Question

For this example I would like to get the first b element from every t (real xml might be more deeply nested). Unfortunately I am limited to xpath 1.0. My initial thought was something like //t//b[position()=1] but i can’t get it to work.

<t>
 <a>
   <b/>
   <b/>
   <b/>
 </a>
</t>
<t>
 <a>
   <b/>
   <b/>
   <b/>
 </a>
</t>

Answer 1

That’s almost right. First note that the predicate [position()=1] is equivalent to simply [1], I’ll use the shorter form from now on. Now by definition

//t//b[1]

is shorthand for

//t/descendant-or-self::node()/child::b[1]

so it will give you all b elements that are nested somewhere inside a t and are the first b child of their respective parent elements. Thus given

<t>
  <a>
    <b attr="1"/>
    <b attr="2"/>
  </a>
  <a>
    <b attr="3"/>
  </a>
</t>

you would get b’s 1 and 3. If you only want the first b inside each t then you need

//t/descendant::b[1]

Given the example XML above this would return only the <b attr="1"/>

You can usually treat .//x as equivalent to descendant::x but this is one of the edge cases that shows up the subtle distinction between the two…