[Versa] function objects

[Uche Ogbuji]

On Thu, 2005-09-22 at 09:57 -0600, Michael Olson wrote:
> > On Wed, 2005-09-21 at 19:58 -0600, Michael Olson wrote:
> >> To revisit this, I like the idea, but syntactically I think we need a
> >> symbol to differentiate these from a qname pattern.  There is no
> >> difference, from the grammars POV between
> >>
> >> ground(rdf:type)
> >>
> >> and
> >>
> >> map(ns:func_name)
> >>
> >> Maybe kinda C'ish and make it
> >>
> >> map (&ns:func_name)
> >>
> >> thoughts?
> >
> > I guess in my thinking of this as a simple function there was no need  
> > to
> > distinguish between the argument to ground.  The argument is just a
> > resource URI, expressed literally, or in qname form.  I would therefore
> > expect to be able to do:
> >
> > ground(uri("http://" + "example" + ".org"))
> >
> > Assuming that we have a conversion function uri() in Versa 2.0 that can
> > convert strings (and other data types?) to a URI.
> >
> > Short answer, I'd expect it to be
> >
> > ground(rdf:type)
> >
> > Maybe I'm missing something?
> >
> 
> I guess I asked the question poorly.  I'm OK with how ground works in  
> the example, what I think we need is a symbol to distinguish a  
> qname-pattern from a function-reference in the grammar because the  
> grammar cannot tell the difference between
> 
> rdf:type
> 
> which, as a writer of a query, probably means a qname pattern, and
> 
> versa:ground
> 
> which, as a writer of a query, most likely means a function-references.
> 
> What I propose is not change anything about the qname-pattern, but add  
> the symbol "&" to the front of function-references so the last example  
> would be
> 
> &versa:ground

Ah.  Gotcha.  And ooh, good call.

Personally, I think we should disambiguate in the other direction.  Now
that we've committed to the idea of functions as first-class objects
(yaay!), I think that the use of qnamed predicates as functions is more
of the special case.  So I would leave function references "bare", and
use a symbol to mark qnamed predicate functions.

Maybe:

@dc:title(<http://foo.com/bar>)

We can reuse @ here without ambiguity, I think.


-- 
Uche Ogbuji                               Fourthought, Inc.
http://uche.ogbuji.net                    http://fourthought.com
http://copia.ogbuji.net                   http://4Suite.org
Articles: http://uche.ogbuji.net/tech/publications/