[ 
https://issues.apache.org/jira/browse/LUCENE-10367?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17475379#comment-17475379
 ] 

LuYunCheng commented on LUCENE-10367:
-------------------------------------

i think when the minimumNumberMatch is ConstantLongValuesSource , it is 
correctly to rewrite to a BooleanQuery which value is less than the 
queries.size(). i do the Unit Test in TestCoveringQuery.java testRandomWand

 

And i do some mini benchmark using  TestCoveringQueryBench.java with different 
constant value, it shows:

 
Constant Value 1:
TEST: WAND with BooleanQuery elapsed 68ms
TEST: NOWAND elapsed 159ms
Constant Value 2:
TEST: WAND elapsed 86ms
TEST: NOWAND elapsed 225ms
Constant Value 3:
TEST: WAND elapsed 80ms
TEST: NOWAND elapsed 169ms
Constant Value 4:
TEST: WAND elapsed 21ms
TEST: NOWAND elapsed 117ms

> Use WANDScorer in CoveringQuery Can accelerate scorer time
> ----------------------------------------------------------
>
>                 Key: LUCENE-10367
>                 URL: https://issues.apache.org/jira/browse/LUCENE-10367
>             Project: Lucene - Core
>          Issue Type: Improvement
>          Components: core/query/scoring, core/search, modules/sandbox
>            Reporter: LuYunCheng
>            Priority: Major
>         Attachments: LUCENE-10367-1.patch, LUCENE-10367.patch, 
> TestCoveringQueryBench.java
>
>
> When using CoveringQuery In Elasticsearch with terms_set query, it takes too 
> much time in CoveringScore and major cost in matain the DisiPriorityQueue: 
> subScorers.
> But when minimumNumberMatch is ConstantLongValuesSource, we can use 
> WANDScorer to optimize it.
>  
> i do a mini benchmark with 1m docs, which code in LUCENE-10367.patch 
> TestCoveringQuery.java  testRandomBench()
> it shows: 
> TEST: WAND elapsed 67ms
> TEST: NOWAND elapsed 163ms
> My testing environment is macBook with Intel Core i7 16GMem. 



--
This message was sent by Atlassian Jira
(v8.20.1#820001)

---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org
For additional commands, e-mail: issues-h...@lucene.apache.org

Reply via email to