Spark-MapReduce编程-自连接(Scala)

关于SQL和Hadoop的实现参考这里 MapReduce编程-自连接

这里用相同的原理，使用spark实现。本人也是刚学Scala，可能写的不好，还请指正。

object SelfUion {
  def main(args: Array[String]) {
    val conf = new SparkConf().setAppName("SelfUnion")
    val sc = new SparkContext(conf)
    val cpFile = sc.textFile("cp.txt")
    //val strs =  Array[String]("a", "b")

    // 1) 生成两张表
    val res = cpFile.flatMap(line => {
      val strs: Array[String] = line.split(" ");
      if (strs.length == 2) Array(strs(0) + " parent_" + strs(1), strs(1) + " child_" + strs(0)) else Array[String]()
    })
      // 2) 转化为key-value形式
      .map(line => {
      val kv = line.split(" ")
      (kv(0), kv(1))
    })
      // 3) 列的相等匹配
      .groupByKey()
      // 4) 解析value，得到结果
      .flatMapValues(values => {
      val childList = new ArrayBuffer[String]();
      val parentList = new ArrayBuffer[String]();
      values.foreach(
        name => {
          if (name.startsWith("child_")) childList += name
          else if (name.startsWith("parent_")) parentList += name;
        })
      for (c <- childList; p <- parentList) yield
        c.substring(6) + " " + p.substring(7)
      }).values
      .saveAsTextFile("selfunion_out")

  }
}

cp.txt的内容为：

Tom Lucy
Tom Jack
Jone Lucy
Jone Jack
Lucy Mary
Lucy Ben
Jack Alice
Jack Jesse
Terry Tom

输出为：

Terry Lucy
Terry Jack
Tom Mary
Tom Ben
Jone Mary
Jone Ben
Tom Alice
Tom Jesse
Jone Alice
Jone Jesse

纯Scala实现

另外，附带上使用Scala的函数式编程实现。注意，下面是用的Scala原有的map，flatMap() 等方法，而不是RDD。

使用一个ArrayBuffer[String]带入输入文件，每一行为一个item.

  def main(args: Array[String]) {
    val list = new ArrayBuffer[String]();
    list+="Tom Lucy"
    list += "Tom Jack"
    list += "Jone Lucy"
    list += "Jone Jack"
    list += "Lucy Mary"
    list += "Lucy Ben"
    list += "Jack Alice"
    list += "Jack Jesse"
    list += "Terry Tom"

    val list2 = list.flatMap(line => {
      val strs: Array[String] = line.split(" ");
      if (strs.length == 2) Array(strs(0) + " parent_" + strs(1), strs(1) + " child_" + strs(0)) else Array[String]()
    } )
    //println("list2 : " + list2)
    val mapout = list2.map(line => line.split(" "))
    //println("mapout : " + mapout)
    //for(item <- mapout) print(item(0) + " " + item(1) + " ; " )
    // println()
    val groupbyout = mapout.groupBy(_(0)) //same as groupByKey, The first element of the Array
    val res = groupbyout.values.flatMap(values => {
      val childList = new ArrayBuffer[String]();
      val parentList = new ArrayBuffer[String]();
      values.foreach(
        names => {
          if (names(1).startsWith("child_")) childList += names(1)
          else if (names(1).startsWith("parent_")) parentList += names(1);
        })
      for (c <- childList; p <- parentList) yield
         Array[String](c,p)
    })

    for(arr <- res){
      println(arr(0).substring(6) + " " + arr(1).substring(7))
    }

  }

可以看出，MR和函数式编程有很多相似之处