Hi Graham, (I'm one of the GeoSPARQL next version editors).
> a) can I use Jena’s GEOSparql stuff with both the feature set and the spatial > index in memory? Can I have the index in mem and the data not, or vice-versa? That's a question for the Jena engineers, sorry. > b) can I mix geometric and non-geometric conditions in one query (select > features with some sort of geometry condition and some sort of non geometry > condition). Yes. The GeoSPARQL part of functions in Jena are usually just FILTER statements to filter in/out geometries. You can use other SPARQL patterns or filters (or text index elements) within the same query as one in which you use the GeoSPARQL FILTER. > c) a typical map has between 50K and 150K features (so that many geoms) — > does anyone have any guess how big the spatial index will be for such a case? Not huge. It matters how detailed your geometries are (precision, number of points etc) but I'd say "normal" size, compared to other spatial indexs in other systems. Smaller than the RDF triples indexes. We have spatial indexes on millions of features working quite easily. > d) how fast are the GEOSparql capabilities? I realize that is a bit of an > open question, but experiments with spatial indexes in H2 in memory show that > i can comfortably do a simple bounding box spatial query (so no other > conditions) and get back 50K results in times of the order of seconds. Pretty fast! Once indexed, GeoSPARQL queries perform like other Java-based spatial indexed systems. I suspect it will work like H2 does on the spatial bit. I would be interested in a report back with specific numbers if you generate them. Cheers, Nick ------- Original Message ------- On Monday, September 25th, 2023 at 12:28, graham <[email protected]> wrote: > Hi > > I am considering using Jena’s GEOSparql capabilities for storing and querying > sailing maps (a map is a collection of features, and each feature has a > geometry and other attributes (sea depth, quality of data metric, etc)). > > Before I do that I was wanting to get a bit of advice to understand my > options a bit: > > a) can I use Jena’s GEOSparql stuff with both the feature set and the spatial > index in memory? Can I have the index in mem and the data not, or vice-versa? > > b) can I mix geometric and non-geometric conditions in one query (select > features with some sort of geometry condition and some sort of non geometry > condition). > > c) a typical map has between 50K and 150K features (so that many geoms) — > does anyone have any guess how big the spatial index will be for such a case? > > d) how fast are the GEOSparql capabilities? I realize that is a bit of an > open question, but experiments with spatial indexes in H2 in memory show that > i can comfortably do a simple bounding box spatial query (so no other > conditions) and get back 50K results in times of the order of seconds. > > thanks > graham
