[
https://issues.apache.org/jira/browse/LUCENE-3246?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13058107#comment-13058107
]
Michael McCandless commented on LUCENE-3246:
--------------------------------------------
bq. As we have now both variants to read/write BitVectors, would it be not a
good idea to automatically use the old encoding for liveDocs, if more than 50%
of all bits are unset?
That seems like a good idea? Ie, handle both sparse set and sparse unset
compactly? Though it should be unusual that you have so many deletes against a
segment (esp. because TMP now targets such segs more aggressively).
We should do this under a new issue (the old code also didn't handle the "many
deletions" case sparsely either, just the "few deletions" case).
> Invert IR.getDelDocs -> IR.getLiveDocs
> --------------------------------------
>
> Key: LUCENE-3246
> URL: https://issues.apache.org/jira/browse/LUCENE-3246
> Project: Lucene - Java
> Issue Type: Improvement
> Components: core/index
> Reporter: Michael McCandless
> Assignee: Michael McCandless
> Fix For: 4.0
>
> Attachments: LUCENE-3246-IndexSplitters.patch, LUCENE-3246.patch,
> LUCENE-3246.patch
>
>
> Spinoff from LUCENE-1536, where we need to fix the low level filtering
> we do for deleted docs to "match" Filters (ie, a set bit means the doc
> is accepted) so that filters can be pushed all the way down to the
> enums when possible/appropriate.
> This change also inverts the meaning first arg to
> TermsEnum.docs/AndPositions (renames from skipDocs to liveDocs).
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
