script.scala

import org.apache.spark.sql.types.DataTypes

def toInt(s: String): Option[Int] = {
        try {
                Some(s.toInt)
        }
        catch {
                case e: Exception => None
        }
}

val df = sqlContext.read.
        format("com.databricks.spark.csv").
        option("header", "true").
        option("inferSchema", "true").
        load("3G.csv").
        withColumnRenamed("管理アドレス種別", "pf").
        withColumnRenamed("IPアドレス", "ip").
        withColumnRenamed("ネットマスク", "mask").
        select(regexp_extract($"pf", "^.*/([^/]+)$", 1).alias("pf"),
               regexp_extract($"ip", "^(\\d+)\\.\\d+\\.\\d+\\.\\d+$", 1).alias("ip0"),
               regexp_extract($"ip", "^\\d+\\.(\\d+)\\.\\d+\\.\\d+$", 1).alias("ip1"),
               regexp_extract($"ip", "^\\d+\\.\\d+\\.(\\d+)\\.\\d+$", 1).alias("ip2"),
               regexp_extract($"ip", "^\\d+\\.\\d+\\.\\d+\\.(\\d+)$", 1).alias("ip3"),
               $"mask").
        withColumn("ip0", $"ip0".cast(DataTypes.IntegerType)).
        withColumn("ip1", $"ip1".cast(DataTypes.IntegerType)).
        withColumn("ip2", $"ip2".cast(DataTypes.IntegerType)).
        withColumn("ip3", $"ip3".cast(DataTypes.IntegerType)).
        withColumn("mask", $"mask".cast(DataTypes.IntegerType))