whatever the entry point exported) to the name provided by output.library.name at whatever scope the bundle was included at.

type: 'var'

export default {
  output: {
    library: {
      name: 'MyLibrary',
      type: 'var',
    },
  },
};

When your library is loaded, the return value of your entry point will be assigned to a variable:

var MyLibrary = _entry_return_;

// In a separate script with `MyLibrary` loaded…
MyLibrary.doSomething();

type: 'assign'

export default {
  output: {
    library: {
      name: 'MyLibrary',
      type: 'assign',
    },
  },
};

This will generate an implied global which has the potential to reassign an existing value (use with caution):

MyLibrary = _entry_return_;

Be aware that if MyLibrary isn't defined earlier your library will be set in global scope.

type: 'assign-properties'

export default {
  output: {
    library: {
      name: 'MyLibrary',
      type: 'assign-properties',
    },
  },
};

Similar to type: 'assign' but a safer option as it will reuse MyLibrary if it already exists:

// only create MyLibrary if it doesn't exist
MyLibrary = typeof MyLibrary === 'undefined' ? {} : MyLibrary;
// then copy the return value to MyLibrary
// similarly to what Object.assign does

// for instance, you export a `hello` function in your entry as follow
export function hello(name) {
  console.log(`Hello ${name}`);
}

// In another script with MyLibrary loaded
// you can run `hello` function like so
MyLibrary.hello('World');

Expose via object assignment

These options assign the return value of the entry point (e.g. whatever the entry point exported) to a specific object under the name defined by output.library.name.

type: 'this'

export default {
  output: {
    library: {
      name: 'MyLibrary',
      type: 'this',
    },
  },
};

The return value of your entry point will be assigned to this under the property named by output.library.name. The meaning of this is up to you:

this['MyLibrary'] = _entry_return_;

// In a separate script
this.MyLibrary.doSomething();
MyLibrary.doSomething(); // if `this` is window

type: 'window'

export default {
  output: {
    library: {
      name: 'MyLibrary',
      type: 'window',
    },
  },
};

The return value of your entry point will be assigned to the window object using the output.library.name value.

window['MyLibrary'] = _entry_return_;

window.MyLibrary.doSomething();

type: 'global'

export default {
  output: {
    library: {
      name: 'MyLibrary',
      type: 'global',
    },
  },
};

The return value of your entry point will be assigned to the global object using the output.library.name value. Depending on the target value, the global object could change respectively, e.g., self, global or globalThis.

global['MyLibrary'] = _entry_return_;

global.MyLibrary.doSomething();

type: 'commonjs'

export default {
  output: {
    library: {
      name: 'MyLibrary',
      type: 'commonjs',
    },
  },
};

The return value of your entry point will be assigned to the exports object using the output.library.name value. As the name implies, this is used in CommonJS environments.

exports['MyLibrary'] = _entry_return_;

require('MyLibrary').doSomething();

:::warning Note that not setting a output.library.name will cause all properties returned by the entry point to be assigned to the given object; there are no checks against existing property names. :::

Module definition systems

These options will result in a bundle that comes with a complete header to ensure compatibility with various module systems. The output.library.name option will take on a different meaning under the following output.library.type options.

type: 'module'

export default {
  output: {
    library: {
      // do not specify a `name` here
      type: 'module',
    },
  },
};

Output ES modules.

This is the more compatibility-oriented ESM library path: Rspack still uses the regular chunk / runtime rendering pipeline and then converts the entry exports into export statements. It fits simpler ESM library output, or cases where you mainly want webpack-style module semantics.

type: 'modern-module'

export default {
  output: {
    library: {
      // do not specify a `name` here
      type: 'modern-module',
    },
  },
};

This configuration enables Rspack's dedicated rendering pipeline for ESM libraries. It emits direct import / export statements and supports library-oriented tree shaking, scope hoisting, code splitting, and output.library.preserveModules.

:::tip module vs modern-module

module stays on the regular chunk / runtime rendering pipeline and is closer to webpack-compatible ESM semantics.
modern-module is handled by a dedicated ESM library plugin that owns chunk rendering, linking, scope hoisting, and splitChunks behavior, making it a better fit for publishable tree-shakable, code-split ESM libraries.
modern-module currently exists as a separate type to fill the ESM library capability gap; once it is mature enough, the long-term direction is to converge it with module into a single library.type.
preserveModules is only available with modern-module.
When output.library is configured and output.module = true, Rspack currently defaults output.library.type to modern-module if you don't specify one explicitly.
Do not mix modern-module with other library types in the same compilation, because it also changes chunk loading, chunk rendering, and scope-hoisting behavior.

:::

type: 'commonjs2'

export default {
  output: {
    library: {
      // note there's no `name` here
      type: 'commonjs2',
    },
  },
};

The return value of your entry point will be assigned to the module.exports. As the name implies, this is used in Node.js (CommonJS) environments:

module.exports = _entry_return_;

require('MyLibrary').doSomething();

If we specify output.library.name with type: commmonjs2, the return value of your entry point will be assigned to the module.exports.[output.library.name].

:::tip Wondering the difference between CommonJS and CommonJS2 is? While they are similar, there are some subtle differences between them that are not usually relevant in the context of Rspack. (For further details, please read this issue.) :::

type: 'commonjs-static'

export default {
  output: {
    library: {
      // note there's no `name` here
      type: 'commonjs-static',
    },
  },
};

Individual exports will be set as properties on module.exports. The "static" in the name refers to the output being statically analysable, and thus named exports are importable into ESM via Node.js:

Input:

export function doSomething() {}

Output:

function doSomething() {}

exports.doSomething = __webpack_exports__.doSomething;

Consumption (CommonJS):

const { doSomething } = require('./output.cjs'); // doSomething => [Function: doSomething]

Consumption (ESM):

import { doSomething } from './output.cjs'; // doSomething => [Function: doSomething]

:::tip This is useful when source code is written in ESM and the output should be compatible with both CJS and ESM. For further details, please read this issue or this article (specifically, this section). :::

:::tip How To Choose A CommonJS library.type

Type	Writes To	Better Fit
`commonjs`	`exports[...]`	Compatibility with legacy configs that intentionally write to the `exports` object
`commonjs2`	`module.exports`	General Node.js / CJS library output; the preferred choice for most new configs
`commonjs-module`	`module.exports`	Currently equivalent to `commonjs2`, mainly for historical webpack config compatibility
`commonjs-static`	`exports.foo = ...`	When you want Node.js ESM to be able to perform named imports from the CJS output

Rspack keeps these spellings primarily for webpack ecosystem compatibility and does not currently deprecate them at the config level. For new projects, prefer commonjs2 or commonjs-static in most cases; reach for commonjs / commonjs-module mainly when migrating existing configs or when you explicitly need exports assignment semantics.

:::

type: 'amd'

This will expose your library as an AMD module.

AMD modules require that the entry chunk (e.g. the first script loaded by the <script> tag) be defined with specific properties, such as to define and require which is typically provided by RequireJS or any compatible loaders (such as almond). Otherwise, loading the resulting AMD bundle directly will result in an error like define is not defined.

With the following configuration:

export default {
  output: {
    library: {
      name: 'MyLibrary',
      type: 'amd',
    },
  },
};

The generated output will be defined with the name "MyLibrary", i.e.:

define('MyLibrary', [], function () {
  return _entry_return_;
});

The bundle can be included as part of a script tag, and the bundle can be invoked like so:

require(['MyLibrary'], function (MyLibrary) {
  // Do something with the library...
});

If output.library.name is undefined, the following is generated instead.

define(function () {
  return _entry_return_;
});

This bundle will not work as expected, or not work at all (in the case of the almond loader) if loaded directly with a <script> tag. It will only work through a RequireJS compatible asynchronous module loader through the actual path to that file, so in this case, the output.path and output.filename may become important for this particular setup if these are exposed directly on the server.

type: 'amd-require'

export default {
  output: {
    library: {
      name: 'MyLibrary',
      type: 'amd-require',
    },
  },
};

This packages your output with an immediately executed AMD require(dependencies, factory) wrapper.

The 'amd-require' type allows for the use of AMD dependencies without needing a separate later invocation. As with the 'amd' type, this depends on the appropriate require function being available in the environment in which the Rspack output is loaded.

With this type, the library name can't be used.

type: 'umd'

This exposes your library under all the module definitions, allowing it to work with CommonJS, AMD, and as global variable. Take a look at the UMD Repository to learn more.

In this case, you need the library.name property to name your module:

export default {
  output: {
    library: {
      name: 'MyLibrary',
      type: 'umd',
    },
  },
};

And finally the output is:

(function webpackUniversalModuleDefinition(root, factory) {
  if (typeof exports === 'object' && typeof module === 'object')
    module.exports = factory();
  else if (typeof define === 'function' && define.amd) define([], factory);
  else if (typeof exports === 'object') exports['MyLibrary'] = factory();
  else root['MyLibrary'] = factory();
})(global, function () {
  return _entry_return_;
});

Note that omitting library.name will result in the assignment of all properties returned by the entry point be assigned directly to the root object, as documented under the object assignment section. Example:

export default {
  output: {
    library: {
      type: 'umd',
    },
  },
};

The output will be:

(function webpackUniversalModuleDefinition(root, factory) {
  if (typeof exports === 'object' && typeof module === 'object')
    module.exports = factory();
  else if (typeof define === 'function' && define.amd) define([], factory);
  else {
    var a = factory();
    for (var i in a) (typeof exports === 'object' ? exports : root)[i] = a[i];
  }
})(global, function () {
  return _entry_return_;
});

You may specify an object for library.name for differing names per targets:

export default {
  output: {
    library: {
      name: {
        root: 'MyLibrary',
        amd: 'my-library',
        commonjs: 'my-common-library',
      },
      type: 'umd',
    },
  },
};

type: 'system'

This will expose your library as a System.register module.

System modules require that a global variable System is present in the browser when the Rspack bundle is executed. Compiling to System.register format allows you to System.import('/bundle.js') without additional configuration and has your Rspack bundle loaded into the System module registry.

export default {
  output: {
    library: {
      type: 'system',
    },
  },
};

Output:

System.register([], function (__WEBPACK_DYNAMIC_EXPORT__, __system_context__) {
  return {
    execute: function () {
      // ...
    },
  };
});

By adding output.library.name to configuration in addition to having output.library.type set to system, the output bundle will have the library name as an argument to System.register:

System.register(
  'MyLibrary',
  [],
  function (__WEBPACK_DYNAMIC_EXPORT__, __system_context__) {
    return {
      execute: function () {
        // ...
      },
    };
  },
);

Other types

type: 'jsonp'

export default {
  output: {
    library: {
      name: 'MyLibrary',
      type: 'jsonp',
    },
  },
};

This will wrap the return value of your entry point into a jsonp wrapper.

MyLibrary(_entry_return_);

The dependencies for your library will be defined by the externals config.

output.library.export

Specify which export should be exposed as a library.

Type: string | string[]
Default: undefined

It is undefined by default, which will export the whole (namespace) object. The examples below demonstrate the effect of this configuration when using output.library.type: 'var'.

export default {
  output: {
    library: {
      name: 'MyLibrary',
      type: 'var',
      export: 'default',
    },
  },
};

The default export of your entry point will be assigned to the library name:

// If your library has a default export
var MyLibrary = _entry_return_.default;

You can pass an array to output.library.export as well, it will be interpreted as a path to a module to be assigned to the library name:

export default {
  output: {
    library: {
      name: 'MyLibrary',
      type: 'var',
      export: ['default', 'subModule'],
    },
  },
};

And here's the library code:

var MyLibrary = _entry_return_.default.subModule;

output.library.auxiliaryComment

Add a comment in the UMD wrapper.

Type: string | { amd?: string, commonjs?: string, commonjs2?: string, root?: string }
Default: undefined

To insert the same comment for each umd type, set auxiliaryComment to a string:

export default {
  mode: 'development',
  output: {
    library: {
      name: 'MyLibrary',
      type: 'umd',
      auxiliaryComment: 'Test Comment',
    },
  },
};

which will yield the following:

(function webpackUniversalModuleDefinition(root, factory) {
  //Test Comment
  if (typeof exports === 'object' && typeof module === 'object')
    module.exports = factory();
  //Test Comment
  else if (typeof define === 'function' && define.amd) define([], factory);
  //Test Comment
  else if (typeof exports === 'object') exports['MyLibrary'] = factory();
  //Test Comment
  else root['MyLibrary'] = factory();
})(self, function () {
  return _entry_return_;
});

For fine-grained control, pass an object:

export default {
  mode: 'development',
  output: {
    library: {
      name: 'MyLibrary',
      type: 'umd',
      auxiliaryComment: {
        root: 'Root Comment',
        commonjs: 'CommonJS Comment',
        commonjs2: 'CommonJS2 Comment',
        amd: 'AMD Comment',
      },
    },
  },
};

output.library.umdNamedDefine

Type: boolean
Default: undefined

When using output.library.type: "umd", setting output.library.umdNamedDefine to true will name the AMD module of the UMD build. Otherwise, an anonymous define is used.

export default {
  output: {
    library: {
      name: 'MyLibrary',
      type: 'umd',
      umdNamedDefine: true,
    },
  },
};

The AMD module will be:

define('MyLibrary', [], factory);

output.library.preserveModules

Type: string
Default: undefined

:::info Only available when library.type is set to modern-module. :::

When enabled, Rspack will preserve the original directory structure of modules within the specified directory. Each module will be emitted as a separate file, maintaining the same relative path as the source.

preserveModules is a special form of bundling — it can be thought of as bundle + splitChunks where each module gets its own chunk. Unlike a bundleless tool, Rspack still applies its full bundling pipeline (tree shaking, scope hoisting, minification, etc.), but splits the output so that each module in the specified directory becomes its own output file.

:::warning Barrel files may not be emitted or exposed as you expect. Since preserveModules is still a form of bundling, Rspack only emits modules that are reachable from the configured entries. Non-entry barrel files (for example, an index.ts that re-exports from other modules) can still be emitted with their re-exports if they are part of the module graph, but only the configured entry exports are exposed through the main library entry. Unused barrel re-exports may be tree-shaken, and a barrel file may not be emitted at all if nothing references it. If you need to expose exports from a barrel file via your library entry, configure it as an entry point. :::

import path from 'node:path';

export default {
  output: {
    library: {
      type: 'modern-module',
      preserveModules: path.resolve('./src'),
    },
  },
};

output.module

Type: boolean
Default: false

Output JavaScript files as module type.

When enabled, Rspack will set output.iife to false, output.scriptType to 'module' and terserOptions.module to true internally.

If you set output.library.type to 'module' or 'modern-module', Rspack will set output.module to true for you automatically.

export default {
  output: {
    module: true,
  },
};

output.path

Type: string
Default: path.resolve(process.cwd(), 'dist')

The output directory as an absolute path.

import path from 'node:path';
import { fileURLToPath } from 'node:url';

const __dirname = path.dirname(fileURLToPath(import.meta.url));

export default {
  output: {
    path: path.resolve(__dirname, 'dist/assets'),
  },
};

</Tab> <Tab label="CJS">

const path = require('node:path');

module.exports = {
  output: {
    path: path.resolve(__dirname, 'dist/assets'),
  },
};

</Tab> </Tabs>

Note that [fullhash] in this parameter will be replaced with a hash of the compilation.

:::warning

The path must not contain an exclamation mark (!) as it is reserved by Rspack for loader syntax.
Relative paths are resolved against context.

:::

output.pathinfo

Type: boolean | 'verbose'
Default: false

Controls whether Rspack adds module-related comments to the generated bundle. These comments are useful for debugging, inspecting build output, and understanding tree-shaking behavior.

true: Enables basic comments, including module path information.
false: Disables all comments, which is the default behavior.
'verbose': Outputs detailed comments, such as module exports, runtime details, tree-shaking information, and bailout reasons. This mode is helpful when diagnosing build issues or performing in-depth bundle analysis.

For example, enabling pathinfo:

export default {
  output: {
    pathinfo: true,
  },
};

The generated bundle will contain comments like:

/*!**********************!*\
  !*** ./src/index.js ***!
  \**********************/

:::warning pathinfo is primarily intended for development and debugging. These comments increase bundle size and may expose details about your source structure. Avoid using it in production. :::

output.publicPath

Type: 'auto' | string | ((pathData: PathData, assetInfo?: AssetInfo) => string)
Default: 'auto' when target is 'web' or 'webworker', undefined otherwise

Set the base URL path prefix for bundled static assets (such as JS, CSS, images, etc.).

export default {
  output: {
    publicPath: '/assets/',
  },
};

See Asset base path for more details.

output.scriptType

Type: 'module' | 'text/javascript' | boolean
Default: false

This option allows loading asynchronous chunks with a custom script type, such as <script type="module" ...>.

:::tip If output.module is set to true, output.scriptType will default to 'module' instead of false. :::

export default {
  output: {
    scriptType: 'module',
  },
};

output.sourceMapFilename

Type: string
Default: '[file].map[query]'

Configure how source maps are named. Only takes effect when devtool is set to 'source-map', which writes an output file.

The [name], [id], [fullhash] and [chunkhash] substitutions from output.filename can be used. In addition to those, you can use substitutions listed under Filename-level in Filename placeholders.

output.strictModuleErrorHandling

Type: boolean
Default: false

Handle error in module loading as per ECMAScript Modules spec at a performance cost.

export default {
  output: {
    strictModuleErrorHandling: true,
  },
};

output.strictModuleExceptionHandling

Types: boolean

Tell Rspack to remove a module from the module instance cache (require.cache) if it throws an exception when it is required.

output.trustedTypes

Type: true | string | object
Default: undefined

Controls Trusted Types compatibility. When enabled, Rspack will detect Trusted Types support and, if they are supported, use Trusted Types policies to create script URLs it loads dynamically. Use when the application runs under a require-trusted-types-for Content Security Policy directive.

It is disabled by default (no compatibility, script URLs are strings).

When set to true, Rspack will use output.uniqueName as the Trusted Types policy name.
When set to a non-empty string, its value will be used as a policy name.
When set to an object, the policy name is taken from the object's policyName property.

export default {
  output: {
    trustedTypes: {
      policyName: 'my-application#rspack',
    },
  },
};

output.trustedTypes.onPolicyCreationFailure

Type: "stop" | "continue"
Default: "stop"

Determine whether to proceed with loading in anticipation that require-trusted-types-for 'script' has not been enforced or to immediately fail when the call to trustedTypes.createPolicy(...) fails due to the policy name being absent from the CSP trusted-types list or being a duplicate.

export default {
  output: {
    trustedTypes: {
      policyName: 'my-application#rspack',
      onPolicyCreationFailure: 'continue',
    },
  },
};

output.uniqueName

Type: string
Default: Derived in the following priority order:
1. The name of output.library
2. The name field in package.json under the context directory
3. '' (empty string)

This option is used to identify a unique name for the current build. It mainly helps prevent runtime code conflicts when multiple Rspack applications run in the same global environment.

output.uniqueName affects the default values of the following options:

output.chunkLoadingGlobal
output.hotUpdateGlobal
output.devtoolNamespace
output.trustedTypes.policyName

To override the default value:

export default {
  output: {
    uniqueName: 'my-package-xyz',
  },
};

output.wasmLoading

Type: false | 'fetch', 'async-node'
Default: 'fetch'

Option to set the method of loading WebAssembly Modules. Methods included by default are 'fetch' (web/webworker), 'async-node' (Node.js), but others might be added by plugins.

The default value can be affected by different target:

Defaults to 'fetch' if target is set to 'web', 'webworker', 'electron-renderer' or 'node-webkit'.
Defaults to 'async-node' if target is set to 'node', 'async-node', 'electron-main' or 'electron-preload'.

export default {
  output: {
    wasmLoading: 'fetch',
  },
};

output.webassemblyModuleFilename

Type: string
Default: '[hash].module.wasm'

Specifies the filename of WebAssembly modules. It should be provided as a relative path within the output.path directory

export default {
  output: {
    webassemblyModuleFilename: '[contenthash:8].wasm',
  },
};

:::tip See Filename placeholders for more information. :::

output.workerChunkLoading

Type: false | 'jsonp' | 'import-scripts' | 'require' | 'async-node' | 'import'
Default: false

The new option workerChunkLoading controls the chunk loading of workers.

:::tip The default value of this option depends on the target setting. For more details, search for "workerChunkLoading" in the Rspack defaults. :::

export default {
  output: {
    workerChunkLoading: false,
  },
};

output.workerPublicPath

Type: string
Default: ""

Set a public path for Worker, defaults to value of output.publicPath. Only use this option if your worker scripts are located in a different path from your other scripts.

export default {
  output: {
    workerPublicPath: '/workerPublicPath2/',
  },
};

output.workerWasmLoading

Type: false | 'fetch-streaming' | 'fetch' | 'async-node' | string
Default: false

Option to set the method of loading WebAssembly Modules in workers, defaults to the value of output.wasmLoading.

export default {
  output: {
    workerWasmLoading: 'fetch',
  },
};