December 03, 2004

Michael Kay compares XSLT and XQuery

In a recent post to xml-dev Michael Kay writes in response to a post questioning the lack of comparisons between XSLT and XQuery:

It seems odd to me that I’ve never seen an article comparing actual XSLT and XQuery code for say moderately complex operations.

Apart from my Knight’s Tour, which Jonathan cited (and which is a very untypical application), I thought I’d try a very simple transformation that I had to do recently. The task is to remove those parts of a document rooted at an element having the attribute secret=”yes”. This is bog-standard XSLT and doesn’t need any 2.0 features:

<xsl:template match="*">
<xsl:copy>
  <xsl:copy-of select="@*"/>
  <xsl:apply-templates/>
</xsl:copy>
</xsl:template>

<xsl:template match="*[@secret='yes']"/>

7 lines, of which 6 are copy-and-paste from many other stylesheets that you’ve written before. (Add two lines if you will for the header and footer).

Here’s an attempt at an XQuery solution:

declare function m:copy($e as element()) as element()? {
  if ($e/@secret = 'yes')
    then ()
    else element {node-name($e)} { 
           $e/@*, 
           for $n in $e/node() return 
               if $n instance of element()
               then m:copy($n)
               else $n
    }
}

document{
  m:copy(doc('input.xml')/*)
}

14 lines of code (twice as much), and less scope for reuse. No great difference in the conceptual difficulty - both require an understanding of recursion - but I think XSLT emerges here as the winner. (And remember this is only XSLT 1.0).

But in fact there’s a bigger issue - the XQuery code is wrong. It loses the namespaces from the source document. Perhaps I’m missing something, but I can’t see any way to solve this. XQuery may be computationally complete, but it’s not actually closed over the data model - there is no way of generating a namespace dynamically. As far as I can see at the moment, that’s a stopper, and this exercise has to be coded in XSLT.

Of course, I’m sure there are other problems where XQuery has the edge.

Michael Kay http://www.saxonica.com/

I think we have a new quote of the day!!! (an area in which sadly has not had much attention since Dimitre submitted the first entry about a week ago)…

Something like…

“XQuery may be computationally complete, but it’s not actually closed over the data model - there is no way of generating a namespace dynamically. As far as I can see at the moment, that’s a stopper, and this exercise has to be coded in XSLT.”

I think it’s time to update the “XSLT Quote of the Day” :)

I’ll do that now…