[
Lists Home |
Date Index |
Thread Index
]
From: "Mitch Amiano" <mamiano@nc.rr.com>
> Using a Schematron approach might be appropriate
> in this case. Depending upon how arbitrary the
> grouping is, it may be a more practical means
> as opposed to permuting all the possible groupings.
I would be a little stronger than that: I think it is generally impractical to
use other schema languages for this: you may have some kind of
combinatorial explosion. Furthermore, constraints such as
that the contents of one branch of a document will determine
the allowed elements in another branch of the document
typically cannot be done by regular-grammar-based systems.
{I say typically, because it also depends on how "ambigiuity" is handled:
some grammar-based systems such as WXS and DTDs say
"based on what the parent is, this is what I expect in the children" while it
is possible to have a schema language that says "this is what I found in the
children, therefore this is what the parent is (or, at least, what the
following siblings may be)". I think this is more how RELAX NG
can act.}
> A better explanation is at
> http://www.topologi.com/public/Schtrn_XSD/Paper.html
> See also
> http://www.topologi.com/products/validator/index.html
And also http://www.ascc.net/xml/schematron
To be more concrete, here is a Schematron rule that
1) constrains that there are 6 abcdef elements
2) sets up simple partial ordering and existance requirement between these elements,
<sch:rule context="x ">
<sch:assert test= "count(.//a | .//b | .//c | .//d | .//e | .//f)=6">
An x should contain directly or indirectly 6 abcdef elements.
</sch:assert>
<sch:assert test=".//f [preceding::e[preceding::d[preceding::c[preceding::b
[preceding::a[ancestor::x=current()]]]]]]">
Under an x, the elements abcdef should follow that order.
</sch:assert>
</sch:rule>
(This fairly terse. There are other ways to factor the constraints which would
allow better diagnostics and better efficiency, where these were an issue.)
Cheers
Rick Jelliffe
|
