David,
I am grateful to you for your very clear and elegant solution!
Wendell,
thank you for your proposition which is almost as good as David's, but
which I personally find more difficult to follow. And it requires the
stylesheet to explicitly state the element names, which might be a
little inconvenient.
You wrote:
> I'm a bit puzzled by your statement that you don't want to "hardwire
> the structure of the tree to create", since that's exactly what the
> style_nesting element does (and fairly neatly too).
What I meant was: Since I am using the "real" meta-stylesheet method (as
opposed to how you characterized your proposal) to generate the
stylesheet which is doing the WordprocessingML transformation, I prefer
to make the style transformation configurable, by specifying the element
equivalents of the styles as well as their nesting in an external XML
document with a structure that could look like the one I showed:
<style_nesting>
<b><i><u/></i></b>
</style_nesting>
So it's certainly somehow hard-wired in XML, but not necessarily in XSL.
Now I have become aware that this opens two ways:
1) Being totally dynamic, i.e. reading in the mapping and hierarchy
description at the operating stylesheet's run time. This is what David's
method does.
2) Being dynamic only in the first step, i.e. generating an operating
stylesheet with the style mapping coded in it. This is one aspect of
Wendell's solution.
I'll try to experiment with both to find the approach which is more
favorable. (And I will have to refine the mapping technique to account
for more complex cases than local name identity.)
Thank you again for your help so far.
Yves