Hacker News Reader

Comment by mkoubaa | Hacker News Reader

I've found that with enough effort and time I could eventually learn any programming concept.

The only exception I've encountered in my career is CSS. I'm no more effective at it than I was when I first tried using it.

View on HN

I feel the same way, and I first wrote CSS in 1998 when it was so new that Netscape misinterpreted even the font size property.

There are now thousands of properties and their values, and browser engines have gone to heroic lengths to produce mostly consistent results from these often conflicting directives. But I still keep looking up basic things because the API is simultaneously sprawling and inexpressive and context-sensitive.

IMO, the core sin of CSS is mixing up three separate concerns: element styling, layout, and compositing. It’s mostly ok for the first. It’s insanely confusing for the second. And it mostly tries to pretend it’s not about the third at all, which leads to weird situations around implicit compositing groups.

We ended up here because CSS started as basically a copy of a handful of Microsoft Word dialog boxes (hence “float” etc.) And as soon as we had got the word processing use case barely working, everybody decided they wanted to build dynamic applications with this same engine instead.

The core sin 0 of CSS is introducing another item-value syntax to the stack when there are regular markup attributes for that exact purpose already. Then it gets worse by taking this simplistic syntactic device ad absurdum, only to then finally also start introducing expressions (calc(), custom properties, media queries). Really, CSS should've been sealed a long time ago, with a low-level API a la Houdini available for programmatic layout from JS rather than bloating the language to become accidental Turing, when these highly dynamic styling features are useful for JS apps anyway. Anyway, considering its complete lack of mental discipline and basic CompSci, the CSS WG at W3C has no business designing languages IMNSHO.

I completely disagree. If CSS were "sealed" it would be absolutely terrible and we'd still be dealing with floats and absolutely positioned elements. The problem with CSS is the weird implicit behavior and edge cases that remain due to compatibility.

Computed properties, custom properties and media queries are well designed.

What CSS needs is a refactor. Make a much smaller Strict-CSS that uses the same code paths in browsers, but provides far less options to the user, sticking to modern tools (grid, flex) with simplified naming schemes, while deprecating old ways of doing (float). That would finally allow a manageable mental model.

That’s one of the great things about react native and react native web. It’s like CSS with about 25 years of crud removed.

Of course there’s some downsides, but it’s refreshing to use.

What do you think can be done about this, at the browser, tooling, or library level?

I solved it by using tailwind css :)

I was like you until about a year or two ago, and then set out to learn it enough to implement a site layout without having to resort to copy and pasting things I don't understand.

It wasn't that hard, once you accept that it'll require it's own study and practice. I put in maybe 20 hours dedicated study and got a lot better out of it.

I did a few video courses on LinkedIn Learning (formerly Lynda.com) through my public library. I did a general CSS course and a flexbox CSS course. Then I set out to design pages I needed without pulling some random blog theme and calling it a day.

I've gotten a bit better at design as well, although it is a much intrinsic field than CSS and requires a lot more study. The biggest difference was once I treated the design stage with respect as its own discipline and started designing in a design tool (figma) instead of "doing it live" by trying to tweak the design in code as I went. It made me 100% focused in that stage at the design problems and the feedback loop got a lot tighter compared to editing and running code to see how it looked.

I admire your dedication. In my case I saw no improvement when I tried to learn it

Because most css without context are useless. A div with width 100% may not stretch depending on it's content and parent css. It's even worse on positioning, margin and responsiveness.

Which is why if you don't interact with it day to day basis, you'll need some trial and error, as well as references and examples.

Opposite to backend or usual programming where `foo = foo + 1` is most of the time do an increment by one (exception apply)

There's a lot of unpredictable stuff too. I was trying to change the border color of a checkbox to match a framework being used elsewhere and for the life of me couldn't figure out why neither setting outline or border changed it. But on the framework's website demo it was doing it just fine.

Combing through the CSS and the computed styles panel in dev tools gave me nothing.

It turned out there is an option to disable the standard OS form theming which isn't communicated at all in dev tools for that element (and I hadn't seen --webkit-apperance used before as the guide says you really shouldn't mess with it): https://developer.mozilla.org/en-US/docs/Web/CSS/appearance

Weeeeeeeell a div is 100% width by default. It will always fill the width of the parent container unless you explicitly change the width.

Afaik div width property is almost useless if the parent set the display as flex, or if the div has no content.

That’s kinda the point of flex though. The width is determined by the flex rules; width 100% doesn’t make much sense in that context.

Not if the parent or ancestors don't have a width set... Hence why you need context.

The parent has _some_ width, even if it’s not explicitly set.

Yes, and I wish 100% would respect that implicit width but it doesn't. Try it

Any chance you could show an example? In theory a block level element can only sit inside another block level element. And those expand to the width of their parent unless you constrain the width.

https://jsfiddle.net/jcaysopd/75/

I have no doubt there is a 'logical reason' for the behaviour but it doesn't stop it being annoying to work with.

Gotcha. That’s because of the 100% vs auto thing. These days everyone just puts “box-sizing: border-box” on everything to make it go away.

This explains a bit more about what you’re seeing

http://stackoverflow.com/questions/28353625/why-does-percent...

100% isn’t the default value, ‘auto’ is. Which has the effect of appearing 100% in many scenarios. But not all.

Fair, but actually that just means that the border and margin etc are pushed outside the container. In either case the div is operating by taking up the full width of the container.

Backend or usual programming without context is also useless. Is foo even mutable? Is it the same type? How does type conversion work here?

CSS and usual programming are identical. If you don't give yourself the time to learn the fundamentals, you'll always run into tedious problems.

I know the fundamentals. Doesn't help :D

Maybe it's time to relearn the fundamentals! web.dev and MDN are great resources to do so. Start with the Box Model.

Fundamentals don’t help when you are 15 divs deep.

You have made some bad decisions if you find yourself 15 divs deep.

There is a surprisingly small amount of content out there on how to competently engineer your CSS (and a lot of it is bad). Most people don't even think it's something you do.

But basic things like making your general rules generic and pushing context descriptively into the HTML go a long way into making your code understandable.

Not my choice, that’s what you get when you inherit an Angular website.

> You have made some bad decisions if you find yourself 15 divs deep.

Open instagram.com in a browser :D

Or even W3C's own website

Do you mean leaning on rules like: .myWidgetClass form{ ... } .myWidgetClass form >input{ ... }

Sure, if you have some general rules that apply to all your widgets. If you are talking about something more specific, you are better with a more specific name.

I've got a feeling some people might consider this to be a rather snarky response, but honestly a lot of trouble comes from assuming you can muddle through CSS, and you can. Unfortunately that means it's easy to always be an expert beginner, and never quite grasp the underlying logic, as obtuse as it may be.

A lot of people learned the fundamentals incorrectly.

A good place to learn how CSS was designed to work is Every Layout at https://every-layout.dev/.

I really benefit from reading comments like this, so I appreciate your honesty. I didn't start learning JavaScript until very recently, even though I was making my own personal websites back when Angelfire, Geocities, and Tripod were still a thing. Instead, I focused on the frontend HTML/CSS combo, constantly learning and implementing new CSS technologies whenever I had an excuse to try them out. Because of that specialization, CSS feels understandable to me. My insecurities about my own abilities often present themselves as a belief that investing time in CSS was the growth of a cheap and relatively useless skill for a web development team. Whenever I see that someone who likely knows incredibly more about JavaScript and other programming languages — things that I feel are far harder to understand than CSS — share about how either HTML or CSS is difficult to use, I feel a little more confident that I do bring some value to the team with whom I work at the office.

CSS is a lot like Regex to me. I learn enough to do what I need to do then immediately forget everything.

if css made as much sense as regex I might actually enjoy front end work - there’s no underlying mathematical theory that explains css

Usually it's not the Automata theory parts of Regex that people find hard. It's the syntax.

Understand the Box Model and you'll understand all of CSS.

Hogwash. Where in the box model are the precedence rules for selectors defined? How does the box model help with page layout and positioning? Animation? Transformations? Pagination?

Sure, the box model is a prerequisite to understand CSS properly. But if you think the box model covers all aspects, you really don't know very much about CSS.

> there’s no underlying mathematical theory that explains css

There's a well-defined algorithm (well, set of algorithms) that explain it, which you can learn if you so choose.

I only do UI for business-facing apps and my own projects. CSS Grid massively improved my frustration with CSS. It makes it much easier to reason about how your page layout works.

I will never ever do another layout without it, stuck together by glue, snot, and magical trial-and-error.

It's one of those things that frameworks had been doing since (before?) 960gs. A Grid really needed to be incorporated into CSS and I'm happy to use it or flexbox for layout purposes depending on my use case.

I’ve still not gotten to grips with CSS grid, mostly because flexbox does what I want so I end up falling back on it at the first hurdle. But the rest of it is very clear to me these days.

IMO CSS Grid is actually much simpler and better designed than Flexbox. I’d recommend taking a proper look when you get a chance.

They solve different problems.

This is the kind of HN comment I come for in these kinds of threads. One person saying one thing (CSS Grid is the one thing keeping me sane!) Another person coming in to say the exact opposite.

I'll be a little more helpful since I wrote a one liner without really explaining anything...

One explanation people like to give is:

CSS Grid is for layouts in 2 dimensions, Flexbox is for layouts along one dimension.

Which can be confusing because you absolutely can use Grid for layouts along one dimension too.

Here is my stupid sounding rule of thumb that I think is more useful: If you want a grid layout use grid, if you want a flexible layout use flexbox.

Grid quite literally splits the area into a grid but what if you want to just say "I want these two elements to be as far apart as possible within the container they are in" (ie a navbar, you want the logo on the left about 20 pixels away from the left of the viewpoert and you want the actual navigation to be on the right with the rightmost element 20px away from the right of the viewport). Flexbox is better for that.

I do often end up using Flexbox for 1-dimensional layouts, but IMO Grid's fr units are a lot more intuitive than Flexbox's combination of flex-basis/flex-grow/flex-shrink. If CSS added a 1-dimensional layout mode based on these principles then I'd switch to it in a heartbeat.

(I'm considering proposing one myself).

What bugs me the most is Flex centering absolutely sucks at handling overflow. Have a flex box with content that is supposed to be centered? Great, looks perfect for the first few items. But then try it on a smaller window or with more items and the content overflowing to the left of the border will be entirely inaccessible.

Allowing scrolling both ways would have been perfect. Switching to left-aligned when content overflows would be acceptable. But completely prohibiting access to half the content when an overflow occurs? Absolutely asinine.

I believe there may be ways to get around this with layers of inner divs and weird `overflow: ` incantations, but that's exactly the kind of bullshit flex was supposed to solve.

See https://developer.mozilla.org/en-US/docs/Web/CSS/align-items. If you need to enable a "safe" mode to prevent data loss, and that isn't supported on any mainstream browsers... you fucked up.

Correct. but grid is simpler to take in on the face. flex needs a bit of understanding.

CSS grid feel like someone asked TikTok to design a rectangle and said yes to every suggestion.

What does this meab?

I think they mean it’s a complicated solution for an uncomplicated problem.

Grid systems existed long before CSS grid and were dead-easy to understand and use (960, Bootstrap etc). CSS grid, by comparison, is completely unintuitive. Sure, it’s definitely more powerful, but the barrier to entry is way higher.

Exactly. "Align things to a grid" seems like a simple concept, but CSS grid seemingly has at least 3 ways to place the top-left corner of an element, and it just gets worse from there. I tried it for a simple layout with 3 elements and quickly decided I prefer flexbox, tables, or even floats.

I think when you come across situations that only grid can solve, then you appreciate it a bit more.

Flexbox complements grid imo, it's not an alternative.

That's fair. I only use CSS for personal/hobby projects.

Rachel Andrew has a very good 3-part tutorial that helped me actually understand what I am doing: https://www.smashingmagazine.com/2020/01/understanding-css-g...

It will take you from a n00b to a Grid pro in no time.

This is a really surprising take for me, grid is by far the most sane of the layout models in my experience.

Most, even 2D layouts map easily into nested 1D flexbox in my head. I feel like a single dimensional approach to layouts is way easier to wrap my head around and design with.

The secret for me was working with something like tailwind etc and in a componentized framework.

the framework part isn’t mandatory, but I suppose very small tweaks with something like tailwind do the trick for me.

I can’t work with css if there are massive classes and nested stuff etc. I guess composition over inheritance

It’s because CSS is a constantly moving target of massive conceptual differences from the last one. Plus all the tooling changes frequently as well.

What options exist today for making CSS easier? (It's been a long time since I've done front-end work)

Seems like there would be some sort of layer on top of it that someone built that makes things more consistent and then will compile down to the inconsistencies and odd behaviours for how CSS is supposed to be implemented.

There's supersets and siblings like scss, sass, less, there's different modes entirely like utility classes (Tailwind) and css-in-js.

I don't think any of those make CSS much easier, though. Quicker to use, easier to write certain things, sure. But you still need to know your way around CSS atleast somewhat before you jump in to some augmentation or replacement.

Same here. I think it's because you typically conquer a language by developing a fairly complete mental model of that language's behavior, but with CSS there are just too many nooks and crannies to get your had around. While it's uncomfortable for me, I find that I do best with CSS when I take a more practical approach and just try to use best practices to accomplish what I want without having the deep understanding that I crave.

I've found that Kevin Powell [1] is a great resource for this approach, and he also helps with the model model aspect as a bonus.

[1] https://www.kevinpowell.co/

I found that Tailwind helped me get more comfortable with CSS. If you put the mechanics of using classes aside, it’s essentially just a well-curated subset of CSS with great documentation. Way more approachable than the sprawl of CSS itself.

The "fun" part of being a frontend dev who naturally gets pushed into fullstack as part of the realities of the job is I have to keep on top of developments in CSS, UI frameworks, Typescript, Vue, testing frameworks, Ruby, Rails, etc. Not to mention the build tooling (Vite/webpack) and deployment / operating system stuff. Fortunately the database stuff rarely changes with Postgres.

I wish I could specialize on any one of these sometimes...but such is life at a startup. It's more interesting being able to do them all anyway and be able to build entire apps yourself. Just occasionally frustrating.

CSS seems like it was “designed” by several teams in an “agile” environment with no plan and simply didn’t consult the people that would actually be using it.

Is there another layout language/engine that you've found more intuitive and flexible?

After a couple hours of using xaml I felt like I learned something. But I didn't spend more time on it.

I've developed apps using XAML for a considerable amount of hours. XAML is significantly more restrictive than CSS. Windows UI toolkits will give you a bunch of pretty high level components, but the primitives for build anything else are extremely poor. It lacks something like flexbox or grid.

Last time I investigated having a uniform grid of items is impossible without falling back to custom C# layout code.

I'm the same. I just use ChatGPT to fix my layouts now.

Colleagues have suggested the same. My strategy has been to try to avoid frontend and when I must do it I make sure everyone involved understands my limitations.

Try tailwind. It can help you get better at CSS itself.

Try XSLT. The only programming language that ever defeated me and I've used quite a few including Haskell.

I found XSLT kind of mind-bending, but also extremely satisfying and quite productive. Like SQL, it’s declarative, so you have to get into the mindset of telling the computer what you want, not what to do.

You may be a smarter person than me, but I couldn't handle it and quite possibly that was partly related to its truly appalling XML-based syntax. I'd like to point out that no other language I'm aware of is so syntacticallyvile that they have to add a new mechanism just to pass parameters around – parameter tunnelling.

https://www.w3.org/TR/xslt20/#tunnel-params

Oh Lord. You’ve just triggered a horrific memory of how in my first job we used XSLT to convert xml files of job applications into sql that was then executed to import them into the db. That truly was mind bending.

XSLT is a functional language. I say writing DTD and stuff like that is harder than XSLT. XSLT is actually pretty fun to write.

If I have a reason to I will. Im generally not motivated to learn things to pad my ego, I need a more tangible reason.

How often do you read the CSS specs?

From start to finish, never. When I had a problem I googled what to do and read the specs of the suggestions that I saw to try to make sense of them. Usually they were not helpful.

Reply on news.ycombinator.com