[
https://issues.apache.org/jira/browse/SPARK-29638?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17240710#comment-17240710
]
Takeshi Yamamuro commented on SPARK-29638:
------------------------------------------
'NaN' is a special value for float/double types, so I think the correct query is
{code:java}
SELECT a, b, SUM(b) OVER(ORDER BY A ROWS BETWEEN 1 PRECEDING AND CURRENT ROW)
FROM (VALUES(1.0,1.0),(2.0,2.0),(3.0,(cast('nan' as
float8))),(4.0,3.0),(5.0,4.0)) t(a,b);{code}
But, it seems the answer is the same with each other;
{code:java}
postgres=# SELECT a, b, SUM(b) OVER(ORDER BY A ROWS BETWEEN 1 PRECEDING AND
CURRENT ROW)
postgres-# FROM (VALUES(1.0,1.0),(2.0,2.0),(3.0,(cast('nan' as
float8))),(4.0,3.0),(5.0,4.0)) t(a,b);
a | b | sum
-----+-----+-----
1.0 | 1 | 1
2.0 | 2 | 3
3.0 | NaN | NaN
4.0 | 3 | NaN
5.0 | 4 | 7
(5 rows)
scala> sql("""
| SELECT a, b, SUM(b) OVER(ORDER BY A ROWS BETWEEN 1 PRECEDING AND CURRENT
ROW)
| FROM (VALUES(1.0,1.0),(2.0,2.0),(3.0,(cast('nan' as
double))),(4.0,3.0),(5.0,4.0)) t(a,b);
| """).show()
+---+---+---------------------------------------------------------------------------------+
| a| b|sum(b) OVER (ORDER BY A ASC NULLS FIRST ROWS BETWEEN 1 PRECEDING AND
CURRENT ROW)|
+---+---+---------------------------------------------------------------------------------+
|1.0|1.0|
1.0|
|2.0|2.0|
3.0|
|3.0|NaN|
NaN|
|4.0|3.0|
NaN|
|5.0|4.0|
7.0|
+---+---+---------------------------------------------------------------------------------+
{code}
So, I will close this. If you reopen this issue, please update the description.
> Spark handles 'NaN' as 0 in sums
> --------------------------------
>
> Key: SPARK-29638
> URL: https://issues.apache.org/jira/browse/SPARK-29638
> Project: Spark
> Issue Type: Sub-task
> Components: SQL
> Affects Versions: 3.0.0
> Reporter: Dylan Guedes
> Priority: Major
>
> Currently, Spark handles 'NaN' as 0 in window functions, such that 3+'NaN'=3.
> PgSQL, on the other hand, handles the entire result as 'NaN', as in 3+'NaN' =
> 'NaN'
> I experienced this with the query below:
> {code:sql}
> SELECT a, b,
> SUM(b) OVER(ORDER BY A ROWS BETWEEN 1 PRECEDING AND CURRENT ROW)
> FROM (VALUES(1,1),(2,2),(3,(cast('nan' as int))),(4,3),(5,4)) t(a,b);
> {code}
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
