Avoiding Babel’s Production Bloat

I once wrote , a 6 chars operator followed by a 3 chars reference, to realize that translated into 424 uncompressed bytes:

"use strict";function _typeof(obj) { if (typeof Symbol === "function" && typeof Symbol.iterator === "symbol") { _typeof = function _typeof(obj) { return typeof obj; }; } else { _typeof = function _typeof(obj) { return obj && typeof Symbol === "function" && obj.constructor === Symbol && obj !== Symbol.prototype ? "symbol" : typeof obj; }; } return _typeof(obj); }typeof any === "undefined" ? "undefined" : _typeof(any);

By no mean this post goal is to discourage you from using the mighty Babel 7 though, which is by far the best transpiler out there. One would be crazy today to trust anything else, including TypeScript, which I start calling TypeSeppuku, and not for its files, just for its resulting code.

This post is about the sometime inevitable bloat transpilers bring in, explaining the why, the what, and the how to avoid such bloat, when possible.

The Three Kinds Of Bloat

To simplify the topic, let’s try to understand what kind of bloat we have:

  • the real syntax sugar bloat, which has the smallest footprint in terms of bloated code, providing simple, safe, and fast fallbacks for older engines
  • the recycled fix bloat, which has the kind of bloat, the first time it’s needed, but it’ll be recycled for every other used
  • the inline syntax polyfill bloat, which is the worst of all kind, ’cause it cannot be optimized in any way and it will bloat your code each time

Every part of JS can be more or less transpiled with one of these kinds, either as single case, or mixed up cases, where recycled polyfills are also an option.

Shall we start with the scariest kind, so that we can have some sort of mild happy ending after all? 😅

The Syntax Polyfill Hell

You can check it live directly, or simply trust me on this, the for/of operator is a curse for any project that needs to transpile it:

for (const value of list);

Accordingly with Closure Compiler Service, the result is 2.18KB of code for doing basically “nothing”.

And the worst part of it, is that the more you copy and paste that loop, the more the bloat grows.

You cannot even blame Babel for this, because is one of those parts of recent specifications that simplifies gazillion tasks so that Babel, if unaware of the kind of is dealing with, has to try all the options: is it an Array? is it Iterable? is it a generator?

Not only this could be true even for engines that support natively , but if I were you, whenever you need to loop over an array, arguments object, static list, or something with an index and a finite , I strongly suggest you to don’t use in the wild.

About the …rest

Another funny, inevitable, bloated transformation is parameters:

function rest(...args) {
return args;

Becoming minumum twice their size every time these are used. But not only above function is basically the equivalent of just:

function rest() { return arguments; }

The performance penalty you pay looping every time over can be relevant, when you invoke functions gazillion times per second, and each function is transpiled as such:

function rest() {
for (var
_len = arguments.length,
args = new Array(_len),
_key = 0;
_key < _len; _key++
) {
args[_key] = arguments[_key];
return args;

Now, maybe we don’t all know that is a special object that might leak into memory, which is the reason transpilers might do this array transformation dance, but any linter rule that forces avoiding arguments, in favor of , is actually a dumb rule if the code will get transpiled.

Generating* Bloat

Generators are another inline polyfill fix, mixed up with some recycling, but basically the following:

function* blow() {}

becomes some unusable code that will make sense only if applied properly.

Add a and a , and be surprised by other 2.18 extra KBs!

Destructuring Arrays

Another lovely, and inevitable, code bloat is generated by array destructuring:

const [a, b] = unknown;

That alone might scare you, but at least here there’s some recycled part of the transpiler that most likely will be reused elsewhere.

Indeed, adding something like after, won’t blow up the final code size any further, it will just reuse the same helpers.

However, there is a better destructuring than array, as example using an object as single parameter:

function ab({a = 1, b = 2}) { return [a, b]; }

or abusing the object:

function ab() {
const {0: a = 1, 1: b = 2} = arguments;
return [a, b];

where both approaches will produce way smaller outcome than:

function ab(a = 1, b = 2) { return [a, b]; }

which is another little monster in terms of bloated size.

An Absurd Class

I cannot be the only one that consciously prefer the good old function way to declare small classes, right?

function Small() {}
// or
class Small {}
// ?

Well, I let you judge the difference, where using results into 298 uncompressed bytes through the closure compiler, something actually irrelevant if we consider 1.49K needed to represent a single .

// 1.49K (almost 10x than ...)
class B extends class A {} {}
// 156 bytes uncompressed
function extend(a, b) {
const sPO = Object.setPrototypeOf;
if (typeof a === 'function')
sPO(a.prototype, b.prototype);
return sPO(a, b);
function A() {}
const B = extend(function B() {}, A);

Add recent private fields in the mix, and reach 1.75KB with ease:

class Top {
#pvt1 = 123;
class Sub extends Top {
#pvt2 = 123;

where will initialize each instance like …

var Sub = function (_Top) {
// extend
_inherits(Sub, _Top);
// the class
function Sub() {
var _getPrototypeOf2;
var _this;
// with instance check
_classCallCheck(this, Sub);
// possible arguments (avoiding leaks)
for (var _len = arguments.length,
args = new Array(_len) ... ) {
args[_key] = arguments[_key];
// possible different return from the super
_this = _possibleConstructorReturn(this, ...);
// private field
_assertThisInitialized(_this)), {
writable: true,
value: 123
// the current context
return _this;
// the current class
return Sub;
// and its `pvt` private field
var _pvt = new WeakMap();

Now imagine if it was you writing such logic per each class, instead of simply defining a within a module and its private for some private reference, as in the following case:

const _ = new WeakMap;function GoodOldES3() {'use strict';
_.set(this, {any: 'value'});
GoodOldES3.prototype.getAny = function getAny() {
return _.get(this).any;
  • the runtime check for is implied by the guard
  • the code is already compatible with IE11 ’cause IE11 supports WeakMap
  • there is absolutely no extra cost in initializing a instance
  • no code outside that module can reach private fields
  • zero production bloat 🎉

… the instanceof curse anyway …

Incapable of knowing upfront if an instance comes from a class that used some modern alchemy to trick the engine, would be another keyword to avoid when possible, ’cause it’ll bring in:

function _instanceof(left, right) {
if (
right != null &&
typeof Symbol !== "undefined" &&
) {
return right[Symbol.hasInstance](left);
} else {
return left instanceof right;

which is yet another recycled bloat, but one that could slow down one of the fastest checks of the JS world, second only to .

OK Then … What Is Safe ?

I’ve sneaked in already most common recycled bloat in previous examples, but it’s time to see what doesn’t really affect the final production code size.

Arrows and Methods are just fine

const o = {
// shortcut
method() {},
// arrows without this / arguments
contextless: () => {}
// even with this / arguments
const arrow = () => [this, arguments];

and so are template literals

These are rightly translated to follow latest ECMAScript standard, and even if bigger than their native counterpart, and somehow slower to interact with, these are just fine ’cause these play well with minifiers and compressors.

function same() {

Be aware that these are not safe in TypeScript if you target IE11, or old environments, and these are different in Babel 6 too.

… but I said already that Babel 7 is the only transpiler you should use, right?

destructuring objects? all good!

const {a, b} = object;

which translates in a super sweet way.

Not arrays though !

Beside bringing a lot of boilerplate for something this short:

const [a, b] = array;
const a = [...b];

specially the spreading one makes no sense when it comes to methods:

// 520 bytes uncompressed
// VS 38 bytes uncompressed
arr.push.apply(arr, values);

I don’t think we always need to be that lazy, right?

How To Avoid The Bloat

There are various ways to accomplish this, and the most obvious one is: do not target ES5 or IE11 unless it’s strictly necessary.

The “how to do that?” part has been already extensively covered in my previous “A Bloatless Web” post, where all it takes is just to adjust your targets or produce multiple bundles and distribute only what’s necessary.

It’s. Deadly. Simple.

However, there are other ways to avoid some bloat ending up in your production code, and here just few of the various solutions.

Exclude Specific Helpers

If you are using , as example, and your are not needing, or using it, to ever deal with , another hard to polyfill and trust variable in the modern specifications, you can exclude that right away.

["@babel/preset-env", {
"exclude": ["transform-typeof-symbol"]

The is probably not needed in Babel 7, but please double check there’s no undesired transformation, usually already the case if the class is already known.

Brute Force Exclusion

If you deal with third parts modules that might have already been transpiled, you can also use something like drop-babel-typeof utility.

Just install it as dev-dependency, and run it as production hook.

… and what about classes ?

Well, there have been alternatives to modern classes declaration for decades now, but the tricky part is when you need to extend builtins.

In that case, if you are targeting ES5, or IE11, your code will be broken, end of the story; you can already start targeting ES6/2015 ’cause until now you never had IE11 users or you’d be drawn by filed bugs and issues.

However, if you are shipping to third parts, and you are worried about class bloat, I can still suggest my good old Classtrophobic work around, and its ES5 compatible counter part.

If you are curious to know more about Classtrophobic, here my previous post.

Finally: Use Common Sense!

New syntax is cool, but if it gets transpiled, users, browsers, and ultimately you, won’t ever benefit from any new feature, slowing down instead, and for no reason, also increasing production code size.

Whenever you can choose a specific pattern, and you are still targeting legacy browsers, choose wisely if it makes sense to use modern syntax penalizing all modern browsers, and slowing down the already, legacy, slow one.

Last, but not least, keep an eye on your transpiled code before it gets minified, and try to figure out if there’s some unnecessary bloat brought in by feaatures you are not even using in your production code.

You now know how to work around, exclude, or deal with it 👋

Web, Mobile, IoT, and all JS things since 00's. Formerly JS engineer at @nokia, @facebook, @twitter.

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store