Re: dynamic-context type-branching

Thanks.
Cheers,
ac

Hi Michael,

I picked the saxon:evaluate() use case because it seemed obvious that anything could come out of it, but I also have one or two other cases where a template/function can output different things, typically a string or a node sequence. All these cases seem to be more concentrated around content query and layout management, where polymorphism contributes to content abstraction. I would certainly support increased orthogonality in handling atomic values, in node selecting/processing contexts (ex: apply-templates).

Thanks,
ac

Michael Kay a icrit :

In schema unaware xslt2 environments, what would be the best way to select action depending on the result type (ex: string, text node(s), attribute(s), element(s)) of saxon:evaluate(), ex:

I think it's basically to do an xsl:choose. However, your note reminded me
to add "apply-templates on atomic values" to the candidate requirements list
for future XSLT versions. It's a little hard to come up with really
convincing use cases, but I think it would make the language more orthogonal
and that's always a good thing. Certainly, polymorphism on atomic values as
well as on nodes seems intrinsically desirable.

Michael Kay
http://www.saxonica.com/