spark StructType的应用，用在处理mongoDB keyvalue

近期在处理mongoDB 数据的时候，遇到了非常奇怪的格式,账号密码的日志都追加在一条记录里面，要取一个密码的时长和所有密码的平均时长就非常繁琐。

用了各种迭代计算，非常困难，而且printschema出来结构也是不规范的。

和同事一起研究后用了StructType 效率非常高。

代码如下：

import java.sql.{DriverManager, ResultSet}

import mongoDb_foundation_data20180913.url
import org.apache.spark.{SparkConf, SparkContext}
import org.apache.spark.sql.types._
import org.apache.spark.sql.SparkSession
object devicests_20180916 {

 // spark-submit --driver-class-path /usr/local/jdk/lib/mysql-connector-java-5.1.46.jar  --class  "devicests_20180916"  /testdata/u3.jar
  val url = "jdbc:mysql://192.168.1.10:3306/saas?user=saas&password=saas2018"
  //val url = "jdbc:mysql://134.175.180.116:3306/saas?user=saas&password=saas2018"
  val conn = DriverManager.getConnection(url)

  def main(args: Array[String]): Unit = {

    val conn = DriverManager.getConnection(url)
    val conf = new SparkConf().setAppName("appName").setMaster("local")
    val sc = new SparkContext(conf)
    val spark = SparkSession.builder().appName("Spark SQL basic example").config("spark.debug.maxToStringFields", "200").getOrCreate()
    spark.sql("use saas")
    import spark.implicits._
    import org.apache.spark.sql.types._
    import org.apache.spark.sql.SparkSession
    val statement = conn.createStatement(ResultSet.TYPE_FORWARD_ONLY, ResultSet.CONCUR_UPDATABLE)

  //  val logData=spark.read.textFile("file:////mysqldata/aasdata/2018-08-17/devices_2018-08-17")

//    val log = spark.read.json(logData)
val prop = new java.util.Properties
//
//    log.createOrReplaceTempView("devicests_states")

//
//
//   // df.write.mode("append").jdbc(url, "saas.devicests_states", prop)
//import org.apache.calcite.adapter
    val schema = new StructType()
      .add("__v", StringType)
      .add("_id", new StructType()
        .add("$oid", StringType))
      .add("device_type", StringType)
      .add("hardware_info", new StructType()
        .add("cid", StringType)
        .add("mac", StringType)
        .add("sn", StringType)
        .add("versions", new StructType()
          .add("app_version", StringType)
          .add("hardware_version", StringType)
          .add("zigbee_version", StringType)))
      .add("model_id", StringType)
      .add("name", StringType)
      .add("nickname", StringType)
      .add("parent", StringType)
      .add("services", ArrayType(StringType))
      .add("states", new StructType()
        .add("onoff_line", StringType)
        .add("passwords",
          // spark 默认将 passwords 视为 struct，不便于使用 explode 和 map_values
          // 需要手动定义为 Map[String, Struct]
          MapType(StringType, new StructType()
            .add("description", StringType)
            .add("id", StringType)
            .add("is_default", StringType)
            .add("name", StringType)
            .add("permission", new StructType()
              .add("begin", StringType)
              .add("end", StringType)
              .add("status", StringType))
            .add("status", IntegerType)
            .add("time", StringType)))
        .add("power", StringType))
      .add("status", IntegerType)
      .add("time", StringType)
      .add("uuid", StringType)

    spark.read.schema(schema)
      .json(s"file:///mysqldata/aasdata/2018-09-12/devices_2018-09-12")
      .createOrReplaceTempView("devices")

    val res = spark.sql(
      """
        |SELECT uuid,
        |    COUNT(passwords.permission) AS count,
        |    AVG(passwords.permission.end - passwords.permission.begin) AS avg
        |FROM
        |    (
        |      SELECT uuid,explode(map_values(states.passwords)) AS passwords
        |      FROM devices
        |    )
        |WHERE
        |    passwords.permission.begin IS NOT NULL
        |    AND passwords.permission.end IS NOT NULL group by uuid""".stripMargin)//.collect.head
    res.write.mode("overwrite").jdbc(url, "saas.res_count_avg", prop)
////
////    val count = Long(res(0))
////    val avg = Double(res(1))

  }
}