Case study: analyzing the Walmart site performance

Walmart is one of the top USA e-commerce retailers. In 2016, they were the second after Amazon by sales.

In e-commerce, the conversion is directly affected by how fast the site loads. For many e-commerce companies, making the site faster by 1 second increased the conversion 1.05, 1.1, or even 1.2 times. That’s because the slower the site, the more users abandon it before it loads, and the lesser is the conversion.

Unfortunately, the Walmart site is pretty slow. In my tests, the content of the product page becomes visible only at the third second:

In comparison, for Amazon, the content gets visible at 1.4 seconds. The customer sees the product they came for twice faster!

Let’s analyze the Walmart’s site and see how we can improve the performance – and help Walmart earn more! I’ll use the Lumia 635 product page as an example.

Fix the invisible text#

The first issue with the page is that it gets more or less rendered at 2.3s, but the text isn’t visible until 3.0s:

This happens because Walmart uses a custom font, and by default, Chrome and Firefox won’t render the text until the font is loaded. This is how it looks live:

See how the page stays without the text for a second?

(Network throttled with the “Fast 3G” preset in Chrome DevTools)

Browsers delay rendering the text to prevent a flash of unstyled text (FOUT). However, this makes the content invisible for longer – and likely decreases the conversion!

To change this behavior, we can add the font-display: optional rule to the @font-face styles. font-display controls how the custom font is applied. In our case, it tells the browser to just use a fallback font if the custom one is not cached:

/* */
@font-face {
font-family: "BogleWeb";
/* ... */
font-display: optional;

Now, when a customer visits the page for the first time, they will see the text immediately, rendered in a fallback font. The browser will download the custom font in the background and use it for subsequent pages. The current page won’t get the custom font – this prevents the FOUT:

Now the text is visible immediately.

(Network throttled with the “Fast 3G” preset in Chrome DevTools. The CSS file was substituted with Fiddler)

Side note: single-page apps#

With font-display: optional, the font won’t be applied until the user reloads the page. Keep this in mind if you have a single-page app: navigating across routes there won’t make the font active.

Optimize JavaScript#

Another issue is that the page downloads around 2 MBs of gzipped JavaScript. That’s a lot:

JavaScript code is minified, so I’m only able to analyze it on the surface. Here’s what I found.

Use defer for the first bundle#

Most of <script> tags on the page have either the async or the defer attribute. This is good because the browser can render the page not waiting for these scripts to download:

The page has more scripts in different places, so that’s just an example

However, one large file – bundle.3p.min-[hash].js, 112.3 kB gzipped – doesn’t have either of these attributes. If it takes a while to download (e.g., the customer is on a bad connection), the page will stay blank until the script is fully loaded. Not cool!

To be honest, the bad connection could delay any non-deferred script, even the smallest one. So I’d try to defer as many scripts as I can

To solve this, add the defer attribute to this script tag too. As soon as all JavaScript that relies on bundle.3p.min-[hash].js is also deferred (which seems to be the case), the code will keep working fine.

Side note: performance marks#

In the screenshot above, there’s code that likely measures the time the bundle takes executing:

<script src="[hash]/v1/standard_js.bundle.[hash].js" id="bundleJs" defer></script>

This code doesn’t work as expected: because of defer, the bundle executes after both of these inline scripts. Just in case somebody from Walmart is reading this.

Load non-important code only when necessary#

Chrome DevTools have the “Coverage” tab that analyzes how much CSS and JS is unused. If we open the tab, reload the page and click around a bit to run the most important JavaScript, we’ll see that around 40-60% of JS still hasn’t executed:

This code likely includes modals, popups and other components that aren’t rendered straight when the customer opens the page. They are a good candidate to be loaded only when actually needed. This might save us a few hundred kBs of JS.

This is how you load components dynamically with React and webpack:

import React from 'react';

class FeedbackButton extends React.Component {
handleButtonClick() {
// ↓ Here, import() will make webpack split FeedbackModal
// into a separate file
// and download it only when import() is called
import('../FeedbackModal/').then(module => {
this.setState({ FeedbackModal: module.default });

render() {
const FeedbackModal = this.state.FeedbackModal;

return <React.Fragment>
<button onClick={this.handleButtonClick}>
Provide feedback!
{FeedbackModal && <FeedbackModal />}

Don’t serve babel-polyfill in modern browsers#

If we look into standard_js.bundle.[hash].js, we’ll notice that it includes babel-polyfill:

Pretty easy to find by searching for “babel”

babel-polyfill weights 32.9 kB gzipped and takes 170 ms to download on Fast 3G:

By not shipping this polyfill in modern browsers, we could make the page fully interactive 170 ms earlier! And this is fairly easy to do:

  • either use an external service that serves polyfills based on User-Agent, like,
  • or build a second bundle without polyfills and serve it using <script type="module">, like in the Philip Walton’s article.

Don’t load polyfills multiple times#

Another problem is that the Object.assign polyfill is served in 3 files simultaneously:

The polyfill is small on its own, but this might be a sign that more modules are duplicated across the bundles. I’d try looking into that if I had access to sources.

Remove Node.js polyfills#

By default, webpack bundles polyfills for Node.js-specific functions when it sees them used. Theoretically, this is useful: if a library relies on setImmediate or Buffer which are only available in Node.js, it will still work in a browser thanks to the polyfill. In practice, however, I’ve seen the following happen:

// node_modules/random-library/index.js
const func = () => { ... };

if (typeof setImmediate !== 'undefined') {
// ↑ Webpack decides that `setImmediate` is used
// and adds the polyfill
} else {
setTimeout(func, 0);

The library is adapted to work in the browser, but because webpack sees that it references setImmediate, it bundles the polyfill.

Node polyfills are small (a few kBs minified), so removing them usually doesn’t make sense. Still, it’s a good candidate to optimize if we were squeezing the last milliseconds from the page. Removing them is super easy (but needs to be tested – what if some code really needs them?):

// webpack.config.js
module.exports = {
node: false,

Decrease the render-blocking CSS#

Apart from JS, page rendering is also blocked by CSS. The browser won’t render the page until all CSS (and JS) files are downloaded.

The Walmart page initially depends on two CSS files. In my tests, the largest of them takes even longer to download than the JS bundle – so it blocks rendering even after the script got downloaded and executed:

Notice how the page stays blank (look into “Frames” in the bottom half of the image) until the CSS is fully downloaded

How to solve this? We can go the way Guardian went in 2013:

  1. Find the critical CSS and extract it into a separate file. “Critical” means “The page looks funny without it”.

    Tools like Penthouse or Critical might be useful here. I’d also tune the result manually to exclude content that’s above the fold but is not very important (e.g., header navigation):

    We can show this a couple seconds later in exchange for faster overall rendering
  2. When serving the initial HTML, only load the critical CSS.
  3. Once the page is more or less downloaded (e.g., when the DOMContentLoaded event happens), dynamically add the remaining CSS:
    document.addEventListener('DOMContentLoaded', () => {
    const styles = ['', ...];
    styles.forEach((path) => {
    const link = document.createElement('link');
    link.rel = 'stylesheet';
    link.href = path;

If we get this right, we’ll be able to render the page several hundred milliseconds earlier.

Remove duplicated styles#

In total, the Walmart page downloads three CSS files: one with the font definitions (BogleWeb.css) and two with the app styles ([hash].css and style.[hash].css). The latter two seemed pretty similar, so I removed all the content except selectors and tried to compare the files.

Guess what? There’re 3400 common selectors among these files – and these selectors mostly have common styles! For the perspective, the first file has around 7900 selectors total, and the second one has around 4400:

The grep command is from StackOverflow

That’s a good area to optimize. This won’t affect time to first paint if we decrease the render-blocking CSS properly, but these CSS files will still load faster!

Add a service worker to cache assets#

The Walmart site is not a single-page app. This means that, on different pages, the customer has to download different styles and scripts. This makes every other page load longer, especially if the customer visits the site rarely.

We can improve that by creating a service worker. A service worker is a script that runs in the background even when the site is closed. It can make the app work offline, send notifications, and so on.

With Walmart, we can create a service worker that caches site resources in the background even before the user needs them. There’re multiple ways to do this; the concrete one depends on the Walmart infrastructure. A good example of one approach is available in the GoogleChrome repo.

Side note: notifications#

With service workers, we also get the ability to send notifications to customers! This should be used with caution – or we can annoy them – but this can increase engagement too. Good examples of notifications are “The product you saved for later got a discount” or “John Ford replied to your question about iPhone 8”.

To learn more, see the WebFundamentals’ guide into web push notifications.

Other ideas#

There’s still a room for further optimizations. Here’re some things that might also help – but we need to confirm that on the real app:

  • Using the local storage for caching large dependencies. The local storage seems to be several times faster than the HTTP cache. We might store large dependencies in the local storage to load them quicker:

    Update: see the Nolan Lawson’s great comment on local storage drawbacks.

  • Improving the time to first byte. Occasionally, the server spends too much time serving static resources. See the long green bars? That’s the time spent waiting for the server:

    These delays are non-deterministic – I’ve seen them pretty often during the analysis, but they keep happening with different resources every time – so this might be a network issue. Still, I’ve noticed them in WebPageTest results too.

  • Enabling Brotli compression. When you download a text resource from a server, the server would usually compress it with GZip and serve the compressed version. The browser will decompress it later, once received. This compression makes the text several times smaller.

    Apart from GZip, there’s also Brotli – a pretty new compression algorithm which compresses text 15-20% better. Right now, all text resources on the Walmart page are compressed with GZip. It makes sense to try Brotli to see if it improves the average download time.

Bonus Increase the product image quality#

That’s kinda related to performance too.

To reduce the size of the images, Walmart compresses them on the server side. The client specifies the image dimensions it expects to receive, and the server sends the appropriate image:[hash].jpeg?odnHeight=&odnWidth=&odnBg=

In most cases, this is great. However, for the primary product images, this has a negative effect. When buying an expensive gadget, I often make a final decision by visiting the product page to see the gadget, to imagine how it looks in my hands. But when I come to the Walmart site, I see a low-quality image with compression artifacts:

See yourself on WebArchive

I’d optimize this part for UX instead of performance – and serve images in a better quality. We can still keep the size difference minimal:

  • Try a different encoding algorithm. WebP is 30% smaller than JPEG given the same compression level. MozJPEG is an optimized JPEG encoder that works everywhere and has significantly less compression artifacts.
  • Use progressive images. Usually, during loading, images are rendered top-to-bottom: you see the top part of the image first, and then it fills

    Use the <picture> tag to stay compatible with different browsers. For example, we could serve WebP for Chrome and JPEG for other browsers:

    <source srcset="[hash].webp?..." type="image/webp">
    <img src="[hash].jpeg?...">
  • Serve Retina images with <source srcset>. Like this:
    srcset="[hash].webp?odnHeight=450&odnWidth=450,[hash].webp?odnHeight=900&odnWidth=900 2x"
    srcset="[hash].jpeg?odnHeight=900&odnWidth=900 2x"

Summing up#

So, to optimize the product page on the Walmart site, we can:

  • Fix the invisible text with font-display: optional
  • Use defer for the large JavaScript bundle
  • Load non-important code with webpack’s import
  • Remove polyfills in modern browsers
  • Decrease render-blocking CSS
  • Remove duplicated styles
  • Add a service worker for caching assets in background

With these tricks, we can render the product page earlier by at least 400-600 ms. If we apply similar improvements to the whole site, we can increase orders by at least 3–6% – and help Walmart earn more.

Thanks to Jacob Groß, Vladimir Starkov, and Anton Korzunov (in no particular order) for reviewing this post.

How to optimize images in webpack

Images take more than a half of the size of an average page:

A pie chart. Title: “Average bytes per page per content type.” Text in the bottom: “Total 3422 kB”. The largest pie chart section: “Images – 1818 kB.”

That’s a lot of traffic! But with webpack, it’s easy to decrease it.

1. Inline small PNG, JPG and GIF images#

Use url-loader to embed small PNG, JPG and GIF images into the bundle.

url-loader converts a file (if it’s smaller than the specified size) into a Base64 URL and inserts this URL into the bundle. This helps to avoid extra image requests (which is useful even with HTTP/2).

The limit of 5-10 KB is OK:

// webpack.config.js
module.exports = {
  module: {
    rules: [
        test: /\.(jpe?g|png|gif)$/,
        loader: 'url-loader',
        options: {
          // Images larger than 10 KB won’t be inlined
          limit: 10 * 1024

2. Inline small SVG images#

Use svg-url-loader to embed small SVG images.

This loader works like url-loader, but it encodes files using the URL encoding instead of the Base64 one. Because SVG is text, the result of the URL encoding is smaller.

The limit of 5-10 KB is also OK:

// webpack.config.js
module.exports = {
  module: {
    rules: [
        test: /\.svg$/,
        loader: 'svg-url-loader',
        options: {
          // Images larger than 10 KB won’t be inlined
          limit: 10 * 1024,
          // Remove quotes around the encoded URL –
          // they’re rarely useful
          noquotes: true,

3. Optimize image size#

Use image-webpack-loader to make images smaller.

This loader compresses PNG, JPG, GIF and SVG images by passing them through optimizers. Since it just pipes images through itself and doesn’t insert them into the bundle, it should be used with url-loader/svg-url-loader.

The default loader settings are OK:

// webpack.config.js
module.exports = {
  module: {
    rules: [
        test: /\.(jpg|png|gif|svg)$/,
        loader: 'image-webpack-loader',
        // Specify enforce: 'pre' to apply the loader
        // before url-loader/svg-url-loader
        // and not duplicate it in rules with them
        enforce: 'pre'


“Why is this important?”
As I’ve said, images take more than 50% of the average page size, and optimizing them is super-easy. Through, at the same time, I’ve rarely seen webpack configs that do it. We seriously should always optimize them.

“What should I do?”
Go and add these loaders to your config.

“Are the any side effects?”
A couple:

  • image-webpack-loader increases the build time, so it’s better to disable it during development (pass the bypassOnDebug: true option to do that)
  • url-loader and svg-url-loader remove extra image requests, but break caching of these images and increase the JS loading/parsing time and memory consumption. If you inline large images or lots of images, you might experience issues. (Thanks to Addy Osmani for noting this.)

Further reading#

How to optimize resizing or scrolling

If you listen for events like mousewheel, resize or scroll, your page could get slow. These events are generated multiple times per second, and if the event handler takes too much time, the browser won’t catch with redrawing the page.

This is what to do if you have such problem.


By default, the event handler is executed on each event. Sometimes, when the handler is taking a long time, it might be acceptable to delay its execution to once per 100-200 ms. This “delaying” can be achieved with the _.throttle utility method.

When is this useful? E.g. if you recalculate the page layout each time the browser resizes. If recalculating layout takes a long time, you could perform it less often.

Before and after:

// Before
window.addEventListener('resize', () => {
  calculateLayout(); // Takes 20 ms

// After
window.addEventListener('resize', _.throttle(() => {
  calculateLayout(); // Still takes 20 ms, but now runs once in 100 ms
}, 100));

passive: true#

passive: true a flag that’s passed into addEventListener. It tells the browser that the event handler will never call event.preventDefault(). Using this knowledge, the browser can start doing the event action (e.g. scrolling) without waiting for the JS code to finish. This makes the page smoother.

Currently, only touchstart, touchmove and mousewheel event handlers could be made passive.

When is this useful? E.g. if you redraw the parallax background on each touch or scroll event. Most likely, it won’t be a problem if the browser starts scrolling before the background finishes redrawing. If so, such event handler could be made passive.

Before and after:

// Before:
window.addEventListener('touchstart', updateBackground)

// After:
window.addEventListener('touchstart', updateBackground, {
  passive: true

Use alternative APIs#

For some common tasks, there are alternative APIs that help to perform the task easier and with better performance:

  • Intersection Observer. It helps if you need to detect when an element goes off the screen or intersects with another element. This API can replace the scroll and resize listeners.
  • window.matchMedia. This API can notify you when a specific media query becomes true. Helps if you need to e.g. update the page layout when the viewport reaches a specific width. It can replace the resize listeners.
  • position: sticky. This CSS property lets you stick an element to the top when it reaches an edge of the viewport. Helps if you need to e.g. make a sticky page header. It can replace the scroll listeners that switch the element from position: static to position: fixed and vice versa.
  • At the moment of July 2017, the API is not yet well-supported. However, there’s a polyfill.

    Resize Observer. It helps if you need to detect when an element gets resized. This API can replace the scroll and resize listeners.

    Thanks to Vitali Kuzmich for mentioning this approach.

Optimize the code itself#

You can also make the event handler itself run faster. It should take not more than 10 ms to run.

To analyze how much time your handler takes, use Web Performance API or the Performance tab in Chrome DevTools:

The optimization approach depends on your code. Try caching some data or skipping some computations. If you use React, you can try dropping parts of the render tree with shouldComponentUpdate.

And this is what won’t work:#

  • requestAnimationFrame. It doesn’t help with slow event handlers and, by my tests, doesn’t bring any benefits for fast ones. Seems like it’s only useful for scheduling JS animations.
  • Event delegation. It optimizes the memory usage, not the speed.

That’s it. Optimize.

Case study: improving a popular library’s size for webpack users

There’s a library called Polished. It’s a utility collection for writing styles in JavaScript.

The polished logo

And it had a problem.


A story in three tweets:

So, this code:

import { opacify, transparentize } from 'polished'; 

generates a much larger bundle than this code:

import opacify from 'polished/lib/color/opacify.js';
import transparentize from 'polished/lib/color/transparentize.js';

even despite the Polished’s bundle is built with ES modules and tree-shaking is enabled.

Let’s find out what causes this.


1. Verify the entry point#

Environment: [email protected] and [email protected] ([email protected] gives the same result)

At first, let’s check that import { ... } from 'polished' picks up a file written with ES exports. If it doesn’t, webpack can’t do any tree-shaking at all.

When you import a package, webpack understands what exact file to use by looking into specific fields in package.json. Polished’s package.json has two of them:

  "name": "polished",
  "description": "A lightweight toolset for writing styles in Javascript.",
  "main": "lib/index.js",  // This one
  "module": "dist/",  // And this one

Webpack prefers module over main. module points to dist/, and this file does have an ES export:

// polished/dist/
export { adjustHue$1 as adjustHue, ... };

This point is OK.

2. Check if there’s unused code that’s unnecessarily kept#

polished/dist/ is written with ES exports. This means that tree-shaking should work properly, and the unused imports shouldn’t be included into the bundle. Then why different imports produce different file sizes?

Side effect is when a function changes something outside of itself – e.g. writes a value to a global variable or initiates a network request

The most possible reason is that polished/dist/ contains some code that’s absent in our polished/lib/... files and that can’t be simply dropped by the tree-shaker. This is the code that could cause side-effects. E.g. if a file includes a top-level function call, the tree-shaker can’t remove the function even if its result isn’t used. The function could be causing side effects, and removing it could break the app.

Let’s compare the bundles that we have after importing polished in two different ways and verify this case.

To do this, I create a package:

# Shell
mkdir polished-test && cd polished-test
npm init -y
npm install polished webpack@2

Add two files that import Polished in two different ways:

console.log() helps finding the index.js file in the bundle + prevents webpack from removing the imports as unused
// index-import-package.js
import { opacify, transparentize } from 'polished';

console.log('polished', opacify, transparentize);

// index-import-files.js
import opacify from 'polished/lib/color/opacify.js';
import transparentize from 'polished/lib/color/transparentize.js';

console.log('polished', opacify, transparentize);

Add a special webpack configuration that emits two bundles:

// webpack.config.js
const webpack = require('webpack');

module.exports = {
  entry: {
    // We’ll compare two different bundles,
    // thus two different entry points
    'bundle-import-package': './index-import-package.js',
    'bundle-import-files': './index-import-files.js',
  output: {
    filename: '[name].js',
    path: __dirname,
  plugins: [
    // We need to run UglifyJS to remove the dead code
    // (this will do tree-shaking), but prevent it
    // from uglifying the code (so it’s easier to read the bundle)
    new webpack.optimize.UglifyJsPlugin({
      // Disable several optimizations so that the bundle
      // is easier to read
      compress: { sequences: false, properties: false, conditionals: false, comparisons: false, evaluate: false, booleans: false, loops: false, hoist_funs: false, hoist_vars: false, if_return: false, join_vars: false, cascade: false },

      // Beautify the bundle after uglifying it
      beautify: true,

      // Don’t rename the variables
      mangle: false,

And run the build:


Now, I have two bundles, each with a different approach to importing stuff. I open them in my editor and switch to the structure view to their content. And here’s what I see:

A comparison between the content of two files. The left file is bundle-import-package.js, it has a lot of functions. The right file is bundle-import-files.js, it has much less functions.

bundle-import-package.js has more methods than bundle-import-files.js. Most likely, they are kept because of calls with side effects. Let’s dig deeper.

3. Find the exact cause of the problem#

So, bundle-import-package.js has a lot of functions that aren’t used but are still included. If we look through the file to see their usages, we’ll see a large snippet of code like this:

// bundle-import-package.js
// ...
function opacify(amount, color) {
    // ...
var opacify$1 = curry(opacify);
function desaturate(amount, color) {
    // ...
function lighten(amount, color) {
    // ...
// ...

Here, desaturate and lighten are those unused functions, and opacify is a function we import in the client code.

This code comes to bundle-import-package.js from polished/dist/ The corresponding code in that file looks like this:

// polished/dist/
// ...
function opacify(amount, color) {
    // ...

var opacify$1 = curry(opacify);

function desaturate(amount, color) {
    // ...

var desaturate$1 = curry(desaturate);

function lighten(amount, color) {
    // ...

var lighten$1 = curry(lighten);
// ...

And this code comes into polished/dist/ from the library sources. This is how it looks:

// polished/src/color/opacify.js
function opacify(amount: number, color: string): string {
  // ...

export default curry(opacify);

// polished/src/color/desaturate.js
function desaturate(amount: number, color: string): string {
  // ...

export default curry(desaturate);

// polished/src/color/lighten.js
function lighten(amount: number, color: string): string {
  // ...

export default curry(lighten);

So what happens here? dist/ is built with Rollup. When the library authors do a build, Rollup grabs all the modules and converts exports (export default curry(lighten)) into variable assignments (var lighten$1 = curry(lighten)).

When we do import { opacify, transparentize } from 'polished', webpack tries to compile dist/ and drop the unused code. It removes the desaturate$1 and lighten$1 variables because they aren’t exported, but it can’t drop the curry(darken) calls because curry could produce side-effects. And because functions like desaturate and lighten are passed into curry(), they are also kept in the bundle.

Screenshot of the editor
This is how you analyze the bundle: open the file structure, find a function that’s absent in the other bundle, and search for its usages


To decrease the bundle size, we should do one of the following things:

Pure function is a function that doesn’t produce side effects
  • tell UglifyJS that it’s safe to remove curry() calls because it’s pure
  • or move currying into the functions instead of wrapping them.
Another option is passing compressor: { pure_funcs: ['curry'] } to the UglifyJS options, but Polished can’t control this

To tell UglifyJS that curry() calls are safe to remove, we have to mark each call with the /*#__PURE__*/ annotation. This way, the minifier will understand that this call is pure and will be able to optimize it:

We can’t just add the /*#__PURE__*/ annotation after export default. Rollup seems to remove comments if they are placed in that position
// polished/src/color/lighten.js
function lighten(amount: number, color: string): string {
  // ...
- export default curry(lighten);
+ const curriedLighten = /*#__PURE__*/curry(lighten);
+ export default curriedLighten;

The second approach is to move currying into the functions body. With it, we should do something like this:

// polished/src/color/lighten.js
- function lighten(amount: number, color: string): string {
-   // method body
- }
+ function lighten(...args) {
+   return applyCurried(function (amount: number, color: string): string {
+     // method body
+   }, args);
+ }
- export default curry(lighten);
+ export default lighten;

I prefer the first approach because it (almost) doesn’t complicate the code.

After adding the /*#__PURE__*/ annotations, minified bundle-import-package.js goes from 16 down to 11.8 kB. But that’s not the end – bundle-import-files.js is still smaller (9.86 kB). This is because there’re a few other places that should be optimized.

I’ll skip the part where I find them and jump right to the solution.

  • Change 1 and 2. Like with curry(), there’re two other places where the export is wrapped into a function. It’s polished/src/helpers/em.js and polished/src/helpers/rem.js. To optimize them, we should similarly add the /*#__PURE__*/ annotations.
  • Change 3. In polished/src/mixins/normalize.js, there’re two global objects that use computed object properties. When they are compiled, Babel transforms them to call the Babel’s defineProperty function. Because of this, UglifyJS can’t remove them. To solve the problem, we should either move these objects into the normalize() function that uses them or wrap them into getter functions.

And, when we apply these additional optimizations, we’ll have this:

                   Asset     Size  Chunks             Chunk Names
  bundle-import-files.js  9.87 kB       0  [emitted]  bundle-import-files
bundle-import-package.js  7.76 kB       1  [emitted]  bundle-import-package

bundle-import-package.js is now even smaller than bundle-import-files.js! Great.

I’ve submitted the pull request.

webpack for real tasks: decreasing front-end size and improving caching

This is the second part of a three-part introduction into webpack:

  1. Bundling front-end and adding compilation
  2. Decreasing front-end size and improving caching (you are here!)
  3. Speeding up build and improving the development workflow

Want to stay tuned for the future posts? Subscribe

Last updated on 21 Jan 2018: replaced the recommended plugin in the Moment.js part.

Task Decrease front-end size#

Given: you have a front-end application. You want to decrease its size to make it load faster.

Let’s see how webpack can help with this.


Minification is when you compress your code by removing extra spaces, shortening variable names, etc. Like this:

Webpack has two approaches to minify the code: the UglifyJS plugin and loaders options. They should be used simultaneously.

The UglifyJS plugin works on the level of the bundle and compresses it after compilation. As you might’ve guessed, it used UglifyJS under the hood. This is how it works:

You write code like this
// comments.js
import './comments.css';
export function render(data, target) {

Webpack compiles it into approximately the following
// bundle.js (part of)
"use strict";
Object.defineProperty(__webpack_exports__, "__esModule", { value: true });
var __WEBPACK_IMPORTED_MODULE_0__comments_css__ =
var __WEBPACK_IMPORTED_MODULE_0__comments_css___default =
__webpack_exports__["render"] = render;

function render(data, target) {

The UglifyJS plugin minifies it into approximately the following
// bundle.js (part of)
"use strict";function r(e,t){console.log("Rendered!")}
var o=n(4);n.n(o);t.render=r

To enable the plugin, add it to the plugins section of the config:

// webpack.config.js
const webpack = require('webpack');

module.exports = {
  plugins: [
    new webpack.optimize.UglifyJsPlugin()

The second approach is loaders options. It allows compressing things that UglifyJS can’t minify. Its point is that some code (e.g. CSS that you import) is compiled as a string which UglifyJS can’t handle:

/* comments.css */
.comment {
    color: black;

// bundle.js (part of)
exports = module.exports = __webpack_require__(1)();
exports.push([module.i, ".comment {\r\n    color: black;\r\n}", ""]);

To minify it, you should configure the loader. Here’s how you do it with css-loader:

// webpack.config.js
module.exports = {
  module: {
    rules: [
        test: /\.css$/,
        use: [
          { loader: 'css-loader', options: { minimize: true } }

Pitfall: ES2015 code#

UglifyJS 2 (which is used in webpack) can’t compile ES2015+ code. This means that if your code uses classes, arrow functions or other new language features, and you don’t compile it to ES5, UglifyJS won’t handle it. In this case, you can use Babili, a Babel-based minifier. See babili-webpack-plugin


Another way to decrease the front-end size is to set NODE_ENV environmental variable to the value “production”.

NODE_ENV is an environmental variable that is commonly used in libraries to detect in which mode the library works – in development mode or on a production server. The library can behave differently based on this variable. For example, React does additional checks and prints warnings when it’s built for development:

// …

if (process.env.NODE_ENV !== 'production') {
  validateTypeDef(Constructor, propTypes, 'prop');

// …

When you’re building your app for production, it’s better to also tell that your libraries. For Node.js libraries, it’s done by configuring the environment and setting the NODE_ENV variable to “production”. For front-end libraries, it’s done by replacing process.env.NODE_ENV with a specific value:

// webpack.config.js
const webpack = require('webpack');

module.exports = {
  plugins: {
    new webpack.DefinePlugin({
      'process.env.NODE_ENV': '"production"'

DefinePlugin takes an object with keys referring to variables to be replaced and values referring to the values that should be substituted. With this configuration, it’ll replace all process.env.NODE_ENV instances with "production", which will make UglifyJS understand that the comparison expression is always false and remove it:

ECMAScript imports#

The next way to decrease the front-end size is to use ECMAScript imports and exports.

When you use ECMAScript imports and exports, webpack becomes able to do tree-shaking. Tree-shaking is when a bundler traverses your whole dependency tree, checks what of them are used, and keeps only the used ones. So, if you use ECMAScript module syntax, webpack can eliminate the unused code:

You write two files where only one export is used
// comments.js
export const commentRestEndpoint = '/rest/comments';
export const render = () => { return 'Rendered!'; };

// index.js
import { render } from './comments.js';

Webpack realizes that commentRestEndpoint is not used and doesn’t generate a separate export point in the bundle
// bundle.js (part of)
(function(module, __webpack_exports__, __webpack_require__) {
  "use strict";
  /* unused harmony export commentRestEndpoint */
  /* harmony export */__webpack_exports__["b"] = render;

  var commentRestEndpoint = '/rest/comments';
  var render = function () { return 'Rendered!'; }

UglifyJS removes the unused variable
// bundle.js (part of)
(function(n,e){"use strict";e.b=r;var r=function(){return"Rendered!"}})

This works even with libraries; the library should also be written with ECMAScript modules.

Pitfall: tree-shaking doesn’t work without UglifyJS#

The less-known fact is that the unused code is removed not by webpack, but by UglifyJS. Webpack just removes export statements for the exports that aren’t used, which makes them possible to be removed by a minifier. Therefore, if you compile your bundle without the minifier, the bundle won’t get smaller.

See how to enable UglifyJS in the “Minification” section.

Pitfall: don’t transpile ECMAScript imports to the CommonJS format#

If you use Babel with babel-preset-env or babel-preset-es2015, check the settings of these presets. By default, they transpile ECMAScript’s import and export to CommonJS’ require and module.exports. Pass the { modules: false } option to disable this:

// webpack.config.js
module.exports = {
  module: {
    rules: [
        test: /\.js$/,
        use: [{'babel-loader', options: {
          presets: [['es2015', { modules: false }]]

Pitfall: complex cases aren’t optimized#

In some complex cases – e. g. when you re-export something (export * from 'file.js'), or when you compile classes with the TypeScript compiler – webpack can’t optimize your bundle. The bad things about this are that the cases when this happens aren’t obvious, and it’s unclear when this will be fixed. Here’s the corresponding GitHub issue: webpack/webpack#2867

Webpack team is working on a solution for re-exports though.


Tested with moment.js 2.18.1

Moment.js is a library for working with dates. By default, when you include it in your app, it takes 217 kB of minified code. That’s huge – the average size of JavaScript on a page was 417 kB in April 2017. The good part, however, is that it can be easily reduced.

165 kB of the size of moment.js is localization files. They’re included even if you don’t use them. This happens because moment.js chooses the localization files dynamically, during runtime:

Webpack doesn’t know which files you’ll need, so it includes all files from the locale directory.

To deal with it, specify the exact files with MomentLocalesPlugin:

// webpack.config.js
const MomentLocalesPlugin = require('moment-locales-webpack-plugin');

module.exports = {
  plugins: [
    // Or: To strip all locales except “en”, “es-us” and “ru”
    // (“en” is built into Moment and can’t be removed)
    new MomentLocalesPlugin({
      localesToKeep: ['es-us', 'ru'],


Lodash is a collection of JavaScript utilities.

Tested with Lodash 4.17.4

When you include Lodash, your bundle grows by 72 KB of minified code. That’s the size of all the 316 Lodash methods. If you use only, like, 20 of them, then approximately 65 KB of the code do just nothing except slowing down the page loading.

Thankfully, Lodash lets you include only the methods you need. The most basic way to do this is to import methods from the files they’re implemented in:

72 KB → 8.27 KB
- import _ from 'lodash';
- _.get();
+ import get from 'lodash/get';
+ get();

This approach might work if you’re starting a project from scratch, but it doesn’t work for existing projects. What you’re gonna do, rewrite all imports? That’s too much work. That’s why I prefer using babel-plugin-lodash and sometimes lodash-webpack-plugin.

babel-plugin-lodash is a plugin for Babel that replaces generic imports with concrete ones during compilation. That is, it does exactly the same thing as depicted in the snippet above:

72 KB → 8.27 KB
// Before babel-plugin-lodash
import _ from 'lodash';
_.get({ a: { b: 5 } }, 'a.b');


// After babel-plugin-lodash
import _get from 'lodash/get';
_get({ a: { b: 5 } }, 'a.b');

lodash-webpack-plugin is a plugin for webpack that modifies Lodash behavior by removing some code and thus cutting the bundle size. For example, _.get by default supports deep paths. If you don’t need this, you can enable lodash-webpack-plugin, which will remove this support:

72 KB → 772 B
// Before babel-plugin-lodash + lodash-webpack-plugin
import _ from 'lodash';
_.get({ a: { b: 5 } }, 'a.b');
// → returns 5


// After babel-plugin-lodash + lodash-webpack-plugin
import _get from 'lodash/get';
_get({ a: { b: 5 } }, 'a.b');
// → returns undefined

Keep in mind, however, that you can’t just enable the plugin and leave it as-is. This plugin changes the Lodash functionality, so your existing code could break. Take a look at the list of features it removes by default.

Pitfall: duplicated Lodash#

There’re two common versions of Lodash: the lodash package and the lodash-es package (which is Lodash with ES exports). If you use the former package, and one of your dependencies uses the latter, you will find yourself having two Lodashes in a single bundle. To avoid this, alias lodash to lodash-es (or vice versa).

An example of a package that uses lodash-es is Redux.

Thanks to Valentin Semirulnik for this tip.


Sometimes you have a large project where some code is compiled with webpack and some code is not. Like a page with crosswords, where the crosswords module is built with webpack, and the site around it is not:

If both pieces of code have common dependencies, you can share the dependencies between them. This is done with the webpack’s externals option which lets you alias module imports to something different.

The common usage is when you have an instance of a library in the global object (like window), and you want to alias the library imports to this instance. In this case, you pass an object mapping the module names to the variable names:

// webpack.config.js
module.exports = {
  externals: {
    'react': 'React',
    'react-dom': 'ReactDOM',

Webpack will replace all module references with variable references.

A less known approach is when the old code doesn’t put the libraries into the global object but loads them with an AMD-compatible loader. In this case, you can compile your webpack as an AMD bundle and alias modules to paths to the libraries:

// webpack.config.js
module.exports = {
  output: { libraryTarget: 'amd' },

  externals: {
    'react': { amd: '/libraries/react.min.js' },
    'react-dom': { amd: '/libraries/react-dom.min.js' },

Webpack will wrap your bundle into define() and make it depend on the libraries from externals:

// bundle.js
define(["/libraries/react.min.js", "/libraries/react-dom.min.js"], function () { … });

Then, the loader will load your libraries along with the bundle. The good thing here is that the libraries will be cached in the browser or in the loader cache – so they won’t be loaded twice.

Σ: Decrease front-end size#

  • Configure minification
  • Pass NODE_ENV=production to the code
  • Use ECMAScript imports and exports
  • Drop unused locales in Moment.js
  • Drop unused methods in Lodash
  • Use externals if you have common libraries

Task Improve caching#

Given: you have a front-end application. You want to cache it better so that the visitor loads it faster and doesn’t re-download the whole app when it’s updated.

Using hash#

The default approach of doing caching is to tell the browser cache a file for a very long time (e.g. a year), and rename the file when changing it to force browser to re-download it:

<!-- Before the change -->
<script src="./index.js?version=15">

<!-- After the change -->
<script src="./index.js?version=16">

Webpack also lets you do such thing. However, instead of versioning a file, it calculates the file hash which you can specify in the bundle name. In this case, each time you change the code, the file name will change, and the browser will re-download it:

// webpack.config.js
module.exports = {
  entry: './index.js',
  output: {
    filename: 'bundle.[chunkhash].js'
       // → bundle.8e0d62a03.js

The only remaining problem is how to get the file name to send it to the client. There are two solutions: HtmlWebpackPlugin and WebpackManifestPlugin.

HtmlWebpackPlugin is a more automated solution. During compilation, it generates an HTML file which includes all compiled resources. If your server logic is simple, then this plugin should be enough for you:

<!-- index.html -->
<!doctype html>
<!-- ... -->
<script src="bundle.8e0d62a03.js"></script>

WebpackManifestPlugin is a more flexible solution which is useful if you have a complex server part. It generates a JSON file with a mapping between file names without hash and file names with hash. You can use this JSON on your server:

  "bundle.js": "bundle.8e0d62a03.js"

Pitfall: hash could change even if the bundle is the same#

The hash could change if you rename a file or compile the bundle under a different OS. This is a bug, and I was unable to find a workaround. You can see the discussion about the bug on GitHub.

Update: the previous version of this part recommended using webpack-chunk-hash as a solution. Turns out, it doesn’t help.

Code splitting#

The next way to improve caching is to split the bundle into smaller pieces.

Imagine you have a large website, and you’re compiling it into a single bundle:

Each time you’re changing a single module, the whole bundle gets recompiled. This means that even if you’re changing the comments module, and a specific user is only visiting the main page, they’ll still have to re-download the code for this page.

If you split your bundle into several pieces – one for the main page and one for the article page – the user will only have to re-download the changed piece of code. Webpack lets you do this. In webpack terminology, these pieces of the bundle are called chunks.

To split the code into chunks, you specify several entry points and do a few other changes. Here’s the optimal webpack config:

You specify multiple entry points, and webpack generates a separate chunk for each point. Each chunk will only include the dependencies it needs
module.exports = {
  entry: {
    homepage: './index.js',
    article: './article.js'
  output: {
You replace a fixed filename with [name]. [name] will correspond to the entry point name
    filename: '[name].[chunkhash].js'
  plugins: [
You add WebpackManifestPlugin and WebpackChunkHash – plugins from the previous section
    new WebpackManifestPlugin(),
    new WebpackChunkHash(),

You add two CommonsChunkPlugins. They let you move some code from existing chunks to new commons chunks.

The first plugin moves all node_modules dependencies to a separate chunk. This allows you update the code without invalidating dependencies.

The second plugin moves webpack’s runtime to a separate chunk. This allows you to update runtime without invalidating other code. Runtime is a webpack’s system code that is responsible for loading the app

    new webpack.optimize.CommonsChunkPlugin({
      name: 'vendor',
      minChunks: m => m.context &&
    new webpack.optimize.CommonsChunkPlugin({
      name: 'runtime',
      minChunks: Infinity,
You add ashedModuleIdsPlugin. By default, each module in webpack has an ID which corresponds to its order. If you add a new module, it can affect other module ids and invalidate the cached chunks. This plugin replaces order-based IDs with hash-based ones
    new webpack.HashedModuleIdsPlugin(),

You add ChunkManifestPlugin.

By default, the webpack’s runtime contains a mapping between IDs of chunks and their names. If you configure the file name to contain the hash, as we did with the filename option, the hash will change with each file change, and so will the runtime.

ChunkManifestPlugin lets you extract this mapping into a separate JSON file. On the server, you’ll need to inline this file into the global webpackManifest variable

    new ChunkManifestPlugin({
      filename: 'chunk-manifest.json',
      manifestVariable: 'webpackManifest'

With this config, webpack will generate 6 files:

Two separate entry points. Each should be loaded on the corresponding pages
File with vendor dependencies and file with webpack runtime
Two manifest files that you’ll need on the server

And this is how often they’ll change:

  • homepage and article – when the app code in these modules changes,
  • vendor – when any dependencies of the app change,
  • runtime – when webpack’s runtime code changes (i.e. rarely and only with new webpack versions),
  • manifest.json – when you add a new chunk – but that doesn’t matter because this file is used in the server,
  • chunk-manifest.json – on any code change – but that doesn’t matter because this file is used in the server.

That’s a bit more files, but it lets you effectively leverage long-term caching.

On-demand code splitting#

The next way to improve caching (and optimize time to first paint) is to load some parts of code on demand.

Imagine you have a page with an article:

When opening this page, the visitor wants to see the content at first. Comments, sidebar and other parts of the page are less relevant to them. However, if you bundle all these blocks into a single file, the visitor will have to wait until the whole file is downloaded – with all the page modules. This isn’t cool.

Thankfully, webpack lets you optimize this by loading code on demand. You can specify that you want to load specific modules dynamically, and webpack will move them to separate chunks and download when they’re required. This is how it works:

You have an article-page.js file. When you compile it, the bundle receives all the code for articles, comments and sidebar
// article-page.js
import { renderArticle } from './components/article';
import { renderComments } from './components/comments';
import { renderSidebar } from './components/sidebar';


To load code on demand, you replace static import with dynamic import() calls. Webpack will move the code from ./comments.js and ./sidebar.js into separate chunks and load them when they’re required
// article-page.js
import { renderArticle } from './components/article';

  .then((module) => { module.renderComments(); });
  .then((module) => { module.renderSidebar(); });

This change will improve the initial loading performance. Also, it will optimize caching because when you change the code that belongs to a specific chunk, other chunks won’t get affected.

The only thing left is to add chunk hashes to their names. This is done with output.chunkFilename option. This option is specific to chunks generated by on-demand code splitting:

// webpack.config.js
module.exports = {
  output: {
    filename: '[name].[chunkhash].js',
    chunkFilename: '[name].[chunkhash].js',

Pitfall: Compiling with Babel#

If you compile this code with Babel with default presets, you’ll have a syntax error: Babel don’t understand import() out of the box. To prevent the error, add the syntax-dynamic-import plugin.

Other solutions#

There are a couple of other solutions that I haven’t worked with but which should also bring benefits with caching:

  • AggressiveSplittingPlugin is a plugin that optimizes your code for HTTP/2 by splitting each chunk into smaller chunks as much as possible. This greatly improves caching on the client side but slightly worsens the gzip compression. See the example in the webpack repository.
  • OfflinePlugin is a plugin that’s usually used for creating offline-ready apps. However, you can use it to improve caching too! The plugin generates a service worker that downloads all the site resources in the background. So when a visitor visits the site and then switches to a different page, they’ll have all the necessary files already cached. See the OfflinePlugin docs.

Σ: Improve caching#

  • Add hash to the name of your resources and make the server tell clients to cache the resources for a long time
  • Split your code into smaller chunks with different entries, on-demand code splitting and AggressiveSplittingPlugin
  • Try caching your resources in the background with OfflinePlugin

The next part of the guide, “Speeding up build and improving the development workflow”, is coming soon. Leave your email to know when it’s out:
(you’ll receive an email about the next part of the guide + a couple of more webpack-related posts if I write them; no spam)