I have two questions that can bot be illustrated with the same base files. I have two XML docs I want to merge and generate different lists.

The first list has the nodes related to the first occurrence of a "node" based on the value of an element within that node.


To illustrate: here's cars1.xml

<?xml version="1.0" encoding="utf-8" ?>
<cars>
           <car owner="BILL" >
        <Manufacturer>Mazda</Manufacturer>
        <Color>Black</Color>
        </car>
           <car owner="Sam" >
        <Manufacturer>Caddy</Manufacturer>
        <Color>white</Color>
        </car>
           <car owner="Paul" >
        <Manufacturer>Audi</Manufacturer>
        <Color>Grey</Color>
        </car>
</cars>

and here's cars2.xml
<?xml version="1.0" encoding="utf-8" ?>
<cars>
           <car owner="Paula" >
        <Manufacturer>Audi</Manufacturer>
        <Color>Purple</Color>
        </car>
           <car owner="Jim" >
        <Manufacturer>Buick</Manufacturer>
        <Color>white</Color>
        </car>
           <car owner="tammy" >
        <Manufacturer>Maserati</Manufacturer>
        <Color>Red</Color>
        </car>
</cars>

The result I'm looking for is a list of cars that is the first instance for any manufacturer after merging cars1 and cars2.

I know how to merge two (or more) XML docs into one using xsl:copy and document().

Can I somehow generate the file below without doing a second transform that looks for not(preceding-sibling::...)?

Here's what I hope to get:

<?xml version="1.0" encoding="utf-8" ?>
<cars>
           <car owner="BILL" >
        <Manufacturer>Mazda</Manufacturer>
        <Color>Black</Color>
        </car>
           <car owner="Sam" >
        <Manufacturer>Caddy</Manufacturer>
        <Color>white</Color>
        </car>
           <car owner="Paul" >
        <Manufacturer>Audi</Manufacturer>
        <Color>Grey</Color>
        </car>
           <car owner="Jim" >
        <Manufacturer>Buick</Manufacturer>
        <Color>white</Color>
        </car>
           <car owner="tammy" >
        <Manufacturer>Maserati</Manufacturer>
        <Color>Red</Color>
        </car>
</cars>

Note that there are only five car nodes- not six- the second "Audi" record is missing. Sorting can take place without problem if helpful.

Is there a way to do this in a single XSLT?

Second question: Using the same files above. Is there a way to preserve all manufacturers listed in cars2, but replace "matching manufacturer" records from cars1 wher they apply7 The desired result in that case would be:
<?xml version="1.0" encoding="utf-8" ?>
<cars>
           <car owner="Paul" >
        <Manufacturer>Audi</Manufacturer>
        <Color>Grey</Color>
        </car>
           <car owner="Jim" >
        <Manufacturer>Buick</Manufacturer>
        <Color>white</Color>
        </car>
           <car owner="tammy" >
        <Manufacturer>Maserati</Manufacturer>
        <Color>Red</Color>
        </car>
</cars>


I've seen some nifty code to do similar sorts of things for matches in both, but can't figure this one out.

All help is appreciated.

Most of the XSLT 1.0 grouping techniques work only on nodes in a single document, so you have to merge the documents first and then do grouping. The main exception is Ken Holman's "variable grouping" technique described at http://sourceware.org/ml/xsl-list/2001-10/msg00933.html

With XSLT 2.0, of course, xsl:for-each-group solves it easily.

Michael Kay
http://www.saxonica.com/
Author, XSLT Programmer's Reference and XPath 2.0 Programmer's Reference

Thanks for saving me some time trying to find a solution that uses only one transform!

Did you/anyone else have any thoughts on how to accomplish the second question?