Re: [xml] building ID/IDREF tables by attaching DTD to an xmlDocPtr??

[Daniel Veillard <veillard redhat com>]

On Tue, Aug 17, 2004 at 04:58:36PM -0500, Graham Clark wrote:
> Hi,
>
> I have an XML file without external or internal subset definitions,
> but which should be valid according to a separate DTD. The DTD
> contains some attributes marked ID and some marked IDREF. I'm trying
> to exploit this fact and use the XPath id() function on the XML file,
> and so was looking for a way to make libxml construct the ID/IDREF
> tables given an xmlDocPtr, and separately, an xmlDtdPtr. If the DTD is
> copied into the XML file (as an internal subset), xmlParseFile results
> in an xmlDocPtr that I can use successfully with the id()
> function. However, if I build a separate xmlDtdPtr, add it to the doc
> as described in Section 10 of the libxml FAQ 
>
>         doc->intSubset = dtd;
>         if (doc->children == NULL) xmlAddChild((xmlNodePtr)doc, (xmlNodePtr)dtd);
>         else xmlAddPrevSibling(doc->children, (xmlNodePtr)dtd);
>
> and then call xmlValidateDocument(), my resulting xpath id() queries
> return empty node sets (adding the DTD and calling
> xmlValidateDocument() was my best guess for making this work).
>
> Stepping through with gdb, I found that when evaluating the XPath
> function, the call to xmlHashLookup() returns an ID with a null attr
> field. xmlAddID (as called by xmlValidateDocument()) creates the ID
> with a name field and a null attribute field because it's "operating
> in streaming mode". This seems to be as a result of the
> xmlValidCtxtPtr, but I'm uncertain of the significance of this. Is the
> approach of calling xmlValidateDocument() doomed, and if so, is there
> a legitimate way to do this short of modifying the XML file itself?

  Hum ... nasty. Obviously it won't work as-is and I can't extend the
xmlValidCtxt without breaking libxml2 ABI so this is not trivial, and would
probably need some code change. Could you try xmlValidateDtd() instead and
not link the dtd to the internal or external subset. I'm afraid it
will fail for the same reason but it would be nice to try that too.

> The best workaround I have at the moment is to use xmlSaveFile() to
> write out the doc with the attached DTD, then read it back in.

  Which is not great, if xmlValidateDtd() won't work either, could
you fill a request for improvement for libxml2 in bugzilla, if possible
provide a self contained example (c + xml + DTD even if basic) so it
makes it easier for whoever might look at implementing the fix,

  thanks,

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/