Show HN: Tailwind Template Directory(tailkits.com) |
Show HN: Tailwind Template Directory(tailkits.com) |
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.
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.
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.
webscrape -> LLM -> generate template -> sell on website
FWIW for a Button, your engineer should just encapsulate all the “ridiculous” styling into a <Button /> element.
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.
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?
It's definitely easier than any alternative, but it's also painful to read through a giant list of classes. There's no getting around that.
I think a better way to put it is that Tailwind optimizes for writes vs reads. It sacrifices the readability of HTML for the writability of CSS.
I still think that's a worthwhile tradeoff, but it's definitely a tradeoff.
And pray you never have to tweak or debug that element.
Hiding away bad code where you don't have to look at it doesn't stop it from being bad.
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.
https://tailwindcss.com/docs/reusing-styles
> 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.
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.
I'm in this war against build steps.
Is there a way to 80/20 Tailwind without the build steps?
I mean, get 80% of the benefits with 20% of the work.
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.
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.
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.
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?
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 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.
My compromise with it is just using @apply in css files:
styles.css:
.button { @apply px-2 py-1 rounded-sm shadow-md hover:shadow-lg font-bold cursor-pointer block; height: 32px; }
.button.create { @apply bg-green-700 text-white; }
template.html:
<button class="button create">...</button>
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.
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.
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.
I wrote about some of my thoughts here about tailwind and the design system it imposes: https://alexmason.me/blog/2024-03-02-tailwind-made-css-great...
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.
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
You can even use it directly in runtime https://unocss.dev/integrations/runtime
(assuming no css-in-js) Your styles are bound to the component, not some global in your app.
States and their representations are controlled in the component themselves
With normal css files and class names, you must jump between files. With tailwind, I see what's going on with a component right in my jawascript
<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?
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 />
<LinkButton variant={"homepage"} />
<LinkButton className="text-xl other-override-here" />
I have found tailwind to be indispensable, and it pairs particularly well with React.
> 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.
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.
nah, I love tailwindcss.
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.
2 - It’s simpler to have styles that depend on js vars/parameters when components handle their own styling.
Encapsulation, in other words.
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
(Filed an issue: https://github.com/muicss/mui/issues/338 )
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>
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.
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.
Anyway the GP said he could see everything without extra files. Which is it?