[
https://issues.apache.org/jira/browse/HIVE-8644?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14188447#comment-14188447
]
Owen O'Malley commented on HIVE-8644:
-------------------------------------
In particular, this impacts the picking of the encoding. When it is calculating
the percentiles, it looks like:
{code}
int perLen = (int) (length * (1.0 - p));
{code}
which for p = 0.9 and length = 10 generates 0.
> ORC rle v2 writer should round when converting from floats to integers
> ----------------------------------------------------------------------
>
> Key: HIVE-8644
> URL: https://issues.apache.org/jira/browse/HIVE-8644
> Project: Hive
> Issue Type: Improvement
> Components: File Formats
> Reporter: Owen O'Malley
> Assignee: Owen O'Malley
> Priority: Minor
>
> The ORC rle v2 writer would do better to round the floating point numbers
> when converting to integers.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)