RE: make aria-owns to rearrange children

IE8 sounds right to me too.

From: James Nurthen [mailto:james.nurthen@oracle.com]
Sent: Tuesday, February 17, 2015 12:40 PM
To: public-pfwg@w3.org
Subject: Re: make aria-owns to rearrange children

I think querySelector and querySelectorAll are available in IE8 but I'm not sure why this is relevant for this discussion as this is new functionality we are discussing.

On 2/17/2015 12:35 PM, Richard Schwerdtfeger wrote:

James, beginning with what version of IE is querySelector supported?


Rich Schwerdtfeger

[Inactive            hide details for James Nurthen ---02/17/2015 11:56:32 AM---I            like the querySelector idea - it would give a great deal]James Nurthen ---02/17/2015 11:56:32 AM---I like the querySelector idea - it would give a great deal of flexibility. On 2/17/2015 9:36 AM, Ale

From: James Nurthen <james.nurthen@oracle.com><mailto:james.nurthen@oracle.com>
To: Alexander Surkov <surkov.alexander@gmail.com><mailto:surkov.alexander@gmail.com>, Dominic Mazzoni <dmazzoni@google.com><mailto:dmazzoni@google.com>
Cc: W3C WAI Protocols & Formats <public-pfwg@w3.org><mailto:public-pfwg@w3.org>
Date: 02/17/2015 11:56 AM
Subject: Re: make aria-owns to rearrange children

________________________________



I like the querySelector idea - it would give a great deal of flexibility.

On 2/17/2015 9:36 AM, Alexander Surkov wrote:
I think I like the idea to use querySelector for aria-owns value (for all idrefs attributes?).

On Tue, Feb 17, 2015 at 12:20 PM, Dominic Mazzoni <dmazzoni@google.com<mailto:dmazzoni@google.com>> wrote:
On Tue, Feb 17, 2015 at 9:09 AM, James Nurthen <james.nurthen@oracle.com<mailto:james.nurthen@oracle.com>> wrote:
One issue with aria-owns is that as soon as you start to use it you end up with a massive number of idrefs.

We discussed this problem a while back in the context of web components, and there were at least two possible solutions I thought could work:

* Expose reflected attributes on an Element, allowing for example: parent.ariaOwns = [child3, child2, child1]; or parent.ariaActiveDescendant = child2;
* Alllow a querySelector string in place of an idref, for example <tr aria-owns="query(.leftcells .rightcells)"> or <div aria-activedescendant="query([tabIndex='0'])">.

Are you still interested in pursuing either of those, or even both? I might like to try implementing one of those behind a flag if the group can come to a tentative consensus.

FWIW, I like Alexander's idea of allowing aria-owns for sibling relationships. I think we should do it either way, and then solve the IDREF problem in general.

- Dominic

Something I would like to be able to do is reference a parent node (with role=presentation) in aria-owns with the effect that its children would be owned.
One place we hit this a lot is in grids with multiple scrollable regions. So, currently, in the following 2 rows table with 2 scrollable regions, we have the following structure.

<tr role="row" aria-owns="r1c1 r1c2 r1c3 r1c4 r1c5 r1c6">
  <td role="gridcell" id="r1c1">
  <td role="gridcell" id="r1c2">
</tr>
<tr role="row" aria-owns="r2c1 r2c2 r2c3 r2c4 r2c5 r2c6">
  <td role="gridcell" id="r2c1">
  <td role="gridcell" id="r2c2">
</tr>
<tr role="presentation">
  <td role="gridcell" id="r1c3">
  <td role="gridcell" id="r1c4">
  <td role="gridcell" id="r1c5">
  <td role="gridcell" id="r1c6">
</tr>
<tr role="presentation">
  <td role="gridcell" id="r2c3">
  <td role="gridcell" id="r2c4">
  <td role="gridcell" id="r2c5">
  <td role="gridcell" id="r2c6">
</tr>

What I would like to be able to (which I would like to have the same result in the API mapping) is the following

<tr role="row" aria-owns="r1c1 r1c2 r1">
  <td role="gridcell" id="r1c1">
  <td role="gridcell" id="r1c2">
</tr>
<tr role="row" aria-owns="r2c1 r2c2 r2">
  <td role="gridcell" id="r2c1">
  <td role="gridcell" id="r2c2">
</tr>
<tr role="presentation" id="r1">
  <td role="gridcell">
  <td role="gridcell">
  <td role="gridcell">
  <td role="gridcell">
</tr>
<tr role="presentation" id="r2">
  <td role="gridcell">
  <td role="gridcell">
  <td role="gridcell">
  <td role="gridcell">
</tr>

Or, taking it one step further, allow aria-owns to reference itself, which would place the natural children of the element into the ownership at that point in the order so I could do:

<tr role="row" aria-owns="t1r1 t2r1" id="t1r1">
  <td role="gridcell">
  <td role="gridcell">
</tr>
<tr role="row" aria-owns="t1r2 t2r2" id="t1r2">
  <td role="gridcell">
  <td role="gridcell">
</tr>
<tr role="presentation" id="t2r1">
  <td role="gridcell">
  <td role="gridcell">
  <td role="gridcell">
  <td role="gridcell">
</tr>
<tr role="presentation" id="t2r2">
  <td role="gridcell">
  <td role="gridcell">
  <td role="gridcell">
  <td role="gridcell">
</tr>

If people like this idea I would be happy to come up with some spec text to allow it.

Regards,
James


On 2/15/2015 8:07 AM, Alexander Surkov wrote:
Hi.

aria-owns is used to define parent-child relationship [1]. The spec says nothing how aria-owns affects on between-siblings relations, i.e. where aria-owns children have to be in the hierarchy. So there's no way for the author to control the ordering of explicit and aria-owns children. The idea [2] was to let aria-owns to refer its own children to change the ordering. For example,

<div role="grid"
<div role="row" aria-owns="c1 c2 c3 c4">
  <div role="gridcell" id="c1">cell1</div>
  <div role="gridcell" id="c2">cell1</div>
</div>
</div>

<!-- somewhere in the DOM -->
<div role="gridcell" id="c3">cell1</div>
<div role="gridcell" id="c4">cell1</div>

In this case role="row" would contain 4 cells in the order specified by aria-owns.

If the idea looks reasonable then the spec could be changed this way

Insert after existing wording:
"The value<http://www.w3.org/WAI/PF/aria/terms#def_value> of the aria-owns attribute<http://www.w3.org/WAI/PF/aria/terms#def_attribute> is a space-separated list of IDREFS that reference one or more elements in the document by ID."

something like:
"The order the referred elements listed in the value should be preserved when their parent-child relationship is set. All explicit unreferred children should be considered followed aria-owns elements."

Thanks.
Alexander.

[1] http://www.w3.org/WAI/PF/aria/states_and_properties#aria-owns
[2] https://bugzilla.mozilla.org/show_bug.cgi?id=1133213