DEV Community

James Palermo
James Palermo

Posted on

Normal to struggle with Tailwind?

I'm trying to experiment with Tailwind. It seems so... Scattered and difficult to keep track of. Should I push through and keep practicing with this method of styling or is normal CSS still going to be around for the foreseeable future?

I like structured and grouped stuff I guess.

Top comments (51)

Collapse
 
tqbit profile image
tq-bit

It seems so... Scattered and difficult to keep track of.

There's a reason why inline styling went out of fashion.

There are some useful tools to keep Tailwind styles under control, such as

Stuff's still quite messy tho. But I'm sure standard CSS will stay.

Collapse
 
darkwiiplayer profile image
𒎏Wii 🏳️‍⚧️

There's a reason why inline styling went out of fashion.

And I still don't understand how or why it came back

Collapse
 
tqbit profile image
tq-bit

I believe due to the way components are defined. Take a simple Vue SFC. There still is a separation of concerns between script, template and style. + they're meant to be reusable, almost like CSS classes are. So much for the theory.

It gets interesting when you try and reuse spacing, colors or even shadows over several components. The moment Tailwind's promise breaks is when I need the same color for, say a navbar AND a button component. Even worse, you probably have those util classes scattered over your whole codebase.

What if the client changes their mind? I want red instead of purple. You'll refactor the all components that use this coloring. Now you might say 'this can be done with a regex replace-statement'. At this very point, the circle closes if you directly applied Tailwind to the component instead of using dedicated classes.

Thread Thread
 
james_palermo_bc208e463e4 profile image
James Palermo

Oh my God, I hadn't even thought of refactoring or redesigns with Tailwind.

That actually caused a legit stress response in me. 😖

Collapse
 
matthewbdaly profile image
Matthew Daly
Thread Thread
 
darkwiiplayer profile image
𒎏Wii 🏳️‍⚧️

That article still only makes it sound like inline styles but a bit more convenient this time.

That still leaves the main problem though: You're inlining your styles in your document. It is inline styles.

Thread Thread
 
matthewbdaly profile image
Matthew Daly

No, it isn't, any more than an approach like BEM is. You're still using classes for styling, but the scope of those classes is generally reduced a little. It's definitely not a one-to-one relationship between Tailwind classes and single CSS rules in every case. And, by using the @apply directive you can easily extract the styles used to a stylesheet.

Fundamentally Tailwind is an abstraction over CSS that provides the following benefits:

  • Limits you to a predefined palette of colours, proportions and so on. You can change this palette if it restricts you, but by limiting the scope of what classes are available for styling, it helps to ensure a greater consistency in your application - for instance, you won't end up with Alice using 3px for padding, Bob using 4px and Chris using 2px
  • Provides consistent modifiers for things like media queries (including dark mode) and state such as focus or hover. Inline styles can't do that
  • Makes it much less likely CSS rules will grow out of control. By encouraging use of the predefined Tailwind classes and discouraging making your own, it makes it easy for tools like PurgeCSS to strip out unwanted styles, resulting in typically smaller CSS bundles in the production application.

It's a particularly good fit for component-based JS libraries like React or Vue where you're actively encouraged to extract common UI sections to their own component. In that context it tends to be extremely quick to style it using Tailwind once you get up to speed, making it extremely useful for prototyping. Further into your project, once you have some styles established, then if you want to reuse them in ways that template partials or separate UI components don't facilitate, then it's easy to use the @apply directive to extract common styles and create your own more conventional classes, without losing the advantages of consistency mentioned above.

Honestly, I thought the same at first and it took a while for it to click, but I'd never go back. I maintain a big legacy project with a huge pile of messy CSS that I inherited and is very difficult to strip out. That would be virtually impossible for that to happen with Tailwind.

Thread Thread
 
darkwiiplayer profile image
𒎏Wii 🏳️‍⚧️

It's definitely not a one-to-one relationship between Tailwind classes and single CSS rules in every case.

That's not the important part though. Whether you're using classes or inline styles, even if one lets you do more things with less typing, the fundamental difference is that you are inlining the actual styling into your HTML.

There may be a CSS document somewhere that defines some more powerful tools in the form of classes, but those aren't styling "rules" in the same sense as with normal CSS; they end up being more similar to CSS properties, in that they just toggle atomic design elements like a colour or the text alignment.

The actual styling is still in your HTML though, so no, it is just plainly wrong to claim utility classes are somehow fundamentally different than inlining CSS using the style attribute. They may be more powerful and convenient, and that might for many be the deciding factor as to why this approach is okay to use while inline CSS is evil; but some sort of defense has to be made. One cannot simply claim that "they're different things" and dismiss all criticism of inlining styles into HTML.

There's a separate discussion to be had about whether enough of the problems with inline CSS aren't present, or whether the most relevant ones still remain, and that discussion can be a lot more nuanced and ultimately up to personal preference and picking the right tool for the job.

Thread Thread
 
matthewbdaly profile image
Matthew Daly

The @apply directive renders much of that moot, though, since once you've settled on a style that works you can easily use that to extract common patterns into reusable classes. It depends on the context you're using it in, and for using component libraries like React it often makes less sense, but certainly if you're using Tailwind in something like Blade or Twig templates then using @apply is more commonplace.

Thread Thread
 
darkwiiplayer profile image
𒎏Wii 🏳️‍⚧️

At that point, haven't you just gone full circle and basically achieved nothing other than introducing two new dependencies?

Thread Thread
 
matthewbdaly profile image
Matthew Daly

No, because Tailwind still works as an abstraction layer. And you can combine the two approaches however you see fit - it doesn't have to be either everything using the utility classes direct or everything using the apply directive.

It's like JSX in that it sounds arse-backwards when you first hear about it, but if you try it then once you get over the hump it starts to make a lot more sense.

Collapse
 
tqbit profile image
tq-bit • Edited

It isn't. But it does feel like it for a lot of classes:

  • flex
  • text-center
  • mb-4

...

Thread Thread
 
machy8 profile image
Vladimír Macháček • Edited

Yes. Utility-First uses classes so the property:values does one thing and can be easily combined and reused.

When you define selector on your own, you are mostly going to duplicate the css property:value in it. Which causes CSS size to grow, you have to use modifiers to change the css on certain places like button button--red. This is less likely to happen with utility first.

More info => dev.to/machy8/comment/1p2jj

Thread Thread
 
darkwiiplayer profile image
𒎏Wii 🏳️‍⚧️

What you call button button--red I just call <button class="danger">, which will need one selector for the button element and one for the .red class, which will work for any element, not just buttons.

When done right, this gives you a complete n * m elements you can style, with every colour applying to every element type (although, of course, in reality some of them will not be implemented because YAGNI).

Thread Thread
 
machy8 profile image
Vladimír Macháček • Edited

My bad, the button--red example was not showing the problem enough.

I am talking about reusing a button in an efficient way. What I mean by that is, when for example a button is used on multiple places and needs various indentation, colors, paddings and alignment.
You can't solve that by just one class. How would you solve that?

Thread Thread
 
darkwiiplayer profile image
𒎏Wii 🏳️‍⚧️

The way you phrase that question doesn't even make sense when we're talking about proper rule-based CSS; what's the context for these different indentations, colours, paddings, etc.? Does it depend on the surrounding elements? are certain buttons special? Is there just no clear design idea and every button just gets positioned by hand?

Collapse
 
machy8 profile image
Vladimír Macháček • Edited

Stylify or Tailwind is NOT inline styling.

Stylify uses selectors like class="color:blue" that are reused, optimized, combined and etc. The same goes for Tailwind and other utility-first frameworks.
Inline styles are style="color: blue;".
More info => dev.to/machy8/comment/1p2jj

Collapse
 
tqbit profile image
tq-bit • Edited

but if you're struggling with Tailwind, one nice alternative made by a fellow member of the DEV Community is Stylify.

I really mean no harm or offense, but this project seems like the worst of both worlds.

  • Additional bundle size mental load for no visible value added
  • No separation of concerns at all

Perhaps somebody can elighten me - why would I do this?

Collapse
 
leob profile image
leob • Edited

You're very kind to say you mean no harm or offense, I'd be a bit more harsh and say that this is just a terrible idea.

Collapse
 
leob profile image
leob

Woa, and how is this supposed to be maintainable? Hard-coded values all over the place, that doesn't really sound like a great idea to me ... you could just as well put a style attribute on your HTML elements and stuff all your CSS styling in there, that's just as horrible but even simpler.

Collapse
 
ben profile image
Ben Halpern

I think it's abundantly normal to struggle with a new paradigm. Tailwind is a pretty simple mental model once you get it, but it's a pretty big departure from how you do CSS right now.

I guarantee it will click soon enough.

 
tqbit profile image
tq-bit • Edited

TL:DR

[...] "separating CSS" is relevant [...] - because it keeps your project maintainable by a team of devs, designers and other team members.

[...] why you think the bundle size will be affected negatively - in fact, judging just by size, I don't think it will. But the value of using this style of applying styles, in my eyes, does not outweigh its costs


I get your points. I'm biased on a few as well, so lots of words would probably not convince. Let me try and elaborate with a few:

About bundle size:
Perhaps phrasing on my first point was poor. With additional bundle size, I meant: It's another package I have to take care of during my dev workflow. I understand your point with atomic component design - it's a pattern where utility CSS frameworks such as Tailwind+PurgeCSS really shine. Because when you develop, you must make compromises between technical cost ( and debt ) vs. value added.

In a nutshell: The value I see at first and second glance does not outweigh the cost of using this package. It's just my opinion. It might as well be wrong and I'm sure there are projects where this will save a lot of time. I just cannot envision them clearly.

About SOC
As per your advice, I did Google "separation of concerns css in js". The first entry that came up led me to the book 'Programming JavaScript Applications by Eric Elliott', where the first sentence reads:

Separation of concerns is the idea that each module or layer in an application should only be responsible for one thing and should not contain code that deals with other things

As per my understanding:

  • HTML = structure
  • CSS = style
  • JS = interactivity

So if you say

"Styling shouldn't be considered a concern"

I say:

"Then neither should interactivity be"

Let's create a single language that does HTML, CSS + JS at once. Even better: Let's all use .jsp for everything again. That'll be fun.

In all honesty: There's a reason why there are whole CSS design systems, such as BEM or SMACCS around. If properly applied, they empower you to scale your product without compromise. Or shift it over to other teams without a lot of struggle.

Or refactor / rewrite a big chunk of your HTML templates. If you scrap your Tailwind-powered views, all the styling is gone for good. If you use a design system, all CSS classes will persist and you can quickly prototype new pages.

Now you might argue: "But this works with Tailwind too!". Sure it does, but building the templates AND the styles at once from scratch doubles the mental payload you might have to carry. If you already have a guiding star, it'll be much easier.

Collapse
 
saravananteslaturing profile image
Saravanan-TeslaTuring

You can use the Tailwind CSS Intellisense. If you have idea about bootstrap then tailwind is easy to understand. You can create any kind of websites without writing single extra CSS. I created 2 landing pages like that only. Final word Tailwind is super dooper

Collapse
 
james_palermo_bc208e463e4 profile image
James Palermo

2 landing pages is, no offense, small enough that I could memorize it.

Where I'm struggling is when you have a lot of pages and all the HTML is heavily... "Cluttered."

Collapse
 
saravananteslaturing profile image
Saravanan-TeslaTuring

It's depends how you are reusing the variables from Tailwind Config. You should have the plan when you are starting the project. Like colors, button sizes, border radius and everything. Only you are going to reuse that. Then whenever new designs you are adding you have to add that in config. Another biggest advantage for tailwind is arbitrary values. You can add particular styles in class itself. But you can't reuse those values again. So if you are creating big application or small application it does not matter. If you are reusing the styles properly, Then outcome will be really nice. Thanks

Collapse
 
machy8 profile image
Vladimír Macháček • Edited

Wow, this is some hell of a fight of arguments here 😆.

I am the author of Stylify and I will try to clarify some points in the thread above and bellow:

Why would anybody use such syntax

  1. Because you don't have tu study anything. It's like CSS instead of randomly named somehow shortened classes. An example: auto-cols-auto is a class from Tailwind. The class is not self explanatory and a dev not using Tailwind daily have to go into the docs or into the dev tools to see what it does. In Stylify you write this grid-auto-columns:auto. Everyone with a bit of knowledge CSS knows what that does.
  2. Another reason is maintainability. What if browsers comes with a property, Tailwind already shortened? An example shrink => flex-shrink: 1; (class from Tailwind). The browsers come with, for example a new shrink: auto. Then they will have to figure out a new name for the new selectory so it makes sense. Which can be confusing.
  3. When using BEM you can end up with <div class="page-section__container page-section__container--full-size page-section__container--without-background"></div>. I can't see how property:value selectors are more bad then this.

Shure, Stylify syntax might not be for anyone. You can however define custom macros for having classes like ml-2 or py-3 if you like it more. It's just a Native Preset that you can ignore and define custom set.

There is plenty of hardcoded values

  • In Stylify, you can configure variables and use them anywhere.
  • It's just up to developer if he is going to have hardcoded values in the code or not.

Why not to put the style directly into the style="" attribute

  • For example a selector like color:red is generated as .color\:red{color:red}. This selector can be reused.
  • When you add a component in Stylify like .button that needs red text, it is generated like this .color\:red,.button{color:red}. The selector is simply attached, reused and the property:value not generated again => This means smaller bundles
  • In production, you can also minify this selectors to => ._zs,_zx{color:red}. This is done even by Medium.com and Facebook.
  • But the minification will make the blocks harder to find when debugging because of the unreadable classes? No. You can use empty class as selector, data or id attributes. Also, in Javascript we normally minify classes and selectors from const myValue to const zx and nobody cares.

Advantages over pure CSS

  • Selectors are dynamically generated => don't have to remember to remove them when removing css and vice versa
  • Selectors are combined and reused => .color\:red,.button{color:red}
  • Selectors are minified from long text-align:left to short _zx
  • You open a template file and a browser and you just type the selectors. You don't have to switch between html and css for removing, renaming and manually combining clases
  • The CSS size doesn't grow exponencialy, because there is a minimum of duplicated css property:value. There is some article about CSS size from Facebook
  • When frontend and backend engineers works on one task, then it's easy for the frontend engineer to tell the backend engineer which classes to add when he just needs to indent or align something. He doesn't have to edit CSS and hope it will work. He just copies selectors.
  • Dynamically generated CSS means less files in the projects => less changes => we normally generate cache, entities, migrations, js bundles and etc. So Stylify dynamically generates CSS. At least from my point of view it's more comfortable and efficient then writting it manually and think about how to pregenerate utilities or combine css files manually.

Bloated templates & maintainability

  • You can define components. So there doesn't have to be any utility in the template.
  • From my experience, you can bloat your templates by a lot of other stuff => conditions, attributes, dynamic attributes, dynamically loaded templates and etc. Blaming CSS selectors is simply wierd.

Separation of CSS from HTML

  • I think, this havily depends on the angle of view
  • Any approach can go wrong
  • Stylify generates CSS files using JavaScript. Selectors are in the HTML/Twig/Latte/Vue/Jsx. Stylify just scan project files and generates CSS. That's all. It doesn't have to interfiere nor be included afterwards.
  • When you decide not to use Stylify anymore, you just run build, get the CSS files that contains CSS similar to what you would normally write, and remove the Stylify from the project.

Although I'm developing Stylify and I like the Utility-First + Components approach and dynamically generated CSS, I understand it doesn't fits everyone. Therefore, it's good that there are more tools that do a good job like Tailwind, Bootstrap, Bulma and various approaches so everyone can use whatever suits them. Different tools and approaches for different needs 🤟.


Thanks @lukeshiru for the persistent and extensive comments in a favor of Stylify.dev ❤️.

Collapse
 
moopet profile image
Ben Sinclair

Normal CSS is going to stay around.

If you have to use something like Tailwind I'd suggest avoiding using any of its utility classes in your HTML. Build your own classes, and include the utilities in them instead, so the HTML doesn't become littered with non-semantic attributes.

This will make it easier to see what's going on, and mean your project can be maintained in the future without resorting to search-and-replace choring.

The way most people use it, and the way most examples show, yes, it's "scattered", but you don't have to do that.

 
leob profile image
leob

Thanks a lot for this extensive answer, I wasn't familiar with Atomic CSS (but I was with Tailwind, and it's more or less the same concept) ...

I see what you mean, same as Tailwind this is to be used in the context of a framework like React or Vue that allows you to create components, hence solving the problems of "repetition" and "consistency" ... you wouldn't want to create a large site with heaps of HTML where you duplicate class=background-color:#023 a hundred times across your codebase.

Thread Thread
 
machy8 profile image
Vladimír Macháček

You can define components, variables and custom macros (like my-2) in Stylify if you need to. Therefore you don't have to have such repetative code.
More => dev.to/machy8/comment/1p2jj

Thread Thread
 
leob profile image
leob

Thanks for explaining that, so with these variables you avoid sprinkling hardcoded RGB values or hardcoded font/padding/margin sizes all over your codebase ... I do understand the advantages of this approach, traditional CSS stylesheets also have their problems.

Collapse
 
shigetorum profile image
Edge

TailwindCss essentially improves inline styling.

If you learn Tailwind you'l find yourself with ease when it comes to writing normal CSS, as many class names are similar in CSS.

Take this example:

<div class="flex items-center justify-center h-screen"> Hello World! </div>
Enter fullscreen mode Exit fullscreen mode

It can easily be converted to:

.class {
  display: flex;
  align-items: center;
  justify-content: center;
  height:100vh;
}
Enter fullscreen mode Exit fullscreen mode

It's also all about practice, eventually you'll get it.

Collapse
 
blindfish3 profile image
Ben Calder

To me Tailwind feels like the CSS equivalent of JQuery: it's a supposedly easier alternative for those who don't want to learn CSS properly 🤣
If you're not comfortable with Tailwind just use CSS or an established preprocessor like SCSS.

Collapse
 
moopet profile image
Ben Sinclair

Adding non-semantic classes to HTML is two steps back.

 
tqbit profile image
tq-bit

First off:

Thanks for going into detail. I really value it when folks discuss in detail [ ... ]

That thanks goes back to you, on top of it (I can't believe I write this): Thank you for staying on topic.

I do understand your points, and while I can't relate, I figure: The concept doesn't have to work for me as long as it does for you.

A few closing words from my side:

You find a really good quote, but I think you might have not interpreted it correctly:

It was the first quote I stumbled across. If I were to pick something more substantial, I'd go for Adam Wathan's article in which he speaks about his transition from "Semantic CSS" to "Utility first".

What even this article is missing: Global CSS variables. I'm talking about the color-brand or text-small that exist above your modular style scope.

Granted, you can extend Tailwind with your own classes. But I don't want to play hide and seek with colors when receiving feedback from a client. Imagine they want each h1 tag in gray-200 colors now instead of gray-300 when:

  • in dark-mode
  • the text is hovered
  • the user is on a mobile device

The only change I'd like to make is the variable --text-color-primary and be done with it. The rest will be handled by SCSS utility functions.

so separating styles or having them "inlined" is pretty much the same from the "outside world"

Can't agree. CSS is, indeed, a layer on top of structured HTML. It can-and imo. has to-be abstracted to make it maintainable. And changeable within a reasonable amount of time. Within a CSS structure, I then separate again-between variables, base styles, component styles, utility styles, and so forth.

This is (almost) also true for components. But extending components by, say variants with Tailwind or Stilify, will undoubtedly lead to unreadable templates.

The same applies to styles, if you need a style to be shared between different modules, then those styles shouldn't be locked inside a single module, and instead, be shared

I think I caught you here. The idea of sharing is only possible if you actually do separate by concerns. For instance, inline styles cannot be shared. Stilify CSS can, because it feels like reverse-engineered inline styles.

then those styles shouldn't be locked inside a single module,

And I'm fully with you: Why declare styles in your HTML while you can do the same in CSS?

 
m4rcoperuano profile image
Marco Ledesma

I see. Although I can use rems, sure, and just change the size of rems - I think specifying an actual unit type in your css class still has a smell for me. It could just be personal preference, since I’m used to abstracting away unit types in class names. But I think the authors goal is to remove the unit abstractions, since one of his pain points of tailwind was remembering class names.

I see that the author put a lot of love into the stylify framework. Reading his post and another of his follow up post, he makes a good argument that it’s a pain to remember class names of other frameworks. I won’t trash his work since the author is passionate, and what works for him works. Just not for me I suppose.

Collapse
 
moose_said profile image
Mostafa Said

I really didn't take anytime learning TailwindCSS. It was simple enough to get used to it in just few days. You would just peak on the docs many times till it sticks in your brain and that's it.

Try to push yourself out of the comfort zone to learn what's new and popular. It became popular for a reason.

Some comments may only be visible to logged-in visitors. Sign in to view all comments.