Can Jena convert a SPARQL endpoint to a Model?

[chenkuncloud]

Such as Fuseki's SPARQL endpoint, can it be converted by Jena to a Model?

This makes it very convenient to use the JAVA API to query.

If not for the time being, will it be considered in the future?

Best Regards

[afs]

Please review the previous discussions:

Issues #1243 and #1246

Networks are much slower than local method calls. In designing a system to work with local data and remote data, Jena has RDFConnection and other APIs. The fine grained model APIs is not the best starting point.

Specifically for arbitrary SPARQL endpoints using only SPARQL standard protocols, blank nodes will be problematic otherwise.

[Claudenw]

Years ago I worked on a project that did something similar. We made remote endpoints appear as part of the graph behind Fuseki. The techniques and results can be explored in the journal article: https://www.researchgate.net/publication/315120429_BioFed_Federated_query_processing_over_life_sciences_linked_open_data

But it still does not provide Model access. Attempting to implement a Model that makes requests across the network will yield a Model that is so slow as to be unusable.

[namedgraph]

The DatasetAccessor interface is now deprecated, but it can be used to abstract local and remote Model access. We have a Service interface (and implementations) that can either be based on a Dataset or a remote endpoints: https://github.com/AtomGraph/Core/blob/master/src/main/java/com/atomgraph/core/model/Service.java

…

On Fri, 13 May 2022 at 17.01, Claude Warren ***@***.***> wrote: Years ago I worked on a project that did something similar. We made remote endpoints appear as part of the graph behind Fuseki. The techniques and results can be explored in the journal article: https://www.researchgate.net/publication/315120429_BioFed_Federated_query_processing_over_life_sciences_linked_open_data But it still does not provide Model access. Attempting to implement a Model that makes requests across the network will yield a Model that is so slow as to be unusable. — Reply to this email directly, view it on GitHub <#1300 (reply in thread)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAGPM5VWALPCQDIVZH5TP7DVJZVEJANCNFSM5V2QLL5A> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[afs]

That is like RDFConnection - fetch-modify-replace (SPARQL Graph Store Protocol) - which isn't what the question asks about.

[chenkuncloud]

Thank you!

I think I asked a very naive question, the graph is still at the remote endpoint, just like a full table query.

I know RDFConnection.fetch() :

try (RDFConnection conn = RDFConnectionRemote.service(endpoint).build()) {
	Model model = conn.fetch(graphName);
	......
}

But if graph is very large, there will be errors: 406 - Not Acceptable :

Exception in thread "main" org.apache.jena.atlas.web.HttpException: 406 - Not Acceptable
	at org.apache.jena.http.HttpLib.exception(HttpLib.java:231)
	at org.apache.jena.http.HttpLib.handleHttpStatusCode(HttpLib.java:161)
	at org.apache.jena.http.HttpRDF.execGetToInput(HttpRDF.java:165)
	at org.apache.jena.http.HttpRDF.httpGetToStream(HttpRDF.java:133)
	at org.apache.jena.http.HttpRDF.httpGetToStream(HttpRDF.java:128)
	at org.apache.jena.sparql.exec.http.GSP.GET(GSP.java:159)
	at org.apache.jena.rdflink.RDFLinkHTTP.get(RDFLinkHTTP.java:394)
	at org.apache.jena.rdflink.RDFConnectionAdapter.fetch(RDFConnectionAdapter.java:146)
	at com.test.Test.main(Test.java:21)

[afs]

406 - Not Acceptable is not necessarily related to graph size. It means the remote end rejected the request. Check the server logs for more details and also check the graph name.

If it is (very) large graph, use SPARQL to query and manipulate it.

[chenkuncloud]

But I wonder if a set of Fuseki special operations can be provided in the future? Like Virtuoso:

VirtGraph graph = new VirtGraph("xxx/sparql", "jdbc:virtuoso://xxx:1111", "dba", "xxx");
VirtModel model = new VirtModel(graph);
......

This will make Fuseki much convenient to operate.

Or that's not what Fuseki had in mind.

[afs]

can be provided in the future

Apache Jena is an open source project. Contributions are welcome, including by Github pull requests.

It needs a private protocol to support blank node labels. The RDF Thrift format provides this feature.

It will be slow as the data scales (as the Virtuoso feature, and it is impractical for client-side SPARQL query).

RDFConnection is a better architecture to abstract away from whether the data is local or remote.

Issues #1243 and #1246

[chenkuncloud]

Thank you. Looking forward to having that feature in the future!

[afs]

Answered by #1246.