[ https://issues.apache.org/jira/browse/LUCENE-8033?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16272817#comment-16272817 ]
Michael Braun commented on LUCENE-8033: --------------------------------------- Sorry [~jpountz] - delete by query. Don't have the snapshot of the sampling handy, but it was spending a large amount of time constructing the FieldInfos, all spent in the adding to byNumber within the constructor - which is dropped in the case of the dense case, though it is used so the FieldInfos are sorted in the dense case too - one would need to do sort on another structure at minimum the way it looks right now, not 100% this would even be faster but hopefully! [~dsmiley] exactly, this was a significant amount of time. > Should FieldInfos always use a dense encoding? > ---------------------------------------------- > > Key: LUCENE-8033 > URL: https://issues.apache.org/jira/browse/LUCENE-8033 > Project: Lucene - Core > Issue Type: Wish > Reporter: Adrien Grand > Priority: Trivial > Labels: newdev > > Spin-off from LUCENE-8018. The dense vs. sparse encoding logic of FieldInfos > introduces complexity. Given that the sparse encoding is only used when less > than 1/16th of fields are used, which sounds uncommon to me, maybe we should > use a dense encoding all the time? -- This message was sent by Atlassian JIRA (v6.4.14#64029) --------------------------------------------------------------------- To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org For additional commands, e-mail: dev-h...@lucene.apache.org