When should a “hidden” element be available to a s

问题:

I hear of advice of making the .hidden class not as a

.hidden { display: none }

but make it width and height of 1, and using clipping, etc, to make the element as if it still exist on screen but the content is just not visible.

But isn't it true that when we use JavaScript to hide something, the purpose of that element is done, and we want it to not show on the screen and not visible to the screen reader either?

Case 1: For example, if it is an input box, and inside the box there is a grey line of text "Enter keyword". When the user actually clicks on, or keydown, (or using the input event), now we have a JavaScript handler to hide the "Enter keyword" text, because it is just a grey prompt inside the input box. In this case, shouldn't the text be completely hidden, using in fact a display: none, so that even screen reader can't read it? (the user have learned enough to start typing, so the prompt shouldn't really still be there to be read by screen reader, right?)

Case 2: If it is a pop up bubble, for error message, or there is a link "enter your email for our subscription", and it will pop up a bubble, then when the bubble is closed, shouldn't the bubble be really closed totally using display: none? The screen reader shouldn't really still be able to read those content out for a bubble that is done and closed.

Case 3: The only case I can think of is a small fraction that the "hidden" should be available to a screen reader is: it is for a bubble for extra information, such as for the product rating (how many stars out of 5), or extra help info that can pop up when the mouse hover over some "?" icon or link. But even in this case, won't screen readers actually read out "link to more info" or "link to show rating", that is in the alt or title of the tag, and not read the pop up info until the user pops it open?

So my question is: shouldn't there be really two types of hidden. One is display: none type that screen reader shouldn't see (make it the .hidden class), and the other is width: 1px; height; 1px that the screen reader can see (or will read out) (make this the .a11y-hidden class), and whether type 2 might be far less often than type 1?

回答1:

Mostly, that's it.

EDIT 2014: I switched to the clipping method from TJK at Yahoo! (because of hidden span into a link where only an <i>con is visible).
I prefer position: relative; left: -5000px over clipping 1x1px, I know that Yahoo! team only use the latter but the aim is the same.
Also I'll call it .visually-hidden (from Twenty Sth theme in WordPress and other CMS and frameworks). .a11y-hidden would mean the contrary: it's only visible (perceivable) by screen readers and in plain HTML.

Tabs content, error messages that didn't occur yet (your Case 2) shouldn't be perceived by anybody. A screen reader user will have to click on a tab to show its content as everybody else.
Note: display: none AND visibility: hidden elements will be ignored by screen readers. And also some of them (VoiceOver on OS X in particular) will ignore an element with height: 0, etc

Relevant article : Invisible Content Just for Screen Reader Users from WebAIM

Case 1: the important part being read out is the label associated with for/id to the form field.
If you use a keyboard to navigate from field to field, this text will have disappeared when you'll have focused it and it won't be visible to users that zoom at x00%. HTML5 introduced the placeholder attribute that has the same role as this prompt (this is NOT a replacement for label, this is a hint but few people read one of the HTML5 doc alas) and can still be passed to screen readers even if it's visually hidden.

Case 3: this information should be coded in HTML in an accessible manner and so you wouldn't need to visually hide information out of the viewport or in a pixel.
alt text is the preferred manner (or even better a visual with some real text by its side, while avoiding too much clutter. I've no training in ergonomics and this will often annoy webdesigners that want to remove everything from a page ;) ).
title attribute should only be used on links and you're never certain a screen reader will choose to read them. It's a personal matter and setting on a per site basis. They can be so annoyed by some sites that they'll disable these titles everywhere... Though, it's still one of the best techniques to add information if for some reason you can't do it otherwise.

How do I use this .visually-hidden class?

• quick access links (to content, navigation and search input) when they don't appear in the mockup. When focused, they'll be brought to screen as in http://www.nanomatrix.fr/ (press Ctrl-L or Cmd-L, tab half a dozen times on Windows and enable tabbing through links in Safari and/or OS X. See on upper left the 3 links)
• on labels when for some (bad) reason, they're not in the mockup I receive. Better have a fix (it's a good one) than doing nothing to improve accessibility when it wasn't thought of beforehand...
• on captions of tables, because most of the times showing them is a no from client, external webdesigner, etc
• on a span with information next to a font icon when it's coded in a performance manner but not very accessible. There are many ways of using a font icon, I'm only speaking of a few use cases

回答2:

Theres a lot going on here that needs to be addressed.

Firstly, and ideally, the grey text in the input box should be handled by the 'placeholder' attribute. The browser will give the user a consistent and expected behaviour, and its also easier for you to code!

Secondly, for accessibility ideally all inputs should always have labels. This is so that the input is described correctly to screen readers. You can hide them if the design doesn't call for them (some websites only use placeholder text, but that doesn't mean they shouldn't) be there) - which brings me to my third point - hiding content on screen but not from screen readers:

{display: none} will hide information on the screen AND from screen readers. It's basically a way of making something disappear - for everyone. This is where the .visuallyhidden class (or similar) comes in - its a way of adding context (extra information) to the DOM that will be picked up by screen readers but not read by sighted users and it's a technique used very often to make pages more understandable to AT users without breaking a layout or design of a page.

Its a good idea to use a tried and trusted class, such as the one inside the HTML5 boilerplate, as getting it wrong could penalise you on Google (for trying to blackhat content) or create problems with some screen readers on some devices (e.g. iOS) - position:left and text-indent is not a good solution for a number of reasons, so stick to the 1px and clipping technique.