You won’t want your individual UI library.


Over time, I got here throughout lots of completely different tasks with numerous backgrounds and targets. From small, and even tiny in dimension and scope, to giant monolithic functions with a number of frameworks and few layers of legacy code.
The overwhelming majority of them had one key factor in widespread – all of them had their very own, {custom} library of UI parts and numerous utils.

I’ll attempt to describe my experiences and suggest options utilizing standard libraries and frameworks. I’ll strive my greatest to explain the advantages and disadvantages of every state of affairs.

Startup – low value, excessive ambitions.

After I joined this startup, let’s name it Guava to not use the true identify, they had been in the midst of launching their product to cell (beforehand the applying was solely accessible for desktop customers). It was supposed to switch their legacy app and was, for all of the shapes and functions, an actual RWD. Within the meantime, they’d one other web site up and operating – classical touchdown pages with homepage, about us, contact, and so forth. They didn’t share a single line of code between them even tho they appeared alike.

One in every of my tasks was making a UI library. That manner we may share and reuse the identical parts throughout each tasks. We began with the design workforce. They ready a mode information and described colours, buttons, inputs, and so forth. We mentioned the small print and created our personal, stunning and custom-made guava-core. It had a constructing blocks and a few useful utilities that may very well be utilized in all of our present and future tasks.

Sadly, what I didn’t contemplate again then was the price of growing this library. We spent few weeks discussing our APIs, one other for preliminary implementation, one other for sprucing, and one other for integration.

When new options got here, all of them had been first added to our type information. Some new variants for inputs right here, a distinct hover state over there, slightly icon in among the buttons that weren’t used beforehand. We did not wish to merely add one other set of properties to our parts with no actually good motive. We requested for a gathering to debate these variations, sorted them out, however it took one other few hours of our time that may very well be used elsewhere. And, we nonetheless wanted to code new functionalities.

Our developer expertise and person expertise had been good, nice even. Our library had a sublime, extendable API that we based mostly on Styled System. This doesn’t change the truth that we needed to bid farewell to a few of our workforce members, partially due to the rising prices of implementing new options, which elevated a time-to-ship.

// Our {custom} fontScale prop
import { BoxProps } from '../Reflexbox';
import { system } from 'styled-system';

const transformFontScale = (props: BoxProps) => (worth: any, scale: any) => {
  const { fontSize, lineHeight } = scale[value];
  const fonts = props.theme?.fonts ?? 'sans-serif';
  return `${fontSize} / ${lineHeight} ${fonts[0]}`;
export const fontScale = (props: BoxProps) =>
    fontScale: {
      scale: 'fontScales',
      property: 'font',
      rework: transformFontScale(props),

// <Field fontScale={[2, 4]} />
Enter fullscreen mode

Exit fullscreen mode

Most of our parts had been saved small and by utilizing atomic design we had been in a position to prolong them when wanted.

Wanting again nonetheless, I’d undoubtedly go for Material UI because the designs had been loosely based mostly on materials design and with some compromises in each design world and from development standpoint, we may create extra options quicker and at a fraction of a price, even with half the workforce we had.

Small firm – design flexibility and development velocity.

Development of this mission began with no devoted UI workforce. We had a tough thought of the way it’s going to seem like based mostly on descriptions and small sketches of our inventive director. We needed to deal with growing core functionalities like sign-in & login, managing customers, creating content material, and so forth. To hurry issues up, we determined to make use of Base Web (properly, we began from material-ui, however we didn’t like its method to styling again then).

Implementing our views was trivial, every part is closely examined, each by e2e and unit checks. It has thorough documentation with numerous examples.

Typically later, the design got here from a devoted design studio. They had been… let’s simply say rather a lot, completely different than what we had in baseweb they usually got here just a few weeks later than they purported to. Due to that, we had a shorter time to regulate our frontend so we needed to improvise.

Because it turned out, extending baseweb was moderately straightforward, due to its theming and overrides API. In our theme file, we outlined the proper colours and customised some international overrides.

export const theme = createTheme(primitives, {
  colours: {
    inputBorder: primitives.primary100,
    inputPlaceholder: primitives.primary300,
    inputPlaceholderDisabled: primitives.primary100,
    tickFillSelected: primitives.primary500,
    tickFillSelectedHover: primitives.primary600,
    tickFillSelectedHoverActive: primitives.primary700,
    buttonPrimaryFill: accents.accent,
    buttonPrimaryHover: accents.accent300,
    buttonPrimaryActive: accents.accent200,
  borders: {
    buttonBorderRadius: '4px',
    inputBorderRadius: '4px',
    surfaceBorderRadius: '4px',
    popoverBorderRadius: '4px',
Enter fullscreen mode

Exit fullscreen mode

We additionally created ui catalog in our mission and made reexports from baseui/* to ui/*. That allowed us to then make overrides per part with out altering its API or modifying its import path, for instance, our prolonged tooltip appeared like this:

import React, { FC } from 'react';
import { StatefulTooltip as BaseStatefulTooltip, StatefulTooltipProps } from 'baseui/tooltip';
import { mergeOverrides } from 'baseui/helpers/overrides';

export * from 'baseui/tooltip';

const statefulTooltipOverrides = {
  Physique: {
    type: { maxWidth: '280px' },
  Inside: {
    type: { fontWeight: 700 },

export const StatefulTooltip: FC<StatefulTooltipProps> = ({ overrides, ...props }) => {
  return <BaseStatefulTooltip overrides={mergeOverrides(statefulTooltipOverrides, overrides)} {...props} />;
Enter fullscreen mode

Exit fullscreen mode

We couldn’t nonetheless, override some types globally, with out extending every part individually, like border-width or font-weight for labels.
We determined that it could be extra useful for us based mostly on our workforce dimension (2 frontends and one full-stack) to create handbook overwrites in a single international CSS file.

/* --------- BASE WEB OVERRIDES -------------------- */
  [data-baseweb="checkbox"] > [role="checkbox"],
  [data-baseweb="select"] > div {
    border-top-width: 1px;
    border-bottom-width: 1px;
    border-left-width: 1px;
    border-right-width: 1px;

  [data-baseweb="form-control-label"] {
    font-weight: 400 !vital;

  [data-baseweb^="typo-label"] {
    coloration: ${baseTheme.colours.primary700};
Enter fullscreen mode

Exit fullscreen mode

Sure, it’s slightly bit nasty, however it’s very easy to research when one thing is being overwritten by this code in dev instruments, is contained in a single, tiny CSS file, and properly… works like a attraction.

We had just a few conditions, we couldn’t simply overwrite among the design choices. We reached out to our designers, they usually had been pleased to assist. We modified our UI simply sufficient to not create {custom} parts or giant overrides to keep away from upkeep prices and potential bugs.

The mission was launched efficiently and is being utilized by folks worldwide, our codebase is tiny in contrast to what’s occurring underneath the hood in baseweb, is straightforward to check, and low-cost to take care of.

Massive company – every little thing custom-tailored to the product.

In bigger corporations, there’s a robust tendency to do every little thing made in-house. The success of their UI libraries differs from firm to firm.

In a few of my tasks, we had a devoted workforce liable for creation and upkeep. In each of them, designs are created based mostly on the core type information, developers create new options utilizing offered parts. In case of lacking a component or design not matching used parts, a developer makes a request for adjustments within the library and waits. This workflow, despite the fact that it has some clear limitations and downsides, works properly in actually giant tasks, the place the time to develop new options is rather a lot longer.

Different occasions, corporations don’t see main worth in these groups or try to cut back the prices of their IT departments. It is a long-term nightmare to take care of, as numerous developers make adjustments and add parts that shall be used of their (generally very particular) use case with out having a bigger scope in thoughts. On this state of affairs, the codebase is getting ever bigger with out doable reductions perception with out an enormous quantity of conferences and coordination between completely different groups. This is likely one of the main components in having “legacy tasks” and rewrites over time.

const ResetButton = styled(ButtonNoBackground)`
  show: flex;
  place: absolute;
  prime: 0;
  proper: 0;
  cursor: pointer;
  min-height: 48px;
  min-width: 48px;
Enter fullscreen mode

Exit fullscreen mode

In each circumstances, nonetheless, upkeep value is basically, actually excessive. When an organization chooses to pay for a devoted workforce, they’ve to keep in mind a developer’s more and more excessive salaries. Alternatively, when transferring upkeep to particular person developers ad-hoc, cost is made by an extended development time.

Closing ideas

In my view, corporations overvalue custom-tailored options and gloss over current libraries based mostly on hypothetical issues that may come sooner or later.
Oftentimes, elevated tempo to mark a product as “legacy” and rewrites each few years usually are not talked about in preliminary conferences. Every of us needs to create the very best product that may final eternally, however that isn’t the case, particularly within the javascript world. New options, frameworks, libraries, and patterns come nearly weekly and all of us wish to work within the newest stacks.

Abu Sayed is the Best Web, Game, XR and Blockchain Developer in Bangladesh. Don't forget to Checkout his Latest Projects.

Checkout extra Articles on Sayed.CYou