[
https://issues.apache.org/jira/browse/HIVE-26683?focusedWorklogId=830758&page=com.atlassian.jira.plugin.system.issuetabpanels:worklog-tabpanel#worklog-830758
]
ASF GitHub Bot logged work on HIVE-26683:
-----------------------------------------
Author: ASF GitHub Bot
Created on: 02/Dec/22 23:22
Start Date: 02/Dec/22 23:22
Worklog Time Spent: 10m
Work Description: scarlin-cloudera commented on code in PR #3800:
URL: https://github.com/apache/hive/pull/3800#discussion_r1038620856
##########
ql/src/java/org/apache/hadoop/hive/ql/udf/generic/GenericUDAFSum.java:
##########
@@ -337,29 +345,35 @@ public GenericUDAFEvaluator
getWindowingEvaluator(WindowFrameDef wFrameDef) {
return null;
}
- return new
GenericUDAFStreamingEvaluator.SumAvgEnhancer<HiveDecimalWritable, HiveDecimal>(
+ return new
GenericUDAFStreamingEvaluator.SumAvgEnhancer<HiveDecimalWritable, Object[]>(
Review Comment:
This matches very similar logic found in GenericUDAFAvg which also uses an
Object[]
The purpose is that we need to keep track of the sum while also keeping
track whether the current window member is a null. If all members in the
window frame are null, we should treat the sum as null instead of 0, which is
how the "preceding" rows behave.
Issue Time Tracking
-------------------
Worklog Id: (was: 830758)
Time Spent: 1h 40m (was: 1.5h)
> Sum over window produces 0 when row contains null
> -------------------------------------------------
>
> Key: HIVE-26683
> URL: https://issues.apache.org/jira/browse/HIVE-26683
> Project: Hive
> Issue Type: Bug
> Components: HiveServer2
> Reporter: Steve Carlin
> Assignee: Steve Carlin
> Priority: Major
> Labels: pull-request-available
> Time Spent: 1h 40m
> Remaining Estimate: 0h
>
> Ran the following sql:
>
> {code:java}
> create table sum_window_test_small (id int, tinyint_col tinyint);
> insert into sum_window_test_small values (5,5), (10, NULL), (11,1);
> select id,
> tinyint_col,
> sum(tinyint_col) over (order by id nulls last rows between 1 following and 1
> following)
> from sum_window_test_small order by id;
> select id,
> tinyint_col,
> sum(tinyint_col) over (order by id nulls last rows between current row and 1
> following)
> from sum_window_test_small order by id;
> {code}
> The result is
> {code:java}
> +-----+--------------+---------------+
> | id | tinyint_col | sum_window_0 |
> +-----+--------------+---------------+
> | 5 | 5 | 0 |
> | 10 | NULL | 1 |
> | 11 | 1 | NULL |
> +-----+--------------+---------------+{code}
> The first row should have the sum as NULL
>
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
