Noticing this a little late, but have a peak at these posts from Chris Hodgson about how the R-Tree fails for variable density GIS data, and his approach to a packing process. Unfortunately, packing is a post-facto process, and it's not clear to me how we would do it with the GiST infrastructure that undergirds the PostGIS database. But it's nice to see a good R-Tree and a reminder of just how ugly they can get under the covers.Friday, August 06, 2010
Take this Tree and Pack it
Noticing this a little late, but have a peak at these posts from Chris Hodgson about how the R-Tree fails for variable density GIS data, and his approach to a packing process. Unfortunately, packing is a post-facto process, and it's not clear to me how we would do it with the GiST infrastructure that undergirds the PostGIS database. But it's nice to see a good R-Tree and a reminder of just how ugly they can get under the covers.
Subscribe to:
Post Comments (Atom)
3 comments:
Vacuum Pack :)
Thanks for the link Paul. I still want to take this further someday.
check out the rtree splitting results that were recently uploaded to the Wikipedia R-tree article.
The Ang-Tan heuristic -- which afaict is what Postgres uses -- is really bad. GiST should be using the R*-tree instead.
Post a Comment