This was a change that was made to match a wrong answer coming from older versions of Hive. Unfortunately I think its too late to fix this in the 1.4 branch (as I'd like to avoid changing answers at all in point releases), but in Spark 1.5 we revert to the correct behavior.
https://issues.apache.org/jira/browse/SPARK-8828 On Thu, Jul 2, 2015 at 11:58 PM, StanZhai <m...@zhaishidan.cn> wrote: > Hi all, > > I have a table named test like this: > > | a | b | > | 1 | null | > | 2 | null | > > After upgraded the cluster from spark 1.3.1 to 1.4.0, I found the Sum > function in spark 1.4 and 1.3 are different. > > The SQL is: select sum(b) from test > > In Spark 1.4.0 the result is 0.0, in spark 1.3.1 the result is null. I > think > the result should be null, why the result is 0.0 in 1.4.0 but not null? Is > this a bug? > > Any hint is appreciated. > > > > -- > View this message in context: > http://apache-spark-developers-list.1001551.n3.nabble.com/SparkSQL-1-4-0-The-result-of-SUM-xxx-in-SparkSQL-is-0-0-but-not-null-when-the-column-xxx-is-all-null-tp13008.html > Sent from the Apache Spark Developers List mailing list archive at > Nabble.com. > > --------------------------------------------------------------------- > To unsubscribe, e-mail: dev-unsubscr...@spark.apache.org > For additional commands, e-mail: dev-h...@spark.apache.org > >
