XSLT 1.0: restrict entries in a nodeset

Question

Being relatively new to XSLT I have what I hope is a simple question. I have some flat XML files, which can be pretty big (eg. 7MB) that I need to make 'more hierarchical'. For example, the flat XML might look like this:

<D0011>
    <b/>
    <c/>
    <d/>
    <e/>
    <b/>
    ....
    ....
</D0011>

and it should end up looking like this:

<D0011>
  <b>
    <c/>
    <d/>
    <e/>
  </b>
  <b>
 ....
 ....
</D0011>

I have a working XSLT for this, and it essentially gets a nodeset of all the b elements and then uses the 'following-sibling' axis to get a nodeset of the nodes following the current b node (ie. following-sibling::*[position()=$nodePos]). Then recursion is used to add the siblings into the result tree until another b element is found (I have parameterised it of course, to make it more generic).

I also have a solution that just sends the position in the XML of the next b node and selects the nodes after that one after the other (using recursion) via a *[position() = $nodePos] selection.

The problem is that the time to execute the transformation increases unacceptably with the size of the XML file. Looking into it with XML Spy it seems that it is the 'following-sibling' and 'position()=' that take the time in the two respective methods.

What I really need is a way of restricting the number of nodes in the above selections, so fewer comparisons are performed: every time the position is tested, every node in the nodeset is tested to see if its position is the right one. Is there a way to do that ? Any other suggestions ?

Thanks,

Mike

Answer 1

Yes there is a way to do it much more efficiently: See Muenchian grouping. If having looked at this you need more help with the details, let us know. The key you'll need is something like:

<xsl:key name="elements-by-group" match="*[not(self::b)]"
   use="generate-id(preceding-sibling::b[1])" />

Then you can iterate over the <b> elements, and for each one, use key('elements-by-group', generate-id()) to get the elements that immediately follow that <b>.

The task of "making the XML more hierarchical" is sometimes called up-conversion, and your scenario is a classic case for it. As you may know, XSLT 2.0 has very useful grouping features that are easier to use than the Muenchian method.

In your case it sounds like you would use <xsl:for-each-group group-starting-with="b" /> or, to parameterize the element name, <xsl:for-each-group group-starting-with="*[local-name() = 'b']" />. But maybe you already considered that and can't use XSLT 2.0 in your environment.

Update:

In response to the request for parameterization, here's a way to do it without a key. Note though that it may be much slower, depending on your XSLT processor.

<xsl:template match="D0011">
   <xsl:for-each select="*[local-name() = $sep]">
      <xsl:copy>
         <xsl:copy-of select="following-sibling::*[not(local-name() = $sep)
               and generate-id(preceding-sibling::*[local-name() = $sep][1]) =
                    generate-id(current())]" />
      </xsl:copy>
   </xsl:for-each>      
</xsl:template>

As noted in the comment, you can keep the performance benefit of keys by defining several different keys, one for each possible value of the parameter. You then select which key to use by using an <xsl:choose>.

Update 2:

To make the group-starting element be defined based on /*/*[2], instead of based on a parameter, use

<xsl:key name="elements-by-group"
   match="*[not(local-name(.) = local-name(/*/*[2]))]"
   use="generate-id(preceding-sibling::*
                           [local-name(.) = local-name(/*/*[2])][1])" />

<xsl:template match="D0011">
   <xsl:for-each select="*[local-name(.) = local-name(../*[2])]">
      <xsl:copy>
         <xsl:copy-of select="key('elements-by-group', generate-id())"/>
      </xsl:copy>
   </xsl:for-each>
</xsl:template>

Answer 2

<xsl:key name="k1" match="D0011/*[not(self::b)]" use="generate-id(preceding-sibling::b[1])"/>

<xsl:template match="D0011">
  <xsl:copy>
    <xsl:apply-templates select="b"/>
  </xsl:copy>
</xsl:template>

<xsl:template match="D0011/b">
  <xsl:copy>
    <xsl:copy-of select="key('k1', generate-id())"/>
  </xsl:copy>
</xsl:template>

Answer 3

This is the fine grained trasversal pattern:

<xsl:stylesheet version="1.0" xmlns:xsl="http://www.w3.org/1999/XSL/Transform">
    <xsl:template match="node()|@*" name="identity">
        <xsl:copy>
            <xsl:apply-templates select="node()[1]|@*"/>
        </xsl:copy>
        <xsl:apply-templates select="following-sibling::node()[1]"/>
    </xsl:template>
    <xsl:template match="b[1]" name="group">
        <xsl:copy>
            <xsl:apply-templates select="following-sibling::node()[1]"/>
        </xsl:copy>
        <xsl:apply-templates select="following-sibling::b[1]" mode="group"/>
    </xsl:template>
    <xsl:template match="b[position()!=1]"/>
    <xsl:template match="b" mode="group">
        <xsl:call-template name="group"/>
    </xsl:template>
</xsl:stylesheet>

Output:

<D0011>
    <b>
        <c></c>
        <d></d>
        <e></e>
    </b>
    <b>
    ....
    ....
    </b>
</D0011>