Spark 算法练习两则

练习一：求最大最小值

在文件中获取数据求浮点数和整数的最大值和最小值

1,2.33,4,1.55,2.56,
55,55,55,23.77,1.88987,0.3324,
22.567,5.5567,7.8895,33

import org.apache.spark.{SparkConf, SparkContext}

object demo01 {
  def main(args: Array[String]): Unit = {
    println("求最大最小值")
    val conf=new SparkConf().setMaster("local").setAppName("minmax")
    val sc = new SparkContext(conf)
    val data= sc.textFile("D:\\resource\\minmax.txt",1)
    var minn=Double.MaxValue
    var maxn=Double.MinValue
    val datap=data.flatMap{x=>x.split("\r\n").mkString("").split(",")}
    var i=1
    val countt=datap.count()
    for (x<-datap){
      val n=x.toDouble
      if(n>maxn)maxn=n
      if(n

练习二：求爷孙关系

janet winnie
winnie poul
sam dida
helen janet
helen jack
jack salon
salon rose
rose tom
jack nicole
sam flitter
janet sam
sam sisi

def ancesor(sc:SparkContext)= {
    val data = sc.textFile("D:\\resource\\ancesor.txt", 2)
    val cps = data.map { x => (x.split(" ")(0), x.split(" ")(1)) }
    val cpss=cps.collectAsMap() //激活RDD
    val to=cps.foreach{cp1=>
      val list=List("")
      cpss.foreach{cp2=>
        if(cp2._1.equals(cp1._2)) {
          println("祖辈："+cp1._1+"，孙辈："+cp2._2)
        }
      }
    }
  }

练习三：排序

对csv文件进行排序

name,score
helen,40
tom,50
mary,69
ben,60
sasa,70
marier,76
dida,78

object demo01 {
  def main(args: Array[String]): Unit = {
    val conf = new SparkConf().setAppName("sort1").setMaster("local")
    val sc = new SparkContext(conf)
    val sqc = new SQLContext(sc)
    //csv文件的第一行为列信息，设置header=true可以读取列信息
    val data2 = sqc.read.option("header", "true").csv("d:\\demo.csv").toDF()
    data2.registerTempTable("tb_score")
    sqc.sql("select * from tb_score order by score desc").show()
  }
}

对普通文件进行排序

hello 2
say 4
dida 5
discuss 6
subway 10

object demo01 {
  def main(args: Array[String]): Unit = {
    val conf = new SparkConf().setAppName("sort2").setMaster("local")
    val sc = new SparkContext(conf)
    val sqc = new SQLContext(sc)
    val data=sc.textFile("d:\\demo2.txt").map(x=>(x.split(" ")(0),x.split(" ")(1).toInt))
    //方法一，直接用内置函数，排序得出降序排序
    sqc.createDataFrame(data).toDF("name","num").orderBy("num").show()
    
    //方法二，转变成临时table进行sql排序，可指定升降序
    sqc.createDataFrame(data).toDF("name","num").registerTempTable("tb_s")
    sqc.sql("select * from tb_s order by num").show()
  }
}

对json文件进行排序

{“id”:1, “name”:“leo”, “age”:18}
{“id”:2, “name”:“jack”, “age”:19}
{“id”:3, “name”:“marry”, “age”:17}

object demo01 {
  def main(args: Array[String]): Unit = {
    val conf=new SparkConf().setMaster("local").setAppName("json")
    val sc=new SparkContext(conf)
    val sqc=new SQLContext(sc)
    val jdata=sqc.read.json("d:/demo3.json").createOrReplaceTempView("demo3")
    sqc.sql("select id,name,age from demo3 order by id").show()
  }
}

练习四：二次排序

hello 2 23
baby 2 44
hello 1 22
hello 3 55
nice 2 58
kitty 3 66
apple 1 44

使用SparkSQL

object demo01 {
  def main(args: Array[String]): Unit = {
    val conf=new SparkConf().setMaster("local").setAppName("second")
    val sc=new SparkContext(conf)
    val sqc=new SQLContext(sc)

    val data=sc.textFile("d:/demo4.txt")
      .map{x=>(x.split(" ")(0),x.split(" ")(1),x.split(" ")(2))}
    sqc.createDataFrame(data).toDF("name","class","score")
      .createOrReplaceTempView("demo4")
    sqc.sql("select * from demo4 as t order by class,score desc").show()

  }
}

使用Spark

//用于排序的类，注意第一行的写法
class record(val clss:Int,val score:Int) extends Ordered[record] with Serializable {
  def compare(other: record): Int = {
    val comp = clss.compareTo(other.clss)
    if (comp == 0) {
      other.score.compareTo(score)
    } else {
      comp
    }
  }
}


//用于处理的Spark
object demo01 {
  def main(args: Array[String]): Unit = {
    val conf=new SparkConf().setMaster("local").setAppName("second")
    val sc=new SparkContext(conf)

    val data=sc.textFile("d:/demo4.txt")
      .map{x=>
        val ele=x.split(" ")
        (new record(ele(1).toInt,ele(2).toInt),x)
      }
    val result=data.sortByKey(true).map(x=>x._2.replace(" ",","))

    result.foreach(println)
}
}

练习五：倒排索引

搜索引擎的索引法则

object demo01 {
  def main(args: Array[String]): Unit = {
    val conf = new SparkConf().setMaster("local").setAppName("second")
    val sc = new SparkContext(conf)

    sc.wholeTextFiles("D:\\resource\\daopai", 2).flatMap{ x =>
      val file = x._1.split("/").last.dropRight(4)
      x._2.split("\r\n").mkString(" ").split(" ")
        .map { x => (x, file) }
    }.map(x=>(x._1,x._2)).groupByKey()
      .map{x=>(x._1,x._2.toList.distinct.mkString(","))}
      .foreach(println)
  }
}

练习六：??