paint-brush
7 Different Ways to Use ES Modules Today!by@transitive-bullshit
17,895 reads
17,895 reads

7 Different Ways to Use ES Modules Today!

by Travis FischerJune 4th, 2018
Read on Terminal Reader
Read this story w/o Javascript
tldt arrow

Too Long; Didn't Read

ES Modules are the future of JavaScript, but unlike many other es@next features that developers have raced to take advantage of mainly thanks to build tools like <a href="https://babeljs.io/" target="_blank">babel</a>, working with ES modules alongside of existing NPM modules is a lot harder to get started with.

People Mentioned

Mention Thumbnail

Companies Mentioned

Mention Thumbnail
Mention Thumbnail
featured image - 7 Different Ways to Use ES Modules Today!
Travis Fischer HackerNoon profile picture

ES Modules are the future of JavaScript, but unlike many other es@next features that developers have raced to take advantage of mainly thanks to build tools like babel, working with ES modules alongside of existing NPM modules is a lot harder to get started with.

The purpose of this tutorial is to provide a thorough set of examples for different ways you can approach writing ES modules, without losing interop with the vast library of mostly commonjs modules that exist on NPM today.

Before getting started, it’s important that you have a solid understanding of the differences between the ES module and CommonJS formats. If you haven’t already, please check out Gil Tayar’s excellent blog post before continuing.


Native ES Modules in NodeJS: Status And Future Directions, Part I_ES Modules are coming to NodeJS. They’re actually already here, behind a --experimental-modules flag. I recently gave a…_medium.com

Goals

Every approach in this tutorial must satisfy the following requirements:

  • Generate a valid npm module
  • Support the same consistent functionality
  • Be usable from both Node.js and browser environments
  • Import at least one existing commonjs module from NPM
  • Import at least one es module source file locally
  • Include at least one unit test

Functionality

The functionality of our example NPM module is a bit contrived, but it should touch on all the major pain points, and trust me, there are a lot of them…

Every approach will define an NPM module with a single default export, async getImageDimensions(input), that takes in an image and returns its { width, height }.

To show how you can bundle modules with slightly different semantics for Node.js and the browser:

  • The node version supports input as a string that can either be a local path, http url, or data url.
  • The browser version supports input as a string URL or an HTMLImageElement.

Both versions return a Promise for { width: number, height: number }.

Approaches

We’ll start with a naive ES module and work our way through a series of increasingly complex example approaches, all of which are intended to define the same, basic module. You can follow along with the source code for each of the 7 approaches on GitHub here!

1. Naive Approach

Our first approach is the most naive possible use of ES modules supporting our functionality. This approach is broken and provided only as an example starting point. Here is core of the Node.js code:

Approach #1 index.js

Approach #1 load-image.js

The functionality is meant to be pretty straightforward, but it’s important to understand because all the following examples will be using the exact same code. Here is the corresponding browser code:

Approach #1 browser.js

Approach #1 browser-load-image.js

In this approach, we use normal .js file extensions for es modules and no transpilation.

It is relatively simple and straightforward but breaks several of our module goals:

  • Not usable from node.js because its main field is an es module whereas it should be a commonjs file.
  • Not usable from the browser via webpack or rollup because its browser field is an es module whereas it should be a commonjs file.
  • The only advantage of this approach is its simplicity, and this may be good enough if you are just working on private modules.
  • Warning: unless you are using strictly local or private modules, we strongly encourage you not to use this approach in practice. It is meant as an example of what not to do when transitioning from commonjs to ES modules, and if you publish a module publicly using this approach, the JavaScript Gods will find and shame you.
  • Unfortunately AFAIK, there is really nothing built into the npm ecosystem which prevents you from publishing broken modules like this, although as ES modules gain popularity over the coming years, I believe this will be addressed.

2. Pure Babel Approach

This approach uses babel with babel-preset-env to transpile all Node.js and browser source files into dist/.

The core of this approach is the build script in package.json:

Approach #2 excerpt from package.json

  • Source files end in .mjs
  • Relatively simple to setup
  • Babel transpiles all source files to ES5 and commonjs
  • Tests are run on the transpiled source, which make debugging slightly harder
  • Currently, our main and browser are commonjs exports that support node >= 4 (or whatever we specify in our babel-preset-env config), whereas the module export is an es module that supports node >=8 due to its usage of async await.
  • Unfortunately AFAIK, package.json engines doesn't support specifying that main supports a certain node version whereas module supports a different module version, and I'd go so far as to say this is a bad practice.
  • To get around this, we could specify the minimum node version to be node >= 8 like we did here or add a second babel step which transpiles the node version to an esm folder, although I find this somewhat clunky.

3. ESM+Rollup Approach

This approach uses esm for Node.js and babel+rollup to compile browser source files.

esm is amazing!

Approach #3 main.js node commonjs entrypoint which loads the esm module.mjs via esm.

Approach #3 ollup.config.js for compiling the browser version to bundled commonjs.

  • Source files end in .mjs (only exception is the commonjs entrypoint main.js)
  • Supports all three targets main module, and browser
  • The only target that is compiled is browser, making debugging the Node.js version easier
  • Also note that we’re requiring esm in the ava unit test config

4. ESM+Webpack Approach

This approach uses esm for Node.js and babel+webpack to compile browser source files. It’s the same as the previous approach, except it switches out rollup for webpack.

Approach #4 webpack.config.js for compiling the browser version to bundled commonjs.

  • Source files end in .mjs (only exception is the commonjs entrypoint main.js)
  • Supports all three targets main module, and browser
  • The only target that is compiled is browser, making debugging the Node.js version easier
  • Also note that we’re requiring esm in the ava unit test config

5. Pure Rollup Approach

This approach uses babel+rollup to compile all Node.js and browser source files. This approach takes the rollup config from approach #3, and takes it one step further by having separate rollup configs for the browser and Node.js targets.

Instead of including the redundant configs, check out the source directly here.

  • Source files end in .mjs
  • Supports all three targets main module, and browser
  • All three targets are compiled via rollup, with Node.js and the browser having two separate configs
  • This is our first module to support node >= 4 (or whatever we specify in our babel-preset-env config) instead of node >= 8
  • Source maps are generated along with the compiled targets

6. Pure Webpack Approach

This approach uses babel+webpack to compile all Node.js and browser source files. This approach takes the webpack config from approach #4, and takes it one step further by having separate webpack configs for the browser and Node.js targets.

WARNING: this approach is currently a broken WIP, and its exports do not behave correctly. All other approaches work as intended.

Instead of including the redundant configs, check out the source directly here.

  • Source files end in .mjs
  • Supports all three targets main, module, and browser
  • Unfortunately, webpack does not support outputting ES module targets (issue).
  • The main and browser targets are compiled, but the module target is unable to be compiled due to this issue.
  • Supports node >= 8, whereas the rollup version is capable of supporting node >= 4 by compiling the module target as well.
  • Unless you have a good project-specific reason to use webpack over rollup, I would strongly recommend using rollup to bundle ES6 module libraries (at least until this webpack issue is addressed).

7. TypeScript Approach

This approach uses typescript to transpile all Node.js and browser source files.

TypeScript Approach #7 index.ts

TypeScript Approach #7 load-image.ts

  • Source files end in .ts
  • Supports all three targets main module, and browser
  • Two compilation passes are necessary, one for targeting commonjs and one targeting esm
  • Currently, commonjs users have to add default to require statements: require('npm-es-modules-7-typescript').default. If you know how to prevent this, please let me know.
  • Resulting module supports node >= 4 (or whatever we specify in our tsconfig.json) instead of node >= 8
  • Note that I’m fairly new to TypeScript, so please let me know if I’m doing anything awkward.

Overall, using TypeScript feels like the cleanest and most future-proof approach if you want to maximize compatibility.

Recommendations

Whew, that was a lot of JavaScripts…

Now to summarize what I’ve learned from creating this breakdown and my practical suggestions for writing your own next-gen NPM modules:

  • If you only care about Node.js compatibility and not browser usage and want something extremely straightforward, then use either use standard commonjs for now or use esm as in Approach #3 without the rollup stuff.
  • If you only care about Node.js compatibility and not browser usage but also want to support older versions of Node.js, then use the Pure Babel Approach #2.
  • If you only care about browser usage, you can likely get by writing ES modules without complicating things with transpilation, because any modern frontend toolchain will take care of building things for you.
  • If you care about both Node.js and browser compatibility, then use either the ESM+Rollup Approach #3 or the ESM+Webpack Approach #4, depending on your preference between rollup and webpack. IMHO, Rollup is a strictly better choice for libraries over webpack because it has a more narrow focus on handling libraries whereas webpack tries to do everything.
  • If you care about both Node.js and browser compatibility and want to support older versions of Node.js, then use the Pure Rollup Approach #5.
  • Finally, if you want to be fully future-proof and increase the quality of your code via static analysis, I’d highly suggest using the TypeScript Approach #7, which handles all the previous use cases and then some.

Conclusion

I hope you’ve found this guide helpful! You can find all the source code, including usable modules for the 7 approaches on GitHub here.

For more info on this topic, check out these great resources:

Have any approaches or suggestions that I left out? Let me know by sharing them in the comments! ❤️

Before you go…

If you liked this article, click the 👏 below, and share it with others so they can enjoy it as well.