[
Lists Home |
Date Index |
Thread Index
]
Michael Kay wrote:
> The standard algorithms for generating a finite-state machine from a
grammar
> suffer from the problem that the number of possible states becomes very
> large when you use a large value of minOccurs or maxOccurs, which
causes an
> increasing use of memory. Saxon's schema processor deals with this by
> treating a large maxOccurs as if you had specified "unbounded" (with a
> warning), which is only marginally better than falling over completely.
> Henry Thompson was promising to publish an improved algorithm but I
haven't
> seen it yet...
>
> I would suggest that you treat this constraint as one of those that
are best
> checked at application level, or using a tool such as schematron.
> Grammar-based validators aren't good at this job.
This note may be of interest:
http://jroller.com/comments/bobfoster/FullSpeedAhead/derivatives_of_bounded_repitition
(sic)
Bob Foster
|
