[jira] [Commented] (SOLR-7818) Distributed stats only calculates with the terms that are present in the last shard of the distributed request

["Anshum Gupta (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/SOLR-7818?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14642279#comment-14642279 ] 

Anshum Gupta commented on SOLR-7818:
------------------------------------

We should merge the terms from shard responses in a set and add that outside of the loop rather than calling {{createNormalizedWeight}} and {{extractTerms}}. This approach could potentially return only local terms, depending upon the query parser.

Also having a test would be nice so we don't regress.

> Distributed stats only calculates with the terms that are present in the last shard of the distributed request
> --------------------------------------------------------------------------------------------------------------
>
>                 Key: SOLR-7818
>                 URL: https://issues.apache.org/jira/browse/SOLR-7818
>             Project: Solr
>          Issue Type: Bug
>    Affects Versions: 5.0, 5.1, 5.2, 5.2.1
>            Reporter: Varun Thacker
>            Assignee: Varun Thacker
>             Fix For: 5.3
>
>         Attachments: SOLR-7818.patch
>
>
> In ExactStatsCache#mergeToGlobalStats we go through the n responses and merge the termStats and colStats. But we keep putting the terms into the map which means only the last shard response terms will get used.
> This can lead to some terms not ending up calculating the distributed idf scores as the last shard might not have contained the term while the other shards actually had those terms.



--
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