No objections here. Just to recap:
The data model interfaces stay as they are, we only introduce the convention 
that we'll never return an XDMSequence from next().

Is that right? 
It seems like we should have a way to check that this is the case (at least at 
runtime in some kind of debug mode).

What do think?

Till

On Jan 5, 2010, at 9:12 PM, Vinayak Borkar wrote:

> Guys,
> 
> 
> Currently there is a value called XDMSequence that can represent a complete 
> sequence in one Java Object. Originally this seemed like a good idea -- it 
> reduces the number of next calls where iterators can send values directly 
> from their producers (The FLWOR from the returnExpr, for eg).
> 
> However, this makes other code a lot more complicated -- Iterators that use 
> the value have to inspect the data and act appropriately if the object is a 
> sequence object. In my opinion, this outweighs the benefits it offers. So I 
> am planning to eliminate the sequence object as a value that can be returned 
> in the iterator stream (next()) call.
> 
> It will still continue to exist to be used with evaluateEagerly().
> 
> Thoughts/Objections?
> 
> Thanks,
> Vinayak

Reply via email to