webpack、rollup 和 parcel 是现代前端模块打包工具,负责将代码转换为浏览器可运行的格式。gulp 是一个任务运行器,用于自动化工作流。browserify 是早期的模块打包工具,允许在浏览器中使用 CommonJS。这些工具解决了代码分割、依赖管理和资源优化的问题,但各自的设计理念和适用场景不同。
在前端工程化领域,选择合适的构建工具直接影响开发效率和项目维护成本。webpack、rollup、parcel、gulp 和 browserify 代表了不同时期的解决方案。虽然它们都能处理代码转换,但核心目标和使用方式有很大区别。本文将从配置、入口处理、插件扩展和维护状态四个维度进行深度对比。
不同的工具对配置文件的依赖程度不同。有的追求零配置,有的追求完全可控。
webpack 需要明确的配置文件,通常命名为 webpack.config.js。
// webpack: webpack.config.js
module.exports = {
entry: './src/index.js',
output: {
filename: 'bundle.js',
path: __dirname + '/dist'
},
module: {
rules: [{ test: /\.css$/, use: 'style-loader' }]
}
};
rollup 也需要配置文件,通常命名为 rollup.config.js。
// rollup: rollup.config.js
export default {
input: 'src/index.js',
output: {
file: 'dist/bundle.js',
format: 'iife'
},
plugins: [require('@rollup/plugin-node-resolve')()]
};
parcel 主打零配置,通常不需要配置文件。
.parcelrc。// parcel: .parcelrc (可选)
{
"extends": "@parcel/config-default",
"transformers": {
"*.js": ["@parcel/transformer-js"]
}
}
gulp 使用 gulpfile.js 定义任务流。
// gulp: gulpfile.js
const { src, dest } = require('gulp');
exports.build = function() {
return src('src/*.js').pipe(dest('dist'));
};
browserify 主要通过命令行或 API 使用,较少使用配置文件。
-t 参数添加转换。# browserify: CLI
browserify src/main.js -o dist/bundle.js -t babelify
处理代码入口的方式决定了工具的灵活性。
webpack 支持灵活的入口定义。
// webpack: 多入口配置
module.exports = {
entry: {
main: './src/main.js',
admin: './src/admin.js'
},
output: {
filename: '[name].bundle.js'
}
};
rollup 通常专注于单入口。
// rollup: 多格式输出
export default {
input: 'src/index.js',
output: [
{ file: 'dist/bundle.cjs.js', format: 'cjs' },
{ file: 'dist/bundle.esm.js', format: 'es' }
]
};
parcel 自动识别入口类型。
# parcel: 自动识别
parcel build src/index.html
gulp 使用 glob 匹配文件。
// gulp: 文件流匹配
exports.styles = function() {
return src('src/scss/**/*.scss').pipe(dest('dist/css'));
};
browserify 指定单个入口文件。
require()。// browserify: API 用法
const browserify = require('browserify');
browserify('./src/main.js').bundle().pipe(require('fs').createWriteStream('bundle.js'));
当默认功能不足时,插件系统决定了工具的上限。
webpack 拥有最庞大的插件生态。
plugins 数组添加功能。// webpack: 插件示例
const HtmlWebpackPlugin = require('html-webpack-plugin');
module.exports = {
plugins: [new HtmlWebpackPlugin({ template: './src/index.html' })]
};
rollup 插件专注于模块转换。
plugins 数组中。// rollup: 插件示例
import terser from '@rollup/plugin-terser';
export default {
plugins: [terser()]
};
parcel 插件系统较新,基于配置。
.parcelrc 指定转换器。// parcel: 自定义转换器
{
"transformers": {
"*.md": ["@parcel/transformer-markdown"]
}
}
gulp 插件基于流式处理。
.pipe() 连接插件。// gulp: 插件链
const uglify = require('gulp-uglify');
exports.minify = function() {
return src('src/*.js').pipe(uglify()).pipe(dest('dist'));
};
browserify 使用 Transform 流。
-t 或 API 添加转换。// browserify: Transform
const b = require('browserify')();
b.transform('babelify');
b.bundle().pipe(require('fs').createWriteStream('bundle.js'));
工具的生命周期直接影响项目长期维护。
webpack 处于活跃维护状态(v5)。
rollup 处于活跃维护状态(v3/v4)。
parcel 处于活跃维护状态(v2)。
gulp 处于维护状态(v4)。
browserify 处于低维护状态。
| 特性 | webpack | rollup | parcel | gulp | browserify |
|---|---|---|---|---|---|
| 核心定位 | 应用打包 | 库打包 | 零配置打包 | 任务运行 | 早期模块打包 |
| 配置难度 | 高 | 中 | 低 | 中 | 低 |
| 模块支持 | CommonJS/ESM | 优先 ESM | 通用 | 文件流 | CommonJS |
| 代码分割 | 强大 | 支持 | 自动 | 不支持 | 不支持 |
| 热更新 | 支持 (HMR) | 支持 | 支持 | 需插件 | 需插件 |
| 当前状态 | ✅ 活跃 | ✅ 活跃 | ✅ 活跃 | ⚠️ 辅助 | ❌ 遗留 |
webpack 是大型应用的安全选择 🏢。如果你需要处理复杂的路由、资源加载和代码分割,它的生态能解决几乎所有问题。虽然配置繁琐,但可控性最高。
rollup 是库开发的标准工具 📦。如果你发布 npm 包,它能生成最干净的代码,且对 Tree-shaking 支持最好。不要用它构建复杂的应用页面。
parcel 是原型开发的神器 🚀。如果你想在几分钟内启动项目且不想配置 Webpack,它是最佳选择。适合中小型项目或内部工具。
gulp 是自动化脚本的补充 🤖。不要用它打包 JS 模块,但可以用它处理图片压缩、文件复制等构建后的任务。
browserify 是历史的一部分 🕰️。除非你正在维护 5 年前的老项目,否则不要在新架构中引入它。现代工具在性能和体验上都远超它。
总结:新应用首选 webpack 或 parcel,开发库首选 rollup,辅助任务用 gulp,避免使用 browserify。
仅在选择维护旧项目时使用 browserify。它是最早让浏览器支持 require() 的工具,但已不再适合新架构。现代打包工具提供了更好的性能和生态支持,新项目应避免使用。
选择 gulp 如果你需要编排复杂的构建任务流程,比如压缩图片、编译 Sass 或部署文件。它不适合单独作为模块打包器,通常与 webpack 或 rollup 配合使用来处理非代码资源。
选择 parcel 如果你想要零配置启动项目。它适合快速原型开发或中小型应用,默认支持多种格式,无需编写复杂的配置文件,能自动处理大多数构建需求。
选择 rollup 如果你主要开发 JavaScript 库或组件。它对 ES 模块支持最好,生成的代码最干净,树摇(tree-shaking)效果优秀,适合需要发布 npm 包的场景。
选择 webpack 如果你构建大型复杂的应用程序。它拥有最强大的插件系统和代码分割能力,适合需要精细控制构建流程、加载各种资源类型的大型企业级项目。
require('modules') in the browser
Use a node-style require() to organize your browser code
and load modules installed by npm.
browserify will recursively analyze all the require() calls in your app in
order to build a bundle you can serve up to the browser in a single <script>
tag.

If you're new to browserify, check out the browserify handbook and the resources on browserify.org.
Whip up a file, main.js with some require()s in it. You can use relative
paths like './foo.js' and '../lib/bar.js' or module paths like 'gamma'
that will search node_modules/ using
node's module lookup algorithm.
var foo = require('./foo.js');
var bar = require('../lib/bar.js');
var gamma = require('gamma');
var elem = document.getElementById('result');
var x = foo(100) + bar('baz');
elem.textContent = gamma(x);
Export functionality by assigning onto module.exports or exports:
module.exports = function (n) { return n * 111 }
Now just use the browserify command to build a bundle starting at main.js:
$ browserify main.js > bundle.js
All of the modules that main.js needs are included in the bundle.js from a
recursive walk of the require() graph using
required.
To use this bundle, just toss a <script src="bundle.js"></script> into your
html!
With npm do:
npm install browserify
Usage: browserify [entry files] {OPTIONS}
Standard Options:
--outfile, -o Write the browserify bundle to this file.
If unspecified, browserify prints to stdout.
--require, -r A module name or file to bundle.require()
Optionally use a colon separator to set the target.
--entry, -e An entry point of your app
--ignore, -i Replace a file with an empty stub. Files can be globs.
--exclude, -u Omit a file from the output bundle. Files can be globs.
--external, -x Reference a file from another bundle. Files can be globs.
--transform, -t Use a transform module on top-level files.
--command, -c Use a transform command on top-level files.
--standalone -s Generate a UMD bundle for the supplied export name.
This bundle works with other module systems and sets the name
given as a window global if no module system is found.
--debug -d Enable source maps that allow you to debug your files
separately.
--help, -h Show this message
For advanced options, type `browserify --help advanced`.
Specify a parameter.
Advanced Options:
--insert-globals, --ig, --fast [default: false]
Skip detection and always insert definitions for process, global,
__filename, and __dirname.
benefit: faster builds
cost: extra bytes
--insert-global-vars, --igv
Comma-separated list of global variables to detect and define.
Default: __filename,__dirname,process,Buffer,global
--detect-globals, --dg [default: true]
Detect the presence of process, global, __filename, and __dirname and define
these values when present.
benefit: npm modules more likely to work
cost: slower builds
--ignore-missing, --im [default: false]
Ignore `require()` statements that don't resolve to anything.
--noparse=FILE
Don't parse FILE at all. This will make bundling much, much faster for giant
libs like jquery or threejs.
--no-builtins
Turn off builtins. This is handy when you want to run a bundle in node which
provides the core builtins.
--no-commondir
Turn off setting a commondir. This is useful if you want to preserve the
original paths that a bundle was generated with.
--no-bundle-external
Turn off bundling of all external modules. This is useful if you only want
to bundle your local files.
--bare
Alias for both --no-builtins, --no-commondir, and sets --insert-global-vars
to just "__filename,__dirname". This is handy if you want to run bundles in
node.
--no-browser-field, --no-bf
Turn off package.json browser field resolution. This is also handy if you
need to run a bundle in node.
--transform-key
Instead of the default package.json#browserify#transform field to list
all transforms to apply when running browserify, a custom field, like, e.g.
package.json#browserify#production or package.json#browserify#staging
can be used, by for example running:
* `browserify index.js --transform-key=production > bundle.js`
* `browserify index.js --transform-key=staging > bundle.js`
--node
Alias for --bare and --no-browser-field.
--full-paths
Turn off converting module ids into numerical indexes. This is useful for
preserving the original paths that a bundle was generated with.
--deps
Instead of standard bundle output, print the dependency array generated by
module-deps.
--no-dedupe
Turn off deduping.
--list
Print each file in the dependency graph. Useful for makefiles.
--extension=EXTENSION
Consider files with specified EXTENSION as modules, this option can used
multiple times.
--global-transform=MODULE, -g MODULE
Use a transform module on all files after any ordinary transforms have run.
--ignore-transform=MODULE, -it MODULE
Do not run certain transformations, even if specified elsewhere.
--plugin=MODULE, -p MODULE
Register MODULE as a plugin.
Passing arguments to transforms and plugins:
For -t, -g, and -p, you may use subarg syntax to pass options to the
transforms or plugin function as the second parameter. For example:
-t [ foo -x 3 --beep ]
will call the `foo` transform for each applicable file by calling:
foo(file, { x: 3, beep: true })
Many npm modules that don't do IO will just work after being browserified. Others take more work.
Many node built-in modules have been wrapped to work in the browser, but only
when you explicitly require() or use their functionality.
When you require() any of these modules, you will get a browser-specific shim:
Additionally, if you use any of these variables, they will be defined in the bundled output in a browser-appropriate way:
You can just as easily create a bundle that will export a require() function so
you can require() modules from another script tag. Here we'll create a
bundle.js with the through
and duplexer modules.
$ browserify -r through -r duplexer -r ./my-file.js:my-module > bundle.js
Then in your page you can do:
<script src="bundle.js"></script>
<script>
var through = require('through');
var duplexer = require('duplexer');
var myModule = require('my-module');
/* ... */
</script>
If you prefer the source maps be saved to a separate .js.map source map file, you may use
exorcist in order to achieve that. It's as simple as:
$ browserify main.js --debug | exorcist bundle.js.map > bundle.js
Learn about additional options here.
If browserify finds a required function already defined in the page scope, it
will fall back to that function if it didn't find any matches in its own set of
bundled modules.
In this way, you can use browserify to split up bundles among multiple pages to
get the benefit of caching for shared, infrequently-changing modules, while
still being able to use require(). Just use a combination of --external and
--require to factor out common dependencies.
For example, if a website with 2 pages, beep.js:
var robot = require('./robot.js');
console.log(robot('beep'));
and boop.js:
var robot = require('./robot.js');
console.log(robot('boop'));
both depend on robot.js:
module.exports = function (s) { return s.toUpperCase() + '!' };
$ browserify -r ./robot.js > static/common.js
$ browserify -x ./robot.js beep.js > static/beep.js
$ browserify -x ./robot.js boop.js > static/boop.js
Then on the beep page you can have:
<script src="common.js"></script>
<script src="beep.js"></script>
while the boop page can have:
<script src="common.js"></script>
<script src="boop.js"></script>
This approach using -r and -x works fine for a small number of split assets,
but there are plugins for automatically factoring out components which are
described in the
partitioning section of the browserify handbook.
You can use the API directly too:
var browserify = require('browserify');
var b = browserify();
b.add('./browser/main.js');
b.bundle().pipe(process.stdout);
var browserify = require('browserify')
browserify([files] [, opts])Returns a new browserify instance.
files and opts are both optional, but must be in the order shown if both are
passed.
Entry files may be passed in files and / or opts.entries.
External requires may be specified in opts.require, accepting the same formats
that the files argument does.
If an entry file is a stream, its contents will be used. You should pass
opts.basedir when using streaming files so that relative requires can be
resolved.
opts.entries has the same definition as files.
opts.noParse is an array which will skip all require() and global parsing for
each file in the array. Use this for giant libs like jquery or threejs that
don't have any requires or node-style globals but take forever to parse.
opts.transform is an array of transform functions or modules names which will
transform the source code before the parsing.
opts.ignoreTransform is an array of transformations that will not be run,
even if specified elsewhere.
opts.plugin is an array of plugin functions or module names to use. See the
plugins section below for details.
opts.extensions is an array of optional extra extensions for the module lookup
machinery to use when the extension has not been specified.
By default browserify considers only .js and .json files in such cases.
opts.basedir is the directory that browserify starts bundling from for
filenames that start with ..
opts.paths is an array of directories that browserify searches when looking
for modules which are not referenced using relative path. Can be absolute or
relative to basedir. Equivalent of setting NODE_PATH environmental variable
when calling browserify command.
opts.commondir sets the algorithm used to parse out the common paths. Use
false to turn this off, otherwise it uses the
commondir module.
opts.fullPaths disables converting module ids into numerical indexes. This is
useful for preserving the original paths that a bundle was generated with.
opts.builtins sets the list of built-ins to use, which by default is set in
lib/builtins.js in this distribution.
opts.bundleExternal boolean option to set if external modules should be
bundled. Defaults to true.
When opts.browserField is false, the package.json browser field will be
ignored. When opts.browserField is set to a string, then a custom field name
can be used instead of the default "browser" field.
When opts.insertGlobals is true, always insert process, global,
__filename, and __dirname without analyzing the AST for faster builds but
larger output bundles. Default false.
When opts.detectGlobals is true, scan all files for process, global,
__filename, and __dirname, defining as necessary. With this option npm
modules are more likely to work but bundling takes longer. Default true.
When opts.ignoreMissing is true, ignore require() statements that don't
resolve to anything.
When opts.debug is true, add a source map inline to the end of the bundle.
This makes debugging easier because you can see all the original files if
you are in a modern enough browser.
When opts.standalone is a non-empty string, a standalone module is created
with that name and a umd wrapper.
You can use namespaces in the standalone global export using a . in the string
name as a separator, for example 'A.B.C'. The global export will be sanitized
and camel cased.
Note that in standalone mode the require() calls from the original source will
still be around, which may trip up AMD loaders scanning for require() calls.
You can remove these calls with
derequire:
$ npm install derequire
$ browserify main.js --standalone Foo | derequire > bundle.js
opts.insertGlobalVars will be passed to
insert-module-globals
as the opts.vars parameter.
opts.externalRequireName defaults to 'require' in expose mode but you can
use another name.
opts.bare creates a bundle that does not include Node builtins, and does not
replace global Node variables except for __dirname and __filename.
opts.node creates a bundle that runs in Node and does not use the browser
versions of dependencies. Same as passing { bare: true, browserField: false }.
Note that if files do not contain javascript source code then you also need to specify a corresponding transform for them.
All other options are forwarded along to module-deps and browser-pack directly.
Add an entry file from file that will be executed when the bundle loads.
If file is an array, each item in file will be added as an entry file.
Make file available from outside the bundle with require(file).
The file param is anything that can be resolved by require.resolve(),
including files from node_modules. Like with require.resolve(), you must
prefix file with ./ to require a local file (not in node_modules).
file can also be a stream, but you should also use opts.basedir so that
relative requires will be resolvable.
If file is an array, each item in file will be required.
In file array form, you can use a string or object for each item. Object items
should have a file property and the rest of the parameters will be used for
the opts.
Use the expose property of opts to specify a custom dependency name.
require('./vendor/angular/angular.js', {expose: 'angular'}) enables require('angular')
Bundle the files and their dependencies into a single javascript file.
Return a readable stream with the javascript file contents or
optionally specify a cb(err, buf) to get the buffered results.
Prevent file from being loaded into the current bundle, instead referencing
from another bundle.
If file is an array, each item in file will be externalized.
If file is another bundle, that bundle's contents will be read and excluded
from the current bundle as the bundle in file gets bundled.
Prevent the module name or file at file from showing up in the output bundle.
If file is an array, each item in file will be ignored.
Instead you will get a file with module.exports = {}.
Prevent the module name or file at file from showing up in the output bundle.
If file is an array, each item in file will be excluded.
If your code tries to require() that file it will throw unless you've provided
another mechanism for loading it.
Transform source code before parsing it for require() calls with the transform
function or module name tr.
If tr is a function, it will be called with tr(file) and it should return a
through-stream
that takes the raw file contents and produces the transformed source.
If tr is a string, it should be a module name or file path of a
transform module
with a signature of:
var through = require('through');
module.exports = function (file) { return through() };
You don't need to necessarily use the through module. Browserify is compatible with the newer, more verbose Transform streams built into Node v0.10.
Here's how you might compile coffee script on the fly using .transform():
var coffee = require('coffee-script');
var through = require('through');
b.transform(function (file) {
var data = '';
return through(write, end);
function write (buf) { data += buf }
function end () {
this.queue(coffee.compile(data));
this.queue(null);
}
});
Note that on the command-line with the -c flag you can just do:
$ browserify -c 'coffee -sc' main.coffee > bundle.js
Or better still, use the coffeeify module:
$ npm install coffeeify
$ browserify -t coffeeify main.coffee > bundle.js
If opts.global is true, the transform will operate on ALL files, despite
whether they exist up a level in a node_modules/ directory. Use global
transforms cautiously and sparingly, since most of the time an ordinary
transform will suffice. You can also not configure global transforms in a
package.json like you can with ordinary transforms.
Global transforms always run after any ordinary transforms have run.
Transforms may obtain options from the command-line with subarg syntax:
$ browserify -t [ foo --bar=555 ] main.js
or from the api:
b.transform('foo', { bar: 555 })
In both cases, these options are provided as the second argument to the transform function:
module.exports = function (file, opts) { /* opts.bar === 555 */ }
Options sent to the browserify constructor are also provided under
opts._flags. These browserify options are sometimes required if your transform
needs to do something different when browserify is run in debug mode, for
example.
Register a plugin with opts. Plugins can be a string module name or a
function the same as transforms.
plugin(b, opts) is called with the browserify instance b.
For more information, consult the plugins section below.
There is an internal labeled-stream-splicer pipeline with these labels:
'record' - save inputs to play back later on subsequent bundle() calls'deps' - module-deps'json' - adds module.exports= to the beginning of json files'unbom' - remove byte-order markers'unshebang' - remove #! labels on the first line'syntax' - check for syntax errors'sort' - sort the dependencies for deterministic bundles'dedupe' - remove duplicate source contents'label' - apply integer labels to files'emit-deps' - emit 'dep' event'debug' - apply source maps'pack' - browser-pack'wrap' - apply final wrapping, require= and a newline and semicolonYou can call b.pipeline.get() with a label name to get a handle on a stream pipeline
that you can push(), unshift(), or splice() to insert your own transform
streams.
Reset the pipeline back to a normal state. This function is called automatically
when bundle() is called multiple times.
This function triggers a 'reset' event.
browserify uses the package.json in its module resolution algorithm, just like
node. If there is a "main" field, browserify will start resolving the package
at that point. If there is no "main" field, browserify will look for an
"index.js" file in the module root directory. Here are some more
sophisticated things you can do in the package.json:
There is a special "browser" field you can set in your package.json on a per-module basis to override file resolution for browser-specific versions of files.
For example, if you want to have a browser-specific module entry point for your
"main" field you can just set the "browser" field to a string:
"browser": "./browser.js"
or you can have overrides on a per-file basis:
"browser": {
"fs": "level-fs",
"./lib/ops.js": "./browser/opts.js"
}
Note that the browser field only applies to files in the local module, and like
transforms, it doesn't apply into node_modules directories.
You can specify source transforms in the package.json in the
browserify.transform field. There is more information about how source
transforms work in package.json on the
module-deps readme.
For example, if your module requires brfs, you can add
"browserify": { "transform": [ "brfs" ] }
to your package.json. Now when somebody require()s your module, brfs will
automatically be applied to the files in your module without explicit
intervention by the person using your module. Make sure to add transforms to
your package.json dependencies field.
When a file is resolved for the bundle, the bundle emits a 'file' event with
the full file path, the id string passed to require(), and the parent
object used by
browser-resolve.
You could use the file event to implement a file watcher to regenerate bundles
when files change.
When a package file is read, this event fires with the contents. The package
directory is available at pkg.__dirname.
When .bundle() is called, this event fires with the bundle output stream.
When the .reset() method is called or implicitly called by another call to
.bundle(), this event fires.
When a transform is applied to a file, the 'transform' event fires on the
bundle stream with the transform stream tr and the file that the transform
is being applied to.
For some more advanced use-cases, a transform is not sufficiently extensible. Plugins are modules that take the bundle instance as their first parameter and an option hash as their second.
Plugins can be used to do perform some fancy features that transforms can't do.
For example, factor-bundle is a
plugin that can factor out common dependencies from multiple entry-points into a
common bundle. Use plugins with -p and pass options to plugins with
subarg syntax:
browserify x.js y.js -p [ factor-bundle -o bundle/x.js -o bundle/y.js ] \
> bundle/common.js
For a list of plugins, consult the browserify-plugin tag on npm.
There is a wiki page that lists the known browserify transforms.
If you write a transform, make sure to add your transform to that wiki page and
add a package.json keyword of browserify-transform so that
people can browse for all the browserify
transforms on npmjs.org.
There is a wiki page that lists the known browserify tools.
If you write a tool, make sure to add it to that wiki page and
add a package.json keyword of browserify-tool so that
people can browse for all the browserify
tools on npmjs.org.
Releases are documented in changelog.markdown and on the browserify twitter feed.
