[
https://issues.apache.org/jira/browse/LUCENE-4342?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
David Smiley updated LUCENE-4342:
---------------------------------
Attachment:
LUCENE-4342_fix_distance_precision_lookup_for_prefix_trees,_and_modify_the_default_algorit.patch
Sure enough, there was an off by one bug in the grid level lookup. Thanks so
much for the test case; I enhanced it quite a bit further. Because
GeoHashUtils is in Spatial4j (quite debatable), some of the bug is still there
where I need to get a fix in for Spatial4j0.3 but in the mean time I simply
inlined the relevant lookup code with it fixed.
I also modified the algorithm in which a specific error distance is derived
from a shape and the percent. It's now half the diagonal distance of opposite
corners of its bounding box.
> Issues with prefix tree's Distance Error Percentage
> ----------------------------------------------------
>
> Key: LUCENE-4342
> URL: https://issues.apache.org/jira/browse/LUCENE-4342
> Project: Lucene - Core
> Issue Type: Bug
> Components: modules/spatial
> Affects Versions: 4.0-ALPHA, 4.0-BETA
> Reporter: Itamar Syn-Hershko
> Assignee: David Smiley
> Attachments:
> LUCENE-4342_fix_distance_precision_lookup_for_prefix_trees,_and_modify_the_default_algorit.patch,
> unnamed.patch
>
>
> See attached patch for a failing test
> Basically, it's a simple point and radius scenario that works great as long
> as args.setDistPrecision(0.0); is called. Once the default precision is used
> (2.5%), it doesn't work as expected.
> The distance between the 2 points in the patch is 35.75 KM. Taking into
> account the 2.5% error the effective radius without false negatives/positives
> should be around 34.8 KM. This test fails with a radius of 33 KM.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
