XOM API for Processing XML with Java

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

A thought has occurred to me while experimenting with Wolfgang's latest 
optimizations. This isn't an optimization one way or the other, really. 
It just happened to come up when I was looking at that code. Consider 
this unusual case:

<element>Text<[![CDATA[data]]>Text<[![CDATA[data]]>Text</element>

A slightly more common variation might be:

<element>
  <[![CDATA[data]]>
</element>

However it's basically a text node that contains one or more CDATA 
sections as well as some text outside CDATA sections. Currently when XOM 
encounters this it represents it as plain text. That is,

<element>TextdataTextdataText</element>

That is, it uses a regular text node and throws away the CDATA sections. 
There's no way XOM can preserve the multiple plain texts and CDATA 
sections because it accumulates all text nodes into a single node on 
parsing. However, instead of using a single Text object here it could 
use a single CDATASection object. That is, it could instead produce this:

<element><[![CDATA[TextdataTextdataText]]></element>

Because the CDATASection class is hidden in the non-public API, this 
really only affects output from toXML and the Serializer. Would this be 
preferable? Or is it better to just output plain text with appropriate 
escapes when mixing CDATA sections with content in non-CDATA sections?

The semantic effect here is (or should be) nil. It's really just a 
question of aesthetic preference.

--
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