Using saxon:parse() on CDATA Section

Belch · June 2nd, 2015, 05:43 AM

Hi, I have some input XML which contains nested XML in a CDATA section, as so:

Code:

<body>
<![CDATA[
<name>Bob</name>
<age>34</age>
]]>
</body>

I would like to be able to use XPath to interrogate the XML inside the CDATA and from what I've read saxon:parse() is the way to go (as we are using the Saxon processor).

Currently in my stylesheet I set a variable as so:

Code:

<xsl:variable name="data">
<xsl:value-of select="//body" />
</xsl:variable>

Note I have tried this with the disable-output-escaping="no" attribute on value-of as well.

I then set another variable as so:

Code:

<xsl:variable name="final" select="saxon:parse($data)" />

Outputting the value of $final just gives me a lot of blank lines.
Outputting the value of $data I can see that the XML is now escaped i.e.:

Code:

&lt;name&gt;Bob&lt;/name&gt;
&lt;age&gt;34&lt;/age&gt;

I cannot unescape the XML before passing it into saxon:parse(), the disable-output-escaping option only seems to affect how it looks rather than how it is treated.

If I add the disable-output-escaping attribute to the output of the $data variable, it is then unescaped.

Is anyone able to advise if I'm missing a step, or if it depends on how the transform is being processed? In this case I'm using Eclipse with Orangevolt and processor Saxon-B 8.9.

Any help appreciated, thanks in advance.

Martin Honnen · June 2nd, 2015, 05:57 AM

Are you sure you don't get a parse error with saxon:parse($data) as your markup lacks a root element? In Saxon 9.6 you could use parse-xml-fragment http://www.w3.org/TR/xpath-functions...e-xml-fragment, but I don't think it exists in 8.9 (which is very old). If you want to successfully parse the data you have with saxon:parse then I would try

Code:

saxon:parse(concat('&lt;root>', $data, '&lt;/root>'))

mhkay · June 2nd, 2015, 06:02 AM

I'm a little surprised this works because according to the spec at http://www.saxonica.com/documentatio...ns/saxon/parse the input to saxon:parse must be well-formed XML, that is, there must be an outermost element that wraps everything else. So I would expect a dynamic error. But 8.9 is a rather old release and I don't have its documentation handy.

I would suggest using the XSLT 3.0 parse-xml-fragment() function in preference: see http://www.w3.org/TR/xpath-functions...e-xml-fragment For that you'll need a more up-to-date Saxon release.

Belch · June 2nd, 2015, 06:06 AM

Sorry, that input XML was just an example.
The actual XML I'm using does have a root element wrapping everything else, hence no error.

Unfortunately I'm limited to XSLT 2.0 in this case.

Martin Honnen · June 2nd, 2015, 06:28 AM

Can you post minimal but complete samples of XSLT and XML that show the problem and allow us to reproduce it? You haven't shown what you do in case of "Outputting the value of $final just gives me a lot of blank lines". Show us the code you have, the output you want and the one you get.

Belch · June 2nd, 2015, 09:30 AM

It appears I have "rubber-ducked" this issue.
As I was posting the actual XML and XSLT causing the issue I realised that I was using a value-of in the wrong place (and therefore getting the values of the parsed XML rather than the nodes).

For reference or in case anyone else makes the same mistake, the variables I'm now setting are:

Code:

<xsl:variable name="requestXML" select="saxon:parse(//xrt:RequestXML)" />
<xsl:variable name="polData" select="$requestXML/PolMessage/PolData[@Type='Input']" />

where $polData is setting the root so further XPath is all relative.

Thanks for your assistance in getting me to this point.