[
https://issues.apache.org/jira/browse/LUCENE-6293?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14336323#comment-14336323
]
Adrien Grand commented on LUCENE-6293:
--------------------------------------
OK I found the reason: Our TimSorter picks minRunLength based on the
description at http://svn.python.org/projects/python/trunk/Objects/listsort.txt
(look for
"Computing minrun") which gives a value between 32 and 65 while Java picks a
value which is always less than or equal to 32. If I hack the test code to use
the same formula as our TimSorter, I can reproduce the bug.
> TimSort bug
> -----------
>
> Key: LUCENE-6293
> URL: https://issues.apache.org/jira/browse/LUCENE-6293
> Project: Lucene - Core
> Issue Type: Bug
> Reporter: Adrien Grand
> Assignee: Adrien Grand
> Priority: Minor
>
> Robert pointed me to
> http://envisage-project.eu/proving-android-java-and-python-sorting-algorithm-is-broken-and-how-to-fix-it/
> yesterday which explains how most implementations of TimSort are broken. We
> should check our TimSorter.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
