apache
[FEATURE REQUEST] Add GeoSPARQL support on arq and tdbquery
Version
4.8
Feature
Dear Jena team, To my understanding fuseki is compiled with GeoSPARQL but not Arq and tdbquery
Is it possible to add this feature to these tools ?
Thanks
Are you interested in contributing a solution yourself?
None
I believe that most of the CLI tools are present in the Fuseki JAR distribution but it's non-obvious in that it requires knowing the package and classname of the tool you want to run e.g.
$ java -cp fuseki-server.jar tdb.tdbquery --help
tdbquery --loc=<path> --query=<query>
I'm pretty sure that GeoSPARQL isn't part of the default Fuseki distribution yet, only the text index is there as an extension. I think I mentioned it some time ago hidden in some other issues, but forgot to keep track of it:
https://github.com/apache/jena/blob/main/jena-fuseki2/jena-fuseki-server/pom.xml
Regarding the confusion, the TO wants to use GeoSPARQL features with tdbquery, but this doesn't work at the moment. Adding a jena-geosparql Uber Jar to the classpath or use tdbquery from the Fuseki fulljar (once jena-geosparql has been added!) will at least make the functions and datatypes being registered and work:
java -cp /tmp/jena-fuseki-fulljar-4.9.0.jar tdb2.tdbquery --loc /tmp/tdb2 '
PREFIX geo: <http://www.opengis.net/ont/geosparql#>
PREFIX geof: <http://www.opengis.net/def/function/geosparql/>
PREFIX uom: <http://www.opengis.net/def/uom/OGC/1.0/>
SELECT * {
BIND("POINT(1.0 1.0)"^^geo:wktLiteral AS ?p1)
BIND("POINT(2.0 2.0)"^^geo:wktLiteral AS ?p2)
BIND(geof:distance(?p1, ?p2, uom:radian) AS ?d)
}'
07:53:56 WARN system :: The “SIS_DATA” environment variable is not set.
------------------------------------------------------------------------------------
| p1 | p2 | d |
====================================================================================
| "POINT(1.0 1.0)"^^geo:wktLiteral | "POINT(2.0 2.0)"^^geo:wktLiteral | 0.024683e0 |
------------------------------------------------------------------------------------
What does not work is probably using a GeosparqlDataset in an assembler file because from what I remember when looking at the code some time ago, tdbquery only considers a TDBDataset?
So my guess once you added a full jar of jena-geosparql (needs Maven shade plugin), the GeoSPARQL functions will work because they are registered via service loader, but the spatial index I guess won't as this is part of the GeosparqlDataset which registers the index to the query context on initialization.
I'm pretty sure that GeoSPARQL isn't part of the default Fuseki distribution yet,
There is jena-fuseki-geosparql, a build of Fuseki with jena-geosparql. It has different command line options which is one factor that stops common integration.
If the dataset is coming from an assembler, the extension mechanism FusekiAutoModule could be used.
That can also include a command line entry point that either (1) generates an assembler for later use (2) writes an assembler that is picked up by the module (3) or creates a datastructure with the details that the module picks up (preference - the latter so in-memory, no disk use at all works.).
tdbquery only considers a TDBDataset?
Yes.
Assembling needs a starting point in the case of more than one description in a file. tdbquery is doing that based on type. The sparql command can take a configuration with any single dataset setup. A feature is needed to specific the starting point of building the dataset.
GeoSPARQL is sufficient different in its feature set and needs that there needs to be a community around it to maintain and evolve the GeoSPARQL support.
See also: #1344 which has several suggestions but hasn't come to a clear consensus on direction.
