This week Sam Magura, an active maintainer of Emotion, a widely-popular CSS-in-JS library for React, published an article about breaking up with CSS-in-JS. As a “pure CSS” guy, I had many questions and comments while reading the article, so I had to document them in this blog post.
It might be worth noting that I never understood why people use React for websites.
Long story short, Sam emphasizes the problems of the CSS-in-JS approach:
- CSS-in-JS adds runtime overhead,
- CSS-in-JS increases your bundle size,
- CSS-in-JS clutters the React DevTools,
- Frequently inserting CSS rules forces the browser to do a lot of extra work, and
- With CSS-in-JS, there’s a lot more that can go wrong, especially when using SSR and/or component libraries.
In short, the CSS-in-JS performance impact of Emotion was significant, so the Emotion team replaced it with Sass Modules and Bootstrap utility classes.
The the first problem
The second problem
“When writing plain CSS, it’s very easy to accidentally apply styles more widely than you intended.”
Locally-scoped styles sound so ridiculous to me. Accidentally applying styles happens when there is no agreement on the project’s naming convention, structure, and architecture.
Imagine someone who only tried Create React App is responsible for writing scalable and maintainable React apps. It doesn’t work that way, does it? The industry should stop ignoring CSS, start educating everyone and acknowledge the CSS developer. It would prevent these kinds of problems in the first place.
The third problem
“...imagine you’re making a list view where each row should have some padding and a border. (...) Several months later when you’ve completely forgotten about the list view, you create another component that has rows. (...) Now the new component’s rows have an unsightly border and you have no idea why!”
Sam uses this example of class
.row to explain the CSS class name conflicts. Any developer experienced in writing CSS will immediately inspect the element and find out what is happening. But even that is not a solution. You should have naming conventions, structure, design systems, and styleguide to prevent these problems.
For example, if you’re using Sass and BEM, you can prevent these problems by splitting your CSS code into multiple files and naming them meaningfully. Every Sass file name should have the component’s name, and every CSS class name should start with the component’s name. No more CSS class name conflicts.
The fourth problem
“If using plain CSS, you might put all of your
.cssfiles in a
src/stylesdirectory, while all of your React components live in
src/components. As the size of the application grows, it quickly becomes difficult to tell which styles are used by each component. (...) A better approach for organizing your code is to include everything related to a single component in same place.”
According to this comment, it is better to make a mess in a single file than to have organized files in separate folders. If developers have issues with finding files, then document it. Make your developers learn and adapt to the project architecture.
The fifth problem
The sixth problem
I’ve been in the web industry for more than ten years. Trends come and go, so stick to the logical stuff. If it smells funny, it will probably become a problem.
I am interested in how you would migrate some CSS-in-JS projects. Is there an easy way to extract all components and make them work in a different environment? I am talking about styling, not functionality. I don’t think there is an easy way to do it.
The seventh problem
“When your components render, the CSS-in-JS library must "serialize" your styles into plain CSS that can be inserted into the document.”
This sentence is the essence of the problem. The end result is plain CSS. The main “benefit” and justification for CSS-in-JS are locally-scoped styles. What a bunch of bull! That is what you get when developers are trying too hard. They overcomplicate the solution instead of using well-adopted and well-tested technologies.
The eighth problem
Not only that, every user loads the component’s CSS code whenever that component appears. And class names usually contain some gibberish class suffix that adds a few bytes, too. Don’t get me started about the caching.
The ninth problem
“Frequently inserting CSS rules forces the browser to do a lot of extra work.”
When I was less experienced, I remember worrying about my CSS performance. But then, I learned that CSS performance is rarely an issue. I don’t know when was the last time when I asked myself about the CSS performance. I think Milica Mihajlija said perfectly: “Don’t worry about the speed of CSS selectors”.
On the other hand, using CSS-in-JS might cause performance issues. So any discussion about it should stop here.
The tenth problem
“To improve the DX around this, we decided to bring in a utility class system. (...) I had already been using Bootstrap for years, so we went with Bootstrap.”
I guess we could expect more Bootstrap and TailwindCSS users. I just wish we had fewer Bootstrap and TailwindCSS users, not more.
There are probably more reasons why I never understood CSS-in-JS, but these are the main ones.
Learn CSS, make your developers learn CSS, or hire CSS and UI developers to do the styling part of the project.
Latest comments (3)
Сongratulations 🥳! Your article hit the top posts for the week - dev.to/fruntend/top-10-posts-for-f...
Keep it up 👍
my 50 cent, after enjoying your article. i do like the js version like styled components in react. for me it was clear by binding the related styled to the component and have them present in the project structure. i also really enjoyed the capability to add values to transforms or transitions on runtime.
i do agree on the some of you points, but honestly, isnt webpack capable to bundle the styles accrodingly and move them correct, so your "overhead" argument among others might not be super valid.
just my toughts. keep on the good article
CSS-in-JS was one of my greatest annoyances when working on a ReactJS project (although not the only one, and I am still not a big React fan). I personally agree with your conclusion,
I hope that utility CSS will be just one of many options, and not the worst either. At least this means using class names and CSS code which can be validated and improved using static analysis tools like stylelint.
So, thanks for your post, and let's stay positive and lead by example.