Github user MaxGekk commented on the issue: https://github.com/apache/spark/pull/21736 I am testing the changes and have found this so far: ``` $ ./bin/spark-shell --master 'local-cluster[1, 1, 1024]' ``` By default the "legacy" behavior is enabled: ``` scala> spark.conf.get("spark.sql.legacy.sizeOfNull") res0: String = true scala> spark.sql("select size(null)").show() +----------+ |size(NULL)| +----------+ | -1| +----------+ ``` Let's store the sql query is a value: ``` scala> val df = spark.sql("select size(null)") df: org.apache.spark.sql.DataFrame = [size(NULL): int] ``` and switch the behavior: ``` scala> spark.conf.set("spark.sql.legacy.sizeOfNull", "false") scala> spark.conf.get("spark.sql.legacy.sizeOfNull") res3: String = false ``` I would expect `null` but got `-1`: ``` scala> df.show() +----------+ |size(NULL)| +----------+ | -1| +----------+ ```
--- --------------------------------------------------------------------- To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org For additional commands, e-mail: reviews-h...@spark.apache.org