[ 
https://issues.apache.org/jira/browse/LUCENE-1476?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12666681#action_12666681
 ] 

Michael McCandless commented on LUCENE-1476:
--------------------------------------------


{quote}
We probably need to support both as implementing top level deleted
docs filtering may have unknown side effects. The user may decide
based on their queries and other variables such as the number of
deleted docs.
{quote}

I agree... and then, if the performance difference is large enough, it
seems like we'll need some simple "search policy" for the interesting
(Boolean) query scorers to pick the best way to execute a query.

This could include which order to visit the segments in (we broached
this in LUCENE-1483, since depending on the query different orders may
perform better).  And when (high vs low) & how (iterator vs random
access) to apply a filter would also be decided by the search policy.

bq. Deprecating isDeleted might be good.

I wonder how this method is used [externally] by applications,
today... I'll go ask on java-user.  And, whether all such uses could
migrate to an iterator API instead without much cost.

bq. Would we need the read only readers?

Good question... I'm guessing there would still be a performance
benefit if the underlying data structures for deletions &
column-stride fields know they cannot change?


> BitVector implement DocIdSet, IndexReader returns DocIdSet deleted docs
> -----------------------------------------------------------------------
>
>                 Key: LUCENE-1476
>                 URL: https://issues.apache.org/jira/browse/LUCENE-1476
>             Project: Lucene - Java
>          Issue Type: Improvement
>          Components: Index
>    Affects Versions: 2.4
>            Reporter: Jason Rutherglen
>            Priority: Trivial
>         Attachments: LUCENE-1476.patch, LUCENE-1476.patch, LUCENE-1476.patch, 
> quasi_iterator_deletions.diff, quasi_iterator_deletions_r2.diff
>
>   Original Estimate: 12h
>  Remaining Estimate: 12h
>
> Update BitVector to implement DocIdSet.  Expose deleted docs DocIdSet from 
> IndexReader.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


---------------------------------------------------------------------
To unsubscribe, e-mail: java-dev-unsubscr...@lucene.apache.org
For additional commands, e-mail: java-dev-h...@lucene.apache.org

Reply via email to