David Carlisle wrote:

Also you are classing attributes as non-empty content
but not namespace nodes so

<B foo=""/> is non empty but
<B xmlns:x="data:,hello"></B> is empty

Ah, good catch David.

Okay, I added namespace::* to the XPath that David provided, to take into account namespaces. Here is the new XPath: empty((node(),@*,namespace::*))

Here are the results I got from running my XSLT program using the new XPath:

Perfect!

Those results are from running SAXON in Oxygen XML.

John and David pointed out I might get different results with another parser. Specifically, another parser might discard comments and the whitespace in <B>  </B>. Such a parser would produce these results:

Is there any way to write an XPath expression that always produces the results shown in the first table, regardless of which parser is used?

/Roger