Detecting CDATA sections with XSLT

[Dave Matthews]

Hi folks,

I'm writing a web-page editing tool for my company which will allow staff (with no "technical" expertise) to maintain their own Intranet sites. The content for each webpage is stored in the form of XHTML in an XML document (which, in turn, is stored in an XML database). So far so good. However the editing tool must allow users to paste in the contents of MS Word documents. I soon discovered that Word does not generate properly-formed HTML, the main problem being that tags that should be nested are often "overlapped" (as my example below shows). My solution is to store this "bad" data as CDATA sections, thereby preventing the finished XML document from being invalidated. My finished XML document looks something like this:

<page id="0001">
<content>
<p>
<i>
<font face="Arial">Properly-formed HTML</font>
</i>
</p>
<![CDATA[<p><i><font face="Arial">The 'i' and 'font' end-tags are wrong and there is no end-tag for 'p'</i></font>]]> <p> <i> <font face="Arial">This is OK.</font> </i> </p> </content> </page>

On retrieving a document for formatting and display within the client browser, my XSL template for the <content> nodes needs to be able to detect whether each of its children can be regarded as proper XML (and, therefore, to transform the it into HTML) or a CDATA section whose contents will simply be passed straight to the browser. So my template needs to look something like this:

<xsl:template match="content">
<xsl:for-each select="*">
<xsl:choose>
<xsl:when test="nodetype(.)=cdata">
<xsl:value-of select=".">
</xsl:when>
<xsltherwise>
<xsl:apply-templates/>
</xsltherwise>
</xsl:choose>
</xsl:for-each>
</xsl:template>

Of course it's the fourth line of code - <xsl:when
test="nodetype(.)=cdata"> - that is giving me problems. Unfortunately I am stuck with a fairly basic XSLT engine that has none of the fancy additional functions MSXML, SAXON or Xalan offer. Try as I might, I can't find a way of getting XSLT to tell when it's dealing with a CDATA section.

(I could simply hold everything as CDATA but in the future I am going to have to interface with other systems that will demand as much content a possible be presented as proper XML/XHTML.)

Any ideas would be very much appreciated!

 

[Crutcher Dunnavant]

Well, what you are asking is impossible (sortof).

CDATA sections are identical, after parsing, to a text node with the contents of the CDATA section, except you don't need things like &amp; or &lt; or the like in the CDATA seciton. It is just there to make putting XML or source code in a text node easier, but once it is _in_ a text node, you loose any structure.

Now, if you were especially enterprising, you _could_ attempt to locate matching pairs of element open/closes within a textnode, and generate an appropriate element from them, but, that's going to be a _lot_ of XSLT, and it's going to take expert level skill to do.

 

[Rolf Magnus]

On retrieving a document for formatting and display within the client browser, my XSL template for the <content> nodes needs to be able to detect whether each of its children can be regarded as proper XML (and, therefore, to transform the it into HTML) or a CDATA section whose contents will simply be passed straight to the browser.

xslt doesn't see CDATA secions. They're converted to text nodes before xslt even sees them. But couldn't you just wrap an element around it?

<page id="0001">
    <content>
        <p>
            <i>
                <font face="Arial">Properly-formed HTML</font> </i> </p> <non-wellformed> <![CDATA[<p><i><font face="Arial">The 'i' and 'font' end-tags are wrong and there is no end-tag for 'p'</i></font>]]> </non-wellformed> <p> <i> <font face="Arial">This is OK.</font> </i> </p> </content> </page>

 

[Dave Matthews]

Thanks for your help, guys. You've confirmed what I was rapidly coming to suspect!

Rolf - your idea of using a "wrapping" element seems an ideal way around the problem - many thanks!