val values = data.map((row:Row)=>(row(indexOfKey),row)) .reduceByKey((x:Row,y:Row)=>if(x.getLong(indexOfTimestamp)>y.getLong(indexOfTimestamp)) x else y) .values
会报错,显示:com.esotericsoftware.kryo.KryoException: Class cannot be created (missing no-arg constructor): org.apache.spark.sql.catalyst.expressions.GenericRowWithSchema
spark1.5的org.apache.spark.sql.catalyst.expressions.GenericRowWithSchema是拥有无参构造器的,
但是现在的spark1.3是通过通过cdh装的,不知道升级为spark1.5好不好弄,中间会不会对现在已经写好的程序有什么影响