[jira] [Commented] (LUCENE-7051) Remove the "estimate match count" optimization from point queries

["Adrien Grand (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/LUCENE-7051?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15262116#comment-15262116 ] 

Adrien Grand commented on LUCENE-7051:
--------------------------------------

After Robert's feedback on LUCENE-7254 I tried to understand better why the cost esimation did not seem to help here. The reason is that IntNRQ is not a good benchmark for DocIdSetBuilder. All IntNRQ queries match MORE than 25% of documents, so the run time is completely dominated by calls to {{FixedBitSet.set}}. I will shortly open a new issue to undo this change.

> Remove the "estimate match count" optimization from point queries
> -----------------------------------------------------------------
>
>                 Key: LUCENE-7051
>                 URL: https://issues.apache.org/jira/browse/LUCENE-7051
>             Project: Lucene - Core
>          Issue Type: Improvement
>            Reporter: Adrien Grand
>            Assignee: Adrien Grand
>            Priority: Minor
>             Fix For: 6.0
>
>         Attachments: LUCENE-7051.patch
>
>
> Point queries try to estimate the number of matches in the visitor so that the doc id set that they build does not have to do it by itself. However, this is incorrect in the multi-valued case and does not seem to buy much (if any) in terms of performance?



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@lucene.apache.org
For additional commands, e-mail: dev-help@lucene.apache.org