With named graph in config:
1. FROM <> doesn't work resp. doesn't terminate
2. Using GRAPH keyword works:
SELECT ?s
WHERE {
GRAPH <http://example.org/wikidata>
{?s a <http://wikiba.se/ontology-beta#Item> }
}
LIMIT 10
OFFSET 20000000
> Thanks Lorenz for your test. I'd like to know much RAM you had available for
> this test. Fuseki+HDT has always been fast for me as well, except this query.
> I have 8GB.
I used 16G but could go up to ~512G in worst case
On 16.12.2017 16:36, Laura Morales wrote:
> Thanks Lorenz for your test. I'd like to know much RAM you had available for
> this test. Fuseki+HDT has always been fast for me as well, except this query.
> I have 8GB.
>
> What I did was *almost* exactly what you did.
>
>> * Have you ever tried to use CURL instead of the GUI?
> Yes, only I used a POST instead of a GET
>
>> * Have you tried any other query in the GUI with any other small HDT
> dataset? It might also be some issue within the GUI + HDT
>
> Yes, I've used other datasets before. BTW other queries on Wikidata return
> quickly, except this one.
>
>> * Do you use the built-in Fuseki of the HDT package which in fact is Fuseki1?
> I use Fuseki 3.5.0 (standalone server, no service)
>
> Basically I did what you did, with one difference: I didn't setup the HDT as
> ja:defaultGraph, but as ja:namedGraph. This seems to make all the difference
> in the world! In fact, I've tried to replicate your test, thus using
> ja:defaultGraph, and it seems to works as expected (it returned quickly)!!!
> If I setup a ja:namedGraph instead, my computer explodes! Can you re-try with
> a named graph? The query is the same, except that you add "FROM
> <http://example.org/wikidata>". Other than this, the only difference is that
> you probably have more RAM than me.
