[ https://issues.apache.org/jira/browse/LUCENE-10189?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17430647#comment-17430647 ]
Robert Muir commented on LUCENE-10189: -------------------------------------- For the SortedSetCase, seems like we want to fix the IW component to use {{DocValues.singleton}}, box it up, and return it from {{SortedSetDocValues getDocValues()}} ? Then the DocValuesConsumer can simply check with {{DocValues.unwrapSingleton}}. The same codepath can work for merge and flush. > Optimize SortedSet/SortedNumeric doc values writers for fields that are > effectively single-valued > ------------------------------------------------------------------------------------------------- > > Key: LUCENE-10189 > URL: https://issues.apache.org/jira/browse/LUCENE-10189 > Project: Lucene - Core > Issue Type: Wish > Reporter: Adrien Grand > Priority: Minor > > I was wondering how much overhead multi-valued doc-value types have over > their single-valued counterparts, so I hacked IndexTaxis to index all > doc-value fields via Sorted(Set|Numeric)DocValuesField instead of > (Sorted|Numeric)DocValuesField and flush times increased by 30%. It should be > easy to automatically detect such cases in the doc values writers? -- This message was sent by Atlassian Jira (v8.3.4#803005) --------------------------------------------------------------------- To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org For additional commands, e-mail: issues-h...@lucene.apache.org