Github user ajithme commented on a diff in the pull request:
https://github.com/apache/spark/pull/22401#discussion_r217038517
--- Diff:
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/aggregate/Average.scala
---
@@ -36,7 +36,13 @@ abstract class AverageLike(child: Expression) extends
DeclarativeAggregate {
}
private lazy val sumDataType = child.dataType match {
- case _ @ DecimalType.Fixed(p, s) => DecimalType.bounded(p + 10, s)
+ /*
+ * In case of sum of decimal ( assuming another decimal of same
precision and scale)
+ * Refer : org.apache.spark.sql.catalyst.analysis.DecimalPrecision
+ * Precision : max(s1, s2) + max(p1 - s1, p2 - s2) + 1
+ * Scale : max(s1, s2)
+ */
+ case _ @ DecimalType.Fixed(p, s) => DecimalType.adjustPrecisionScale(s
+ (p - s) + 1, s)
--- End diff --
well i tested as per your suggestion with my PR :
```
sql("create table if not exists table1(salary decimal(2,1))")
(1 to 22).foreach(_ => sql("insert into table1 values(9.1)"))
sql("select avg(salary) from table1").show(false)
+-----------+
|avg(salary)|
+-----------+
|9.10000 |
+-----------+
```
which is expected result and i don't see a overflow as divide will readjust
precision. Can you test with my patch for a overflow specifically in case of
average.?
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
