XSLT 2.0 Idea: third argument for key()

[Jeni Tennison <mail@xxxxxxxxxxxxxxxx>]

Hello everyone,

In brief: I'd like to suggest having a third argument for key(), a node
set, such that the key only retrieves nodes with the relevant key value
that are within this node set.

In detail: There are two related limitations on using key() as it stands.
The first is the way it interacts with document() [nodes are only retrieved
by key() within the document that the current node is in].  The second is
in restricting the nodes retrieved by key() to a certain subset of the
document.

As an example of the first, say you want to sort values in an arbitrary
order based on some data defined within your stylesheet:

<foo:order>
  <item>first</item>
  <item>second</item>
  <item>third</item>
  <item>fourth</item>
</foo:order>
<key name="order"
     match="foo:order/item"
     use="." />

It is difficult to use the position of these nodes in order to sort
elements within the source because the only way it could plausibly be
expressed is:

<ol>
  <xsl:for-each select="runner">
    <xsl:sort select="count(key('order', @rank)/preceding-sibling::item)"
              data-type="number" />
    <li><xsl:value-of select="name" /></li>
  </xsl:for-each>
</ol>

and in this, the current node when key() is called is always the node in
the source ('runner' element), and therefore the key() only retrieves nodes
from the source, never from the stylesheet.

[David Carlisle's pointed out that you can get around this using an
extension element like saxon:function that allows you to define extension
functions in XSLT to retrieve the required value.]

With a third argument on key, you could do:

<xsl:variable name="items" select="document('')/foo:order/item" />

<ol>
  <xsl:for-each select="runner">
    <xsl:sort
      select="count(key('order', @rank, $items)/preceding-sibling::item)"
      data-type="number" />
    <li><xsl:value-of select="name" /></li>
  </xsl:for-each>
</ol>

Here's an example of the second area in which this would be useful,
restricting nodes to a known node set: say I had a document with a massive
dataset like:

<athletes>
  <country id="GRB">
    <athlete event="100 metres">...</athlete>
    <athlete event="200 metres">...</athlete>
    <!-- another 400 athletes -->
  </country>
  <country id="ROI">
    <athlete event="100 metres">...</athlete>
    <athlete event="800 metres">...</athlete>
    <!-- another 400 athletes -->
  </country>
  ...
</athletes>

I want to put the details of the athletes in a table, with columns being
countries and rows being events, so I use a key to index on the event, and
access the data for the table using:

  key('athletes', @event)[generate-id(parent::country) =
                          generate-id(current()/parent::country)]

or something similar.  It would be a lot cleaner to do:

  key('athletes', @event, parent::country/child::athlete)

[or key('athletes', @event, ../*) for brevity]

(i.e. search for only the athletes that are children of the parent::country
of the current node).

A third argument on key() could not only enable you to expand the search to
other documents, but also to restrict it within the current one - to choose
which nodes to take into consideration.

Any thoughts?  Objections?  Implementation issues?

Jeni

Jeni Tennison
http://www.jenitennison.com/


 XSL-List info and archive:  http://www.mulberrytech.com/xsl/xsl-list