RE: REDUCED tests

-------- Original Message --------
> From: Lee Feigenbaum <>
> Date: 12 November 2007 18:21
>
> Seaborne, Andy wrote:
> >
> >
> > Seaborne, Andy wrote:
> > > Lee Feigenbaum wrote:
> > > > I added the following tests.
> > > >
> > > > syntax-sparql5/manifest#syntax-reduced-01
> > > > syntax-sparql5/manifest#syntax-reduced-02
> > >
> > > ARQ passes the syntax tests.
>
> Great.
>
> > >
> > > >
> > > > reduced/manifest#reduced-1
> > > > reduced/manifest#reduced-2
> > > > The latter two are rdf:type'd as both mf:QueryEvaluationTest
> >
> > ARQ now passes these tests.
>
> Great.
>
> Can anyone else try these tests out and report results, please?
>
> > > Some people may find it a nuisance to have two types on a test.
> > > Maybe a qualifier would be better in the style of requirements and
> > > notable features.
>
> A nuisance? That didn't occur to me - why is that? (I don't have a
> strong feeling for how to model it, so I'm happy to change it.)

It's the change that catch existing code out.  Code has to change but this but this might mean that more cdo chnages - just guessing but it may affect the main "find a test" code.

One approach is to search the file for rdf:types.  Having found via "QueryEvaluationTest", new code is needed to look for a second type.  Having two can be a surprise to code that current assumes one - and because we have only have one up to now, such code may well exist.  And remember the searching may not be in SPARQL - after all, this is SPARQL test code and you've got to start somewhere!

What happens if it has two types that imply different tests? :-)  These aren't subtypes.  Run two tests?  Is QueryEvaluationTest really a supertype of

        Andy

>
> > > My plan at the moment is to ignore this and ask the query if it is a
> > > REDUCED syntax query.
> >
> > My testing approach is to turn both query answers and test results
> > into unique forms and compare.  This does not catch an implementation
> > of REDUCED that introduces new, extra duplicates.
>
> I'll pretend I didn't hear that...
>
> Lee
>
> > >
> > >  > and also as
> > > > mf:ReducedCardinalityTest. The intended semantics of the latter
> > > > is:
> > > >
> > > > + The given mf:result for a mf:ReducedCardinalityTest is the
> > > > results as if the REDUCED keyword were omitted. To pass a
> > > > mf:ReducedCardinalityTest, an implementation must produce a result
> > > > set with each solution in the expected results appearing at least
> > > > once and no more than the number of times it appears in the
> > > > expected results. Of course, there must also be no results
> > > > produced that are not in the expected results.
> >
> >     Andy

Received on Monday, 12 November 2007 18:35:23 UTC