[
https://issues.apache.org/jira/browse/IMPALA-9722?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17104612#comment-17104612
]
ASF subversion and git services commented on IMPALA-9722:
---------------------------------------------------------
Commit f0e4b611734064e984bc3a208c76bcbf53a08d6c in impala's branch
refs/heads/master from Tamas Mate
[ https://gitbox.apache.org/repos/asf?p=impala.git;h=f0e4b61 ]
IMPALA-9722: Consolidate avg_size calculation in PerColumnStats
This change refactors the 'avg_size' calculation, 'total_width' will be
storing the sum of row widths till a Finalize call calculates the
average column size.
Testing:
- Added unit test to verify the aggregation result
Change-Id: Iae1efb1c568c67dff6f25887c2ea2b8b249eea4b
Reviewed-on: http://gerrit.cloudera.org:8080/15878
Tested-by: Impala Public Jenkins <[email protected]>
Reviewed-by: Csaba Ringhofer <[email protected]>
> Consolidate unused total_width and the way avg_width is computed in
> PerColumnStats
> ----------------------------------------------------------------------------------
>
> Key: IMPALA-9722
> URL: https://issues.apache.org/jira/browse/IMPALA-9722
> Project: IMPALA
> Issue Type: Bug
> Components: Backend
> Reporter: Tamas Mate
> Assignee: Tamas Mate
> Priority: Major
> Labels: statistics
> Fix For: Impala 4.0
>
>
> PerColumnStats.total_width is an unused field in
> [incr-stats-util.cc|https://github.com/apache/impala/blob/master/be/src/exec/incr-stats-util.cc#L134]
> while the comment says that it is the result of {{(avg_width * num_rows)}}.
> Later [similar
> multiplication|https://github.com/apache/impala/blob/master/be/src/exec/incr-stats-util.cc#L159]
> can be found in the Update method, however the result is stored in
> {{avg_width}}.
> {code:cpp}
> avg_width += (new_avg_width * num_new_rows);
> {code}
> We should consolidate these fields and the calculations.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
