[
https://issues.apache.org/jira/browse/LUCENE-9887?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17310552#comment-17310552
]
ASF subversion and git services commented on LUCENE-9887:
---------------------------------------------------------
Commit 9a9676e03b64e3a4c054160e8b20959ab4d424e5 in lucene-solr's branch
refs/heads/branch_8x from liupanfeng
[ https://gitbox.apache.org/repos/asf?p=lucene-solr.git;h=9a9676e ]
LUCENE-9887: fix error param use in RadixSelector
> error param use in RadixSelector
> --------------------------------
>
> Key: LUCENE-9887
> URL: https://issues.apache.org/jira/browse/LUCENE-9887
> Project: Lucene - Core
> Issue Type: Improvement
> Components: core/other
> Affects Versions: 8.8
> Environment: None
> Reporter: liupanfeng
> Priority: Trivial
> Labels: patch
> Attachments: LUCENE-9887.patch, LUCENE-9887.patch
>
>
> There is a param use error in
> *org.apache.lucene.util.RadixSelector#select(int, int, int, int, int).*
> What is we expected in this method is:
> if the range becomes narrow or when the maximum level of recursion has been
> exceeded, then we get a fall-back selector(it's a IntroSelector).
> *So, we should use the recursion level(param f) compare to LEVEL_THRESHOLD.
> NOT the byte index of value(param d).*
> effect:
> This bug will not affect the correctness of the program. but affect
> performance in some bad case. In average, RadixSelector and IntroSelector are
> all in linear time. This bug will let we choose a fall-back selector too
> early, then the constant of O(n) will be bigger.
>
> other evidence:
> # In comments, said we use recursion level (f) not byte index of value(d).
> # if *d* is right, then the *param f* could be deleted because of it was
> not used by any method.
> verification:
> # It also can select right value if i change d -> f.
> # I did some benchmark works. but the result was unstable on random data.
>
> Thanks for your read. I'm new of lucene. So please reply me if I am wrong. Or
> fix it in future.
>
> I will do benchmark. But I can't promised the result is better. If you need
> the result. Ask for me.
>
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
