You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@lucene.apache.org by "Adrien Grand (Jira)" <ji...@apache.org> on 2019/09/02 13:21:00 UTC
[jira] [Commented] (LUCENE-8956) QueryRescorer sort optimization
[ https://issues.apache.org/jira/browse/LUCENE-8956?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16920853#comment-16920853 ]
Adrien Grand commented on LUCENE-8956:
--------------------------------------
I was thinking that we could verify that we have the right hits by rescoring twice, once with topN=random().nextInt(numDocs) like in your patch, and another time with topN=numDocs, then make sure that the first topN hits are the same in both cases (CheckHits#checkEquals might help).
> QueryRescorer sort optimization
> -------------------------------
>
> Key: LUCENE-8956
> URL: https://issues.apache.org/jira/browse/LUCENE-8956
> Project: Lucene - Core
> Issue Type: Improvement
> Components: core/query/scoring
> Reporter: Paul Sanwald
> Priority: Minor
> Attachments: LUCENE-8956.patch
>
>
> This patch addresses a TODO in QueryRescorer: We should not sort the full array of the results returned from rescoring, but rather only topN, when topN is less than total hits.
>
> Made this optimization with some suggestions from [~jpountz] and [~jimczi], this is my first lucene patch submission.
--
This message was sent by Atlassian Jira
(v8.3.2#803003)
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@lucene.apache.org
For additional commands, e-mail: dev-help@lucene.apache.org