Found the solution: Creating the TypeInformation in Scala and then passing it to Java where it is used explicitly in the returns method resolved the issue.
Op Maandag, 01/06/2015 om 16:59 schreef Pieter-Jan Van Aeken: >From the Java API, I get an instance of_ DataSet_ which I wrap with the method available in _org.apache.flink.api.scala.package. _This works fine when I print out the result, but when I try to write as CSV, I get an exception: java.lang.IllegalArgumentException: requirement failed: CSV output can only be used with Tuple DataSets. It seems that an instance of _DataSet[scala.Tuple2[Long, Long]]_ is somehow considered different from an instance of_ DataSet[(Long, Long)]_ Are there any suggestions on how to fix this problem? Note that I cannot hardcode the conversion as the tuple size may vary or it might be an entirely different custom class. Regards, Pieter-Jan
