Just out of curiosity - if I find an OSM geometry (say, a way) from e.g. a SpatialDatabaseRecord, is there any easy way to get to its underlying OSM points, e.g. via an iterator? The Encoder/Decoder could probably do that job and return this without the client needing to know details about the graph layout, right?
Cheers, /peter neubauer GTalk: neubauer.peter Skype peter.neubauer Phone +46 704 106975 LinkedIn http://www.linkedin.com/in/neubauer Twitter http://twitter.com/peterneubauer http://www.neo4j.org - Your high performance graph database. http://www.thoughtmade.com - Scandinavia's coolest Bring-a-Thing party. On Wed, Jan 19, 2011 at 1:45 AM, Craig Taverner <[email protected]> wrote: > Hi Bryce, > > While I had this on my mind, I decided to look into the code and I went > ahead developed the first approach. So now we index osm nodes in the r-tree > also (in addition to ways). By default, only nodes that have tags are added > to the index, because these are seen as 'points of interest', while all > other nodes are assumed to only be interesting as part of a way. However, I > added a new method to the OSMImporter that allows you to turn on indexing of > all nodes, regardless of whether they are points of interest or not. > > Just replace the old call to osmImporter.importFile(batchInserter,filePath) > with instead osmImporter.importFile(batchInserter,filePath,true). The last > parameter turns on indexing of all nodes. I've attached an image from the > AWE product (based on uDig) that shows a large number of dynamic layers > exported from the OSM model of Malmö, and you can see all nodes are visible. > This is the output from the TestDynamicLayers unit test, run with the > boolean set to true. > > If you are curious, the Neo Technology office is actually included in this > image :-) > > I have just pushed the code to github, so you need to get the latest from > there. > > Regards, Craig > > On Tue, Jan 18, 2011 at 9:27 PM, Craig Taverner <[email protected]> wrote: > >> Hi Bryce, >> >> Currently the OSM model in Neo4j Spatial only indexes the ways, which means >> that when you do spatial searches on the layers, you are always talking >> about the ways. Whether this suites you or not depends on your specific use >> case. You could search for the closest ways, and the use the >> SpatialTopologyUtils to snap to the point node on the way that is closest to >> your point of interest, which may give you the result you want. >> >> If you are specifically interested in only searching on nodes, not ways, >> then there are two options, both requiring some refactoring of the OSM model >> (ie. you need to make changes to Neo4j Spatial). >> >> - In the OSMImporter code, simply add all nodes as Point geometries to >> the index. This is simple to code and will get what you want, but might >> not >> lead to the most efficient index, since both the ways and the nodes they >> are >> composed of get indexed in the same index. >> - Enhance the OSM model to support multiple indexes, for nodes, ways >> and relationships. This requires much more code, but opens the possibility >> to enhance the performance by allowing the index to be optimized for the >> geometries and densities seen in the dataset. >> >> The reasons we have not done this yet are: >> >> - Existing use cases have not required it (but we are reaching cases >> that do now) >> - It is not clear yet which of the above options is the best overall >> (one index or several). Originally I believed in the multiple-index >> approach, mostly because points can be indexed with very different indexes >> than geometries (eg. we can even place lucene behind the point index, >> which >> we cannot do for geometries). But at the moment I am leaning back towards >> the single index approach (simpler code). >> >> So, I hope the approach of searching for close ways, and then using the >> topology utils to snap to the closest point is the right solution for you. >> Otherwise considering contributing to the updates necessary to complete the >> point index :-) >> >> Regards, Craig >> >> On Tue, Jan 18, 2011 at 6:08 PM, bryce hendrix <[email protected]>wrote: >> >>> Peter, >>> >>> I think my problem is more of understanding how the graph is built from >>> OSM >>> data. The method of find the closest edge gives me the way geometry, but >>> (please correct me if I am wrong) I need to find the closest OSM node >>> (Node >>> with "lat" and "lon" properties) in the geometry. >>> >>> Thanks, >>> Bryce >>> >>> On Tue, Jan 18, 2011 at 2:05 AM, Peter Neubauer < >>> [email protected]> wrote: >>> >>> > Bryce, >>> > are you looking for some kind of "snapping" to the nearest geometry? I >>> > was doing something like that, Craig did a test on that, see >>> > >>> > >>> https://github.com/neo4j/neo4j-spatial/blob/master/src/test/java/org/neo4j/gis/spatial/TestSpatialUtils.java#L53 >>> > , >>> > there is some basic support for that in >>> > >>> > SpatialTopologyUtils.java. >>> > >>> > Does that help to get started? Feel free to add more utilities and >>> > maybe a routing example, when you get it working. >>> > >>> > Cheers, >>> > >>> > /peter neubauer >>> > >>> > GTalk: neubauer.peter >>> > Skype peter.neubauer >>> > Phone +46 704 106975 >>> > LinkedIn http://www.linkedin.com/in/neubauer >>> > Twitter http://twitter.com/peterneubauer >>> > >>> > http://www.neo4j.org - Your high performance graph >>> database. >>> > http://www.thoughtmade.com - Scandinavia's coolest Bring-a-Thing party. >>> > >>> > >>> > >>> > On Tue, Jan 18, 2011 at 5:50 AM, bryce hendrix <[email protected]> >>> > wrote: >>> > > Craig, >>> > > >>> > > I've ran into a bit of a stumbling block. I am attempting to do find a >>> > > simple route using A* from 2 nodes on 2 different ways. What is the >>> best >>> > way >>> > > to find the points closest to 2 reference points which are on ways? >>> > Assuming >>> > > I've got those nodes, is there anything special I have to do, other >>> than >>> > > setting up the estimate and cost evaluators? >>> > > >>> > > If I can figure this out, I'll submit my example back to you guys. I'm >>> > > excited about this stuff, but I seem to be discovering the limits of >>> the >>> > > docs. >>> > > >>> > > Bryce >>> > > >>> > > On Sun, Jan 16, 2011 at 7:11 PM, Craig Taverner <[email protected]> >>> > wrote: >>> > > >>> > >> Great that they all work :-) >>> > >> >>> > >> Good luck with the rest of the project and keep us posted, we're >>> > interested >>> > >> in any feedback on the API. >>> > >> >>> > >> (getting lat/long mixed up is one of those things we all keep doing, >>> I'm >>> > >> pretty sure I did it once as recently as last month ... ;-) >>> > >> >>> > >> On Sun, Jan 16, 2011 at 11:52 PM, bryce hendrix < >>> [email protected] >>> > >> >wrote: >>> > >> >>> > >> > Craig, Peter, >>> > >> > >>> > >> > Its useful if I get the latitude and longitude in the correct order >>> > for >>> > >> the >>> > >> > Point. Ugh. I've found >>> > >> > that SearchPointsWithinOrthodromicDistance, SearchClosest, >>> > >> > and SpatialTopologyUtils.findClosestEdges all work for me. Looks >>> like >>> > my >>> > >> > project is well on its way now, thanks for the help. >>> > >> > >>> > >> > Bryce >>> > >> > >>> > >> > On Sun, Jan 16, 2011 at 3:16 PM, Craig Taverner <[email protected]> >>> > >> wrote: >>> > >> > >>> > >> > > The SearchPointsWithinOrthodromicDistance basically does a search >>> on >>> > a >>> > >> > > rectangular bounding box, and then inside the result set filters >>> by >>> > >> > > distance >>> > >> > > from the center. The filter probably works only on points as >>> implied >>> > by >>> > >> > the >>> > >> > > class name. >>> > >> > > >>> > >> > > The SpatialTopologyUtils class has a method findClosestEdge, >>> which >>> > will >>> > >> > do >>> > >> > > what you are looking for. If you call it without a distance >>> value, >>> > it >>> > >> > will >>> > >> > > take 1% of the total span of your layer as the search window, so >>> if >>> > >> this >>> > >> > > does not make sense for your data (eg. your layer covers a small >>> > area, >>> > >> as >>> > >> > > you hinted at), then pass in the distance in units of the >>> coordinate >>> > >> > system >>> > >> > > of the layer (probably WGS84, degrees, if you are using only OSM >>> > data). >>> > >> > Try >>> > >> > > it out and let us know. >>> > >> > > >>> > >> > > See: >>> > >> > > >>> > >> > > - >>> > >> > > >>> > >> > > >>> > >> > >>> > >> >>> > >>> https://github.com/neo4j/neo4j-spatial/blob/master/src/main/java/org/neo4j/gis/spatial/SpatialTopologyUtils.java >>> > >> > > - >>> > >> > > >>> > >> > > >>> > >> > >>> > >> >>> > >>> https://github.com/neo4j/neo4j-spatial/blob/master/src/test/java/org/neo4j/gis/spatial/TestSpatialUtils.java >>> > >> > > >>> > >> > > >>> > >> > > On Sun, Jan 16, 2011 at 11:19 AM, Peter Neubauer < >>> > >> > > [email protected]> wrote: >>> > >> > > >>> > >> > > > Bryce, >>> > >> > > > I think (Craig, correct me if I'm wrong) you need to have a >>> Point >>> > >> > > > layer to be able to do that search. The default OSM layer is >>> > >> > > > containing a lot of geometries, so I think you first should >>> define >>> > a >>> > >> > > > layer on top of the full imported one, then search. I did >>> > something >>> > >> > > > like that in another spike, see >>> > >> > > > >>> > >> > > > >>> > >> > > >>> > >> > >>> > >> >>> > >>> https://github.com/popdevelop/snapplr/blob/master/server_java/src/main/java/com/geosnappr/TaginfoImporter.java#L312 >>> > >> > > > >>> > >> > > > The layer is defined with something like >>> > >> > > > >>> > >> > > > >>> > >> > > >>> > >> > >>> > >> >>> > >>> https://github.com/neo4j/neo4j-spatial/blob/master/src/test/java/org/neo4j/gis/spatial/TestDynamicLayers.java#L26 >>> > >> > > > on top of the imported full data layer. >>> > >> > > > >>> > >> > > > Does that help? >>> > >> > > > >>> > >> > > > Cheers, >>> > >> > > > >>> > >> > > > /peter neubauer >>> > >> > > > >>> > >> > > > GTalk: neubauer.peter >>> > >> > > > Skype peter.neubauer >>> > >> > > > Phone +46 704 106975 >>> > >> > > > LinkedIn http://www.linkedin.com/in/neubauer >>> > >> > > > Twitter http://twitter.com/peterneubauer >>> > >> > > > >>> > >> > > > http://www.neo4j.org - Your high performance >>> graph >>> > >> > > database. >>> > >> > > > http://www.thoughtmade.com - Scandinavia's coolest >>> Bring-a-Thing >>> > >> > party. >>> > >> > > > >>> > >> > > > >>> > >> > > > >>> > >> > > > On Sat, Jan 15, 2011 at 10:58 PM, bryce hendrix < >>> > >> > [email protected]> >>> > >> > > > wrote: >>> > >> > > > > I'm pretty new to neo4j, so please excuse me if this is a >>> FAQ. >>> > >> > > > > >>> > >> > > > > I exported OSM data for a city from the OSM site, then >>> imported >>> > it >>> > >> > > using >>> > >> > > > the >>> > >> > > > > OSMImporter. I can see the layer via the webserver, so I know >>> if >>> > >> got >>> > >> > > > > imported okay. Now I would like to find the way nearest to a >>> > >> > coordinate >>> > >> > > > via >>> > >> > > > > the Java API, but I'm not really sure how to do that. I've >>> tried >>> > >> > using >>> > >> > > > > SearchPointsWithinOrthodromicDistance, but the results of the >>> > query >>> > >> > are >>> > >> > > > > always empty. Can someone give me some tips, or provide a >>> simple >>> > >> > > example? >>> > >> > > > > >>> > >> > > > > Thanks in advance, >>> > >> > > > > Bryce >>> > >> > > > > _______________________________________________ >>> > >> > > > > Neo4j mailing list >>> > >> > > > > [email protected] >>> > >> > > > > https://lists.neo4j.org/mailman/listinfo/user >>> > >> > > > > >>> > >> > > > _______________________________________________ >>> > >> > > > Neo4j mailing list >>> > >> > > > [email protected] >>> > >> > > > https://lists.neo4j.org/mailman/listinfo/user >>> > >> > > > >>> > >> > > _______________________________________________ >>> > >> > > Neo4j mailing list >>> > >> > > [email protected] >>> > >> > > https://lists.neo4j.org/mailman/listinfo/user >>> > >> > > >>> > >> > _______________________________________________ >>> > >> > Neo4j mailing list >>> > >> > [email protected] >>> > >> > https://lists.neo4j.org/mailman/listinfo/user >>> > >> > >>> > >> _______________________________________________ >>> > >> Neo4j mailing list >>> > >> [email protected] >>> > >> https://lists.neo4j.org/mailman/listinfo/user >>> > >> >>> > > _______________________________________________ >>> > > Neo4j mailing list >>> > > [email protected] >>> > > https://lists.neo4j.org/mailman/listinfo/user >>> > > >>> > _______________________________________________ >>> > Neo4j mailing list >>> > [email protected] >>> > https://lists.neo4j.org/mailman/listinfo/user >>> > >>> _______________________________________________ >>> Neo4j mailing list >>> [email protected] >>> https://lists.neo4j.org/mailman/listinfo/user >>> >> >> > > _______________________________________________ > Neo4j mailing list > [email protected] > https://lists.neo4j.org/mailman/listinfo/user > > _______________________________________________ Neo4j mailing list [email protected] https://lists.neo4j.org/mailman/listinfo/user
