[jira] [Updated] (LUCENE-6329) Calling score() should be ok even if needsScores is false

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

     [ https://issues.apache.org/jira/browse/LUCENE-6329?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Adrien Grand updated LUCENE-6329:
---------------------------------
    Attachment: LUCENE-6329.patch

I had started seeing if we could implicitely wrap with ConstantScorer but there seem to be a couple of places that call score() when needsScores is false. Here is another alternative that still stops calling matches() after the first matching clause is found but now also pretends there is a single matching clause so that score() won't fail. 

This is the smallest viable fix I can think of for https://builds.apache.org/job/Lucene-Solr-NightlyTests-5.x/775/. But we should probably think about better fixes for the future and avoid calling score() when it's not needed.

> Calling score() should be ok even if needsScores is false
> ---------------------------------------------------------
>
>                 Key: LUCENE-6329
>                 URL: https://issues.apache.org/jira/browse/LUCENE-6329
>             Project: Lucene - Core
>          Issue Type: Bug
>            Reporter: Adrien Grand
>            Assignee: Adrien Grand
>            Priority: Minor
>             Fix For: Trunk, 5.1
>
>         Attachments: LUCENE-6329.patch
>
>
> Currently if you use approximations on a disjunction scorer and call score() while needsScores was set to false, then you might run into exceptions.
> The reason is that when scores are not needed then the scorer doesn't maintain the list of scorers positioned on the current doc ID and you can eg. end up with exhausted scorers in this list.



--
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