XSLT 2.0 and XPath 2.0 Programmer's Reference, 4th Edition (140 page)

If there is more than one attribute set named
A1
these must be merged first (taking import precedence into account), and then the merged contents must be substituted into
B
.

Then the same process is applied to
A2
. The referenced attribute sets are expanded in order. The attributes that result from expanding
A1
are output before those that result from expanding
A2
. This means that attributes from
A2
take priority over those from
A1
if there is a clash, because of the rule that when several attributes have the same name, the last one wins.

In turn, the attribute set
B
must be fully expanded before it is merged with any other attribute set called
B
. That is, the processor must replace the references to attribute sets
A1
and
A2
with an equivalent list of

instructions before it merges this
B
with other attribute sets of the same name.

When
B
is expanded, the attributes derived from
A1
and
A2
will be output before the attributes
p
,
q
, and
r
, so if expanding
A1
and
A2
generates any attributes called
p
,
q
, and
r
, these will be overwritten by the values specified within
B
(
percy
,
queenie
, and
rory
).

Duplicate attribute names or attribute set names do not cause an error. If several attributes have the same name, the one that comes last (in the order produced by the merging rules given above) will take precedence.

Usage

The most common use of attribute sets is to define packages of attributes that constitute a display style, for example a collection of attributes for a font or for a table. They are often used when generating XSL-FO.

A named attribute set is used by referring to it in the
use-attribute-sets
attribute of the

or

element or in the
xsl:use-attribute-sets
attribute of a literal result element or, of course, in the
use-attribute-sets
attribute of another

. The first three cases all create an element node and have the effect of adding the attributes in the named attribute set to that element node. Any attributes added implicitly from a named attribute set can be overridden by attribute nodes added explicitly by the invoking code.

An attribute set is not simply a textual macro. The attributes contained in the attribute set each have a
select
attribute or sequence constructor to define the value, and although this will often return a fixed value, it may also, for example, declare variables or invoke other XSLT instructions such as

and

.

The rules for the scope of variables, described under

on page 500, are the same as anywhere else, and are defined by the position of the definitions in the source stylesheet document. This means that the only way to parameterize the values of attributes in a named attribute set is by reference to global variables and parameters: There is no other way of passing parameters to an attribute set. However, the value of the generated attributes may depend on the context in the source document. The context is not changed when the attribute set is used, so the context item (
.
) as well as the context position and size are exactly the same as in the calling instruction.

Examples

The following example defines an attribute set designed for generated HTML

This attribute set can be used when generating an output element, as follows:

  …

  …

  …

   2

   cols

Example: Using an Attribute Set for Numbering

This is a rather untypical example, designed to show that attribute sets are more powerful than you might imagine. Suppose you want to copy an XML file containing a poem, but with the

elements in the poem output in the form

within the stanza.

Source

The source file
poem.xml
has the following structure (I'm only showing the first stanza):

Stylesheet

The stylesheet
number-lines.xsl
copies everything unchanged except the

elements, which are copied with a named attribute set:

Output

The output (again showing only the first stanza) looks like this: