On Thu, Jan 16, 2025 at 5:36 PM Andy Seaborne <a...@apache.org> wrote: > > > > On 15/01/2025 23:50, Martynas Jusevičius wrote: > > Hi, > > > > Is it possible to get DESCRIBE results as quads? > > No. It's a single model.
Does it have to be though? :) Why shouldn't a quad description be possible? I have a use case for that: for every triple rendered in a UI I want to know which named graph it came from, so that I can override the navigation of external URI resources (<https://example.org/Denis> being external to the webapp on <https://localhost:4443/>) or resources described in multiple graphs, and lead the user to the named graph(s) that contain them. > > > I tried requesting > > text/trig but simply got the data in the default graph: > > > > $ curl http://localhost:3031/ds/ -d 'query=DESCRIBE > > <https://example.org/Denis>' -i -H "Accept: text/trig" > > application/trig Right, my bad. But I re-ran with application/trig and got the same result. > > > HTTP/1.1 200 OK > > Date: Wed, 15 Jan 2025 23:36:46 GMT > > Fuseki-Request-Id: 552 > > Content-Type: text/trig > > text/turtle? No, Fuseki actually returned text/trig :) > > > Transfer-Encoding: chunked > > > > <https://example.org/Denis> > > a <http://schema.org/Person> ; > > <http://schema.org/name> "Denis Villeneuve" ; > > <http://schema.org/url> <https://www.imdb.com/name/nm0898288/> . > > > > The result I'm looking for can be obtained using CONSTRUCT ... GRAPH > > ... WHERE [1] syntax: > > > > $ curl http://localhost:3031/ds/ -d 'query=CONSTRUCT { GRAPH ?g { > > <https://example.org/Denis> ?p ?o . } } WHERE { GRAPH ?g { > > <https://example.org/Denis> ?p ?o } }' -i -H "Accept: text/trig" > > HTTP/1.1 200 OK > > Date: Wed, 15 Jan 2025 23:36:10 GMT > > Fuseki-Request-Id: 550 > > Content-Type: text/trig > > Transfer-Encoding: chunked > > > > <https://localhost:4443/8a440293-253e-42f0-bdd6-3e7fc8acb337/> { > > <https://example.org/Denis> > > a <http://schema.org/Person> ; > > <http://schema.org/name> "Denis Villeneuve" ; > > <http://schema.org/url> > > <https://www.imdb.com/name/nm0898288/> . > > } > > > > Just for curiosity's sake I also tried requesting Turtle with > > CONSTRUCT ... GRAPH ... WHERE and got an empty result: > > > > $ curl http://localhost:3031/ds/ -d 'query=CONSTRUCT { GRAPH ?g { > > <https://example.org/Denis> ?p ?o . } } WHERE { GRAPH ?g { > > <https://example.org/Denis> ?p ?o } }' -i -H "Accept: text/turtle" > > HTTP/1.1 200 OK > > Date: Wed, 15 Jan 2025 23:43:04 GMT > > Fuseki-Request-Id: 554 > > Content-Type: text/turtle > > Transfer-Encoding: chunked > > > > Looking at the above, the DESCRIBE behavior with quads feels inconsistent > > to me. > > I get that the result triples can simply be placed into the default > > graph when quads are requested. But does that really make sense? > > IMO DESCRIBE should behave like CONSTRUCT ... GRAPH ... WHERE when > > quads are requested (i.e. return quads), and return triples when > > triples are requested (as it always has done). > > How do you associate the resources to be described with a graph? The same way as CONSTRUCT GRAPH does? If we compare CONSTRUCT GRAPH and DESCRIBE results, we can see that the <https://example.org/Denis> is actually stored in a named graph. Are you saying that a resource description cannot be quads? I.e. cannot be split across multiple named graphs or even combine default and named graphs? I don't see why not. What DESCRIBE does to produce triples is simply strip the graph URIs from the quads (in this case <https://localhost:4443/8a440293-253e-42f0-bdd6-3e7fc8acb337/>), which is lost information *when quads are requested*. > > > > > Martynas > > > > [1] https://jena.apache.org/documentation/query/construct-quad.html >
