使用LocalDateTime的Spark序列化錯誤

Question

代碼-

val rdd=sc.textFile("/tmp/abc.csv")
rdd.first.split(",").zipWithIndex
val rows=rdd.filter(x => !x.contains("ID") && !x.contains("Case Number"))
val split1=rows.map(x => x.split(","))
split1.take(3)
import java.time._
import java.time.format._
val format=DateTimeFormatter.ofPattern("MM/dd/yyyy h:m:s a")
val dates=split1.map( x => LocalDateTime.parse( x(2) , format))

錯誤：

org.apache.spark.SparkException：無法在org.apache.spark.util.ClosureCleaner $ .ensureSerializable（ClosureCleaner.scala：304）上序列化的任務

Answer 1

解決這個問題的比較丑陋的方法是在匿名函數中推送格式初始化：

split1.map(x => 
  LocalDateTime.parse(x(2), DateTimeFormatter.ofPattern("MM/dd/yyyy h:m:s a")))