[
https://issues.apache.org/jira/browse/LUCENE-7069?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15241326#comment-15241326
]
Robert Muir commented on LUCENE-7069:
-------------------------------------
like i say, worst case linear scan :)
I just don't think we should add trappy/slow operations when there are still so
many traps in basic functionality. And using a distance sort is fine for use
cases like that (where unlike deleted docs, filtering is not "bound" by merging
or some other process, its arbitrary). It is already designed to handle that
situation, prevent any trappy/crazy worst-cases, etc.
> Add LatLonPoint.nearest to find closest indexed point to a given query point
> ----------------------------------------------------------------------------
>
> Key: LUCENE-7069
> URL: https://issues.apache.org/jira/browse/LUCENE-7069
> Project: Lucene - Core
> Issue Type: Improvement
> Reporter: Michael McCandless
> Assignee: Michael McCandless
> Fix For: master, 6.1
>
> Attachments: LUCENE-7069.patch, LUCENE-7069.patch, LUCENE-7069.patch,
> LUCENE-7069.patch
>
>
> KD trees (used by Lucene's new dimensional points) excel at finding "nearest
> neighbors" to a given query point ... I think we should add this to Lucene's
> sandbox as:
> {noformat}
> public static Document nearest(IndexReader r, String field, double lat,
> double lon) throws IOException
> {noformat}
> I only implemented the 1 nearest neighbor for starters ... I think we can
> easily generalize this in the future to K nearest.
> It could also be generalized to more than 2 dimensions, but for now I'm
> making the class package private in sandbox for just the geo2d (lat/lon) use
> case.
> I don't think this should go into 6.0.0, but should go into 6.1: it's a new
> feature, and we need to wrap up and ship 6.0.0 already ;)
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
