Re: [xml] XPath performance

["Karel Michek" <intra611+xml gmail com>]

Thank you for the advice. The performance is great. The only problem
is that the DOM tree needs to build in the memory and xml:id does not
seem work to work in this case (so to make the test work the XML
document was serialized and parsed back).

The following fragments from different e-mails can be relevant to this problem:

   >> DOMDocument::getElementByID() doesn't work with xml:id
   >> if the DOMDocument is being created in code.

   > And will never work that way. IDs would need to be set
   > using the DOMElement setIdAttribute methods, but these
   > aren't currently implemented as there are some libxml
   > issues surrounding this.

and

   > setIDAttribute - thats on the TODO list. Were some previous libxml
   > issues to be resolved before that could be implemented (so it will be
   > added).

So I wander if it is currently possible to build DOMDocument in the
code and and make
IDs work (or will that be possible)?

Thank you,
Karel


On 11/6/06, Daniel Veillard <veillard redhat com> wrote:
> On Sun, Nov 05, 2006 at 10:22:46AM +0100, Karel Michek wrote:
> > Hi,
> > When accessing XML tree that has structure outlined bellow and which
> > contains hundreds of thousands of "doc" nodes and using something like
> >
> > xmlXPathEvalExpression(/job/doc[100000]/page/meta/meta_type/meta1, xpathCtx);
> >
> > the performance is not acceptable. I wander if there is any way to
>
>   the siblings are stored in a doubly linked list. There is no array or
> database index on name tag, sorry there is no way to avoid walking up the
> full sibling list in libxml2.
>
> >  <doc id="1">
> > <doc id="2">
>
> id('100000')/page/meta/meta_type/meta1 may work if the id attribute was defined
> as an XML ID. Using xml:id would do that automatically without DTD.
> In XSLT defining doc/@id as a key would also allow to index and have an instant
> lookup.
>
>   Those are really basic XML and XSLT solution design, and not specifically
> dependant on libxml2, nothing replace learning about the technology in question.
>
> Daniel
>
> --
> Red Hat Virtualization group http://redhat.com/virtualization/
> Daniel Veillard      | virtualization library  http://libvirt.org/
> veillard redhat com  | libxml GNOME XML XSLT toolkit  http://xmlsoft.org/
> http://veillard.com/ | Rpmfind RPM search engine  http://rpmfind.net/