XML, Namespaces and XSLT, Oh, My!

mphare · August 3rd, 2010, 03:02 PM

First, I've been through Dave Pawson's excellent collection of namespace related information. If the answer to my question is in there.. I missed it.

I'm using Saxon9.0.06J, although I have OxygenXML 11, so there is a newer Saxon in there if I can figure out how to use from the command line.

I am writing a stylesheet to transform one XML file into another with a different schema. I need namespaces, including a default namespace, on the result XML file, so I've added them to the stylesheet so they will be declared in the result file. But, this is causing me problems.

I've boiled down the problem to as simple form as I can get it.

The Stylesheet:

Code:

<?xml version="1.0" encoding="UTF-8"?>
<xsl:stylesheet 
        xmlns:xsl="http://www.w3.org/1999/XSL/Transform" 
        version="2.0"
        xmlns="http://MyDefaultNS"
    >
    <!--
        xmlns="http://MyDefaultNS">
        
        We need the default namespace decl in the stylesheet element
        so it will appear on the root element decl of the result file and put the default NS in scope for the result document. This is needed so it will pass validation against the schema.
        
    -->
    <xsl:template match="/">
        <xsl:text>&#x0a;Test of Subtree&#x0a;</xsl:text>
        
        <xsl:variable name="testST">
            <SubElem1>
                Value for Sub Element 1
            </SubElem1>
            <SubElem2>
                Value for Sub Element 2
            </SubElem2>
        </xsl:variable>
        
        <xsl:value-of select="$testST/SubElem1"></xsl:value-of>
    </xsl:template>
    
</xsl:stylesheet>

The problem is, value of $testST/SubElem1 is not returning anything.
I use these subtrees throughout my stylesheet to pre-process things and
none of it works since I added the default 'xmlns' statement to the stylesheet declaration.

If I remove the default NS decl, then I get "Value for Sub Element 1" as expected from the 'value-of' statement.

What did adding this default NS statement do to my stylesheet?

Thanks!

mphare · August 3rd, 2010, 04:07 PM

Well, #48 of Dave's top 63 reason why I don't understand namespaces provided me a clue.

I changed the definition of the temporary tree to

Code:

<xsl:variable name="testST">
            <SubElem1 xmlns="">
                Value for Sub Element 1
            </SubElem1>
            <SubElem2 xmlns="">
                Value for Sub Element 2
            </SubElem2>
        </xsl:variable>

and it works.

So, my question becomes.. why?

Why can't the temporary results be part of the globally defined default namespace just like anything else? I guess they are and that's the problem.. So, why is it a problem and what's the best solution? I 'could' go through my stylesheet and add the xmlns="" decl to all the temporary elements, but that seems like a brut-force solution. I'm still missing something in the XML/NS/XSLT mix.

It seems the alternative is to remove the global NS decl and place local default NS decls on all elements I write to the results, which also seems a bit brut-force.

Thanks

mhkay · August 3rd, 2010, 04:21 PM

A default namespace declaration in the stylesheet DOES apply to unqualified element names in literal result elements, but does NOT apply to unqualified element names in a path expression.

As for the question why? - meaning why did the WG decide to design the language that way - I think that demands historical research to discover, and in my experience, explaining why a committee made a decision can be difficult even when you were present at the relevant meetings, which in this case I wasn't. (I could interpret the question as being "can you advance any arguments in favour of doing it this way", but that's a different question.)

mphare · August 3rd, 2010, 04:54 PM

Thanks Michael.

The why question wasn't so much a why did they do it this way, that's history and it's done. More, I assume it was done this way for a purpose, and so there must be a solution to my problem. XSLT/XPath is a fascinating language to program in. I can spend hours cooking up solutions to a problem, then discover there is a much better way by either changing my logic, or by learning a new facet of the language. Usually, all I have to do is stop thinking like a C programmer for a minute or two.

For this problem, the solution I came up with is to either give all my temporary elements an empty NS (xmlns=""), or to add a NS to all my queries (which I am still researching).

If that(those) is(are) the solution(s), then I'll proceed. I really just want to be sure that was the best solution and that I was not going about this the wrong way missing some magic construct that makes this all better.

Thanks again!

samjudson · August 4th, 2010, 03:24 AM

An alternative in XSLT 2.0 is to add the xpath-default-namespace attribute to the stylesheet element - this changes the default namespace in your XPATH expressions from the null namespace to the one you define.

Usually this would be the main namespace in your input XML, not the one in your output XML as in your example above.

mphare · August 4th, 2010, 02:27 PM

I tried xpath-default-namespace, it fixed the temporary results trees, but it had the adverse affect of causing all my xpath directives pointing to the source document to fail for the same reason (wrong NS)

The easy solution was to add a xmlns="" directive to all the variable definitions that are building temporary trees causing all the temp tree to be in the null NS, which is what the xpath statements are expecting.

Thanks!