[xquery-talk] RE: XQuery - Grammar? Implementations?

[Michael Kay]

> I am not sure your criticism is justified in its harshness. 
> At least if it is the case, as I believe, that implicit 
> conversions may indeed interfere with optimization.

Well, for starters, I think that usability is a much more important goal in
language design than optimizability. 

And I don't think they are in contradiction. Usability demands consistency
and orthogonality, and these are the very features of a language which help
to make it optimizable.

> 
> Declarations, on the other hand, offer a different situation, 
> because they are not "called": only one explicit conversion 
> per declaration may become necessary, and the expression to 
> be converted is a top level expression, which can be 
> comfortably wrapped in the appropriate conversion call. 

It can be done just as comfortably by the compiler as by the user. I don't
see why making the user do atomization "by hand", rather than having the
compiler do it on the user's behalf, makes for a better language; and I
don't see why variable declarations should be treated differently from any
other context where there is a supplied expression and a required type.

It's interesting to note that Java allows *more* implicit conversions in a
variable declaration than in a function call.

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