[ https://issues.apache.org/jira/browse/SPARK-11194?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Yin Huai updated SPARK-11194: ----------------------------- Description: Right now, we stack a new URLClassLoader when a user add a jar through SQL's add jar command. This approach can introduce issues caused by the ordering of added jars when a class of a jar depends on another class of another jar. For example, {code} ClassLoader1 for Jar1.jar (A.class) | |----- ClassLoader2 for Jar2.jar (B.class depending on A.class) {code} In this case, when we lookup class B, we will not be able to find class A because Jar2 is the parent of Jar1. > Use a single URLClassLoader for jars added through SQL's "ADD JAR" command > -------------------------------------------------------------------------- > > Key: SPARK-11194 > URL: https://issues.apache.org/jira/browse/SPARK-11194 > Project: Spark > Issue Type: Improvement > Components: SQL > Reporter: Yin Huai > Assignee: Yin Huai > > Right now, we stack a new URLClassLoader when a user add a jar through SQL's > add jar command. This approach can introduce issues caused by the ordering of > added jars when a class of a jar depends on another class of another jar. > For example, > {code} > ClassLoader1 for Jar1.jar (A.class) > | > |----- ClassLoader2 for Jar2.jar (B.class depending on A.class) > {code} > In this case, when we lookup class B, we will not be able to find class A > because Jar2 is the parent of Jar1. -- 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