Re: Overloading functions in XSLT and XQuery

But Michael, this makes no sense.

When would overloading by arity ever produce a result inconsistent with
overloading by type? Overloading by type always starts with the premise that
the two functions have the same name and same number of arguments -- and
then looks to the types of the arguments to distinguish which function is
invoked.  Therefore, by definition, overloading by arity is
forwards-compatible with overloading by type.

Note also that you¹ve already defined overloading by arity for functions in
the built-in function namespace.  Therefore, any future overloading rules
must be compatible with overloading by arity, or else you just create an
inconsistent mess for implementers in which the built-in functions use a
different set of overloading rules than other functions.  (Oh wait, you¹re
already there.)


Based on technical merits alone, there¹s simply no reason to exclude
overloading user-defined functions by arity in this version of XQuery.
There are, however, plenty of other kinds of reasons (time-to-market,
inertia, personal preference, inability to achieve consensus, etc.).


Cheers,
Michael (also speaking for himself)

On 11/3/03 4:05 PM, "Michael Rys" <mrys@microsoft.com> wrote:

> The reason is that the current built-in overloading with arity should continue
> to work if we go to overloading based on the full signature information. Once
> we allow the users to overload, I am concerned that we get cases where
> overloading with arity will result in a different resolution than overloading
> based on the typed signature. So instead of having to deal with this in (for
> XQuery) the first version, we wanted to postpone this feature.
> 
>  
> 
> Best regards
> 
> Michael (speaking for himself)
> 
>  
> 
>  
> 
> 
> From: Kay, Michael [mailto:Michael.Kay@softwareag.com]
> Sent: Sunday, November 02, 2003 8:17 AM
> To: Noe Michejda; Michael Rys; public-qt-comments@w3.org
> Subject: RE: Overloading functions in XSLT and XQuery
> 
>  
> 
>> > 
>> > RE: Overloading functions in XSLT and XQuery> XQuery 1.0
>> > decided to only allow overloading of XQuery's own built-in
>> > functions and not provide overloading
>>> > > capabilities for the user-defined functions.
>>> > > This explains the perceived contradiction below.
>> > 
>> > But XSLT allows overloaded user-defined functions.
>> > This will be very inconvinient for products allowing using of
>> > XSLT libraries in XQuery. And to some degree event for people
>> > porting libraries from XSLT to XQuery.
>> > 
> 
> Personally, I agree. I can't see any logical objection to overloading by arity
> (two functions with the same name but different numbers of arguments). I think
> it's very odd that the function calling mechanism in XQuery allows for it, but
> that user-defined functions have to be uniquely named.
> 
> Michael Kay 
>