If the results from TDB and from Lucene have to be joined, that can cause some 
overhead, but I am not familiar enough with that tooling to see from your query 
whether that is a potential issue.

---
A. Soroka
The University of Virginia Library

> On Dec 23, 2016, at 11:48 AM, Samur Araujo <[email protected]> wrote:
> 
> Hi Andy, I run the query many times and it still slow.
> 
> I observed that when I index the data directly on lucene/solr (version
> 5.5.3) the query takes 9ms.
> 
> Is it the lucene version (4.1) used by Fuseki that is slow or are there
> other potential overhead?
> 
> On 23 December 2016 at 17:42, Andy Seaborne <[email protected]> wrote:
> 
>> Quite possibly. jena-spatial is a lightweight solution using an external
>> index to using geo data - it is not GeoSPARQL.
>> 
>> (Just running a query once will incur a lot of start-up costs.)
>> 
>>    Andy
>> 
>> 
>> On 22/12/16 14:39, Samur Araujo wrote:
>> 
>>> Hi all,
>>> 
>>> I loaded geonames(40 million triples) into Fuseki and I indexed the data
>>> into lucene.
>>> 
>>> The query below takes 4 seconds to execute. While a similar SQL one into
>>> postgis takes 13 ms.
>>> 
>>> PREFIX spatial: <http://jena.apache.org/spatial#>
>>> PREFIX rdfs: <http://www.w3.org/2000/01/rdf-schema#>
>>> 
>>> SELECT ?place
>>> {
>>>    ?place spatial:withinCircle (32.55668 -117.12865    1  "km"  ) .
>>> 
>>> 
>>> }
>>> 
>>> SQL
>>> 
>>> select  geonameid from geoname where ST_Intersects(the_geom,
>>> ST_Buffer(ST_SetSRID(ST_MakePoint(-117.12865, 32.55668
>>> ),4326)::geography,
>>> 1000)::geometry)
>>> 
>>> Do I need to make any extra configuration on Fuseki to improve the
>>> performance?
>>> 
>>> 
>>> Best,
>>> 
>>> 
> 
> 
> -- 
> Senior Data Scientist
> Geophy
> www.geophy.com
> 
> Nieuwe Plantage 54-55
> 2611XK  Delft
> +31 (0)70 7640725
> 
> 1 Fore Street
> EC2Y 9DT  London
> +44 (0)20 37690760

Reply via email to