A couple CSS tricks for HTML Dialog elements | Hacker News Reader

A couple CSS tricks for HTML Dialog elements

cassidoo.co

163 pointssurprisetalk45 comments

datalist is one I stumbled upon and blew me away. https://developer.mozilla.org/en-US/docs/Web/HTML/Element/da...

It's not a replacement for select as you still need an input to tie it to but it seems to handle filtering a list of options nicely.

Also, if you have two selects with the same list in it, you can do it once with datalist and have two inputs, say a list of clients with client_a and client_b for inputs.

I don't quite care for how it displays the value, like if you put the ID as the value and the client name in the option element, you can filter by the ID or the name but the input will show the ID only.

My frown continues to deepen at Apple's UI backslide as they crib more and more junk from iOS/iPadOS. I'm on Sonoma 14.5, and this is Safari 17.5.

macOS Safari's <select>: https://imgur.com/a/05YWDCc

macOS Safari's <datalist>: https://imgur.com/a/4f3JwuA

There are SO, SO, SO many things wrong with Safari's datalist element here. Esc doesn't close it (close from keyboard by switching tabs...). There's no hover effect on the options. The active background color is more saturated than the system's accent color (typical for iPadOS/Catalyst junk). There's no left/right margin, and no border radius, on the options. Option text isn't vertically centered. The font is different (it seems differently aliased? Perhaps just larger). The datalist element itself lacks the same border-radius of select. On select elements, selection does not wrap (down arrow with the last option active); on datalist it does.

Here's an egregious one - when you zoom in with Cmd-+ a few times, this is how the <select> element looks: https://imgur.com/a/Vpu536j

And this is <datalist>: https://imgur.com/a/JrfXLW9

Argh! I used to revere Apple for sweating the details. Their UI/UX quality inspired me to become a frontend dev.

Today, they ship things that wouldn't pass Q/A at my worst jobs.

It does make you wonder, Safari recently had a burst in features where they modernised and even overtook Chromium/FF in some features, and then in the past year or so it’s languishing again.

I do wonder if the metrics show the average person downloads Chrome straight away so they’re just not investing heavily in it? I mean anyway, who browses traditional websites any more, right…?

They've purposefully underinvested in Safari to force developers to create native apps for their platforms where Apple makes a sizeable cut of all sales and subscriptions rather than allowing developers to create a web-app that could have done the same thing where the developers reap all the rewards for their work.

The only reason they had that burst of activity is that they needed to quickly catch up and save face in an attempt to prove to EU regulators that they weren't hampering developers.

The EU didn't buy it and forced Apple to open up their devices to allow alternate app stores and browsers on their devices in the EU.

> They've purposefully underinvested in Safari to force developers to create native apps for their platforms where Apple makes a sizeable cut

Can't speak to how accurate this is, but for WebXR, this hits the nail on the head. Purposeful stagnation on supporting it and thus indirectly bringing down the whole point of the standard, pushing of their own AppStore bound ARKit, and when they released Apple Vision Pro it's magically supported again, because I guess they needed content that badly.

Yeah, the more I dig into it the more I see it's not all that great. It has some potential for some things but not necessarily for an autocomplete list.

https://jsfiddle.net/nhu4zef2/

This is one occurs in every browser when if you want to have a list but send the ID for the item instead of the value, it shows the value in the list and you can search by the ID or the value but the result in the input shows just the ID. User's should be not be required to know the ID of something. Like say a list of clients where user's know them by name but necessarily by ID but the database links them by ID.

This just got so much worse with scrollbars.

Select: https://imgur.com/a/fi1SPBJ

Datalist: https://imgur.com/a/sTiQhPF

This looks exactly like an iOS control now. The multiply effect on the scrollbar is comically out of place.

I've tried using datalist with text inputs and it never quite worked out from a UX perspective. Users would always complain about weird quirks with how it populates & clears values. A normal <select> element with an "Other" option + conditional input element is much more predictable.

Yeah, for autocomplete it has a weird UX feel to it.

Some of the other examples like using it with the slider and the color picker seem like they're useful.

What a great find! I'll definitely be thinking through where this is appropriate to use.

It looks like the UI would clash with the browser's autofill recommendations.

Doesn't seem to work on Firefox Android

https://bugzilla.mozilla.org/show_bug.cgi?id=1535985

The trick to prevent scrolling by setting overflow: hidden unfortunately results in visual page jumping for me.

The reason is I have macOS set up to always show scroll bars, instead of hiding them. At least one browser (I forget which, but I test on Safari, Firefox and Chrome) doesn’t have a disabled scroll bar but removes it altogether. This makes the page wider and causes it to reflow and move.

Does anyone know how to keep the scroll bar onscreen, just not enabled?

It's been a while since I've tested this, but an explicit overflow-y: scroll used to keep the scrollbar there, so that when you needed to change the property, the user-agent controls wouldn't pop in or out.

yes, "html { overflow-y: scroll }" works.

Website authors can set a scrollbar gutter.

scrollbar-gutter: stable; to those unfamiliar. https://developer.mozilla.org/en-US/docs/Web/CSS/scrollbar-g...

Man, this makes me feel old.

This is only recently supported by all major browsers thanks to the interop efforts. pre-2024 browser versions will not support this

Freezing the page isn't so simple, as overflow: hidden messes up things like the sticky header on that page. I had so much trouble with it, I decided to just let users scroll and hide the modal during scrolling: https://radogado.github.io/n-modal/

This is a neat piece of modern CSS:

  body:has(dialog[open]) {
    overflow: hidden;
  }

https://caniuse.com/css-has confirms the has() selector has had widespread browser support since December 2023.

YMMV / be careful with this, body:has() and html:has() can be extremely expensive (and introduce severe lag visible to the user) if you have dynamic components on the page that are constantly altering the DOM (ex. react/vue apps)

Inert should be used instead of overflow. Achieves the same thing but is also compliant with accessibility in a way overflow isn’t

I still can see my scrollbar and scroll with inert?

Its intended to stop interaction[0] of background elements. It can be used as part of the solution to stop the background scrolling.

Per MDN When implementing modal dialogs, everything other than the <dialog> and its contents should be rendered inert using the inert attribute.[1]

`body[inert] { overflow: hidden; }`

This would be better, and is what I was getting at. I can't edit the other comment unfortunately.

[0]: https://developer.mozilla.org/en-US/docs/Web/HTML/Global_att...

[1]: https://developer.mozilla.org/en-US/docs/Web/HTML/Element/di...

Your original message was "Inert should be used instead of overflow" which is incorrect because `inert` doesn't affect scrolling of the viewport. Your CSS rule example is a good way to demonstrate using the presence of an `inert` attribute on the body to determine when `overflow: hidden` should be applied to the body.

That section of the MDN article is somewhat confusing but if the dialog is opened using the `.showModal()` method, there's no need to add an `inert` attribute yourself, the browser automatically makes the rest of the page inert.

If a <dialog> that's meant to be modal is opened not using `.showModal()`, say by making it a `popover` and the `popovertarget` of a button, then you might set `inert` yourself (and remove it when the <dialog> is closed). However, you can't simply do <body inert> if that <dialog> is inside the <body> because then the dialog itself would be inert.

I was on mobile. I apologize my comment was insufficient

I used this same approach in a recent web app and it worked great. You can also use scrollbar-gutter: stable, which disables scrolling but maintains the preserved space to avoid content reflows.

Be careful using:

    dialog::backdrop {
        backdrop-filter: blur(2px);
    }

If there is frequently updating content on the page like a video, it can kill CPU performance. Restream does this and it punishes my M3 macbook air.

It seems you can use the transform3d trick to kick it to the GPU to help fix it.

I'd want to pause any video playing while a dialog with such a backdrop is open.

In the case of Restream (or cases like their "stage"), thats not always doable. What about superfluous hero animations that marketing sites love to do these days? Pausing some generic canvas rendering logic is usually more annoying than you would want.

Is it weird that I expected the post to actually have running examples?

From the article:

> If you want to see a decent quick example of them in action, you can check out my game Jumblie and click the Settings gear button at the top.

It has the backdrop filter but it doesn't prevent page scrolling.

BTW, MDN's data on Safari's support for the unprefixed `backdrop-filter` property is wrong, it still sometimes requires using `-webkit-backdrop-filter` (works in iOS Safari 18.2.1, doesn't work in Safari 18.2 on macOS 14.7.1).

I have read the article, and saw the link, I simply thought it is so simple to actually add an example since the post itself is a web page.

That would require adding JavaScript to the page and some people don't do that in their blog, especially when it's published using a static site generator.

A demo of each trick would be nice to see

Thanks for the article!

I remember I had a hard time trying to stretch dialog to full screen for mobile devices, but it actually didn't want to work. The code was something like this:

dialog { position: absolute; top: 10px; right: 10px; bottom: 10px; left: 10px; }

You can shorthand the last four declarations with a single `inset: 10px;` (or maybe `inset: .625rem;`?

That's cool, actually I didn't know about this one :)

this is most likely due to the absolute positioning. position: absolute will use the top-left corner of the closest ancestor that is "positioned" as the origin for it's layout [1]. If you want that origin to be the top-left corner of the viewport, use position: fixed.

[1] https://developer.mozilla.org/en-US/docs/Web/CSS/Containing_...

In addition to `position: fixed`, shouldn't it be top, left, height, width, instead of top, left, bottom, right? In the second case, it would follow the top and left instructions then take the necessary amount of space, ignoring right and bottom?

For absolutely positioned elements, if you set all four sides then the height and width will be automatically computed.

OT: I really want to see marquee tag have native UI scrolling via CSS property. The many websites that use this horizontal scrolling layout with JS makes me want to cry.

160 likes for this?

Reply on news.ycombinator.com