Did you get this working in the end Maarten? I have the same problem with the way 'intersects' works and Jilles's solution doesn't work for me; possibly due to the 'tree_levels' accuracy for quad tree. As a kind of workaround, I was thinking that you could draw 2 'envelope' geo_shape intersection; one vertical and one horizontal. Using a boolean filter you could reduce your results to only the polygons which intersect BOTH intersections. It would look something like this: <https://lh3.googleusercontent.com/-NkNXyNjCLwQ/U6wBLrRIbNI/AAAAAAAAJkU/383LpZNwVzE/s1600/bbox.png> Here's an example of the query: https://gist.github.com/missinglink/ce25f1f9f7554ff1010f Unfortunately I couldn't get this working either, do you think this may be a simpler workaround for the 'intersects' issue? -P On Tuesday, 5 March 2013 14:17:01 UTC+1, Jilles van Gurp wrote: > > Actually, I was wrong again. I did make it work in the end. > > If others want to replicate my experiment, I pasted a few sample queries > and responses as well as the data and mapping I indexed here: > http://pastebin.com/QpXvYK5B > > The first query shows that a query with a circle of 1m radius returns > Berlin and Mitte for a circle near Rosenthalerplatz. So, the 1m circle > polygon intersects with both polygons that fully contain the circle polygon. > > If I increase the radius to 300 meter it also finds the nearby pois. There > is a problem with accuracy though. I noticed that 100 meter doesn't return > any of the pois. even though I know that it should have returned a few. If > I increase to 150 it returns the Prenzlauerberg polygon but still no POIs. > At 157 still the same. At 158, I suddenly get 8 results. > > I might be able to improve the accuracy by playing with the levels in the > mapping. > > So what might be happening is that you are running into this accuracy > issue here as well. If you haven't already, upgrading to a post 0.90-beta1 > snapshot build might help you. That would get you the changes made a few > days ago to the geo_shape implementation. > > Jilles > > > > > On Monday, March 4, 2013 6:44:10 PM UTC+1, Maarten Janssen wrote: >> >> Or even better a new ShapeRelation should be produced called CONTAINS or >> IS_PART_OF where you get do the other way around. In our case (I think we >> need the same approach) we could go top down and match first all the >> polygons, and then use indeed a contains method, but in my opnion it should >> be part of the spatial search capabilities. >> >> >> On Mon, Mar 4, 2013 at 6:33 PM, Jilles van Gurp <jilles...@gmail.com> >> wrote: >> >>> Hmm, I just tried this and it seems you are right. Very annoying. I >>> guess was assuming too much about intersects. >>> >>> I guess you could turn things around and use a really large circle that >>> for sure includes all the polygons you care about and then do a filter on >>> polygon containment after you get the results. My library offers a contains >>> function. But that is far from ideal. >>> >>> This sounds like it might be a bug though since technically a large >>> feature should intersect with any small feature contained by it. >>> >>> Jilles >>> >>> >>> >>> >>> -- >>> You received this message because you are subscribed to a topic in the >>> Google Groups "elasticsearch" group. >>> To unsubscribe from this topic, visit >>> https://groups.google.com/d/topic/elasticsearch/bDUQKTK0oe0/unsubscribe?hl=en-US >>> . >>> To unsubscribe from this group and all its topics, send an email to >>> elasticsearc...@googlegroups.com. >>> For more options, visit https://groups.google.com/groups/opt_out. >>> >>> >>> >> >> >> >> -- >> Maarten Janssen >> CTO Infohubble >> >> Email: mjan...@infohubble.com >> URL: www.infohubble.com >> Mobile: +31620006790 >> Card: eee.am/mjanssen >> >> *Address:* >> >> Infohubble BV >> Gebouw Sevilla >> Entrada 304 >> 1096 ED Amsterdam >> The Netherlands >> > -- You received this message because you are subscribed to the Google Groups "elasticsearch" group. To unsubscribe from this group and stop receiving emails from it, send an email to elasticsearch+unsubscr...@googlegroups.com. To view this discussion on the web visit https://groups.google.com/d/msgid/elasticsearch/efa8d361-fc4f-46fb-9e99-d40910caa395%40googlegroups.com. For more options, visit https://groups.google.com/d/optout.
