Spark分析sogou日志， RDD

import com.hankcs.hanlp.HanLP
import org.apache.spark.rdd.RDD
import org.apache.spark.{SparkConf, SparkContext}
import scala.collection.immutable.StringOps
import scala.collection.mutable


object SoGou_WC {
  def main(args: Array[String]): Unit = {
    //准备环境
//    val conf: SparkConf = new SparkConf().setAppName("SouGou_WC").setMaster("local[*]")
    val conf: SparkConf = new SparkConf().setAppName("SouGou_WC")   //SouGou_WC集群模式
    val sc = new SparkContext(conf)
    sc.setLogLevel("WARN")
    //读取数据
//    val data: RDD[String] = sc.textFile("data/input/SogouQ.sample")
    val data: RDD[String] = sc.textFile(args(0))
    //封装数据
    val sogouDataRDD: RDD[SogouData] = data.map(line => {
      val arr: Array[String] = line.split("\\s+")

      SogouData(
        arr(0),
        arr(1),
        arr(2),
        arr(3).toInt,
        arr(4).toInt,
        arr(5),
      )
    })
    //处理数据
    val res1: Array[(String, Int)] = sogouDataRDD.flatMap(record => {
      val str: StringOps = record.data3.replaceAll("\\[|\\]", "")
      import scala.collection.JavaConverters._
      HanLP.segment(str).asScala.map(_.word)
    })
      .filter(word => {
        if(word.equals("+") || word.equals(".")){
          false
        }
        else
          true
      })
      .map((_, 1))
      .reduceByKey(_ + _)
      .sortBy(_._2, false)
      .take(10)

    val res2: Array[((String, String), Int)] = sogouDataRDD.flatMap(record => {
      val str: StringOps = record.data3.replaceAll("\\[|\\]", "")
      import scala.collection.JavaConverters._
      val words: mutable.Buffer[String] = HanLP.segment(str).asScala.map(_.word)
      val userid: String = record.data2
      words.map(word => (userid, word))
    })
      .filter(tuple => {
        if (tuple._2.equals("+") || tuple._2.equals(".")) {
          false
        }
        else
          true
      })
      .map((_, 1))
      .reduceByKey(_ + _)
      .sortBy(_._2, false)
      .take(10)

    val res3: Array[(String, Int)] = sogouDataRDD.map(record => {
      val time: String = record.data1
      val time1: String = time.substring(0, 5)
      (time1, 1)
    })
      .reduceByKey(_ + _)
      .sortBy(_._2, false)
      .take(10)

    //输出结果数据
    res1.foreach(println)
    res2.foreach(println)
    res3.foreach(println)

    sc.makeRDD(res1).coalesce(1,true).saveAsTextFile(args(1))
    sc.makeRDD(res2).coalesce(1,true).saveAsTextFile(args(2))
    sc.makeRDD(res3).coalesce(1,true).saveAsTextFile(args(3))

//    sc.makeRDD(res1).coalesce(1,true).saveAsTextFile("data/output1")
//    sc.makeRDD(res2).coalesce(1,true).saveAsTextFile("data/output2")
//    sc.makeRDD(res3).coalesce(1,true).saveAsTextFile("data/output3")

//    System.setProperty("HADOOP_USER_NAME","node")   //权限

    //关闭环境
    sc.stop()
  }
  case class SogouData(
                        data1 : String,
                        data2 : String,
                        data3 : String,
                        data4 : Int,
                        data5 : Int,
                        data6 : String
                      )

}

标签：arr,word,String,val,sogou,RDD,sc,Spark,true
来源： https://www.cnblogs.com/tan2022/p/14962765.html