John Gentilin <[EMAIL PROTECTED]> wrote:
> plus there is more than one
> column-header element (one for each column)
Yup. My mistake.
> I was looking into overriding the traverser but it was not clear to
> me how to do it without the having the NodeTest available.
In the future you'll be able to do this fairly safely. The traversers and
iterators have two ways of getting the next node: one that knows about an
extended ID, and one that does not. Right now, the forms that use the
extended ID aren't used much. In the future, they'll be used most of the
time. (long story... we have to map extended IDs used in the stylesheet to
extended IDs used in the DTM). The ExtendedID is the basic node-test.
> Each column (column-header) has its own set of
> attributes if we moved them, then we need some strange naming conventions
> such as column-label-001 ??
Yup. I'm not sure what I was thinking.
>> I was thinking about adding another level of hierarchy such as:
> <sql>
> <metadata>
> <column-header/>
> <column-header/>
> </metadata>
Yup. Sounds good to me. But then stylesheets that use the existing
extensions have to change. I suspect this is something we can live with
right now, since the SQL extensions have always been presented as
experimental.
Still, there's a general problem about making streaming implementations of
DTMs be able to handle this sort of thing. I wonder if a better approach
than just handing in the ExtendedID, is to pass in a NodeTest interface,
and get rid of the methods that don't take a NodeTest altogether. It would
simplify the traverser interface. Bit-o-work, though. Something to think
about...
-scott
