[
https://issues.apache.org/jira/browse/LUCENE-6537?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14579075#comment-14579075
]
Alan Woodward commented on LUCENE-6537:
---------------------------------------
I think it's scoring changes. The benchmark is getting the top ten hits,
ranking them by score, merging any docs that have the same score into a group,
and then counting the groups. What's happened here is that doc 3979685's score
has increased (presumably because NSO is now finding an extra Span in that
document that was being discarded by the eager shrink-to-smallest-fit algorithm
before), and it has pushed doc 85504 out of the top 10. But 85504 was part of
a group of three docs with identical scores, so the number of score groups has
increased by one.
I'm not sure what the point of doing the score-grouping is though? It seems a
pretty arbitrary thing to be checking?
> Make NearSpansOrdered use lazy iteration
> ----------------------------------------
>
> Key: LUCENE-6537
> URL: https://issues.apache.org/jira/browse/LUCENE-6537
> Project: Lucene - Core
> Issue Type: Improvement
> Reporter: Alan Woodward
> Priority: Minor
> Attachments: LUCENE-6537.patch
>
>
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org
For additional commands, e-mail: dev-h...@lucene.apache.org
