thibault volpert <no*******@stop.invalid> writes:
Hi all,
Here is an element of my the DTD I wrote :
<!ELEMENT uMeshPart (coordinateSystem?, nodeList?,
edgeList?, faceList?, uElementList?,
uElementGroup*, group*)>
Now my question : I dont want to have any order between
the elements of a <uMeshPart>, but as i wrote it it seems
to have one. What can i change in the definition of the
uMeshPart to avoid having this order between elements.
A. Is the order in which things appear in the input significant?
(That is, does it have a meaning for the processor?) If so, then what
you want is the equivalent of the SGML & operator:
<!ELEMENT uMeshPart (coordinateSystem? & nodeList? &
edgeList? & faceList? & uElementList? &
uElementGroup* & group*)>
This does insist that all the occurrences of uElementGroup appear
together, and ditto for the group elements.
In XML Schema you can just about do this, except that the XML Schema
'all' connector doesn't allow maxOccurs > 1; if I needed to do this
with XML Schema, I'd replace the uElementGroup* and group* with
uElementGroups and groups elements, which wrap the multiply occurring
uElementGroup and group elements. In XML Schema 1.1 some people are
hoping to relax the requirement that maxOccurs be at most 1; that
might help you here.
An alternative is to allow the DTD to over-generate and write
<!ELEMENT uMeshPart (coordinateSystem | nodeList |
edgeList | faceList | uElementList |
uElementGroup | group)*>
and make the application check to make sure that only uElementGroup
and group occur more than once. I don't advise this, as it will
inevitably lead to dirty data.
If you really want to allow any order, you can write it in XML DTD
notation, as Klaus Johannes Rusch has pointed out, but the result will
be very hard to understand; there are thirty-two different sequences
in which the five elements which can occur at most once could occur,
and you have to list them all.
B. If, on the other hand, the sequence of items has no significance
for your application, then the content model you have written is quite
correct: it shows very clearly that the sequence in which
coordinateSystem, nodeList, etc., occur in the document can have no
significance for the application.
-C. M. Sperberg-McQueen
World Wide Web Consortium
