Build vs Buy: Component Libraries edition

February 4th, 2021 — 6 min read

by George Bakos
by George Bakos
No translations available.Add translation

When you started your last software project, did you write your own programming language first? Did you build your computer from scratch? Of course not. That would be ludicrous. It would be a waste of time (and in sustainable business that means money). It's not your core competency. It's not your value proposition. You'd run out of money before you finished.

When you started working on that user preferences page of your app (or whatever the latest feature you implemented was), did you just grab something off the shelf? It's very possible your database of choice has a pre-built UI interface for simple CRUD actions. Why didn't you just use that? Probably because your customers expect a professional, modern, and consistent design and experience throughout the app and that pre-built UI couldn't do that for your customers.

Those are examples of the extremes of build vs buy in the software business. There's gonna be a middle-ground between those two lines of thinking that's optimal for the success of your project (and your career/the business).

The open source world is filled with amazing libraries and frameworks that can help us get our software built and delivered to users quickly. Libraries like ReachUI, MaterialUI, and even some company-built component libraries like Adobe's React Spectrum are mature and battle-tested (and importantly: accessibility-tested). Encouragingly in the UI space, there are even paid options that are becoming available (for example TailwindUI and Remix). For many things you'll find yourself engaged in as a UI developer, there are existing solutions for the problems you'll face.

What I love about this is that it means there are fewer things I have to spend my time doing that aren't my core competency as a developer. Just like it would be ludicrous for me to build my own programming language, it would be just as foolish for me to build my own modal implementation when such great ones exist already.

I can hear some of you complaining already, so let me address the elephant in the room. Sometimes you'll find that you just don't like the way a component was implemented and you're bending over backward to fit your use case into their API. I would suggest that when what you're dealing with is a mature and established solution for the kind of UI you're trying to build, that you pause a moment. Consider whether what bothers you about the implementation is really a legitimate issue (performance and/or accessibility problems for example), or if it's just a matter of familiarity. Often we can have knee-jerk reactions to APIs that are different from what we're used to and it would be a mistake to let those first impressions control whether we embraced the new (React + JSX comes to mind for me).

Please keep in mind that I'm not saying you should never build a component library for your company. Most companies should do that (heck, where do you think most of the existing solutions came from?). But for goodness sake, if you're thinking about building your own accordion, tabs, or combobox, please reconsider. These are nontrivial to do well. Personally I blame the web platform for this problem (style-able versions of these common components should be baked-into the platform), but that's irrelevant to this discussion. Find a mature and battle-tested version of these kinds of low-level components and use that instead. Your company is not as much of a unique snowflake as you might like to think.

The kinds of components that should be included in your company's component library are the ones that wrap the lower-level primitive components for the common use cases in your application. And no, I'm not saying: "make a wrapper, then you can swap out the implementation" (this rarely works in practice). What I'm saying is you take the capabilities of the components you've chosen and you limit what people can do to promote consistency in your UI and codebase as much as is practical. Additionally, there are probably UI elements in your application that are pretty unique to your application, but are a composition of several other lower-level components. Those are also good candidates for a component library of your own.

So sure, you can have a component library that includes things like buttons, layout, typography, form elements, etc. But you don't have to build every one of these things from scratch. The other day, I saw a button component with thousands of lines of code. I mean, these things can get surprisingly complex if you let them, but for goodness sake, please just focus on your business.

I know it can be a lot of fun to build everything yourself. When I was at PayPal, I worked on a React component library and I built the button from scratch (but I also wrapped some open source stuff for other components). So I've been there. But not all companies have the resources to dedicate an engineer or two solely on building UI components. At some point, you've gotta put those components into something useful to the people who pay the bills. And those people don't pay for your internal component library doc's site (even if it does have runnable in-browser code examples).

I really don't want to make it sound like having a component library of some kind is a waste of time. It is not a waste of time. But you've just got to balance your excitement for building the perfect abstraction for a <Divider /> with the practicality of the sustainability of your business. Keep in mind the core-competency of your business and don't just go for the "build" option because it's a fun technical challenge, but because it's actually necessary for your business to succeed.

Whatever mix of build and buy you go for, I wish you the best of luck. 👍

Epic React

Get Really Good at React

Illustration of a Rocket
Kent C. Dodds
Written by Kent C. Dodds

Kent C. Dodds is a JavaScript software engineer and teacher. Kent's taught hundreds of thousands of people how to make the world a better place with quality software development tools and practices. He lives with his wife and four kids in Utah.

Learn more about Kent

If you found this article helpful.

You will love these ones as well.