hive-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Gopal V (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HIVE-13089) Rounding in Stats for equality expressions
Date Sat, 20 Feb 2016 03:03:18 GMT

    [ https://issues.apache.org/jira/browse/HIVE-13089?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15155335#comment-15155335
] 

Gopal V commented on HIVE-13089:
--------------------------------

{code}
source/itests/qtest/../../ql/src/test/results/clientpositive/annotate_stats_deep_filters.q.out
124c124
<               Statistics: Num rows: 300 Data size: 2392 Basic stats: COMPLETE Column
stats: COMPLETE
---
>               Statistics: Num rows: 280 Data size: 2232 Basic stats: COMPLETE Column
stats: COMPLETE
126c126
<                 Statistics: Num rows: 300 Data size: 2392 Basic stats: COMPLETE Column
stats: COMPLETE
---
>                 Statistics: Num rows: 280 Data size: 2232 Basic stats: COMPLETE Column
stats: COMPLETE
{code}

in case we lose the logs by monday.

> Rounding in Stats for equality expressions
> ------------------------------------------
>
>                 Key: HIVE-13089
>                 URL: https://issues.apache.org/jira/browse/HIVE-13089
>             Project: Hive
>          Issue Type: Bug
>          Components: Statistics
>    Affects Versions: 2.1.0
>            Reporter: Jesus Camacho Rodriguez
>            Assignee: Jesus Camacho Rodriguez
>         Attachments: HIVE-13089.patch
>
>
> Currently we divide numRows(long) by countDistinct(long), thus ignoring the decimals.
We should do proper rounding.
> This is specially useful for equality expressions over columns whose values are unique.
As NDV estimates allow for a certain error, if countDistinct > numRows, we end up with
0 rows in the estimate for the expression.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Mime
View raw message