XOM API for Processing XML with Java

[Elliotte Harold <elharo AT metalab.unc.edu>]

Andrew Thompson wrote:


> * cosmetically, we've lost the newlines after the comment

You're not providing sufficient information to be sure, but it sounds 
like these comments come from the document prolog. If so, it's the 
parser that loses the newlines, not XInclude.

> * the div has picked up an xmlns="" which makes it fail w3 validation.  
> This is a Xerces bug, I think... because I've seen it when using XSLT  
> before

That's weird. It's probably something that's required in your case due 
to things you haven't told us, but it might be my bug. Hard to tell 
without seeing the full example. I'd be surprised if it were a Xerces 
bug though.

> * The killer is this  
> xml:base="file:///path/on/my/local/harddisk/website%20bits/sidebar.xml"
>
> EHR, I notice you didn't mention this in your article @  
> http://cafe.elharo.com/xml/xinclude/ but from a quick read of the  
> XInclude spec this seems to be expected (XInclude spec, section 4.5.5).

That's correct. It is expected. XOM is allowed to compute relative URIs 
instead. So far I don't implement that, and it's not exactly clear what 
that would mean if I did. (I could do it if the parent element had an 
xml:base attribute that contained an absolute URI, but if the document 
is simply written out to an arbitrary location, the only way to get the 
base URI right is to write an absolute URI).

> This is horrible from a workflow perspective though... Quite apart from  
> it also failing w3 validation, I want to resolve XIncludes on my local  
> system before uploading these files to my web server. So I really don't  
> want references to my local filesystem littered all over these  
> documents. Even if today's tools ignore xml:base, I don't want my local  
> disk names published!

You should perhaps ask the W3C for an option to skip base URI fixup.

> In an attempt to bring thing back on topic, is there anything I can do  
> in XOM to try to suppress this stuff? (the base and the xmlns?). Is  
> there a better XML parser to use for this than the Xerces that's  
> bundled with JDK 1.4.2?

The xmlns stuff is weird. See if you can produce a simple example of 
that. I suspect it's either my bug or it's necessary in your context.

You could write a post-process that stripped the xml:base attributes. It 
would be easiest to do this using XSLT or (in XOM 1.1) XPath:

Nodes nodes = doc.query("//*[@xml:base]");
for (int i = 0; i < nodes.size(); i++) {
  Element e = (Element) node;
  e.removeAttribute("xml:base", Namespace.XML_NAMESPACE);
}


--
Elliotte Rusty Harold  elharo AT metalab.unc.edu
XML in a Nutshell 3rd Edition Just Published!
http://www.cafeconleche.org/books/xian3/
http://www.amazon.com/exec/obidos/ISBN=0596007647/cafeaulaitA/ref=nosim