Chris’ Corner: Tricks With CSS

There are plenty of very legit reasons you’d want to have a scrolling element start out scrolled to the bottom, and stay scrolled to the bottom (as long as a user hasn’t scrolled back up). As ever, you could do this with JavaScript, as JavaScript can adjust scroll positions of elements. There is a way to do this primarily with CSS now that the anchor-overflow property exists, and I think it’s an extremely great CSS trick.

There is another way though! Kitty Giraudel covers it in CSS-only bottom-anchored scrolling area. The base of the trick is quite simple and requires no additional elements. You just set flex-direction: column-reverse; and then put the HTML inside in reverse order. So the stuff you want pinned at the bottom visually you put at the top of the element. In a way, this makes sense to me, as the thing you want to read first is at the top.

Element with scrolling pinned to bottom (as long as you add the stuff at the visual-bottom to the top of the DOM). Think of a chat interface.

But there is an accessibility concern that Kitty notes. It “creates a disconnect between the visual order and the DOM order, which can be confusing for screen-reader users.” I’d want to verify that with a screen reader user I think (probably applies mostly to people who use a screen reader and have some vision). But it’s a good point and a classic problem that comes up any time you use CSS to position things in such a way they appear visually differently than the source order suggests. I’m sure you can imagine the akwardness of focus states jumping around the screen unpredictably.

The thing that makes all this so news-worthy to me is that CSS is working on a solution for this that I didn’t know about:

reading-order: normal | flex-visual | flex-flow | grid-rows | grid-columns | grid-order

In our case, we could use reading-order: flex-visual to align the way sighted users and screen-reader users consume our feed.

So we’ve reversed the order using flexbox, but we can make the elements still read top-to-bottom (visual order) by forcing it with this property. I might argue, again, that in this case, users might want to read bottom-to-top. But at least you’ve got options now.

And this reader-order stuff is generally interesting. Like if you use flexbox and totally mess with where the flex items are placed with the order property, placing, for instance, the 7th item in the 2nd place, and the 19th item in the 1st place, updating the reading order to flex-visual will be great. I notice there is no grid-visual though, which is curious, since you can mess with the order of grid just the same.


Jonathan Snook has a play with the idea of lenticular cards. Those are those ridged plastic novelty cards that have two different images you can see depending on the angle you look at it from. Or more!

Since Apple released Live Photos, I’ve always felt like they could be used to create a similar effect and yet, no photo app that I’ve seen has implemented it, from what I’ve come across.

I enjoyed playing with the demo on mobile (where the DeviceOrientation API is a thing):

I love the experimentation spirit here. Like thinking of something you think should exist, but doesn’t seem to in an obvious way, then building it anyway.

Yair Even Or had the idea that a box-shadow could be cool if it… wasn’t actually a shadow, but was a blur instead.

The implementation is that perfect tornado of cleverness that appeals to me. It’s not incredibly complicated, but it requires usage of a number of different CSS features that you not think about immediately. In the end, it’s:

  • Place a pseudo element behind the element, a specified amount larger than the original element.
  • Blur the background with that pseudo element using backdrop-filter.
  • This doesn’t “fade out” the effect like a box-shadow would naturally, so instead, two masks are used to fade out the effect (vertical and horizontal).
  • Mask compositing is used to combine the masks.

I think the two masks are needed because of the rectangular nature of the element. I’d be tempted to try it with a single radial-gradient, but I think you’d lose blurring near the corners.


Dan Wilson always does a good job looking at new CSS features and the possibilities they unlock. Particularly the new features that are a bit esoteric, or seem to be at first glance, like math functions.

In The New CSS Math: pow(), sqrt(), and exponential friends, Dan looks at those new CSS functions (and a few more), and point out some somewhat practical things they can do. For example, a typographical system where the header sizes aren’t a straight multiple of one another, but are grown on a curve. Or simulating an easing effect by animating a number linearly, but having the movement distance calculated by a pow() on that number. There is even a function now that makes quick work of the Pythagorean theorem.

If you’re into this stuff, Dan looked at rem() and mod() here, which are similar methods for determining what is left over when you divide a number into another number. Is 9 divisible by 3? Yes, and you can know if the remainder is 0. But in web design, you could do things like figure out how many 125px grid columns could fit into 693px of space, if you needed to.

Dan has looked at trig functions as well, and shortly after that, Hypersphere looked at simulating randomness in CSS with those functions. The sin() function, for example, modulates from -1 to 1. So by farting around with that and incorporating a seed value, you can build pretty darn random looking CSS output:


The can’t-miss link recently is Ahmad Shadeed’s An Interactive Guide to CSS Container Queries. His interactive guides are always outstanding. This one is full of practical examples of where container queries are useful.