Re: A question regarding the semantics of multi-triple term reifications and querying

Hi Thomas,

On Mon, 2025-09-15 at 10:30 +0200, Thomas Lörtsch wrote:
> Hi Olaf,
> 
> thank you for your help!
> 
> > On 11. Sep 2025, at 17:51, Olaf Hartig <olaf.hartig@liu.se> wrote:
> > 
> > Hi Thomas,
> > 
> > On Thu, 2025-09-11 at 17:16 +0200, Thomas Lörtsch wrote:
> > > Since reifiers can refer to multiple triple terms, do queries for annotations on a triple need to filter out
> > > eventual
> > > multi-triple term annotations? Take for example the following graph.:
> > > 
> > >    :r rdf:reifies <<( :s :p :o )>> ,
> > >                   <<( :x :y :z )>> ;
> > >        :b :c .
> > > 
> > > Querying for all annotations on <<( :s :p :o )>> might naively query for all annotations on reifiers that
> > > rdf:reify
> > > <<( :s :p :o )>>, e.g.:
> > > 
> > > 
> > >    select ?pred ?obj where {
> > >      _:x rdf:reifies <<( :s :p :o )>> ;
> > >          ?pred ?obj
> > >    }
> > > 
> > > However, in the above example that would return the annotation ':r :a :b',
> > 
> > Assuming you meant ':r :b :c', then yes, one of the solution mappings in the result of this query over the graph
> > given
> > above, is { ?pred -> :b, ?obj -> :c }.
> > 
> > In addition to this one, there are two more solution mappings in the result:
> >  { ?pred -> rdf:reifies, ?obj -> <<( :s :p :o )>> } 
> > and
> >  { ?pred -> rdf:reifies, ?obj -> <<( :x :y :z )>> } 
> 
> Right.
> 
> > That's exactly what the query asks for. If you want to retrieve data about a reifier that reifies only <<( :s :p :o
> > )>>
> > and nothing else---at least not within the graph that you are querying---then you need to specify that as an
> > additional
> > condition in your query.
> 
> IIUC specifying that additional condition requires something like 
> 
>     FILTER NOT EXISTS { 
>         ?reifier rdf:reifies ?otherTriple . 
>         FILTER (?otherTriple != <<( :s :p :o )>> ) 
>     }

Correct.

> > Notice also that this is not specific to reifiers at all! As an example, consider the graph
> > 
> >  :s :p 1 .
> >  :s :p 2 .
> >  :s :q :y .
> > 
> > and the query
> > 
> >  SELECT ?v WHERE {
> >    _:b :p 1 .
> >    _:b :q ?v
> >  }
> > 
> > Also here, you get a solution mapping { ?v -> :y }.
> 
> Indeed! That seems to suggest that in a multi-part reifier the combination of individual triple terms is to be
> understood as OR, not AND,

No, I don't think so. Look again at my example RDF graph with predicates :p and :q above. I wouldn't say that the value
that :s has for property :p is 1 or 2. Instead, I would say that :s has two (known) values for :p, namely 1 and 2.

-Olaf


> and that my intuition was wrong. I agree that this should not be surprising, given the general design of RDF.
> 
> I think this topic should be part of introductory material.
> 
> 
> Thanks again,
> Thomas
> 
> 
> > -Olaf
> > 
> > 
> > > but ':r' doesn’t just reify <<( :s :p :o )>>, it reifies <<( :s :p :o )>> AND <<( :x :y :z )>>. I at least
> > > consider
> > > that a difference (but I’m not quite sure if the specs do too).
> > > 
> > > OTOH, filtering out multi-triple term reifiers like ':r' from a result set is pretty tedious, especially if it
> > > would
> > > have to be done on a regular basis.
> > > 
> > > Best,
> > > Thomas
> > 
>

Received on Monday, 15 September 2025 08:45:04 UTC