XSLT 2.0 and XPath 2.0 Programmer's Reference, 4th Edition (381 page)

Effect

So what exactly does the
/
operator do?

I will explain this in terms of an expression
E1/E2
, where
E1
and
E2
are arbitrary expressions. This expression is evaluated as follows:

• E1
  is evaluated to produce a sequence of nodes; let's call this
  S1
  . If the result of
  E1
  contains an atomic value, a type error is reported.
  
• For each node in
  S1
  , the expression
  E2
  is evaluated. The context for evaluating
  E2
  has this node from
  S1
  as the context node. It also has the position of this node in the sequence
  S1
  as the context position, and the number of nodes in
  S1
  as the context size, but in practice it's very rare to write an expression on the right-hand side of
  /
  that depends on the context position or size.
  
• Each time
  E2
  is evaluated, it produces a sequence which must either consist entirely of nodes, or entirely of atomic values (if it doesn't, a type error is reported). For the moment, we're only interested in the case where it delivers nodes; the other case is described in Chapter 10 (page 644). All the nodes produced when
  E2
  has been evaluated once for every node in
  S1
  are bundled together into a single sequence. Duplicate nodes are then removed, and the remaining nodes are sorted into document order.
  
• The resulting sequence of nodes forms the result of the path expression
  E1/E2
  .
  

The most common kind of expression to use on the right hand side of
/
is an
axis step
. We'll describe axis steps in detail later in this chapter, on page 606. The syntax also allows a
FilterExpr
, which in effect means any expression, except that if it contains operators that bind less tightly than
/
(an obvious example is
|
) then it must be written in parentheses.