[
https://issues.apache.org/jira/browse/LUCENE-7199?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Karl Wright resolved LUCENE-7199.
---------------------------------
Resolution: Fixed
Fix Version/s: 6.x
master
Committed a change to address this. For most polygons, a correct pole is now
chosen in between 1 and 5 iterations. Some rare ones need up to 23 iterations.
Still, far better than before, which sometimes went through 500 iterations to
find the pole.
[~mikemccand], if you are so inclined, it would be great to perf test the
polygon non-pre-built query case again to see if this made a substantial
difference.
> Improve algorithm for randomly finding an interior point for Geo3D polygons
> ---------------------------------------------------------------------------
>
> Key: LUCENE-7199
> URL: https://issues.apache.org/jira/browse/LUCENE-7199
> Project: Lucene - Core
> Issue Type: Improvement
> Components: modules/spatial3d
> Affects Versions: master
> Reporter: Karl Wright
> Assignee: Karl Wright
> Fix For: master, 6.x
>
>
> [~mikemccand]'s benchmark for polygon search shows that constructing Geo3D
> polygon queries is quite expensive compared to other technologies. My belief
> is that this is due largely to how an interior point for determination of
> clockwise/counterclockwise are found: this is currently searched for randomly
> across the entire globe.
> I suspect that we could replace this algorithm with a random algorithm that
> uses one of the polygon's edge points and looks randomly within a small
> distance of that point. This would greatly reduce the number of failed
> attempts to find a "pole" (as I call it).
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
