lucene-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Aaron McKee <>
Subject Re: Lucene Spatial
Date Fri, 25 Sep 2009 16:29:14 GMT

This all sounds like great work, thank you for putting in the time! I 
ended up retooling a fair bit of Spatial myself (especially the Tiering 
stuff which was, perhaps, a bit questionable in places). One of the 
bigger concerns I had throughout the package, after delving into it, was 
the overuse of 2d projections and planar math; this is fine for small 
regions, but tends to cause noticeable distortions as you scale up to 
large areas. There was also almost no handling of meridian traversing, 
although it looks like Mark Miller has started adding some code to help 
that (thank you!). These are sort of esoteric concerns, but fairly vital 
if the solution is to scale to meet global needs. I didn't find the JTS 
libraries useful as a geometric basis for basically these reasons, 
unfortunately, although there are likely portions that could be useful 
(e.g. simple features polydata parsing). I've tried to move a fair bit 
of my code to be slightly more 3d intelligent, handle different shapes 
better (bounding boxes natively, at least, with hooks for arbitrary 
polygons), and work better with meridians. To be honest, it's been an 
almost complete rewrite. I'm trying to get management to settle on an 
open source policy (currently, it's all just for internal purposes) and 
hopefully I'll be able to post the bits that may be useful, although I'm 
happy to say it sounds like you're making much of this work obsolete and 
I can't wait to see the patch. =)


gento0nz wrote:
> Hi,
> Since my contribution in LUCENE-1732 I have been working on re-writing 
> most of the spatial contrib, particularly the Cartesian Tier code.  I 
> have added javadoc, tests and and have removed many of the classes 
> which were either deprecated, unused or unnecessary.  I have also 
> tried to define easier extension points for adding things like JTS 
> support, and different data formats (whether it be geohashes or 
> traditional latitude and longitude).  Much of the work has been done 
> with Solr integration in mind with the overall goal being to create 
> enterprise ready spatial search support.
> I am also planning on collaborating with some of the developers from 
> the JPL to improve the range the geographic functions supported and to 
> provide efficient searching with even the most massive and varied of 
> data sets.
> I have every intention to contribute this as another patch soon.
> Thanks,
> Chris

To unsubscribe, e-mail:
For additional commands, e-mail:

View raw message