RE: question about subtyping and type annotations

Right. and I think that from the paragraph about casting that you quoted 
it is clear
that you get the type annotation of the type into which you are casting.

It seems the most natural semantics in a number of ways. Notably it is 
useful is the
user wants to explicitely set the type annotation, as in:

xs:byte(1)

or in case of casting from another type:

xs:integer("1")

I think people will find it confusing if you change the type annotation 
behind their
back.

I also remember that one feature that was discussed at some point for 
XQuery was
something like Expr instance of only xs:integer, where you want to check 
for the type
annotation but not one of its derived types. That kind of feature won't 
really make sense
if we let implementation pick up a derived type arbitrarily.

The rules about type substitutability is not about giving implementations 
the freedom
to chose the type annotation. I think the language is pretty deterministic 
in that sense.
When you validate or cast you get a specific type annotation. When you 
pass the
value to the function you preserve the existing type annotation. That 
should helps 
interoperability.

- Jerome



"Michael Kay" <mhk@mhk.me.uk> wrote on 02/24/2006 11:57:55 AM:

> > Section 3.12.5 Constructor Functions seems to be pretty clear 
> > about it:
> 
> That only says they are equivalent to casts. Are casts exempt from the
> general rule of substitutability, which says that any expression can 
return
> a value whose dynamic type is a subtype of the required type?
> 
> Michael Kay
> http://www.saxonica.com/
> 
>

Received on Saturday, 25 February 2006 01:56:14 UTC