Can we use a different <dl> for each <dt>, <dd> pairs? Will there be any problems with accessibility if we do it? Use-case being to simplify some Component API. For example:
<!-- Can we do this -->
<div class="container">
<dl>
<dt>First Name</dt>
<dd>Jeff</dd>
</dl>
<dl>
<dt>Last Name</dt>
<dd>Bezos</dd>
</dl>
</div>
<!-- instead of this? -->
<div class="container">
<dl>
<dt>First Name</dt>
<dd>Jeff</dd>
<dt>Last Name</dt>
<dd>Bezos</dd>
</dl>
</div>
In this case, it doesn't matter because definition lists suck in general for accessibility purposes. See my (editorial) comment in this answer, How to properly use aria selectors for definition lists in puppeteer
A definition list does not have a "list" role by default so different screen readers interpret them differently. Because of this, whether you have separate <dl>s or not won't have a huge impact.
"Real" lists, <ul> and <ol>, do make a difference. If you have a group of things that are related and are contained in one list, then it should only be one <ul> rather than a series of <ul>s. A screen reader conveys metadata information about lists such as the number of elements in the list. You get that information "for free" with your eyes if you're sighted. You can generally tell how many items are in a list (if it's not too long) whereas a screen reader user doesn't have that "seeing" option so the screen reader can tell them it's a list with 5 items, but only if it's a single <ul> with 5 <li> elements.
Let's say you had something like a TV show listing, where you had a show title, and a show description. You want the listing to be accessible for people with disabilities as well.
Would it make more sense to use a definition list:
<dl>
<dt>...title...</dt><dd>...description...</dd>
...
</dl>
Or an unordered list with headings?
<ul>
<li><h3>...title...</h3><p>...description...</p></li>
...
</ul>
Which makes more semantic sense and will respond better to screen readers? (knowing that they can both be styled the same way)
If you are using HTML 4.01, you shouldn't use dl as it's defined as "definition list" (and your example does not consist of terms and their definitions). If you are using HTML5, the use of dl is fine, because the definition of dl changed.
Using headings inside of li might be a bit problematic regarding the document outline. The scope of a heading would include the start of the next li: <li><!--scope start--><h3>title</h3><p>description</p></li><li><!--scope end--><h3>…. By using section (resp. article), this could be avoided.
So, for HTML5, I think the following ways are possible:
dl
<dl>
<dt>Title1</dt>
<dd>Description1</dd>
<dt>Title2</dt>
<dd>Description2</dd>
</dl>
That would be my favorite, if you only want to provide title and description for each show (if not, see the last example).
ul + section
<ul>
<li>
<section>
<h1>Title1</h1>
<p>Description1</p>
</section>
</li>
<li>
<section>
<h1>Title2</h1>
<p>Description2</p>
</section>
</li>
</ul>
I don't like that very much. The list isn't adding much here, so why not omit it? (see next example)
headings only
<section>
<h1>Title1</h1>
<p>Description1</p>
</section>
<section>
<h1>Title2</h1>
<p>Description2</p>
</section>
Instead of section the article element might be possible, too.
You could also omit section (or article) and use headings only (in the case of section it wouldn't change the meaning); in that case you'd need to apply the correct heading level.
headings + dl
If you want to provide additional metadata (maybe in the future), I'd go with the following markup:
<section>
<h1>Title1</h1>
<dl>
<dt>Description</dt>
<dd>…</dd>
<dt>Rating</dt>
<dd>…</dd>
<dt>Time</dt>
<dd>…</dd>
<dt>Length</dt>
<dd>…</dd>
</dl>
</section>
I prefer the former. First, it seems to make more sense to me just based on the content.
But that's me. I think the markup should reflect the document structure, and since (as you say) the CSS can style it either way, why not make the markup reflect the content? A list containing items that contain a header for a title, followed by a description seems a bit of overkill to me.
But, hey. You know what they say about opinions.
In this case using a Definition list makes much more sense. Aside from this though, is it really necessary to use a list at all? It may make more sense just to use your Heading tags appropriately on the page wit a tag (x= 2-6) and have everything apply under the header of that. TV Shows in specific it may not make sense to use a "List" to display them with definitions or anything else. Again, they can be styled however, so i'm only worried about sematics with this.
Hope this helps
Zach
This is not another general 'tables versus div elements for general layout' type question, like the "why not use tables for layout" question.
I'm working on a timetable/calendar project and I have always assumed that a calendar would be an example of when to use to a use a table. Although, after a quick look at Google Calendar's structure, it seems it consists of a table, containing a <td> for each column and within each column, an event is a <div> with definition list inside.
Why is this beneficial?
My own ideas:
Tables may be more troublesome to style, nicely & compactly, when there are multiple varying length events beginning at the same time (beginning in the same <td>). Possibility of unwanted whitespace.
Harder to update tables when user adds event after the page is loaded, e.g. with JavaScript (because the row/colspan of the table headers might have to change)
If tables were used, the width of x-axis/top headers & cells, and the height of y-axis/left headers and cells, would be matched automatically. Could be tough to manage this without tables.
Does any of this matter? Should tabular data always be stored in actual tables?
The following is a simplified example of a Google Calendar column:
<td> <!-- column -->
<div> <!-- start event -->
<dl>
<dt>START TIME – END TIME </dt>
<dd>EVENT TITLE</dd>
</dl>
</div> <!-- end event -->
</td> <!-- end column -->
The following is a full example:
<td class="tg-col"> <!-- column td -->
<div id="tgCol0" class="tg-col-eventwrapper" style="height:1008px;margin-bottom:-1008px;"> <!-- column div -->
<div class="tg-gutter">
<div class="ca-evp130 chip " style="top:588px;left:-1px;width:100%;"> <!-- start event div -->
<dl class="cbrd" style="height:35px;border-color:#9FC6E7;background-color:#E4EFF8;color:#777777;">
<dt style="background-color:;">START TIME – END TIME <i class="cic cic-dm cic-rcr" title="Recurring event"> </i></dt>
<dd><span class="evt-lk ca-elp130">EVENT TITLE</span></dd>
<div><!-- start masks -->
<div class="mask mask-top" style="border-color:#9FC6E7;background-color:#E4EFF8;"> </div>
<div class="mask mask-bottom" style="border-color:#9FC6E7;background-color:#E4EFF8;"> </div>
<div class="mask mask-left" style="height:38px;border-color:#9FC6E7;background-color:#E4EFF8;"> </div>
<div class="mask mask-right" style="height:38px;border-color:#9FC6E7;background-color:#E4EFF8;"> </div>
</div><!-- end masks -->
<div class="resizer"> <!-- start resizer -->
<div class="rszr-icon"> </div>
</div> <!-- end resizer -->
</dl>
</div> <!-- end event div -->
</div>
</div> <!-- end column td -->
<div id="tgOver0" class="tg-col-overlaywrapper"></div> <!-- column overlay div -->
</td> <!-- end column td -->
Edit:
Don't forget to include why Google Calendar is structured as it is, e.g. why does Google Calendar have a table but only use it for the columns?
Personally, I'd go with div's instead of tables. That's not to say table's are entirely wrong, it's just that div's can be much more flexible when it comes to styling them, especially if you're adding other elements (such as a meeting that might span 2 dates) etc.
Div's would also help in a fluid layout more so than a table might.
It both is and isn't tabular data, I guess it depends on how far you're taking it's functionality, and layout.
I think there are two main reasons to use a table for displaying calendars:
Rows with variable height cells are simpler in tables, though the same thing can be accomplished with floats and clearfixes. The table approach is more likely to not break (and certainly more compatible with ancient browsers), and is probably more efficient for the browser to render.
The flow of multi-day events is very hard to manage with CSS, but it's pretty simple through the use of colspan (even if it does produce relatively hideous and non-semantic markup).
(I'm asking a question elsewhere on StackOverflow wondering whether there's a reasonably robust and elegant way to achieve these ends without tables. See: HTML markup for multi-day calendar events)
Adam, that's a fantastic question.
I think a calendar is the perfect use for tables actually. You're right, tables are harder to style in most senses, but when you think about the real issue, it's what you prefer. Sure, you could technically build a lot of modern websites with tables rather than divs and it'd be really tough, but there's a time and a place for everything. It comes down, in my opinion, to preference and if you can write something with less markup, then that's what you should use... even if it's considered bad practice by modern standards.
My vote is for something as square and unchanging as a calendar... go with tables if that's the cheaper solution.
Let's say I have the following layout on some pages:
Title: Some Title
Author: Some Author
Author Date of Birth: Date of birth
Notes:
Left side text is right-aligned and bold.
Used on several different pages.
The right side can contain input controls at some point.
What would be the most appropriate method to apply in this situation? I can think of a few options (assume CSS applied in external style sheet):
Table
Simple, easy, but I'm not sure this would be considered a good use of tables.
<table>
<tr>
<td>Title</td>
<td>Some Title</td>
</tr>
</table>
Div + Classes
I feel like this is a case of divitis and classitis rolled into one.
<div class="information">
<div class="title">Title</div><div class="value">Some Title</div>
</div>
Container Div
This feels more like the right path but I'm not sure.
<div class="information">
<strong>Title</strong> <span>Some Title</span>
</div>
Suggestions?
I think a good semantic choice here is the dl (description list) element.
http://developers.whatwg.org/grouping-content.html#the-dl-element
<dl>
<dt>Title</dt>
<dd>Some Title</dd>
<dt>Author</dt>
<dd>Some Author</dd>
<dt>Author Date of Birth</dt>
<dd>Date of birth</dd>
</dl>
Use a table, this is one of the few instances where using a table actually isn
't all that wrong. You're not using it for layout but text markup.
Then apply a class to every first column and in css make that class have text-align: right; which applies to that column.
I think you're right in you saying you want to use a table, but don't want to use. In this case i don't think a table is correct either. I personally only use tables if i need to organize data nicely. Because you have so much, a bunch of floated divs everywhere is more of a hassle then just using a table.
Because this is only two columns I would say use use two divs with floats or use two spans, instead of strong use a span and then style it with css.
This is clearly a good example of when to use tables.
It's tabular data.
Use ths for the first columns cells to be able to style it.
I even think ths are bold by default. Not sure about all browsers though so wouldn't hurt to style them bold to be sure :).
A table is a means of arranging data in rows and columns.
And this is what you are doing.
Definition list would be semantically correct.
<dl>
<dt>Title:</dt>
<dd>Some Title<dd>
<dt>Author:</dt>
<dd>Some Author</dd>
<dt>Author Date of Birth:</dt>
<dd>Date of birth</dd>
</dl>
See W3C for more details - http://www.w3schools.com/tags/tag_dl.asp
This is a question I have been struggling with for a while. What is the proper way to mark up name/value pairs?
I'm fond of the <dl> element, but it presents a problem: There is no way to separate one pair from another - they have no unique container. Visually, the code lacks definition. Semantically, though, I think this is the correct markup.
<dl>
<dt>Name</dt>
<dd>Value</dd>
<dt>Name</dt>
<dd>Value</dd>
</dl>
In the above code, it is difficult to properly offset the pairs visually, both in code and rendered. If I wanted to, for instance, but a border around each pair, that would be a problem.
We may point to tables. It could be argued that name-value pairs are tabular data. That seems incorrect to me, but I see the argument. However, the HTML does not differentiate the name from the value, except in position, or with the addition of class names.
<table>
<tr>
<td>Name</td>
<td>Value</td>
</tr>
<tr>
<td>Name</td>
<td>Value</td>
</tr>
</table>
This makes much more sense from a visual standpoint, both in code and in CSS. Styling the aforementioned border is trivial. However, as mentioned above, the semantics are fuzzy at best.
Thoughts, comments, questions?
Edit/Update
Perhaps this was something I should have explicitly mentioned in relation to structure, but a definition list also has the problem of not semantically grouping the pairs. The ordering and implicit border between a dd and a dt is easily understood, but they still feel slightly off to me.
Thanks for this interesting question. There are few more things to consider here.
What is a pair? Two elements together. So we need a tag for this.
Let's say it is pair tag.
<pair></pair>
The pair contains the key, and the corresponding value:
<pair><key>keyname</key><value>value</value></pair>
Then, we need to list the pairs:
<pairlist>
<pair><key>keyname</key><value>value</value></pair>
<pair><key>keyname</key><value>value</value></pair>
</pairlist>
The next thing to consider, is the display of the pairs.
The usual layout is the tabular one:
key value
key value
and the optional separator, which is usually colon:
key : value
key : value
The colons can be easily added via CSS, but this certainly won't work in IE.
Case described above is the ideal one. But there is no valid HTML markup to fit in this easily.
To sum up:
dl is semantically closest, for simple cases of key and value, but is hard to apply visual styles
(eg. to display the pairs inline or to add red border to just hovered pair). The case which fits most for dl is glossary. But this is not the case we discuss.
The only alternative I can see in this case is to use table, like this:
<table summary="This are the key and value pairs">
<caption>Some notes about semantics</caption>
<thead class="aural if not needed">
<tr><th scope="col">keys</th><th scope="col">values</th></tr>
</thead>
<tbody class="group1">
<tr><th scope="row">key1</th><td>value1</td></tr>
<tr><th scope="row">key2</th><td>value2</td></tr>
</tbody>
<tbody class="group2">
<tr><th scope="row">key3</th><td>value3</td></tr>
<tr><th scope="row">key4</th><td>value4</td></tr>
</tbody>
</table>
One more:
<ul>
<li><strong>key</strong> value</li>
<li><strong>key</strong> value</li>
</ul>
or:
<ul>
<li><b>key</b> value</li>
<li><b>key</b> value</li>
</ul>
or, when the keys may be linked:
<ul>
<li>key1 value</li>
<li>key1 value</li>
</ul>
The key and value pairs are usually stored in database, and those usually store tabular data,
so the table would fit best IMHO.
What do you think?
Following the specification (and further details) provided by Alexandr Antonov: use dl, dt, dd, and optionally div.
A combination of dl, dt, and dd is semantically fine for key-value pairs:
<dl>
<dt>Key1</dt>
<dd>Value1</dd>
<dt>Key2</dt>
<dd>Value2</dd>
</dl>
For easier styling or parsing, divs can be used as children of dl to group the key-value pairs (and makes dt and dd be grandchildren of dl):
dl { display: table; }
dl > div { display: table-row; }
dl > div > dt, dl > div > dd { display: table-cell; border: 1px solid black; padding: 0.25em; }
dl > div > dt { font-weight: bold; }
<dl>
<div>
<dt>Key1</dt>
<dd>Value1</dd>
</div>
<div>
<dt>Key2</dt>
<dd>Value2</dd>
</div>
</dl>
XHTML 2 introduces the ability to group terms and definitions using the di element
<!-- Do not us this code! -->
<dl>
<di>
<dt>Name</dt>
<dd>John</dd>
</di>
<di>
<dt>Age</dt>
<dd>25</dd>
</di>
</dl>
X/HTML 5 vs XHTML 2 > Enhancement To Definitions Lists
Unfortunately though, XHTML 2 is dead and HTML5 doesn't have di element
So, you can combine ul > li with dl > dt + dd :
<ul>
<li>
<dl>
<dt>Name</dt>
<dd>John</dd>
</dl>
</li>
<li>
<dl>
<dt>Age</dt>
<dd>25</dd>
</dl>
</li>
</ul>
I think a definition list is probably a bad idea. Semantically, they are used for definitions. Other key-value lists will often differ from definition titles and descriptions.
A table is one way to go, but what about an unordered list?
<ul>
<li class="key-value-pair">
<span class="key">foo</span>
<span class="value">bar</span>
</li>
</ul>
dl {
display: grid;
grid-template-columns: auto auto;
}
dd {
margin: 0
}
<dl>
<dt>key</dt>
<dd>value</dd>
<dt>key</dt>
<dd>value</dd>
</dl>
I used <dl> <dt> <dd> and styled them with grid
This is not my preferred solution, but it is a clever abuse of the semantic element:
Use a new <dl> per <dt>/<dd> pair:
<div class="terms">
<dl><dt>Name 1</dt><dd>Value 1</dd></dl>
<dl><dt>Name 2</dt><dd>Value 2</dd></dl>
</div>
An example with css floats and red border on hover:
dt:after { content:":"; }
dt, dd { float: left; }
dd { margin-left: 5px }
dl { float: left; margin-left: 20px; border: 1px dashed transparent; }
dl:hover { border-color: red; }
<div class="terms">
<dl>
<dt>Name 1</dt>
<dd>Value 1</dd>
</dl><!-- etc -->
<dl><dt>Name 2</dt><dd>Value 2</dd></dl>
<dl><dt>Name 3</dt><dd>Value 3</dd></dl>
<dl><dt>Name 4</dt><dd>Value 4</dd></dl>
<dl><dt>Name 5</dt><dd>Value 5</dd></dl>
<dl><dt>Name 6</dt><dd>Value 6</dd></dl>
</div>
it is difficult to properly offset the pairs visually, both in code and rendered. If I wanted to, for instance, but a border around each pair, that would be a problem.
Others before me have dealt (quite well I think) with the problem of providing visual definition in code. Which leaves the problem of rendering and CSS. This can be done quite effectively in most cases. The biggest exception is placing a border around each set of dt/dds, which is admittedly extremely tricky - perhaps impossible to style reliably.
You could do:
dt,dd{ border:solid; }
dt{ margin:10px 0 0 0; border-width:1px 1px 0 1px; }
dd{ margin:0 0 10px 0; border-width:0 1px 1px 1px; padding:0 0 0 10px; }
dd::before{ content:'→ '; }
Which works for key-value PAIRS, but presents problems if you have multiple dds within one "set" like:
<dt>Key1</dt>
<dd>Val1.1</dd>
<dd>Val1.2</dd>
<dt>Key2</dt>
<dd>Val2.1</dd>
<dd>Val2.2</dd>
However, border-styling limitations aside, doing anything else to associate sets (backgrounds, numbering, etc.) is quite possible with CSS. There's a nice overview here: http://www.maxdesign.com.au/articles/definition/
Another point I think is worth noting, if you're looking to optimally style definition-lists to provide visual separation - CSS's automatic-numbering features (counter-increment and counter-reset) can come in quite handy: http://www.w3.org/TR/CSS2/generate.html#counters
Of course, you could just use HTML Custom elements. ( spec )They're completely valid HTML5.
Unfortunately, browser support isn't that great, but if rarely do we care about such pedestrian things as browser support when we're dealing with semantics. :-)
One such way you could represent it:
After registering your brand new fancy element like so:
var XKey = document.registerElement('x-key');
document.body.appendChild(new XKey());
You write markup like so:
<ul>
<li>
<x-key>Name</x-key>
Value
</li>
</ul>
The only condition that HTML Custom elements have are that they need a dash in them. Of course, you can now be super creative... if you're building an auto parts warehouse, with lists of parts and serial numbers:
<ul>
<li>
<auto-part>
<serial-number>AQ12345</serial-number>
<short-description>lorem ipsum dolor...</short-description>
<list-price>14</list-price>
</auto-part>
</li>
</ul>
You can't get much more semantic than that!
However, there IS a chance I have gone too far into semantic-shangri-la-la (a real place) and might be tempted to scale it back to something a little more generic:
<ul>
<li class='auto-part' data-serial-number="AQ12345">
<p class='short-description'>lorem ipsum dolor...</p>
<p class='list-price'>14</p>
</li>
</ul>
Or perhaps this ( if I needed to style and show the key visually )
<ul>
<li class='auto-part'>
<x-key>AQ12345<//x-key>
<p class='short-description'>lorem ipsum dolor...</p>
<p class='list-price'>14</p>
</li>
</ul>
Hmm. dt/dd sound best for this and it is possible to offset them visually, although I do agree it's more difficult than for a table.
As for source code readability, how about putting them into one line?
<dl>
<dt>Name</dt> <dd>Value</dd>
<dt>Name</dt> <dd>Value</dd>
</dl>
I agree it's not 100% perfect, but seeing as you can use space characters for indentation:
<dl>
<dt>Property with a looooooooooong name</dt> <dd>Value</dd>
<dt>Property with a shrt name</dt> <dd>Value</dd>
</dl>
it might be the nicest way.
Except for the <dl> element, you've pretty much summed up all the options of HTML: limited ones.
However you're not lost, there is one option left: using a markup language that can be translated to HTML. A very good option is Markdown.
Using a table extension that some markdown engines provide, you can have code like this:
| Table header 1 | Table header 2 |
-----------------------------------
| some key | some value |
| another key | another value |
This means that you need a build step to translate the Markdown to HTML of course.
This way you get meaningful markup (table for tabular data) and maintainable code.
How about just placing a blank line between each pair?
This requires no special handling for long values.
<dl>
<dt>Name</dt>
<dd>Value</dd>
<dt>Name</dt>
<dd>Value</dd>
<dt>Name</dt>
<dd>Value</dd>
</dl>
What about using lists?
Ordered:
<ol>
<li title="key" value="index">value</li>
…
</ol>
Or use <ul> for an unordered list, and skip the value="index" bit.
The line from spec:
Name-value groups may be terms and definitions, metadata topics and values, questions and answers, or any other groups of name-value data.
I assume use of elements <dl>, <dt> and <dd>.
I do like Unicron's idea of putting them all on one line, but another option would be to indent the value below the definition name:
<dl>
<dt>Name</dt>
<dd>Value</dd>
<dt>Name</dt>
<dd>Value</dd>
</dl>
This way might be a little easier on the eye in ridiculously long definitions (although if you're using ridiculously wrong definitions then perhaps a definition list isn't what you really want after all).
As for the rendering on screen, a fat bottom margin applied to the dd is plenty visual separation.