XSLT 2.0 and XPath 2.0 Programmer's Reference, 4th Edition (31 page)

• A generated sequence can also be returned by a function. Rather than defining the sequence of month elements in a variable, it could equally well be defined in a function.
  

• You could then refer to the third month as
  m:months()[3]
  . Note that user-defined functions must always be in a namespace, to avoid conflicts with system-defined functions.
  
• There is a subtle difference between using a variable and a function to capture the sequence: the variable will return the same elements every time, whereas the function will create new elements every time you call it. For example, this means that the expression
  $months[3]
  is $months[3]
  will be
  true
  , while
  m:months()[3] is m:months()[3]
  will be
  false
  . (The
  is
  operator in XPath 2.0 tests whether the values of the two operands are references to the same node.) In practice, you would want to use a function only if you were supplying parameters, because the data that came back depended on the parameters in some way.
  

Text nodes appearing within a sequence constructor are copied to the result sequence when the sequence constructor is evaluated. However, text nodes that consist entirely of whitespace will be ignored, unless the
xml:space
attribute is used on an enclosing element to define them as significant.

For more information on the treatment of whitespace see Chapter 3, page 141.

Nested Sequence Constructors

Suppose the template rule actually looks like this:

Here the

element contains two child elements,

and

. These are both literal result elements, so they are evaluated by copying them from the stylesheet to the result sequence.

Evaluating the

element in turn causes the sequence constructor within the

element to be evaluated. This sequence constructor contains a single literal result element, the