[
https://issues.apache.org/jira/browse/LUCENE-6759?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14725402#comment-14725402
]
Karl Wright commented on LUCENE-6759:
-------------------------------------
The computation that determines the perpendicular plane that is intersected
with the original plane to find the X or Y bounds is as follows:
(1) Create a perpendicular plane which includes the X axis (or Y axis,
depending)
(2) Evaluate that plane at the two Z points that represent the maximum
excursions in Z for the plane intersecting the planet
(3) Take the average of the evaluations, and construct a new plane that is
parallel to the first perpendicular plane but offset by that amount
In this case, the following is produced:
{code}
[junit4] 2> zPoint: [X=0.09847149376598374, Y=-0.07921875599763584,
Z=-0.9897798734262737]
[junit4] 2> zPoint: [X=0.7398138828430078, Y=-0.5951685429682101,
Z=0.31625095037048134]
...
[junit4] 2> computing X bound
[junit4] 2> not degenerate; B=-0.5302912804708979;
C=-0.5331856607581867
[junit4] 2> originPlane.evaluate(zpoint)=-0.6418043015460514
[junit4] 2> originPlane.evaluate(zpoint)=0.6450052856365084
{code}
So, the D value being computed is the average of two fairly large numbers,
which are largely canceling each other out. Unfortunately, however, that kind
of floating-point calculation leads to a fairly high error value, which I
believe to be the source of the problem we're seeing.
I will ponder if there's a better way to compute what we need here.
> Integrate lat/long BKD and spatial 3d, part 2
> ---------------------------------------------
>
> Key: LUCENE-6759
> URL: https://issues.apache.org/jira/browse/LUCENE-6759
> Project: Lucene - Core
> Issue Type: New Feature
> Reporter: Michael McCandless
> Attachments: LUCENE-6699.patch, LUCENE-6699.patch, LUCENE-6699.patch,
> LUCENE-6699.patch, LUCENE-6699.patch, LUCENE-6699.patch, LUCENE-6699.patch,
> LUCENE-6699.patch, LUCENE-6699.patch, LUCENE-6699.patch, LUCENE-6699.patch,
> LUCENE-6699.patch, LUCENE-6699.patch, LUCENE-6699.patch, LUCENE-6699.patch
>
>
> This is just a continuation of LUCENE-6699, which became too big.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
