In the styled-components instance, our class name gives us no identifying information as to the state of the button, while reading our BEM-notated class names gives us an insight into what our intended styling is. With a simple package such as classnames, you can easily coordinate the programmatic control of class names. Taking it one step further, you can simply roll your own to further reduce overhead.
What does styled-components aim to solve?
On their website, styled-components provides a bulleted list of problems they aim to solve. I’d like to break down and address each item.
- Automatic critical CSS: styled-components keeps track of which components are rendered on a page and injects their styles and nothing else, fully automatically. Combined with code splitting, this means your users load the least amount of code necessary.
- Code splitting is not specific to styled-components, it’s a feature provided by webpack. If you’re working with React, you’re working with webpack and can enable this feature regardless of how you write CSS.
- Only loading styling required for rendered components also is not unique to styled-components. Assuming your teams are adhering to best practices (more on their implementation later) you’re scoping CSS to a given component and have no component-specific styling which is applied from a global scope. To capture your component-specific CSS, simply leverage style-loader, an npm package with nearly 10 million weekly downloads. Style-loader allows you to simply import your CSS in your React component. Style-loader can be further extended in your webpack configuration to allow for SCSS precompilation, the programmatic application of vendor prefixes, and any other piped functionality.
- No class name bugs: styled-components generates unique class names for your styles. You never have to worry about duplication, overlap or misspellings.
- The concern of unique class names or scoped CSS is simply mitigated in React by leveraging a displayName attribute and scoping everything in your component under this. You can enforce this practice with ESLint, stylelint (CSS linter), and BEM notation.
- The misspelling of class names is certainly a real concern. Ideally, you’re not writing code and pushing it to production, so you’re most likely catching these errors during testing. That said, there is certainly some increase in developer effort from having no autocompletion support for class names in your text editor. I’d argue that CSS Modules is a better solution to solve this though.
- Easier deletion of CSS: it can be hard to know whether a class name is used somewhere in your codebase. styled-components makes it obvious, as every bit of styling is tied to a specific component. If the component is unused (which tooling can detect) and gets deleted, all its styles get deleted with it.
- How much code are you writing where you lose track of the CSS for a given component? If you’re writing so much CSS for a component, it probably means your component is too complicated.
- How does styled-components make it easy to identify code which can be deleted? You check what components you’ve imported, right? You’d just search for these with regular CSS. In my experience this would be a marginal gain.
- In your code base, I’d suspect to see three sections of CSS. The first, and most global would be whatever your equivalent of a CSS reset is. In this file, you’d be applying CSS targeted at native elements. I wouldn’t expect to find any class names or ID selectors here. The second file would be global helper classes. These classes would provide useful classes to control element positioning, structure, or a few other restricted attributes. This file would clearly be identified as a global file and it would be implied that modification of any of these helper classes would have broad-sweeping consequences. Finally, I would suspect to see a file with component-scoped CSS. This would contain component-specific styling and would be scoped to the displayName of the component. If you find yourself needing CSS in a React view, you’ve probably done something wrong. That’s a topic for another discussion though.
- Simple dynamic styling: adapting the styling of a component based on its props or a global theme is simple and intuitive without having to manually manage dozens of classes.
- Is manually adding classes actually a negative? I find significant benefits in seeing actual class names on components when debugging in developer tools. Following BEM notation, I can easily see what I’m trying to achieve at a given point. For example, why is a button red? With styled-components you’d have to go back and look at the logic. With SCSS following BEM notation, I might see something like .button__primary–danger which is indicative of such a scenario. Linked is an example of a button with many conditional classes. This has worked fine in many large-scale production environments for me. I believe that seeing all of the presentational logic in one neat object helps determine the intended goal of each class.
- Following a programmatic class name approach, you’re still applying styling based off of passed props and I suspect the effort is comparable – it’s just a matter of where. Following styled-components, you’d apply a bit of conditional logic such as a ternary operator driven off of a prop to control an attribute. Following a SCSS/BEM approach, you might use a util to programmatically control the application of classes based on props, such as with classnames.
- Painless maintenance: you never have to hunt across different files to find the styling affecting your component, so maintenance is a piece of cake no matter how big your codebase is.
- Any enterprise-level code base will have practices in place which mitigate this, even with vanilla CSS. Styled-components can extend classes anyways which spreads CSS across files causing similar issues.
- If you’re using scoped SCSS (enforced through linting), your CSS will anyways be scoped. Additionally, with React we typically only write CSS for a component.
- Automatic vendor prefixing: write your CSS to the current standard and let styled-components handle the rest.
- We can do this automatically with SCSS through webpack. Furthermore, we have more granularity when configuring this through webpack. Bundled or embedded tools work well for small projects, but typically fail to scale.
At what sacrifice?
I appreciate self-documenting code and as you’ve probably seen by now, I’m a big fan of linting everything. Leveraging styled-components means there’s no easy way to lint your CSS. Perhaps you’ve never even thought to lint your CSS, but there’s a world of value to be found there. You can enforce everything from the sorting order of attributes to the restriction of units or adherence to brand colors. This guiding structure provided by linting CSS is essential in the mitigation of technical debt when working with large teams in large code bases.
I’d like to compare the readability two approaches:
JSX + BEM + SCSS
While completely opinionated (what things aren’t opinionated in coding?) I’d argue that while the styled-components approach is much smaller, it can become much more difficult to read. Taking a JSX+BEM+SCSS approach groups things into nice little buckets. You’ll find all of the programmatic class names for your component in one easy spot and you can easily review what logic controls a given class name. The same goes for the CSS itself. We’re following BEM-structured SCSS which allows us to see our nice little blocks of attributes for each conditional class. There’s no time spent searching for a given attribute from a long list which may be sorted randomly.
Yes, my styled-components example is a worst-case scenario. You can drastically improve its readability by taking several steps, but how do you enforce this? To my knowledge, there aren’t any linting rules which prevent developers from using styled-components like in the example above. You’re relying on code reviews for manual intervention. This is not a sustainable practice. People make mistakes and may not catch everything. This also opens it up to developer interpretation and arguments. Linting rules aim to mitigate opinionated reviews by capturing consensual opinions in written rules. Relying on developers to enforce best practices will lead to divisions and time wasted on arguing what’s “best”.
Off the rails
Having read my previous sections, I’d suspect you’ve managed to infer my viewpoint on styled-components. To be clear, I believe styled-components does provide benefits through a low cost of entry, but I do not believe it is necessary if you have experienced developers to help establish the best practices required to sustain your code base. React does not care if you do a poor job of structuring your code base. It offers no help or guidance. This gives great flexibility, but also provides a situation for things to quickly jump off the rails. If you’re a startup or building out a product with less experienced developers and simply need to get things off the ground, that’s where styled-components shines. It’s a simple module which is easily dropped in, propagating out its rules throughout your code base. Styled-components gives inexperienced teams the ability to move fast without incurring loads of technical debt. It teaches best practices and ensures a quality product. As with anything in life, you’ll soon find that nothing in life comes for free. It’s important to understand what styled-components aims to solve.
Which should I use?
- You find yourself needing to get a project off the ground with limited time and resources
- Your development team is less experienced and may have trouble implementing a robust approach through SCSS/BEM coupled with the necessary linting rules
- You are developing a small-to-medium offering which will
- You’ll have a small number of co-located resources working in the code base
JSX + BEM + SCSS:
- You’re looking to build out a large-scale product
- You have direct or indirect access to skilled developers capable of placing the required infrastructure
- You’re looking to control with high granularity every step in the build process
- You’ll have a large number of resources working in the code base, potentially working in a distributed environment where highly consistent CSS is critical to developer efficiency and reduction of technical debt
To help us assist you, check out Infostretch’s UX Design Services or get in touch directly using the form below.