A lot of Tailwind's value comes not from the utility classes, but from the design system they enforce on the user, which is the most valued feature Tailwind offers, in my opinion.
I’ve always had the idea of producing a CSS design system with just a bunch of preset CSS variables to use like ‘margin-left: var(—-p2);’, as it’s close to how I end up using Tailwind lately; ‘@apply ml-2;’
Why use @apply for this? The classes are already simple and standardized. Now if I want to modify a component style I have to go find the css file with the @apply rule, which is how css normally used to work & kind of defeats the point of Tailwind IMO. I also have to grep the whole codebase and make sure nobody was using `.card-bg` or whatever in an unexpected spot. Plus you have to deal with a mixture of bespoke CSS classes and tailwind ones. Seems like a hassle!
It’s indeed a hassle if you were to mix your own CSS classes and Tailwind, but I don’t do that. Instead I use Tailwind for its design system and set of defaults.
My own CSS is classic BEM style, so very component like.
In hindsight I could easily replace Tailwind and just produce a set of CSS variable as a design system.
Yeah, the advantage of limiting colours and sizes to preset multiples made life a lot nicer for a while. Very easy to keep things looking consistent without needing to re-reference specific values.
But they keep making it more and more easy to do things like p-[7px] which rather defeats the point and there seems to be talk about changing the scales to just using px/rem numbers which feels like a complete abandonment of the idea.
Customise the tailwind config for your design system, add additional CSS in a css file for specific needs. Generating arbitrary values on the fly just muddies the system back up again.
Exactly. Bootstrap always had Utility Classes but they lost to Tailwind due to the focus on design and great examples/templates that Tailwind team focussed on early on and then the market took over.
I have been using 3.5 sonnet to generate me tailwind components it does a pretty good job, and then with clsx, tailwind-merge and tailwind-variants things become pretty maintainable. I reach-out to shadcn-svelte when I need components which would take some extra tuning.
Same. 3.5 Sonnet seems particularly good at React and Tailwind. I'm moving away from other frameworks and libraries to React and Tailwind because of this.
I suppose various network effects have always hindered diversity, but I fear LLM training will severely hinder the adoption of new languages, frameworks, and growth of smaller technical communities. Businesses will choose to use the most automated solution rather than the best tool for the job.
Maybe the most automated solution will become the meaning of 'best tool for the job'.
The only positive I can see from this is that it should push maintainers towards stabilization faster as large API changes would result in huge productivity losses. Already true but perhaps more clear when the LLM starts producing failing tests.
This comment got downvoted because of link. I'm a slient reader for years, this account was registered days ago and not related to langcss. Langcss is a good product allow users generate UI with words. I should stand out and speak for good project and show my support.
Tailkits offers a lot of website templates that are written in tailwind, which are great to use when developing front-end pages, and even the free templates look great!
I really was hoping that tailwind was a fad which died but like everything JavaScript it's here to stay sadly
I recently found myself staring at a button element with more classes than I have ever seen like <button class="bg-blue-500 hover:bg-blue-600 text-white font-bold py-2 px-4 rounded-md shadow-md">Click me!</button>.
It’s like the eccentric artist who insists on using every color in the palette. After been forced to use it for almost 2 years I still don't see the point. Your HTML files? Bloated. Your sanity? Questionable. And those cryptic class names? Why God did we have to do this.
The point is that it’s ridiculously fast to build and extremely easy to maintain. It’s also not hard to read after, idk, a week of working in it? The only people I’ve seen struggle beyond that are people who dug in their heels and dogmatically decided to struggle with it.
FWIW for a Button, your engineer should just encapsulate all the “ridiculous” styling into a <Button /> element.
I strongly disagree on the maintainability. Mistakes are easy to make and have been made when comparing a list of 20 class names across a dozen elements. Its only "maintainable" if you ignore the class list when reading. I review my local evangelists' tailwind code and from their mistakes its clear they're not reading the class list.
It reminds me of how some people claim they're great at multitasking and can text & drive with no issues - meanwhile not even being aware of what they're missing.
Tailwind has its uses, but it should not be the default choice.
I’ve experienced the exact same thing. Most TW evangelists I know are admittedly not that great at CSS. Those who are generally prefer a library like MUI instead, which offers the same maintainability benefits, a far more consistent UI, but with a higher learning curve.
That's what I like about tailwindcss - unlike MUI and co, it lets you follow sane design rules laid by tailwind and yet make the website look different from the rest.
Consistency of UI entirely depends on the user of tailwind.
As an end user (not developer) I'd take just about any design system over MUI. It's so clunky/heavy, has an abundance of useless white space, and just feels really outdated.
I'd love to love MUI, but in my experience: MUI + React + Typescript just moves the toil of doing anything away from the style concerns and makes it a purely Typescript puzzle.
Strongly disagree, Tailwind should be a first choice for anyone reaching for their first utility/css framework. Tailwind imposes a very minimally opinionated styling system, that can easily be broken out of when a developer needs to.
> FWIW for a Button, your engineer should just encapsulate all the “ridiculous” styling into a <Button /> element.
Sure thing.
What I don't understand though is why not put a Button.css file next to the Button element file, and describe all the button styles in there. What's so hard about building or maintaining that?
Soon after, you're going to need slightly adjust that button, you will add some helper classes. Then there will be a case when base style of that button doesn't work with new button style, and now you have 2 bases for a button. You might have semantically named those buttons, but 3 months later that no longer makes sense, and now you're afraid to rename it.
Tailwind lets you style elements in the same file as the element, you don't have to switch between `button.<whatever you use>` and `button.css` to know what is going on.
> Soon after, you're going to need slightly adjust that button
The proper thing to do, in my mind, is to gradually set up a collection of such buttons (and other components), and document their properties. A design system, if you will. You shouldn't have that many button variations.
This should also help with the "3 months later" case. Once you've catalogued your buttons, you can reason about them, update their styles, rename them as you please, and so on, while maintaining consistency of buttons across your site.
You lost me at ‘maintain’. It’s definitely not easier than just updating a single .button class, and I’m not going to create reusable components out of everything in my project, the <Button /> approach doesn’t work for everything.
You’re assuming I’m going to create a component out of any element on the page that needs styling. In my project I could just have a small handful of these scattered across hundreds of files:
<table class=“datatable”>
If I used tailwind utility classes, how do I update my data table styling without having to search and replace across all of my files? Assuming you could even search and replace, the utility classes could be in a different order. Not saying it’s not doable, but it seems like a real pain in the ass. What am I missing?
Open the browser tools, select the element, view the styles and look what class cause the problem. Nothing else would you do, if you had a problem in styles css file.
Then I’m failing to see why it’s “bad” other than the verbosity/visual noise, which does in fact go away if you hide it (this is the point of abstraction in general, of course)
It's adequate for some folks who work in the same codebase, and define their own style of work. It's been a real pain for me who comes in to various codebases for short periods of time and have to read/digest/learn every single project's unique take on buttons.
class="btn btn-primary" is also pretty easy to read. But in the last couple weeks I've have to deal with both class="bg-blue-500 hover:bg-blue-600 text-white font-bold py-2 px-4 rounded-md shadow-md" style buttons and <LinkButton4 /> and <LinkButton7 /> components, both of which still seemed to be affected by another wrapping container that decided to "text-xs" everything below it, but which wasn't obvious. Trying to simply add a 'button' that had the same styles around it was far more work than it should have been.
No doubt, I'm somehow in a minority of switching between projects/codebases more often than other people. But this was much less of a problem with bootstrap and related definitions.
> your engineer should just encapsulate all the “ridiculous” styling into a <Button /> element.
When you want, say, most things to share some common color/style definitions, you will still end up repeating a lot of stuff (same long string for buttons, and for headers, etc). The tailwind system gives you the option of @apply, but the tw docs are also relatively strongly against it, for ... I can't really understand the reason why.
> If you’re going to use @apply, use it for very small, highly reusable things like buttons and form controls — and even then only if you’re not using a framework like React where a component would be a better choice.
There's some reasons listed on that page justifying "don't use @apply!" but... they don't ring all that useful to me. They do say "well, you could do it for reusable things like buttons", but those are the very things that most tailwind examples use longhand examples for.
> Yes, HTML templates littered with Tailwind classes are kind of ugly. Making changes in a project that has tons of custom CSS is worse.
There's a balance between the 'ugly' (their words) and 'tons of custom css'. Everyone's balance is different, I get it, but I'm usually having to come in after someone who was jazzed about tailwind has moved on and left me the 'ugly' (their words) mess. If every link should have "text-dark font-bold font-xs"... @apply that with an understandable name. Or document your design decisions...? But that conflicts with "ridiculously fast to build". Documenting them in an '@apply' way would let me see how you want some styles grouped together without having to scan through dozens of your components looking for a pattern.
> without having to scan through dozens of your components looking for a pattern
Wouldn't you have to scan disparately located CSS files anyway? How many projects have you seen with nonsensical or clobbered classnames or conflicting rules?
Really for a lot of designs I think you can just eyeball it. If `px-4` looks about right, then use it. Otherwise you can find a similar component and hijack the exact padding if you want, or get it from the design.
You probably think @apply makes more sense because you know and like traditional CSS. You want to apply tailwind classes and condense them into a single reusable one. Tailwind asks you to think in a different way. The classes are aptly named (mostly) and granular (one class per css directive), so my knowledge of CSS felt very transferable.
<LinkButton4 />
Imo should be
<LinkButton variant={"homepage"} />
Leveraging something like Class Variance Authority or more simply:
Again, at least part of the core issues I have is coming in to other projects where each person thinks in a 'different way'. And almost all of them are simply copy/pasting loads of tailwind extended strings of stuff, then semi randomly tweaking until something looks 'ok', then moving on.
> scan disparately located CSS files anyway
usually, I encounter CSS in one of two ways:
1. a main base css file or set of files (from a framework, maybe), and then an override css file. Single override file can get hairy to look through, admittedly, but it's typically in one spot. I rarely come across projects with dozens of standalone css files.
2. 'style-in-components' - primarily in vue, because vue sfc encourages scoped styling - structure, code, style in a single file. Whether that's bootstrap styling ('btn btn-primary', etc) or tailwind doesn't make too much of a difference.
Where I've hit more maintenance issues is coming in to projects with tailwind in say, react or vue - with the 'style in components' - is that "text-xs px-1 rounded-sm bg-gray-200 inline-block text-gray-600 mb-1" repeated across multiple components, is harder to find when I want to replicate something vs just "class='tooltip'"
Of course, we're all doing it wrong, and if we all just 'think different' all the time, and spend loads of time rewriting "wrong" stuff from other people - things will be just fine. But the world I live in, and the world I see a lot of people in - is not greenfield, but brownfield, and we have to live with the decisions others made. Tailwind does not make my life any easier that way, and generally introduces more cognitive overhead earlier in a project when it comes to visual stuff.
Every time something with tailwind is posted there is always a rehash of this response. I believe most of this frustration comes from using tailwind in the raw. There are a few essential libraries available that amplify the power of tailwind by magnitudes: tailwind-merge for last-specified wins, and class-variance-authority to reduce your class declarations in coherent reusable variants.
Tailwind empowers the expressiveness of styling. I steadfastly counter that the statement that it is the eccentric artist who insists on using every color in the palette. By design, tailwind implements constraints on what can be utilized by it's configuration file. Sure, you can absolutely break out with arbitrary classes but the spirit is akin to a Crayola box of 64 crayons and you have to work with what you have in front of you.
The largest benefit is that I can read my JSX/HTML and understand the presentation of the elements. Nothing is cryptic. I don't have to dive into another file where I can lose cognitive load. DX is enhanced. Critically stated: tailwind has been nothing but a massive boon for web dev since its inception.
Your last argument is a perfect reason why tailwind isn't a fad. Its a standard where, if you've worked with it, you know exactly what these classes will do.
Saying "the classlist becomes long" is not an argument to avoid using tailwind, especially if you already have a component based framework. I'd rather read a tailwind classlist than having to spend time setting up my own half-arsed class list.
Yes, CSS variables. Defining a preset of variables on :root for sizes and color ranges goes a long way.
The design philosophies behind Tailwind are sound, but like so many I dislike the technical implementation. I feel like there is a lot of halo effect happening with Tailwind where judgement of technical merits is influenced by judgement on visual merits.
You may use CDN build of UnoCSS (16k GH stars) https://unocss.dev/integrations/runtime which generates styles on the fly for 80% of benefits with 0% build steps (even with regular HTML)
It does have quirks compared to build variant, so if you use a framework, using it in a build step may be better, it'll still be much lighter then tailwind
They’re as cryptic as ls, cat, pwd. They have well defined semantics, sensible default, and tooling that can warn if your properties are conflicting.
In fact, with regular css you need to target elements by coming up with arbitrary class names, like .banner-main-content. Those are more cryptic imo, and completely arbitrary. (Naming things is problem #2, remember)
Do I wish there was a better way than space-separated attrs inside the human hostile xml-like format that is html? Yes. Do I want to go back to having to write custom selectors? No. 80-95% of the time utility classes is a better fit, especially with component based frameworks.
Have you used MUI before? No need for utility classes, no need for writing custom selectors, everything ties back to the theme, and everything is consistent. I don’t believe that “80-95% of the time utility classes are better” when super popular alternatives like MUI exist.
Tailwind introduces some solutions while introducing other problems. It has its place when used correctly, but I’ve never seen a single TW codebase implemented properly.
I’ve been writing CSS for 25 years and have gone through just about everything at this point—inline/attribute styles, CSS, SCSS/SASS, LESS, stylus, Bourbon, Tailwind, Bootstrap, Ant, Semantic UI, MUI, etc. And tbh, MUI is the closest I think we’ve gotten towards rapid iteration while solving specificity issues, the problem of selector generation, proper a11y and state styling (e.g. button focus styling), and inconsistency issues, but again—only when architectured correctly.
No, React-specific? Keep in mind Tailwind is not bound to a specific component engine and works fantastic in e.g. Svelte which I'm using.
That said, it looks good I guess? But I fail to see what's so unique. Tailwind is just a lower level preprocessor that can be used with unstyled components (which is definitely not unique - there's lots of traction outside the React ecosystem as well..) From what I can see, it looks like MUI also has a utility class type of approach, or is it baseUI?
In either case, my point is that plain CSS with class selectors is not great, and that very often it makes more sense to bundle markup and styles. Well-thought-out interactions between the component- and style/theming systems is also an important part, and Tailwind clearly plays well with different component systems in the ecosystem. Other than that I don't have any interest in dying on some Tailwind-specific hill. I'm happy to see it improve or be superseded, should that be necessary.
> Tailwind [...] has its place when used correctly, but I’ve never seen a single TW codebase implemented properly.
Yes but ironically that's not necessarily a bad thing. That's the expected result of tech that is flexible enough to let you do what you need, including painting yourself into a corner. The most common mistake in framework design I've seen is "one to rule them all" - where you have to play according to restrictive rules within, or suffer greatly to break out. Tailwind is a la carte, incremental, and plays nice with other paradigms, which isn't to be underestimated.
Last commit was four years ago and it has some weird bugs, eg this page tells me "Warning: Your browser does not support WebComponents" when Mobile Safari most definitely does. https://www.muicss.com/docs/v1/webcomponents/buttons
MUI (and Joy UI) is a lot different than a utility CSS framework. It has <Box>, which is a utility CSS component, and sx, which is a theme-aware way of doing inline styles.
The difference is that it comes with a <Button> component (and of course more), which has different variants, and you can set default props and default styles at the theme level.
So where an outlined button in a utility CSS world is class="... ... ... ...", or figuring out a way to be able to express that with class="button button-outlined", in MUI and Joy it's <Button>
So that’s a mid blue button that gets slightly darker on hover, with bold white text, it’s got a little more horizontal padding than vertical padding and has slight corner rounding and a slight drop shadow.
And I can see all of that in context where the button is declared rather than having to go look at another file and see what everything is. Oh and I know that none of those TW classes conflict or over write any properties, unlike two in house styles.
A few points on that. First, in 25 years of front end dev at scale I can't think of a time when a site I've worked on has ever actually made a site wide change to the styling without it being a big deal. It's always a lot more work than just changing some CSS because you have to consider things like whether you actually want to change every button, what background color the buttons are sat on, what text color it has, whether the color you choose should be applied everywhere or just in dark|light mode, and so on. Site wide changes are a big deal and the actual code is a relative small part of that.
Secondly, with the componentization of web apps this is largely quite simple from a code perspective. All the other complexities remain but changing a Tailwind styled button isn't much hard than changing a CSS class.
Lastly, designers are moving to design tokens applied in Tailwind config or CSS vars. This decouples the colors from the classes and turns them into config that's exported from an app like Figma, with tools to automate that export as a git PR if you're feeling technical.
These things were hard to do. People in front end tech have thought about that and built tools to make it a lot easier.
There are solutions of course. But they quickly break down when you're iterating quickly, copying in code from libraries like flowbite, etc and want to go back and restyle later.
A better approach that doesn't seem to exist would be some after-the-fact automated cleanup that could at least restructure the class list on attributes to e.g. sort class names alphabetically, replace where you copied text-blue-* instead of text-primary-* to fix issues like that, or other tools to aid refactoring.
Having just dipped my toes into tailwind and getting back up to speed on CSS, its shortcut syntax to base CSS. Putting all of that right on the element makes it easy for anyone to understand what is going on.
After seeing web tech come and go over the years, I have a distaste for most tech that transpiles or similar things. In my experience, transpiling increases the likelihood of painting yourself into a corner eventually. It complicates the tech stack by adding more dependencies and it appears a goal with a lot of frameworks is to avoid actually writing HTML and CSS. Sure the goal of using techniques so you don't have to repeat code is there too but how many abstractions are actually necessary to achieve that?
Some time ago i thought the same, but using it a while change my mind. You can create and edit quick things, especially if some elements should have another styling. I don't miss the times when you specify every elements styling, especially nested components where you have to be carefull.
Whether web-app or website, nobody create them in 2024 from scratch without any tools. And these tools can render your HTML with minimal classnames and optimize you css-file with needed tailwind styles.
Primary i use it in web-app, and i don't care about dozens of classnames. The tools optimize it.
> I really was hoping that tailwind was a fad which died but like everything JavaScript it's here to stay sadly
I feel with Tailwind it's worse somehow.
At least heavy javascript frameworks are commonly ridiculed by back-end or full-stack developers here on HackerNews; but with Tailwind, they embraced it just as firmly as front-end developers have. It is common to hear from back-end or full-stack dev that all they need is just jQuery or plain vanilla; but far more rare to hear same sentiment expressed about plain old (or rather plain new) CSS.
I would say I have basic skills (and interest frankly) in web dev, and I find all the HTML and CSS to look pretty messy anyway. At least with Tailwind I can have the mess in one place and try and understand it there, rather than needing to track down and interpret which CSS class is causing what and how to modify it without breaking something else.