You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@lucene.apache.org by "Michael McCandless (JIRA)" <ji...@apache.org> on 2010/07/10 22:48:49 UTC

[jira] Updated: (LUCENE-2531) FieldComparator.TermOrdValComparator compares by value unnecessarily

     [ https://issues.apache.org/jira/browse/LUCENE-2531?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Michael McCandless updated LUCENE-2531:
---------------------------------------

    Attachment: LUCENE-2504.patch
                LUCENE-2504-3x.patch

Attached patches, for 3.1 and 4.0.

This also cleans up the String/TermOrdValComparator, absorbing the convert method directly into setBottom, and doing away with some dead code.  With this fix we no longer compare by value at all if index is a single segment (I think!).

> FieldComparator.TermOrdValComparator compares by value unnecessarily
> --------------------------------------------------------------------
>
>                 Key: LUCENE-2531
>                 URL: https://issues.apache.org/jira/browse/LUCENE-2531
>             Project: Lucene - Java
>          Issue Type: Improvement
>          Components: Search
>            Reporter: Michael McCandless
>            Priority: Minor
>             Fix For: 3.1, 4.0
>
>         Attachments: LUCENE-2504-3x.patch, LUCENE-2504.patch
>
>
> Digging on LUCENE-2504, I noticed that TermOrdValComparator's compareBottom method falls back on compare-by-value when it needn't.
> Specifically, if we know the current bottom ord "matches" the current segment, we can skip the value comparison when the ords are the same (ie, return 0) because the ords are exactly comparable.
> This is hurting string sort perf especially for optimized indices (and also unoptimized indices), and especially for highly redundant (not many unique values) fields.  This affects all releases >= 2.9.x, but trunk is likely more severely affected since looking up a value is more costly.

-- 
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: dev-unsubscribe@lucene.apache.org
For additional commands, e-mail: dev-help@lucene.apache.org