Skip to content
This repository has been archived by the owner on Aug 8, 2019. It is now read-only.

code splitting

Jump to bottom
Ricardo Marques edited this page Mar 23, 2017 · 32 revisions

For big web apps it's not efficient to put all code into a single file, especially if some blocks of code are only required under some circumstances. Webpack has a feature to split your codebase into "chunks" which are loaded on demand. Some other bundlers call them "layers", "rollups", or "fragments". This feature is called "code splitting".

It's an opt-in feature. You can define split points in your code base. Webpack takes care of the dependencies, output files and runtime stuff.

To clarify a common misunderstanding: Code Splitting is not just about extracting common code into a shared chunk. The more notable feature is that Code Splitting can be used to split code into an on demand loaded chunk. This can keep the initial download small and downloads code on demand when requested by the application.

Defining a split point

AMD and CommonJs specify different methods to load code on demand. Both are supported and act as split points:

CommonJs: require.ensure 

require.ensure(dependencies, callback)

The require.ensure method ensures that every dependency in dependencies can be synchronously required when calling the callback. An implementation of the require function is sent as a parameter to the callback.

Example:

require.ensure(["module-a", "module-b"], function() {
  var a = require("module-a");
  // ...
});

Note: require.ensure only loads the modules, it doesn't evaluate them.

AMD: require

The AMD spec defines an asynchronous require method with this definition:

require(dependencies, callback)

When called, all dependencies are loaded and the callback is called with the exports of the loaded dependencies.

Example:

require(["module-a", "module-b"], function(a, b) {
  // ...
});

Note: AMD require loads and evaluate the modules. In webpack modules are evaluated left to right.

Note: It's allowed to omit the callback.

ES6 Modules

tl;dr: Webpack doesn't support ES6 modules; use require.ensure or require directly depending on which module format your transpiler creates.

Webpack 1.x.x (coming in 2.0.0!) does not natively support or understand ES6 modules. However, you can get around that by using a transpiler, like Babel, to turn the ES6 import syntax into CommonJs or AMD modules. This approach is effective but has one important caveat for dynamic loading.

The module syntax addition (import x from 'foo') is intentionally designed to be statically analyzable, which means that you cannot do dynamic imports.

// INVALID!!!!!!!!!
['lodash', 'backbone'].forEach(name => import name )

Luckily, there is a JavaScript API "loader" specification being written to handle the dynamic use case: System.load (or System.import). This API will be the native equivalent to the above require variations. However, most transpilers do not support converting System.load calls to require.ensure so you have to do that directly if you want to make use of dynamic code splitting.

//static imports
import _ from 'lodash'

// dynamic imports
require.ensure([], function() {
  let contacts = require('./contacts')
})

Chunk content

All dependencies at a split point go into a new chunk. Dependencies are also recursively added.

If you pass a function expression (or bound function expression) as callback to the split point, webpack automatically puts all dependencies required in this function expression into the chunk too.

Chunk optimization

If two chunks contain the same modules, they are merged into one. This can cause chunks to have multiple parents.

If a module is available in all parents of a chunk, it's removed from that chunk.

If a chunk contains all modules of another chunk, this is stored. It fulfills multiple chunks.

Chunk loading

Depending on the configuration option target a runtime logic for chunk loading is added to the bundle. I. e. for the web target chunks are loaded via jsonp. A chunk is only loaded once and parallel requests are merged into one. The runtime checks for loaded chunks whether they fulfill multiple chunks.

Chunk types

Entry chunk

An entry chunk contains the runtime plus a bunch of modules. If the chunk contains the module 0 the runtime executes it. If not, it waits for chunks that contains the module 0 and executes it (every time when there is a chunk with a module 0).

Normal chunk

A normal chunk contains no runtime. It only contains a bunch of modules. The structure depends on the chunk loading algorithm. I. e. for jsonp the modules are wrapped in a jsonp callback function. The chunk also contains a list of chunk id that it fulfills.

Initial chunk (non-entry)

An initial chunk is a normal chunk. The only difference is that optimization treats it as more important because it counts toward the initial loading time (like entry chunks). That chunk type can occur in combination with the CommonsChunkPlugin.

Split app and vendor code

To split your app into 2 files, say app.js and vendor.js, you can require the vendor files in vendor.js. Then pass this name to the CommonsChunkPlugin as shown below.

var webpack = require("webpack");

module.exports = {
  entry: {
    app: "./app.js",
    vendor: ["jquery", "underscore", ...],
  },
  output: {
    filename: "bundle.js"
  },
  plugins: [
    new webpack.optimize.CommonsChunkPlugin(/* chunkName= */"vendor", /* filename= */"vendor.bundle.js")
  ]
};

This will remove all modules in the vendor chunk from the app chunk. The bundle.js will now contain just your app code, without any of its dependencies. These are in vendor.bundle.js.

In your HTML page load vendor.bundle.js before bundle.js.

<script src="vendor.bundle.js"></script>
<script src="bundle.js"></script>

Multiple entry chunks

It's possible to configure multiple entry points that will result in multiple entry chunks. The entry chunk contains the runtime and there must only be one runtime on a page (there are exceptions).

Running multiple entry points

With the CommonsChunkPlugin the runtime is moved to the commons chunk. The entry points are now in initial chunks. While only one initial chunk can be loaded, multiple entry chunks can be loaded. This exposes the possibility to run multiple entry points in a single page.

Example:

var webpack = require("webpack");
module.exports = {
  entry: { a: "./a", b: "./b" },
  output: { filename: "[name].js" },
  plugins: [ new webpack.optimize.CommonsChunkPlugin("init.js") ]
}
<script src="init.js"></script>
<script src="a.js"></script>
<script src="b.js"></script>

Commons chunk

The CommonsChunkPlugin can move modules that occur in multiple entry chunks to a new entry chunk (the commons chunk). The runtime is moved to the commons chunk too. This means the old entry chunks are initial chunks now. See all options in the list of plugins.

Optimization

There are optimizing plugins that can merge chunks depending on specific criteria. See list of plugins.

  • LimitChunkCountPlugin
  • MinChunkSizePlugin
  • AggressiveMergingPlugin

Named chunks

The require.ensure function accepts an additional 3rd parameter. This must be a string. If two split point pass the same string they use the same chunk.

require.include

require.include(request)

require.include is a webpack specific function that adds a module to the current chunk, but doesn't evaluate it (The statement is removed from the bundle).

Example:

require.ensure(["./file"], function() {
  require("./file2");
});

// is equal to

require.ensure([], function() {
  require.include("./file");
  require("./file2");
});

require.include can be useful if a module is in multiple child chunks. A require.include in the parent would include the module and the instances of the modules in the child chunks would disappear.

Examples

For a running demo see the example-app. Check Network in DevTools.

webpack 👍

Clone this wiki locally