[ https://issues.apache.org/jira/browse/SPARK-16071?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Xiangrui Meng updated SPARK-16071: ---------------------------------- Description: Several bugs have been found caused by integer overflows in Tungsten. This JIRA is for taking a final pass before 2.0 release to reduce potential bugs and issues. We should do at least the following: * Raise exception early instead of NegativeArraySize * Document clearly the largest array size we support in DataFrames. To reproduce one of the issues: {code} val n = 1e8.toInt // try 2e8, 3e8 sc.parallelize(0 until 1, 1).map(i => new Array[Int](n)).toDS.map(_.size).show() {code} Result: * n=1e8: correct but with slow * n=2e8: NegativeArraySize exception * n=3e8: NegativeArraySize exception but at a different location was: Several bugs have been found caused by integer overflows in Tungsten. This JIRA is for taking a final pass before 2.0 release to reduce potential bugs and issues. We should do at least the following: * Raise exception early instead of NegativeArraySize * Document clearly the largest array size we support in DataFrames. To reproduce one of the issues: {code} val n = 1e8.toInt // try 2e8, 3e8 sc.parallelize(0 until 1, 1).map(i => new Array[Int](n)).toDS.map(_.size).show() {code} Result: * n=1e8: correct but with slow * n=2e8: NegativeArraySize exception > Not sufficient array size checks to avoid integer overflows in Tungsten > ----------------------------------------------------------------------- > > Key: SPARK-16071 > URL: https://issues.apache.org/jira/browse/SPARK-16071 > Project: Spark > Issue Type: Bug > Components: SQL > Affects Versions: 2.0.0 > Reporter: Xiangrui Meng > Assignee: Yin Huai > Priority: Critical > > Several bugs have been found caused by integer overflows in Tungsten. This > JIRA is for taking a final pass before 2.0 release to reduce potential bugs > and issues. We should do at least the following: > * Raise exception early instead of NegativeArraySize > * Document clearly the largest array size we support in DataFrames. > To reproduce one of the issues: > {code} > val n = 1e8.toInt // try 2e8, 3e8 > sc.parallelize(0 until 1, 1).map(i => new > Array[Int](n)).toDS.map(_.size).show() > {code} > Result: > * n=1e8: correct but with slow > * n=2e8: NegativeArraySize exception > * n=3e8: NegativeArraySize exception but at a different location -- This message was sent by Atlassian JIRA (v6.3.4#6332) --------------------------------------------------------------------- To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org For additional commands, e-mail: issues-h...@spark.apache.org