[ https://issues.apache.org/jira/browse/SPARK-26991?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16782128#comment-16782128 ]
Jungtaek Lim commented on SPARK-26991: -------------------------------------- The outcome would be * "invalid" if there's reasonable reason for Java side to use `returnNullable = true` * new patch if there's no reason for Java side to use `returnNullable = true` separately and can follow Scala side Btw, is it discouraged to file an issue for TODO task? I agree this wouldn't be an issue after investigation: this is just not to miss for TODO - given this issue is opened to public, other than me can investigate this. > Investigate difference of `returnNullable` between > ScalaReflection.deserializerFor and JavaTypeInference.deserializerFor > ------------------------------------------------------------------------------------------------------------------------ > > Key: SPARK-26991 > URL: https://issues.apache.org/jira/browse/SPARK-26991 > Project: Spark > Issue Type: Task > Components: SQL > Affects Versions: 3.0.0 > Reporter: Jungtaek Lim > Priority: Major > > This issue tracks the effort on investigation on difference between > ScalaReflection.deserializerFor and JavaTypeInference.deserializerFor, > especially the reason why Java side uses `returnNullable = true` whereas > `returnNullable = false`. > The origin discussion is linked here: > https://github.com/apache/spark/pull/23854#discussion_r260117702 -- This message was sent by Atlassian JIRA (v7.6.3#76005) --------------------------------------------------------------------- To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org For additional commands, e-mail: issues-h...@spark.apache.org