XPath: How to check multiple attributes across similar nodes

If I have some xml l开发者_开发知识库ike:

    <root>
   <customers>
        <customer firstname="Joe" lastname="Bloggs" description="Member of the Bloggs family"/>
        <customer firstname="Joe" lastname="Soap" description="Member of the Soap family"/>
        <customer firstname="Fred" lastname="Bloggs" description="Member of the Bloggs family"/>
        <customer firstname="Jane" lastname="Bloggs" description="Is a member of the Bloggs family"/>
   </customers>
 </root>

How do I get, in pure XPath - not XSLT - an xpath expression that detects rows where lastname is the same, but has a different description? So it would pull the last node above?

How do I get, in pure XPath - not XSLT - an xpath expression that detects rows where lastname is the same, but has a different description?

Here's how to do this with a single XPath expression:

   "/*/*/customer
        [@lastname='Bloggs'
       and
        not(@description
           = preceding-sibling::*[@lastname='Bloggs']/@description
            )
        ]"

This expression selects all <customer> elements with attribute lastname equal to "Bloggs" and different value of the attribute description.

The selected nodes are:

<customer firstname="Joe" lastname="Bloggs" description="Member of the Bloggs family"/>
<customer firstname="Jane" lastname="Bloggs" description="Is a member of the Bloggs family"/>

/root/customers/customer[@lastname='Bloggs'
  and not(@description = preceding-sibling::*[@lastname='Bloggs']/@description)
  and not(@description = following-sibling::*[@lastname='Bloggs']/@description)]

It would perform better doing it in steps, though.