[XSL-LIST Mailing List Archive Home] [By Thread] [By Date]

Re: [xsl] Using memory addressing to retrieve a value vice using a software string library to retrieve a value


Subject: Re: [xsl] Using memory addressing to retrieve a value vice using a software string library to retrieve a value
From: "David Carlisle d.p.carlisle@xxxxxxxxx" <xsl-list-service@xxxxxxxxxxxxxxxxxxxxxx>
Date: Fri, 20 Nov 2015 12:43:18 -0000

On 20 November 2015 at 12:13, Costello, Roger L. costello@xxxxxxxxx
<xsl-list-service@xxxxxxxxxxxxxxxxxxxxxx> wrote:
> Hi Folks,
>
> I want to retrieve "west".
>
> Which of these is faster?
>
> -------------------
> Approach #1
> -------------------
> The <edge> element contains text:
>
>         <edge>garden west door</edge>
>
> "west" is retrieved using this XPath:
>
>         substring-before(substring-after(., ' '), ' ')
>
> Note: assume that <edge> is the context node.
>
> -------------------
> Approach #2
> -------------------
> The <edge> element contains markup:
>
>         <edge>
>             <garden/>
>             <west/>
>             <door/>
>         </edge>
>
> "west" is retrieved using this XPath:
>
>         *[2]/name()
>
>
> I believe that Approach #2 is much, much faster. Do you agree?
>
> As I see it, Approach #2 is simply a memory addressing task (which computers
do very well) to obtain <west/> and then output the symbols at that memory
address. Approach #1, on the other hand, requires the use of software string
libraries, which, I imagine, results in hundreds or thousands of machine
instructions. In fact, I would imagine that Approach #2 is hundreds or
thousands of times faster than Approach #1. Do you agree?
>
> /Roger

In addition to what Mike said,

<edge>garden west door</edge>

is two nodes (an element and text)


    <edge>
            <garden/>
            <west/>
            <door/>
    </edge>

is four or eight nodes (depending on whether you have white space text
nodes or not) so it depends a lot on your internal object model, but
the second one might be a lot bigger. If you run out of memory then
speed isn't really the issue. But the second form could take less
memory than the first if the element names are repeated often and some
kind of name pool is used for element names but not text content.

Nothing really that you can say in general.

David


Current Thread