[jira] [Updated] (LUCENE-7103) further optimize LatLonPoint.newDistanceSort
[ https://issues.apache.org/jira/browse/LUCENE-7103?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Hoss Man updated LUCENE-7103: - Fix Version/s: (was: 6.0) master (7.0) Manually correcting fixVersion per Step #S5 of LUCENE-7271 > further optimize LatLonPoint.newDistanceSort > > > Key: LUCENE-7103 > URL: https://issues.apache.org/jira/browse/LUCENE-7103 > Project: Lucene - Core > Issue Type: Improvement >Reporter: Robert Muir > Fix For: 6.1, master (7.0) > > Attachments: LUCENE-7103.patch > > > This comparator creates bounding boxes to avoid calling haversin(), so its no > longer a hotspot for most use cases. > But in the worst case, it could still get called many times. This could be > because the user had a massive top N value, or because the incoming data is > sorted or mostly sorted by decreasing distance, etc. > We can optimize the worst case by not invoking a full haversin, just using > something that is rank-equivalent. -- This message was sent by Atlassian JIRA (v6.3.4#6332) - To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org For additional commands, e-mail: dev-h...@lucene.apache.org
[jira] [Updated] (LUCENE-7103) further optimize LatLonPoint.newDistanceSort
[ https://issues.apache.org/jira/browse/LUCENE-7103?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Robert Muir updated LUCENE-7103: Attachment: LUCENE-7103.patch simple patch. We "do the rest" of the haversin calculation only when we need the final value: value() and compareTop(). > further optimize LatLonPoint.newDistanceSort > > > Key: LUCENE-7103 > URL: https://issues.apache.org/jira/browse/LUCENE-7103 > Project: Lucene - Core > Issue Type: Improvement >Reporter: Robert Muir > Attachments: LUCENE-7103.patch > > > This comparator creates bounding boxes to avoid calling haversin(), so its no > longer a hotspot for most use cases. > But in the worst case, it could still get called many times. This could be > because the user had a massive top N value, or because the incoming data is > sorted or mostly sorted by decreasing distance, etc. > We can optimize the worst case by not invoking a full haversin, just using > something that is rank-equivalent. -- This message was sent by Atlassian JIRA (v6.3.4#6332) - To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org For additional commands, e-mail: dev-h...@lucene.apache.org
