On Thursday 12 Dec 2002 3:26 am, Gunther Schadow wrote:
Hi Michael and other XSLT engine builders,
among all this lazy evaluation and optimization how about piping SAXON
events right into templates building only partial trees? Suppose a
transform doesn't use global forward access with XPath but only
accesses the local area in the current node and nodes seen before.
Could be a bit faster that way, and more memory efficient? Are you
guys thinking of doing this?
There has been some interesting results coming from publish/subscribe systems
that use XPath as a selection language, e.g. XFilter, although event based
XPath is so very tricky I would not hold your breath. Many of the commonly
used, publically available processors are lagging quite away behind the best
performers (orders of magnitude differences are common) which would indicate
there is plenty of head room in current models for most processors.
Kev.
XSL-List info and archive: http://www.mulberrytech.com/xsl/xsl-list