RE: [xml-dev] XML aggregation question?

["Michael Kay" <mike@xxxxxxxxxxxx>]

> Based on all the comments thus far as well as reading some of 
> the articles/documentation on eXist, it would seem that an 
> XML database is really the only viable choice if I want to 
> keep my data as XML and still provide aggregated views across 
> the instances based on values of attributes (or other 
> expressions using XPath and/or XQuery).


This isn't absolutely true. For example, the W3C XSLT test suite (not
published, unfortunately) is a collection of over 5000 XML documents held in
filestore, and it's quite feasible to run queries in Saxon (using either
XSLT or XQuery) across this collection. For example, to count how many of
the stylesheets in the collection specify version="1.0" on the
xsl:stylesheet element, use:

count(collection('file:///c:/xslts_1_0_0/TestInputs?select=*.xsl;recurse=yes
')
         [(xsl:stylesheet|xsl:transform)/@version='1.0'])

What you don't get with this approach is performance. There's no database
load operation, so there are no collection-level indexes: the system works
its way through the directory parsing and testing each individual document.
But it's still very useful (and surprisingly fast) for the occasional ad-hoc
search.

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