Re: question about Parser Liaison and Documents

[Paul Levin <pl...@bii.com>]

Scott,
    Thank you for the information.

    Thanks again,
    Paul

Scott Boag/CAM/Lotus wrote:

> (Seems like this should have gone to the xalan list, which may be why I
> missed it...)
>
> > I need to be able to instantiate a Document without parsing, so it looks
> like my only
> > choice is the Xerces classes, even though they are slower than the DTM.
> Again, please comment.
>
> Yes, you need to use the Xerces liaison... liaison's tightly couple to a
> given DOM implementation.
>
> The only real reason the Xerces dom liaison is a disadvantage in terms of
> speed is node-order sorting and on-the-fly namespace resolution, since the
> DOM is already parsed, and incrementality and memory use (in terms of the
> size of the DOM, since it is already constructed) aren't factors for your
> case.  The node order sorting we hope to solve with a new search algorithm
> we are working on.  The namespace resolution might be better with some of
> the work they've been putting into the Xerces DOM lately (I have to go back
> and re-investigate this).   For certain operations (non-union selections),
> the DOM will actually be a tiny bit faster.
>
> So, my advice would be to go ahead and use the Xerces liaison, and the
> performance differences that are applicable to your case will be resolved
> over the next few weeks.
>
> -scott
>
>
>                     Paul Levin
>                     <plevin@bii.c        To:     xerces-dev@xml.apache.org
>                     om>                  cc:     (bcc: Scott Boag/CAM/Lotus)
>                                          Subject:     Re: question about Parser Liaison and Documents
>                     01/31/00
>                     01:51 PM
>                     Please
>                     respond to
>                     xerces-dev
>
>
>
> Can someone please comment on this.  I posted this last Thursday, and no
> one has made any comments.
>     Thanks,
>     Paul
>
> Paul Levin wrote:
>      I realize that the Document implementation class used, and the
>      ParserLiaison class used, are tightly coupled.
> I notice, in the latest Xalan distribution, that there are 2 significant
> Liaison/Document packages --
>
> |-------------------------------------------+---------------------------------------|
> | ParsserLiaison Class                      | Document Impl Class                   |
> |-------------------------------------------+---------------------------------------|
> | org.apache.xalan.xpath.xdom.XercesLiaison | org.apache.xerces.dom.DocumentImpl    |
> |-------------------------------------------+---------------------------------------|
> | org.apache.xalan.xpath.dtm.DTMLiaison     | org.apache.xalan.xpath.dtm.DTMProxy   |
> |-------------------------------------------+---------------------------------------|
>
>      I am trying to understand some pros and cons of these 2 choices.
>           It seems that the DTM classes are supposed to be faster than the
>           Xerces classes.
>           The Xerces DocumentImpl class has a constructor (so a document
>           can be made without having to parse a file); where as the
>           DTMProxy class has no constructor.
>      Would people please comment on these findings, and correct me where I
>      am wrong,
>
>      I need to be able to instantiate a Document without parsing, so it
>      looks like my only choice is the Xerces classes, even though they are
>      slower than the DTM.  Again, please comment.
>
>          Thanks