W3C home > Mailing lists > Public > www-style@w3.org > March 2009

Re: Selecting an element combining nth-of-type() and a class/ID

From: Tab Atkins Jr. <jackalmage@gmail.com>
Date: Mon, 30 Mar 2009 10:36:17 -0500
Message-ID: <dd0fbad0903300836w5dcecc7es2310df27b77a7e65@mail.gmail.com>
To: Giovanni Campagna <scampa.giovanni@gmail.com>
Cc: "www-style@w3.org list" <www-style@w3.org>
On Mon, Mar 30, 2009 at 10:22 AM, Giovanni Campagna
<scampa.giovanni@gmail.com> wrote:
> 2009/3/30 Tab Atkins Jr. <jackalmage@gmail.com>:
>> On Mon, Mar 30, 2009 at 8:07 AM, Charles <landemaine@gmail.com> wrote:
>>> What if you want it to match another pseudo class with also an ID?
>>> For instance: li:nth-match(2, :nth-last-of-type#today.new)
>>> We need a similar flexibility...
>>
>> I'm not sure what that would match, even if you had the hypothetical
>> ability to do that.  Can you give an example?
>>
>> ~TJ
>>
>>
>
> Theoretically, the :nth-match works as follows, in the following (invalid) tree.
> <ul>
> <p class="new">a</p>
> <li class="new">b</li>
> <li id="today">c</li>
> <q>d</q>
> <li class="new" id="today">e</li>
> <p class="new" id="today">f</li>
> <li class="new" id="today">g</li>
> </ul>
> 1) First take all direct childs that matches the second part of the
> nth-match(), that is "#today.new:nth-last-of-type" (rewritten in
> common order). This means:
> 1.1) Take the elements with ID #today: "c", "e", "f" and "g".
> 1.2) Take the subsets of those which have class .new: "e", "f" and "g".
> 1.3) Take their element types: "li" and "p"
> 1.4) Find which of them is the last in the set of elements with the
> same type: we have "f" and "g". "e" is not, because it is followed by
> another <li> (it doesn't matter if the latter <li> matches #today.new)
> 1.5) The resulting list is then "f" and "g"
> 2) Among the results of the first match, find which is number two: in
> our case it is "g".
> 3) Among the results of the step 2 (just one node), find those which
> match "li": only "g"
>
> It is a li element that is the second element being the last of its
> type and having class "new" and id "today".
> If <q> (element "c") had class "new" and id "today", we would have had
> no match from the selector.

All right, that works.  I forget that CSS doesn't care about html's
"ids must be unique in a document" restriction.

> Anyway, I'm not sure what advantage you can get from such a complex selector.

I don't think there's anything *wrong* with such a complex selector -
it doesn't pose any conceptual problems.  You grab a bunch of <li>s,
specialize the matched group by the provided selector, then select the
2nd one. But I wouldn't shed a tear if it wasn't allowed, either
(however, defining that to be not allowed would likely be too complex,
unless we went overboard and said that only simple selectors are
allowed).

~TJ
Received on Monday, 30 March 2009 15:45:10 GMT

This archive was generated by hypermail 2.3.1 : Tuesday, 26 March 2013 17:20:17 GMT