[
https://issues.apache.org/jira/browse/LUCENE-7669?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15852718#comment-15852718
]
Robert Muir commented on LUCENE-7669:
-------------------------------------
{quote}
I do think it is bad that we become so slow in this case? Why should we
penalize people searching around the Fiji islands
{quote}
This is incorrect. It is not "so slow". The difference is 6%.
The outer bounding box is entirely unnecessary: each individual sub-polygon
(which cannot cross the dateline) already has its own bounding box in the
KD-tree. The outer bounding box around the whole thing is just a
micro-optimization that saves a kd-tree lookup before checking yet another
bounding box. Please don't add rare optimizations to a micro-optimization that
is already unnecessary :)
Removing it completely, not optimizing it, is the correct direction. Maybe the
kd-tree lookup just needs some cpu speedup or something like that.
> Rectangle.fromPolygon could compute smaller bounding boxes
> ----------------------------------------------------------
>
> Key: LUCENE-7669
> URL: https://issues.apache.org/jira/browse/LUCENE-7669
> Project: Lucene - Core
> Issue Type: Improvement
> Reporter: Adrien Grand
> Priority: Minor
> Attachments: LUCENE-7669.patch
>
>
> Currently it computes the smallest bounding box that does not cross the
> dateline. However allowing to cross the dateline could allow to create
> smaller bounding boxes. For instance, because of that, the bounding box of
> the Russia polygon has a width of 360 longitude degrees. By allowing
> rectangles that cross the dateline, we could get a polygon whose width is
> only 171 longitude degrees. This is useful combined with LUCENE-7661 since it
> means the grid would have higher resolution.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
