The point here was not to find the “right” way to do it, but to find how far XPath can be abused in an attempt to use it.

That’s why I wrote somewhere in the post that “using the right tool for the job is just no fun” and that “practically-minded people would have long moved to another query language or at least to version 2.0 of XPath”.

William

Thanks for the Groovy code Paul (Paul has more XPath/Groovy examples here for those interested).

BTW, if anyone wants to play with this graph query but doesn’t have an XPath tool handy, you can use this one. Then use http://stage.vambenepe.com/pages/xpath-brain-teaser.xml as the URL to the XML file to load the document that I use in this example.

<xsl:variable name=”a” select=”/doc/person[@name='alan']/g[not(.=preceding-sibling::g)]“/>
<xsl:variable name=”p” select=”/doc/person[@name='paul']/g”/>
<xsl:variable name=”i” select=”/doc/person[@name='ivan']/g”/>

<xsl:copy-of select=”$a[.=$p][.=$i]“/>

/doc/person[@name='alan']/g
[.=../../person[@name='paul']/g]
[.=../../person[@name='ivan']/g]”

which using a [][] form rather than [ []] doesn’t have quite such bad behaviour even if evaluated naively.

And adding as you suggested [not(.=preceding-sibling::g)] if you need to filter out duplicates)

Of course if your xpath is being hosted by xslt or any other API that allows you to use variables it’s easy to avoid the potentially expensive ../../person steps

but it’s quite likley that an xpath engine will rewrite the first form to the second anyway, inferring local variables to take the constant expression out of the loop.

David

def root = new XmlParser().parseText(xml)
Set groups = root.person.g.collect{ it.text() }
assert groups.findAll{ groupId ->
root.person.g.findAll{
it.text() == groupId
}.’..’.'@name’.containsAll(['alan', 'paul', 'ivan'])
} == ['2', '4']

john