Fwd: Proposal: getElementsBySelector()

Whoops, I meant to send this to the public list:

Begin forwarded message:

> Resent-From: member-webapi@w3.org
> From: Maciej Stachowiak <mjs@apple.com>
> Date: February 6, 2007 3:06:36 PM PST
> To: Charles McCathieNevile <chaals@opera.com>
> Cc: Web API WG <member-webapi@w3.org>
> Subject: Re: Proposal: getElementsBySelector()
>
>
>
> On Feb 6, 2007, at 11:53 AM, Charles McCathieNevile wrote:
>
>> On Tue, 06 Feb 2007 17:41:03 +0530, Maciej Stachowiak  
>> <mjs@apple.com> wrote:
>>
>>> On Feb 5, 2007, at 10:18 PM, Charles McCathieNevile wrote:
>>>> <annevk@opera.com> wrote:
>>>>> ... Given that, I propose we rename .getAll() to
>>>>> .getElementsBySelector() and drop .get() (on both Document and
>>>>> Element).
>>>>>
>> ...
>>>>> If there are no strong objections I'll implement this in the
>>>>> specification.
>>>>
>>>> Not having heard strong objections, and having had support for
>>>> getElementsBySelector() that is at least as strong as anything
>>>> else, I think (with
>>>> my chair's hat) this can be taken as the current resolution of the
>>>> naming debate.
>>>>
>>>> Which would also resolve ISSUE-110.
>>>>
>>>> Any objections?
>>>
>>> I stated my objection already in my message on this thread from
>>> January 28th.
>>
>> Yep. Thank you.
>
> I mentioned this because you said you hadn't heard any strong  
> objections, so I wasn't sure if you missed mine.
>
>> I am trying to find out if you are one lone voice of reason
>> actually speaking for a multitude who didn't answer for  
>> themselves, or just
>> marching to the beat of a different drum, or somewhere in between...
>>
>> It would also be helpful to have an idea of how strongly you  
>> object on each of
>> the points (name and having a single method).
>
> I have no more or less problem with the name than previous name  
> proposals - I already made my big honkin' list of suggested shorter  
> names.
>
> I do object to having a single method because it will lead to worse  
> performance when you really just want one element, even given  
> Bjoern's interesting optimization idea (which I think is a  
> problematic approach in its own way since it increases code  
> complexity, and potentially slows down DOM mutations a bit, where  
> the whole idea of using a StaticNodeList was to keep list access  
> and DOM manipulations from interfering with each other.
>
> I think it should at least in principle be possible to implement  
> gEBS to be as fast as getElementById when given an id selector and  
> requesting one result, but I don't think that is doable with  
> Bjoern's design, at the very least due to the extra allocation and  
> also because when you look for a single element you are highly  
> likely to mutate it in some way and therefore make Bjoern's  
> proposed mutation ineffective.
>
> Is there a major benefit to having a single method that outweighs  
> the performance considerations? I think we should have a better  
> reason for deciding this than "we feel like it".
>
> Adding list back to Cc, I assume it was omitted by accident.
>
> Regards,
> Maciej
>
>
>
>
>