"Arndt Jonasson" <arndt.jonasson@g...> wrote in message 
news:3d8ef72b-e938-482d-9c1c-43583584331a@z......
>
> Let's say we have a schema (maybe expressed in XML Schema, but not
> necessarily so), that allows this instance document:
>
> <top>
>  <txt>This is text</txt>
>  <books>
>    <book>Tarzan</book>
>    <book>Harry Potter</book>
>  </books>
> </top>
>
> The text /top/txt may be empty, and the element /top/books may have no
> children, so this instance document is also allowed:
>
> <top>
>  <txt/>
>  <books/>
> </top>
>
> I now want to write an XPath expression that selects all nodes that do
> not have children in the schema. It would always select /top/txt and
> it would never select /top/books, even in the second example above.
>

This is possible in XPath 2.0 if the schema has separate types for all cases 
of element that must not have children-elements.

Then one can use the so called ElementTest, which is defined in the 
following way:

     ElementTest    ::=    "element" "(" (ElementNameOrWildcard ("," 
TypeName "?"?)?)? ")"


One of the possible XPath 2.0 expressions will be something like the 
following:

             //element(*,Type1) | //element(*,Type2) | ... | 
//element(*,TypeN)


where Type1, Type2, ..., typeN  are all the schema types that define 
elements that cannot have children-elements.

Probably substitution groups can be used so that all types above can be 
derived from a single abstract type, let's say "ChildlessElement".

Then the expression would be simply:

              //element(*,ChildlessElement)


Of course, to be able to evaluate such XPath 2.0 expressions one must have a 
full-blown XPath 2.0 implementation (either a Schema-Aware XSLT 2.0 
processor, or an XQuery processor)

For more information see the XPath 2.0 spec:

  http://www.w3.org/TR/xpath20/#doc-xpath-ElementTest

  http://www.w3.org/TR/xpath20/#id-element-test


Cheers,
Dimitre Novatchev