css-modules VS @artsy/fresnel

One aspect of CSS modules that I truly appreciate is its ability to compress class names into very short hashes. This feature allows me to keep my CSS selectors as long and descriptive as needed, while still compressing them into concise three or four character hashes. It aligns with my rule for CSS: selectors should be written for human readability, but compressed for machine efficiency.

Use methodologies such as BEM, and technologies including CSS modules, CSS-in-JS, and Shadow DOM to isolate the styles of each micro-application and prevent conflicts, thus ensuring reliable encapsulation and modularity.

For many years, Culture Amp took the second option, and distributed shared components without compiled CSS. This meant that every app that consumed shared components needed to include the necessary CSS build tooling – at that time CSS Modules and node-sass – with a compatible version and configuration. This was relatively easy to set up, but over time proved difficult to maintain. When node-sass was deprecated in favour of (the much faster but slightly incompatible) Dart Sass, this demanded a difficult lock-step migration across all those codebases, which we have yet to achieve. And as new applications have switched to Tailwind for their own styles, they've had to continue to maintain those old build tools in parallel for the shared components' styles.

So CSS modules are a form of 3rd-party CSS-in-JS, where what you import are the class names, which are then usually obfuscated etc at compile time, and all the actual style declarations are (usually) compiled into a single css file or tag as part of the bundling process. You can read the og docs on'em here, and you've probably seen'em used in React like:

import styles from "./styles.css";

function Example(){
    return (
        

            Hello
        
    );
}

They predate the ability to import non-js files in vanilla by a good while, and rely on the compile process to translate your .css files into .js files that can be imported using whichever loader you use in your bundler.

Import assertions are a vanilla way to import non-js files by telling the browser how to import them; assert { type: "css" } says to treat the file as CSS and create a CSSStyleSheet, assert { type: "json" } says to treat the file as JSON and create a JSON object - and hopefully assert { type: "html" } will hopefully arrive soon and create a #document-fragment or something similar.

Hope that clears it up!

Extensions of CSS: for example, Sass, Less, Tailwind, CSS Modules, to make stuff look a certain way on your own.

The components are styled with CSS modules. When building the library, these styles will get transformed to normal CSS style sheets. This means that the consuming application will not even be required to support CSS modules. (In the future I want to extend this tutorial to use vanilla-extract instead.)

NextJS comes with built-in support for CSS Modules which allows you to scope your styles locally in individual components without worrying about name collisions or messing up other parts of the codebase.

CSS modules are not to be confused with mixins, as they serve the opposite purpose. While mixins are components or functions to be reused globally, modules are style sheets with a local scope used in a similar way as styled components in React.

CSS Modules CSS Modules is a pre-processing step: by default, styles are scoped locally to the current component, and the transpiler ensures no conflicts.

This! If you want to limit double rendering afterwards you could look into something like fresnel to automatically unmount unused breakpoints’ components after hydration.

the SSR default I'm guessing is the fancy one, it feels like due to SSR the most flexible way is to use media queries so it is correct on the first paint. maybe components would be better than hooks, something like https://github.com/artsy/fresnel

This article is extracted from this github discussion, which goes into greater detail on how to achieve this with the libraries Dripsy (a component design system) and Fresnel (SSR w/ media queries).

Also, take a look at a library @artsy/fresnel, which uses a similar approach to solve CLS issues in server-side rendered applications.

If you want both SSR and RNW, then you are not so lucky. But you can read my research of the options you have, in this discussion: How to combine React Native Web (RNW) + Responsive styles (media queries) + NextJS Server-Side Rendering (SSR), to get SEO on the web. If you don't want to make a compromise like the current options, there is a library called Fresnel, which has some promise to afford this combination. But it is not quite there yet. So please consider contributing to Fresnel in the RNW+NextJS compatibility issue or the useMedia() hook issue, so that Dripsy could be upgraded to use Fresnel again to support SSR.

Something I starred a while ago: https://github.com/artsy/fresnel

Uses actual CSS breakpoints on web, to become responsive, even when using SSR. Uses Fresnel to achieve this. CSS breakpoints on web will rely on the mobile device's user agent for most cases. But since this can be inaccurate, it also uses pre-emptive server-side breakpoint generation. Basically, if the mobile user-agent cannot be accurately detected, Fresnel renders all media query breakpoints on the server (a bit of extra work for the server, since it might lead to rendering more components). So the client's browser will receive all breakpoints on first render and can immediately start rendering according to the correct ones. Instead of waiting for React to rehydrate before only then running the media queries with CSS-in-JS, which would have given latency and potentially unwanted visual side-effects. This becomes most acute when you are using SSR, and thus don't want to wait for rehydration on the client to start showing the responsively laid out content.

@artsy/fresnel - Lib to create media queries components in JSX.