- tabindex=0 doesn’t affect ordering, does it?
- why do you need to listen for events at the document level?
not that i disagree with the article, but some arguments didn’t seem right.
replies(2):
Of course it does. tabindex=0 doesn't sort naturally into the automatic tabindex order, it sorts AFTER everything. So you are jumping through all the other tabindex elements, then you are jumping back to all tabindex=0.
The issue isn't with tabindex=0 specifically, but fucking with tabindex in general. People go down that path, and start putting that shit on everything, like it's Frank's Red Hot.
And in my experience, the same folks who use div's instead of button's are the ones who don't know better and start throwing tabindex around.
"why do you need to listen for events at the document level?"
Not events generally, keydown events specifically, which do not fire on child elements of the document.
It's generally advised not to set tabindex to anything but 0 or -1 and let the document order dictate tab order.
https://developer.mozilla.org/en-US/docs/Web/API/Element/key...
Yes, but often you have elements with taborder > 0.
> It's generally advised not to set tabindex to anything but 0 or -1 and let the document order dictate tab order.
Only if document order is sane. Usually with modern websites it isn't, document order is a broken notion if you can position elements at will and e.g. put navigation at the bottom of a document but move it to the top by CSS. Which is actually a recommendation that some people make for acessibility...
What you usually want to do is assign a sensible taborder > 0 to the one form element that the user is probably currently using. Otherwise, he will pointlessly tab through search, menus, cookie bars and a ton of other pointless stuff first.
Are you sure? I have a 17 year old HTML tool written using plain, vanilla JavaScript where keydown on a child element seems to have been working as expected.
https://susam.net/quickqwerty.html
https://github.com/susam/quickqwerty/blob/1.2.0/quickqwerty....
Nice article, by the way!
You can just as easily apply the same tabindex to a div though.
> Only if document order is sane. Usually with modern websites it isn't...
Well that's the real problem, all your non-interactive content (like text) is going to be out of order too. You're just adding to the confusion if buttons and other inputs are in a different order from the content they're associated with.
> Otherwise, he will pointlessly tab through search, menus, cookie bars and a ton of other pointless stuff first.
The proper way of dealing with this is a "Skip to Main Content" element:
See this fiddle https://jsfiddle.net/483uqjnp/
(again, I do not condone building your own <button>, just pointing this out)
No, it isn't the proper way. That only works if you can see the skip link and know to press enter. Otherwise you will tab straight into the navigation. So possibly useful for screen readers, but completely useless for most keyboard users. Yet another stupid webdev workaround for a selfimposed problem.
What you should do is autofocus the first form element (if there is a form), give it tabindex=1 and number the other form elements in a sensible ascending tabindex order. Otherwise, proper semantic markup is sufficient, even for screen readers.
You mention using it to skip a nav header or something similar, but (1) you have autofocus for that if you need it (and you don't usually need it), and (2) the pattern where you provide a "jump to content" button is typically a better approach, because it allows the user to decide what they want to do when they arrive on your site, rather than you deciding for them. If the "jump to content" button behind visible when focused and is the first element on the screen, then it works perfectly for screen readers and other keyboard users, and you don't need to handle taborder manually.
There are always exceptions to these sorts of rules, and some times tabindex might be necessary, but I've not yet come across a case where the problem couldn't be solved better in a different way.
And yes, this is an acceptable solution according to the W3C:
https://www.w3.org/WAI/WCAG21/Techniques/general/G1
Your solution of focusing the first form element is pretty idiosyncratic. It's better to follow WAI patterns, because patterns have predictable behavior. Otherwise, keyboard users will have to learn how to interact with your website from scratch, instead of just following the same pattern they're used to from other sites.
Your original comment was to use tabindex to skip search, menu bars, breadcrumbs, etc. and for that there are better options.
- It only counts as "fucking with tabindex" if you give it a value that's not 0 or -1. You should give that specific disclaimer, because there are uses for tabindex=0 other than reimplementing <button>.
- Divs can definitely receive keydown events. If I go to an arbitrary web page, pick a div and run `div.tabIndex = 0;` + `div.addEventListener('keydown', console.log);`, I see those events coming through when I have the div keyboard-focused.
- "Run your code, somehow..." I think just calling `notRealBtn.click()` is the best option.
- Stupid but semi-interesting nitpick: 'keydown' is good for enter, but you should be listening to 'keyup' for the space bar. That's how real <button>s work anyway.
- The 'keyup' listener should call event.preventDefault() to prevent the default behavior of the space bar scrolling the page.
> So you are jumping through all the other tabindex elements
This part is correct (for elements with an explicit positive tabindex), which is why specifying an explicit positive tabindex is considered a code smell. If you don’t specify a tabindex on an element that’s focusable by default, it behaves like tabindex=0.
Try it:
data:text/html,<button>foo</button><i tabindex=0>bar</i><button>baz</button>I can imagine screen readers would deal particularly poorly with this behaviour, because the user would be dropped in a field without any of the context of the page, just the field label to work with. However, I've not been able to test that out properly.
I don't think the behaviour you're describing is anywhere near as common as you think it is, and I suspect it would make a page less accessible for a number of kinds of users.