[jira] Commented: (LUCENE-912) DisjunctionMaxScorer.skipTo has bug that keeps it from skipping

["Paul Elschot (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/LUCENE-912?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12501827 ] 

Paul Elschot commented on LUCENE-912:
-------------------------------------

I'm not sure, but the patch allows to increment all generators right after the first time, and that might cause it to make a mistake in the score computation for its first doc.

It might be worthwhile to make DisjunctionMaxScorer use the same specialized queue/heap as DisjunctionSumScorer.
Factoring this out could lead to a common superclass that provides e.g. an array of Scorers that match the disjunction.
At the same time the score computations of the subscorers could be delayed until an actual score value is needed.
That is a rather drastic approach, but the specialized queue/heap of DisjunctionSumScorer might also help performance of DisjunctionMaxScorer


> DisjunctionMaxScorer.skipTo has bug that keeps it from skipping
> ---------------------------------------------------------------
>
>                 Key: LUCENE-912
>                 URL: https://issues.apache.org/jira/browse/LUCENE-912
>             Project: Lucene - Java
>          Issue Type: Bug
>    Affects Versions: 2.0.0, 2.1
>            Reporter: Hoss Man
>             Fix For: 2.2
>
>         Attachments: dismax_skipto.patch
>
>
> as reported on the mailing list, DisjunctionMaxScorer.skipTo is broken if called before next in some situations...
> http://www.nabble.com/Potential-issue-with-DisjunctionMaxScorer-tf3846366.html#a10894987

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


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