Thank you for taking the time to write such a clear response Martin. Its quite nice to see an Apache project doing due diligence on complying with the geo standards. The current lucene-spatial module and dependencies are lacking in this area so there is mutual interest in leveraging the SIS sister project to fill many of these gaps. It keeps lucene free from overlapping capabilities and clear to focus on the search mission.
I'm finishing up some compressed QuadTree improvements for Lucene spatial and will be shifting my focus to experimenting with what initial spatial geometry capabilities can be replaced by SIS that's currently provided by JTS and S4J. After that investigation is complete I like your proposed approach of retrofitting geo3d to the ISO 19107 model and investigating how it can be integrated with SIS. I think this exercise alone will expose the work needed to bring SIS to lucene while introducing an experimental 3d package to the SIS core-referencing module. One additional question (to you or the group): in quickly browsing the source code (primarily the referencing module) I did not find any spatial relation implementation (e.g., DE4/9IM). Its highly likely I overlooked, but these relations will become important for the spatial search problem. If not currently provided this will likely be the first step required to bringing SIS to lucene-spatial. On Mon, Apr 6, 2015 at 10:29 AM, Martin Desruisseaux < [email protected]> wrote: > Hello Nicholas > > Thanks you for pointing us to LUCENE-6196, I was not aware of it. > > After reading LUCENE-6196 threads and browsing quickly through the > source code, I got the feeling that your proposal - to temporarily add > Geo3D into Lucene-spatial and progressively see which classes could find > a home in SIS - is a good approach. For the reason mentioned in > LUCENE-6196 (code is tuned for the search problem), but also because a > different conceptual model may be considered. > > We are strongly interested to a 3D geometric package, and we indeed plan > to provide one. But this long term project may happen too slowly for > Lucene-spatial. The reason is that we try to put a strong emphasis on > international standards, both from ISO (International Organization for > Standardisation) and from OGC (Open Geospatial Consortium). Those > organisations have an international standard for 3D geospatial > geometries, which is ISO 19107 (Geographic information — Spatial > schema). ISO 19107 is also the model behind the GML (Geographic Markup > Language) format. Unfortunately ISO 19107 is reputed very complex. We > know at least 2 Ph.D students who made their thesis about implementing > ISO 19107 [1], but we have not yet been able to make their work > operational. > > Fortunately, ISO 19107 authors are aware of the standard complexity and > are working on a new versions which brings great simplifications. This > new revision is not yet completed, but is quite advanced. I think that > we should be able to start an experimental implementation soon if we > have volunteer time. One possible approach that I would like to propose > would be to pick Geo3D classes one-by-one and see how they can be > retrofitted in the ISO 19107 model. One a class has been ported into SIS > in an ISO 19107 way, Lucene-spatial may consider replacing their old > class by the new one if they wish. Would that make sense? > > Maybe the above proposal would imply one issue that Lucene-spatial > developers will want to debate: if the above approach was taken, it > would imply a dependency from Lucene-spatial to SIS, which may be a > large dependency (expect ~2 Mb for the "coordinate by referencing" + > "geometry" + "metadata" parts). Whether it would be worst the cost > depends on the amount of functionalities of interest to Lucene-spatial. > If Lucene-spatial used only the Geo3D objects ported to the ISO 19107 > model, that may be an overkill. But if Lucene-spatial is interested in > map projections, including parsing projection definitions from Well > Known Text (WKT) or fetching them from the industry-standard EPSG > database, or in reading geometries from file formats (Shapefile, GML), > then I think SIS would be worth. > > Martin > > > [1] http://w1.cirrelt.ca/~jena/files/DiplThesisJena07.pdf > > Le 06/04/15 14:36, Nicholas Knize a écrit : > > There is a proposed contribution for a 3D planar geometry package to be > > added to Lucene-spatial. See the description and package here: > > https://issues.apache.org/jira/browse/LUCENE-6196 > > > > Its quite a nice contribution that has stalled due to some back and forth > > on whether this belongs in Lucene or a Computational Geometry project. > > Since its primary function is 3D planar Geometry (in Spherical or > Euclidean > > lat/lon space), and the SIS .geometry package has no 3D functionality, it > > seems like a natural fit for SIS. > > > > I thought I'd reach out to the group to gauge interest in the > > contribution. After speaking with other Lucene contributors it seems > > temporarily adding it to Lucene-spatial until it can find a permanent > home > > in its natural habitat provides the best value to the Geo community. > > > > Interested in thoughts. > > > > - Nick > > > >
