Re: [xml] Canonical XML

[Daniel Veillard <veillard redhat com>]

On Thu, Feb 28, 2002 at 10:47:13AM -0800, Aleksey Sanin wrote:
> Please, find examples bellow. The brief description is that AFAIK
> if we evaluate XPath expression against document then the document
> should be modified at least as follows:
>     1) default attributes added to element nodes
  
  Done at parse time in xsltproc.c with
    xmlLoadExtDtdDefaultValue = XML_DETECT_IDS | XML_COMPLETE_ATTRS;
    you also need
    xmlSubstituteEntitiesDefault(1)
  just do the same to the parser context when parsing a document for
processing XPath

>     2) namespaces from ancesstors are propagated to the node
> These new added nodes could be included in the resulted XPath nodes set for
> further processing.

   Namespaces are propagated to children. But namespace nodes have
no parent in libxml2 XPath implementation. This property is actually
expected to be removed in XPath 2.0 I was told.

> > >    1) "parent::ietf:e1" does not include any e1 namespaces
> > Give an example, sounds strange.
> Example:
> ------------
> ## xml doc:
> <doc xmlns="http://www.ietf.org"; xmlns:w3c="http://www.w3.org";>
>    <e1 a=b>
           ^This is not XML ! I assume it is actually a="b"

>       <e2 xmlns="" />
>    </e1>
> </doc>
>
> ## xpath expr:
> (//. | //@* | //namespace::*)
> [
>    self::ietf:e1 or (parent::ietf:e1 and not self::e2)
> ]

  To compile such an expression you must provide a mapping for 
the ietf namespace prefix. What is its namespace name ? How did you pass
it to the XPath evalution ? I assume you did this.

Running the beginning of the XPath query on the corrected document brings:
orchis:~/XML -> xmllint --shell tst.xml
/ > ls (//. | //@* | //namespace::*)
--n        3 doc
t--        4     
-a-        3 e1
t--        1  
t--        4     
t--        7        
--n        0 e2
t--        4     
t--        1 b
t--        7        
--n        0 e2
t--        4     
t--        1  
n        1 default -> http://www.ietf.org
n        1 w3c -> http://www.w3.org
n        1 xml -> http://www.w3.org/XML/1998/namespace
n        1 default -> 
/ > 

  So the namespace nodes inherited are found, but the test "parent::ietf:e1"
on the namespaces nodes returned will not work as expected using the XPath-1.0
semantic.

> ## CVS version of file xpath.c, function xmlXPathNextParent() :
>    
>     case XML_NAMESPACE_DECL:   
>                                  /*  
>                                   * TODO !!! may require extending 
> struct _xmlNs with           
>                                   * parent field
>                                   * C.f. Infoset 
> case...                                   
>                                    */                                
>                                    return(NULL);

  A libxml2 XML_NAMESPACE_DECL node is NOT an XPath namespace node. It
is a node defining the namespace, not reflecting that the 

> > >    2) namespaces from the parent element are not included
> > >    in the child's namespaces axis
> >
> >    Hum, strange. I'm pretty sure it works well, I even have some XSLT test
> > cases for those.
> See the example above.

  That's not a good caracterization of the problem.

> > >    3) default attributes from inner DTD are not included
> > >    in the result nodes set

  xmlLoadExtDtdDefaultValue = XML_DETECT_IDS | XML_COMPLETE_ATTRS;
          you also need
  xmlSubstituteEntitiesDefault(1)

Daniel

-- 
Daniel Veillard      | Red Hat Network https://rhn.redhat.com/
veillard redhat com  | libxml Gnome XML XSLT toolkit  http://xmlsoft.org/
http://veillard.com/ | Rpmfind RPM search engine http://rpmfind.net/