Re: [xml-dev] When did you vanish, o' CDATA section wrapper?

[David Carlisle <davidc@nag.co.uk>]

On 21/08/2012 18:13, Costello, Roger L. wrote:
> Hi Folks,
>
> I opened a document in Notepad and it showed me this:
>
> <?xml version="1.0" encoding="UTF-8"?> <Document> <![CDATA[ if (a <
> b) & (b < c) then x = 23; ]]> </Document>
>
> I will refer to this set of characters
>
> <![CDATA[]]>
>
> as the "CDATA section wrapper"
>
> That is, it is the CDATA section sans content.
>
> I gave the document to an XML parser. The XML parser ingested the
> document and then gave its output to an XML application (e.g., an
> XSLT processor or an XML Schema validator).
>
> document ---> XML Parser ---> XML Application
>
> Somewhere between the time the XML parser ingested the document and
> the time the XML application ingested the output of the XML parser,

most likely the CDATA section wasn't recorded after the initial parse 
(as others have said XSLT works off an XDM view of the data not DOM and 
XDM doesn't record CDATA sections.

> the CDATA section wrapper was removed (and reserved characters inside
> the CDATA section were escaped).

the characters are not escaped in the internal model "<" is the single 
character "<". When the result of the XSLT is serialised back to an XML 
document "<" in content will need to be escaped (however they were 
entered).

>
> document ---> CDATA section wrapper present ---> XML Parser --->
> CDATA section wrapper absent ---> XML Application
>
> Precisely when did the CDATA section wrapper get removed?
>
> I believe the DOM API contains methods for accessing a CDATA section,
> right?

Most XSLT engines don't use the DOM (and even those like msxsl that do 
natively use a DOM model the DOM API isn't accessible from XSLT.
>
> The XML parser ingests the document and builds a DOM tree, right?

Not usually a DOM, but it is usually an in-mempry tree of some sort.
>
> So the CDATA section wrapper is still in the DOM tree, right?

Not usually. Think of the choice of using <![CDATA[<]]> and &lt; like 
the choice of using attribute value quotes a="b" or a='b'  the two forms 
look different but the model doesn't record which form is used and so an 
identity transform may use either form, the original is not preserved.
>
> XML Parser:
>
> ---> ingest ---> build DOM ---> remove CDATA section wrapper now?
> ---> do what? --->
>
> /Roger
>
>

David



-- 
google plus: https:/profiles.google.com/d.p.carlisle