Re: [xml] xml:space and xml:lang problem

[Daniel Veillard <veillard redhat com>]

On Wed, Jun 08, 2005 at 04:45:10PM -0400, Rob Richards wrote:
> The xmllint example was just a way to exemplify the issue using xmllint. 
> I was loading xml using dom methods and told it to stip out whitespaces.
> This works fine, except the xml:space attribute when set to preserve 
> should prevent the removal of whitespaces in its context.
> 
> When the xml is output it looks like:
> <?xml version="1.0"?>
> <doc><e1><e2 xml:space="preserve"><e3 id="E3"/></e2></e1></doc>
> 
> while it should look like:
> <?xml version="1.0"?>
> <doc><e1><e2 xml:space="preserve">
>         <e3 id="E3"/>
>      </e2></e1></doc>
> 
> which it does using 2.5.x as well as 2.6 with the patch.

  okay that was my guess, sounds fine.

> Hadn't played with the xml:lang, but saw it has the same problem as 
> xml:space. The code in xmlParseAttribute was carried over to 
> xmlParseAttribute2, but within xmlParseAttribute2, the attribute is 
> broken out into prefix and name, so the tests xmlStrEqual(name, BAD_CAST 
> "xml:space") and (xmlStrEqual(name, BAD_CAST "xml:lang")) will never be 
> true like they are in xmlParseAttribute.

  hum, right ...

> As far as moving code around, I made a test just for 
> xmlStrEqual(*prefix, BAD_CAST "xml") and if that is true will fall into 
> the block to check for name equal to space or lang. figured it would be 
> faster rather than testing everything twice when the prefix may not even 
> be correct in the first place.

  string interning should even allow to reduce that to a pointer comparison
with ctxt->str_xml 

> The logic is the same as that used in xmlParseAttribute, which does only 
> check lang if pedantic is used. The change was just to get the checks to 
> actually kick in as they would never be true in xmlParseAttribute2.

  thanks for the explanations.

Daniel

-- 
Daniel Veillard      | Red Hat Desktop team http://redhat.com/
veillard redhat com  | libxml GNOME XML XSLT toolkit  http://xmlsoft.org/
http://veillard.com/ | Rpmfind RPM search engine http://rpmfind.net/