asfimport

Ignacio Vera (@iverase) ([migrated from JIRA](https://issues.apache.org/jira/browse/LUCENE-10396?focusedCommentId=17559739&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-17559739)) If I understand you correct, you mean leveraging the inverted index to get the first document per term. I tried that and my conclusion...

Robert Muir (@rmuir) ([migrated from JIRA](https://issues.apache.org/jira/browse/LUCENE-10396?focusedCommentId=17559769&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-17559769)) do we have any idea of the comparison? I'm just curious because it seems like doing TermsEnum.next() and getting first doc ID should be...

Robert Muir (@rmuir) ([migrated from JIRA](https://issues.apache.org/jira/browse/LUCENE-10396?focusedCommentId=17559772&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-17559772)) Also if performance changes are minor between the two solutions, perhaps we could speed terms/postings up for the sorted case to close the gap....

Ignacio Vera (@iverase) ([migrated from JIRA](https://issues.apache.org/jira/browse/LUCENE-10396?focusedCommentId=17559825&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-17559825)) Here is the code I am using when using postings/TermsEnum from the inverted index which might be totally wrong / inefficient as unfortunately I...

Robert Muir (@rmuir) ([migrated from JIRA](https://issues.apache.org/jira/browse/LUCENE-10396?focusedCommentId=17559848&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-17559848)) The slowness is probably the lookupOrd? Can you avoid this? Just next() the termsenum to move on to the next ord.

Robert Muir (@rmuir) ([migrated from JIRA](https://issues.apache.org/jira/browse/LUCENE-10396?focusedCommentId=17559854&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-17559854)) I'd also modify the call to `termsEnum.postings()` to be `termsEnum.postings(postingsEnum, PostingEnum.NONE)`. Depending on your data, it might not do anything, but you don't need...

Ignacio Vera (@iverase) ([migrated from JIRA](https://issues.apache.org/jira/browse/LUCENE-10396?focusedCommentId=17559856&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-17559856)) > The slowness is probably the lookupOrd? Can you avoid this? Just next() the termsenum to move on to the next ord. Not really...

Ignacio Vera (@iverase) ([migrated from JIRA](https://issues.apache.org/jira/browse/LUCENE-10396?focusedCommentId=17559859&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-17559859)) More exactly, the SortedDocValues iterator might have been advanced to a different ordinal.

Robert Muir (@rmuir) ([migrated from JIRA](https://issues.apache.org/jira/browse/LUCENE-10396?focusedCommentId=17559928&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-17559928)) I don't understand why you are starting from an ordinal at all? it seems a bit of an XY problem. Doesn't the task start...

Robert Muir (@rmuir) ([migrated from JIRA](https://issues.apache.org/jira/browse/LUCENE-10396?focusedCommentId=17559936&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-17559936)) In fact, i don't see any need to involve docvalues at all for this feature. You can go thru terms dict and just read...