[JENA-1424] LOAD ... INTO GRAPH relies too much on filename extension

• Type: Bug
• Status: Closed
• Priority: Minor
• Resolution: Fixed
• Affects Version/s: Jena 3.5.0
• Fix Version/s: Jena 3.6.0

Environment

Ubuntu 16.04

• Assignee: Andy Seaborne
• Reporter: Osma Suominen
• Created: 14/Nov/17 14:42
• Updated: 17/Dec/17 19:46
• Resolved: 18/Nov/17 09:48

Description

I tried to perform this SPARQL Update on an empty TDB:

LOAD <http://api.finto.fi/rest/v1/yso/data> INTO GRAPH <http://www.yso.fi/onto/yso/>

but got this error from tdbupdate:

$ tdbupdate --loc tdb --update=load-yso.ru 
org.apache.jena.update.UpdateException: Attempt to load quads into a graph
        at org.apache.jena.sparql.modify.UpdateEngineWorker.visit(UpdateEngineWorker.java:146)
        at org.apache.jena.sparql.modify.request.UpdateLoad.visit(UpdateLoad.java:64)
        at org.apache.jena.sparql.modify.UpdateVisitorSink.send(UpdateVisitorSink.java:46)
        at org.apache.jena.sparql.modify.UpdateVisitorSink.send(UpdateVisitorSink.java:26)
        at org.apache.jena.atlas.iterator.Iter.sendToSink(Iter.java:546)
        at org.apache.jena.atlas.iterator.Iter.sendToSink(Iter.java:553)
        at org.apache.jena.sparql.modify.UpdateProcessorBase.execute(UpdateProcessorBase.java:59)
        at arq.update.execOneFile(update.java:105)
        at arq.update.execUpdate(update.java:81)
        at arq.cmdline.CmdUpdate.exec(CmdUpdate.java:63)
        at jena.cmd.CmdMain.mainMethod(CmdMain.java:93)
        at jena.cmd.CmdMain.mainRun(CmdMain.java:58)
        at jena.cmd.CmdMain.mainRun(CmdMain.java:45)
        at tdb.tdbupdate.main(tdbupdate.java:37)

So basically Jena is complaining that I'm trying to load quads into a graph. But that's not true. The URL specified in the LOAD actually performs content negotiation and then redirects (302) either to an RDF/XML or a Turtle serialization (with the proper Content-type headers), both are graphs not quads.

The problem seems to be that UpdateEngineWorker checks the URL specified in the LOAD as if it were a filename and throws an exception if its file extension doesn't match the known graph formats. In this case there is no extension so it won't match.

The check was introduced 6 months ago in this commit:
https://github.com/apache/jena/commit/931a437bb49fecdb1cb70a5e6225e27141dec86c#diff-d0b3b8995c502712dac778f5bb61bc9dR146

If I use the URL that the above URL redirects to, which contains a .ttl file extension, loading works fine:

LOAD <http://api.finto.fi/download/yso/yso-skos.ttl> INTO GRAPH <http://www.yso.fi/onto/yso/>

But this means that the LOAD ... INTO GRAPH ... command cannot be used with arbitrary Linked Data URIs, just ones that happen to contain a file extension like .ttl or .rdf or .nt.

Issue Links

• links to: GitHub Pull Request #311