HyukjinKwon commented on code in PR #37137:
URL: https://github.com/apache/spark/pull/37137#discussion_r917171663
##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/aggregate/linearRegression.scala:
##########
@@ -278,7 +278,8 @@ case class RegrSlope(left: Expression, right: Expression)
extends DeclarativeAgg
covarPop.mergeExpressions ++ varPop.mergeExpressions
override lazy val evaluateExpression: Expression = {
- If(covarPop.n === 0.0, Literal.create(null, DoubleType), covarPop.ck /
varPop.m2)
+ If(Or(covarPop.n === 0.0, varPop.m2 === 0.0),
Review Comment:
Will remove `covarPop`.
BTW, I think this is a bug fix. Such null behaviour is consistent everywhere
in other aggregate functions that use divisions, see also
```scala
scala> spark.conf.set("spark.sql.ansi.enabled", true)
scala> spark.range(1).selectExpr("sum(null)").show()
+---------+
|sum(NULL)|
+---------+
| null|
+---------+
```
It's more null propagation that failed before. ANSI documentation doesn't
mention about null handling from my cursory look so I think we can just keep ti
consistent within our codebase
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
