joda DateTime格式导致火花RDD函数中的空指针错误

Question

异常消息如下

我的代码如下：

import org.apache.hadoop.conf.Configuration
import org.apache.spark.rdd.RDD
import org.apache.spark.SparkContext._
import org.apache.spark.{ SparkConf, SparkContext }
import org.joda.time.DateTime
import org.joda.time.format.{ DateTimeFormat, DateTimeFormatter }




object DateTimeNullReferenceReappear extends App {

  case class Record(uin: String = "", date: DateTime = null, value: Double = 0.0) 

  val cfg = new Configuration
  val sparkConf = new SparkConf()
  sparkConf.setAppName("bourne_exception_reappear")
  val sc = new SparkContext(sparkConf)

val data = TDWSparkContext.tdwTable(   // this function just read data from an data warehouse
  sc,
  tdwuser = FaceConf.TDW_USER,
  tdwpasswd = FaceConf.TDW_PASSWORD,
  dbName = "my_db",
  tblName = "my_table",
  parts = Array("p_20150323", "p_20150324", "p_20150325", "p_20150326", "p_20150327", "p_20150328", "p_20150329"))
  .map(row => {
    Record(uin = row(2),
      date = DateTimeFormat.forPattern("yyyyMMdd").parseDateTime(row(0)),
      value = row(4).toDouble)
  }).map(x => (x.uin, (x.date, x.value)))
  .groupByKey
  .map(x => {
    x._2.groupBy(_._1.toString("yyyyMMdd")).mapValues(_.map(_._2).sum)   // throw exception here
  })

//      val data = TDWSparkContext.tdwTable(  // It works, as I don't user datetime toString in the groupBy 
//      sc,
//      tdwuser = FaceConf.TDW_USER,
//      tdwpasswd = FaceConf.TDW_PASSWORD,
//      dbName = "hy",
//      tblName = "t_dw_cf_oss_tblogin",
//      parts = Array("p_20150323", "p_20150324", "p_20150325", "p_20150326", "p_20150327", "p_20150328", "p_20150329"))
//      .map(row => {
//        Record(uin = row(2),
//          date = DateTimeFormat.forPattern("yyyyMMdd").parseDateTime(row(0)),
//          value = row(4).toDouble)
//      }).map(x => (x.uin, (x.date.toString("yyyyMMdd"), x.value)))
//      .groupByKey
//      .map(x => {
//        x._2.groupBy(_._1).mapValues(_.map(_._2).sum)
//      })

  data.take(10).map(println)

}

Answer 1

您需要禁用Kryo，使用Kryo JodaTime序列化器，或者避免序列化DateTime对象，即传递long。