Re: Sequences in XSLT 2.0..

On Tue, 15 Mar 2005 22:16:21 -0000, Michael Kay <mike@xxxxxxxxxxxx> wrote:
> > although perhaps they might have been better
> > >still if they had allowed sequences of sequences)
> >
> 
> >I guess they decided to do just one thing at a time....
> 
> There has been a proposal to add sequences of sequences, and it was
> rejected. It's always hard to summarize why a particular decision was taken,
> but it's probably fair to say that some people felt it was a bad idea
> intrinsically and others felt that the time was not right.
> 
> There are many interesting things you could do with sequences of sequences.
> It might come up in future as part of an XQuery solution to grouping
> problems. If you think about it, xsl:sort allowed you to process things in
> sorted order, and sequences allow you to save things in sorted order.
> Similarly, xsl:for-each-group allows you to process things in groups, and
> sequences of sequences would allow you to save things in groups. However,
> there's also a feeling that you'd then have two completely separate ways of
> building trees: the XML way, and the "sequences of sequences" way, and that
> this might be somewhat redundant.
> 
> Michael Kay
> http://www.saxonica.com/

I find the lack of "sequence of sequences" one of the major things
that are definitely missing in the XPath data model.

This prevents FXSL from providing some very elegant solutions. 

One example is that in the presence of this feature every function
could be represented by a function, having just one argument -- the
sequence of all its arguments.

This could make much more easy a unified approach towards processing
higher-order functions.


Implementing sequence of sequences in a later XPath version may also
be difficult, because it will conflict with the current ideology.



Cheers,
Dimitre Novatchev