From: Michael Kay <mike@sa...> - 2009-01-29 13:22:16
> I am interested in how saxon is using xml schema to increase
> performance. Is there any documentation about this issue?
There is some information (rather brief, I'm afraid) in my recent paper for
the Data Engineering Bulletin:
See section 3.6.
I see the major benefits of schema-awareness as being in robustness / ease
of debugging, rather than performance per se. As the paper suggests, the
performance impact is highly variable, and can sometimes be negative. The
biggest win is when the search space for a path such as .//a can be reduced
to a small part of the document; however, users with sufficient engineering
nous to make their code schema-aware also tend to have the good sense to
write their paths more selectively in such cases!
One thing not really discussed in the paper is storing typed values. Current
releases of Saxon only store the string value, and materialize the typed
value on demand. I'm currently experimenting with caching the typed value,
and it's showing significant benefits for some queries.
Get latest updates about Open Source Projects, Conferences and News.