Comment by burlesona | Hacker News Reader

Comment by burlesona | Hacker News Reader

I'm ambivalent about Tailwind, but I've used it a lot. I will say the hypothetical advantages are mostly the following:

1. It's a step function over CSS units. This is the biggest strength, just standardizing that your design uses padding of 2, 4, 8px, but not 1px, 3px, or 1.23123em :). It provides more steps than you need, but still it's good that the core of Tailwind is a design system with defined unit and color variables.

2. Some of the utility classes are very helpful. Even as someone who likes writing CSS, it's nice to not need to give something a custom classname just because I want to put margin-top on it. class="mt-4", done.

I think the problem is Tailwind goes too far and tries to replace EVERYTHING with a stack of utility classes.

This works okay in extremely componentized web apps. It's a nightmare if your UI isn't highly componentized. I've seen projects where you make a button by copy pasting this ~80 character string of tailwind classes all over the place, and then changing the color names if you need to. Good luck fixing that when the designer decides that we don't want any buttons to have rounded corners anymore.

Personally I think the best parts of Tailwind are captured in Pollen[1], but I do wish it came with a subset of utility classes for colors, font sizes, margin, padding, and text alignment. I think the hard part is defining which subset is the right subset... I doubt you could find strong agreement from a large majority of developers on that.

1. https://www.pollen.style

>This works okay in extremely componentized web apps. It's a nightmare if your UI isn't highly componentized. I've seen projects where you make a button by copy pasting this ~80 character string of tailwind classes all over the place, and then changing the color names if you need to. Good luck fixing that when the designer decides that we don't want any buttons to have rounded corners anymore.

That app is done wrong. If you are using the same styles to represent a button you should use postcss and do

.myButtonClass { @extend: (80 tailwind classes here) }

So now we have to add this new postcss thing. Modern web dev is bandaids all the way down.

postcss is literally how tailwind works. If you are using tailwind, you already have postcss. I get it, you don't like modern frameworks. Don't use it.

Or make a Button component in your framework of choice.

    function Button = ({ children }) => <button className={80 tailwind classes here}>{children}</button>

    <Button>Create</Button>
    <Button>Edit</Button> // Same style
    <Button>Delete</Button> // Same style

Even then, I would still recommend using the postcss extend tag. Much easier to read that way. Also, postcss doesn't require any frameworks, you can still do everything with plain ol' html and javascript.

No, we don't want all of the buttons to look identical.

Some are big and some are small; some are bold and primary and some are muted and secondary; some have icons; some have shadows; some are disabled, etc, etc.

It's easy to make them identical. The challenge is to be as flexible as necessary in a mature application, while minimizing verbosity and complexity.

In my experience Tailwind hurts more than it helps here. It forces you to use your component system to abstract things which otherwise wouldn't warrant the extra level of indirection.

How would you have done it without TailWind? You'd still have to create classes for each of them in plain css anyway, except it would probably be in a different .css file. The inherent complexity pertaining to your app is not taken away, it just moved the places.

My point is that plain CSS is already good at this. So then you don't need to make more components with their own interface to learn.

    <button class="btn large shadow">Edit</button>

    .btn {
      ...default button styles
    }

    .btn.large {
      font-size: larger;
    }

    .btn.shadow {
      box-shadow: 0 0 8px #0004;
    }

but now to understand what a single class might do to an elements style i have to check for ever possible combination .e.g

.btn.large {} .field.large {} .fieldgroup.large {} .title.large {} .link.large {} .large {}

to find all the elements that the .field.large declaration impacts we need to enumerate all the elements that have both .field and .large in any order.

we've coupled out html and css files

You should look more into design systems. It's not only for Figma files but can structure your whole frontend. To extend on the example above, you can do something like :

  <Button variation="primary" size="large">Primary Button</Button>
  <Button variation="secondary" size="medium">Medium</Button>
  <Button variation="tertiary" size="small">Small</Button>

If you a 100 different types of buttons I can see your problem, if it is more like 5 then you just create them first and then makes the code more DRY.

For example when using React you can create 5 button components first and the refactor your code to make the 5 buttons use/call a generic button. Or have 2 generic buttons if the combinations are hard to handle. The thing with atomic CSS like Tailwind is that it very easy to quickly create a few different buttons.

Doesn't that negate most of the benefits of using Tailwind in the first place?

Not at all. You design and iterate on the button using the individual classes, then you swap them out for a concise class once you are done. It's how tailwind is designed to be used.

https://tailwindcss.com/docs/reusing-styles

Although I realize in looking this up in the documentation to show you, it's actually @apply and not @extend you are supposed to use.

Using @apply is the ultimate last resort though.

Yes, there is a right way to do it. But in the code I've seen in the wild, people are often not doing it right and those 80 char strings of utility classes are on the low side compared to some of what I've seen. I'm skeptical of Tailwind not for its own sake but because a lot of (most?) shops do not have the discipline to use it effectively. And what you are arguing here is basically a no true Scotsman defense.

I wonder if there's an ESLint rule out there to limit how many tailwind classes you throw onto one thing...

I agree. I used Tailwind for roughly six months on a few projects before I noticed that I was only using the basic utility classes. I didn't like all the screen sizes and state "bloat" classes so I decided to take the best parts (margins, padding, background, font, text alignment) and create a simple CSS framework.

For anyone interested, you can find it at: https://brixi.dev/

Reply on news.ycombinator.com