Re: [rdf4j-dev] GeoSPARQL (lack of) support

[Jeen Broekstra <jeen.broekstra@xxxxxxxxx>]

PR available with updated GeoSPARQL docs, see https://github.com/eclipse/rdf4j-doc/pull/64 .

Also very pleased to report that our GeoSPARQL support is a lot better than I initially feared, and with a few minor updates we applied it's now close to full support (at least in terms of supported function calls). Though to be fair we need a more rigorous set of tests for the various available operators I think.

We're still blocked from final release by some outstanding legal reviews (for one thing it turns out that JLine was never officially approved before), but I'll try and schedule another milestone build this week.

Cheers,

Jeen

On Mon, Aug 6, 2018 at 9:33 AM, Jeen Broekstra <jeen.broekstra@xxxxxxxxx> wrote:

Thanks Bart,

Fwiw I'm diving in a bit deeper myself, both on the implementation and the docs, and have a draft of an update ready locally. I'll put up a pr this week.

Cheers,

Jeen

On Sun, 5 Aug. 2018, 23:48 Bart Hanssens (BOSA), <bart.hanssens@xxxxxxxxxxxx> wrote:

Hi Jeen,

the GeoSPARQL support is indeed minimal...

I guess I was the last one updating that specific part of the documentation,

so I can take another look at it when I'm back from holidays mid-August.

Best regards

Bart

---

From: rdf4j-dev-bounces@xxxxxxxxxxx <rdf4j-dev-bounces@xxxxxxxxxxx> on behalf of Jeen Broekstra <jeen.broekstra@xxxxxxxxx>
Sent: Tuesday, July 31, 2018 1:59:34 AM
To: rdf4j developer discussions
Subject: [rdf4j-dev] GeoSPARQL (lack of) support

 

I'm worried about our support for GeoSPARQL.

I personally never looked into this much before, but recently decided to do some digging, hoping to improve our documentation on RDF4J GeoSPARQL support.

As far as I can tell, our GeoSPARQL support is really minimal. The only geo function I managed to run is geof:distance, and even that failed when using it as part of a math _expression_ (e.g. dividing by 1000 to turn meters into kilometers). This is all the more worrying because that is the actual example query in our current documentation. I'm hoping that I have misunderstood something and we actually do support more than this, but I'd be grateful for some pointers.

I've put up a PR with a test case for this in rdf4j-testsuite, and tried to see about a fix in the LuceneSail, but to be fair I'm having a hard time figuring out the code. Is any of you available to invest a little time in understanding our GeoSPARQL support a little better and helping either fixing bugs or just improving the documentation?

At this point we need to either up our game in terms of documentation and support, or no longer claim we support this, because as it stands, claiming RDF4J "does GeoSPARQL" is false advertising, IMHO. 

Regardless of anything else, our GeoSPARQL documentation is out of date because it still talks about spatial4j 0.4, and as not being included due to licensing problems, which is no longer the case (we include spatial4j 0.6). It also mentions JTS, but I have no idea how that relates to anything else (except that it seems to add polygon support but I don't understand how we would make that work). FWIW JTS itself is now also a LocationTech project so if we want, we can now add it as a dependency.

Jeen 

_______________________________________________
rdf4j-dev mailing list
rdf4j-dev@xxxxxxxxxxx
To change your delivery options, retrieve your password, or unsubscribe from this list, visit
https://dev.eclipse.org/mailman/listinfo/rdf4j-dev