[ 
https://jira.codehaus.org/browse/MINDEXER-69?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Igor Fedorenko updated MINDEXER-69:
-----------------------------------

    Attachment: 0001-MINDEXER-69-reuse-rewritten-query-for-all-artifacts.patch

DefaultIteratorResultSet executed Query#rewrite(IndexReader) for each 
ArtifactInfo/MatchHighlightRequest. This was always redundant, but in Lucene 
3.0.3 (and likely earlier) Query#rewrite was relatively inexpensive and 
redundant calls did not cause obvious problems.

In Lucene 3.6.1 Query#rewrite performance comparably with earlier version for 
one indexing context but becomes slower as number of indexing contexts involved 
in the search grows.

The proposed solution is to invoke Query#rewrite only once for each 
MatchHighlightRequest and then reuse the same rewritten query for all 
ArtifactInfo
                
> Severe keyword serach performance regression 
> ---------------------------------------------
>
>                 Key: MINDEXER-69
>                 URL: https://jira.codehaus.org/browse/MINDEXER-69
>             Project: Maven Indexer
>          Issue Type: Bug
>    Affects Versions: 5.0.0
>            Reporter: Igor Fedorenko
>         Attachments: 
> 0001-MINDEXER-69-reuse-rewritten-query-for-all-artifacts.patch
>
>
> I have a performance regression test evaluates performance of keyword search 
> from 15 indexes of various sizes. Everything else being equal, I see ~2.3 
> times performance drop going from maven indexer 4.1.2 to 5.0.0. I'll attach 
> proposed patch with analysis of the problem shortly.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators: 
https://jira.codehaus.org/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira

        

Reply via email to