xsl:for-each-group element selects a sequence of nodes and/or atomic
values and organizes them into subsets called groups. There are four possible ways of defining the
group-by: This groups together all items having the same value for a grouping key. The grouping key may have multiple values (a sequence of values) in which case the item is added to more than one group.
group-adjacent: This groups together all items having the same value for a grouping key, provided that they are also adjacent in the input sequence. This is useful when you need to wrap a new element around a sequence of related elements in the source documents, for example a consecutive sequence of
<bullet>elements. In this case the grouping key must be single-valued.
group-starting-with: This processes the items in the supplied sequence in turn, starting a new group whenever one of the items matches a specified pattern. This is useful, for example, when matching an
<h2>element and its following
group-ending-with: This processes the items in the supplied sequence in turn, closing the current group whenever one of the items matches a specified pattern. This is useful when matching a sequence of items in which the last item in the group carries some distinguishing attribute such as
Saxon implements the
xsl:for-each-group instruction in full.
For examples of using the instruction, see
the XSLT 2.0 specification.
In XSLT 3.0, the capabilities of the
are extended by virtue of the fact that the pattern used in
group-ending-with can now match atomic values as well as nodes.
XSLT 3.0 introduces the ability to bind variables to the current group and the current grouping key
as an alternative to using the functions
The advantage of using variables is that they have static scope, so the optimizer has more chance to see
what the stylesheet is doing with each group. Streamed processing is only possible if the new variables
are used, as this enables the processor to determine whether or not the use of the current group is