[
https://issues.apache.org/jira/browse/LUCENE-5828?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14063416#comment-14063416
]
Uwe Schindler commented on LUCENE-5828:
---------------------------------------
Thanks Robert! I expected that already, just had no time to benchmark (came
back from lunch a minute ago). The slice wrapper is not used in any core
directories, so maybe we should leave it as it is. I added a warning in the
previous issue, but for already slow IndexInputs like HDFS I don't care.
I will close this issue as won't fix. The bounds checks are the issue. We could
make it faster by let it delegating all read calls, but I don't think its worth.
> Improve BufferedIndexInput.SlicedIndexInput to directly delegate
> ----------------------------------------------------------------
>
> Key: LUCENE-5828
> URL: https://issues.apache.org/jira/browse/LUCENE-5828
> Project: Lucene - Core
> Issue Type: Bug
> Reporter: Uwe Schindler
> Assignee: Uwe Schindler
> Attachments: LUCENE-5828.patch, LUCENE-5828.patch
>
>
> Currently we have some IndexInputs which do not have a separate slice()
> implementation. Those are using BufferedIndexInput#wrap(), which extends
> BufferedInputStream. This is nonsense, because the underlying IndexInput is
> in most cases already Buffered, and if not, its not needed.
> This issue will change SlicedIndexInput to directly delegate, correcting
> offsets. The new class SlicedIndexInput will be top level and public.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
