Spark 递归解析Json

package com.aisainfo

import org.apache.spark.sql.{Column, Dataframe, SparkSession}
import org.apache.spark.sql.types._
import org.apache.spark.sql.functions._


object Test {
  def main(args: Array[String]): Unit = {
    val spark: SparkSession = SparkSession.builder().appName("test").master("local[2]").getOrCreate()
    val linesDF: Dataframe = spark.read.json("E:\work_space\work_space\work_asiainfo\java\scala\kafka_spark_hive_json\src\main\scala\com\aisainfo\json")


    val columns1: List[Column] = getAllColumns(linesDF).toList
    linesDF.createOrReplaceTempView("test")
    spark.sql(s"select ${columns1.mkString(", ")} from test").show()


    println(columns1.mkString(", "))
    println(linesDF.schema.fields.mkString(", "))
    //{"asd":{"device_id": 0, "device_type": "sensor-ipad", "ip": {"fgh":123, "456":456}},"qwe":"zxc"}
    //linesDF.select("asd.ip.fgh").show()

  }

  def getAllColumns(df: Dataframe) = {
    df.schema.fields.flatMap { data: StructField =>
      recursiveSolution(data)
    }
  }

  private def recursiveSolution(data: StructField, prefix: String = ""): List[Column] = {
    data match {
      case column_info if column_info.dataType.isInstanceOf[StructType] => {
        column_info.dataType.asInstanceOf[StructType].fields.flatMap { field =>
          if (prefix != "") recursiveSolution(field, s"$prefix.${column_info.name}")
          else recursiveSolution(field, s"${column_info.name}")
        }.toList
      }
      case column_info => {
        //json字段名称
        if (prefix != "") List(col(s"$prefix.${column_info.name}"))
        //json路径拼接字段名称
        //if (prefix != "") List(col(s"$prefix.${column_info.name} as ${prefix.replace(".", "_")}_${column_info.name}"))
        else List(col(s"${column_info.name}"))
      }
    }
  }
}