Details
-
Sub-task
-
Status: Closed
-
Major
-
Resolution: Won't Fix
-
None
-
None
-
None
-
Spark 1.5 release
Description
Spark SQL's Decimal class should use Java's BigDecimal instead of Scala's, since this removes a layer of object allocation and works around an issue where Scala's BigDecimal.hashCode() can lead to OOMs (see https://issues.scala-lang.org/browse/SI-6173).