1. 学科最受欢迎老师解法补充
day21中该案例的解法四还有一个问题，就是当各个老师受欢迎度是一样的时候，其排序规则就处理不了，以下是对其优化的解法
实现方式五
FavoriteTeacher5
package com._51doit.spark04
import org.apache.spark.{Partitioner, SparkConf, SparkContext}
import org.apache.spark.rdd.RDD
import scala.collection.mutable
object FavoriteTeacher5 {
def main(args: Array[String]): Unit = {
val isLocal = args(0).toBoolean
//创建SparkConf，然后创建SparkContext
val conf = new SparkConf().setAppName(this.getClass.getSimpleName)
if (isLocal) {
conf.setMaster("local[*]")
}
val sc = new SparkContext(conf)
//指定以后从哪里读取数据创建RDD
val lines: RDD[String] = sc.textFile(args(1))
//对数据进行切分
val subjectTeacherAndOne = lines.map(line => {
val fields = line.split("/")
val subject = fields(2).split("[.]")(0)
val teacher = fields(3)
((subject, teacher), 1)
})
//计算所有的学科，并收集到Driver端
val subjects: Array[String] = subjectTeacherAndOne.map(_._1._1).distinct().collect()
//paritioner是在Driver端被new出来的，但是他的方法是在Executor中被调用的
val partitioner = new SubjectPartitioner3(subjects)
//根据指定的key和分区器进行聚合(减少一次shuffle)
val reduced: RDD[((String, String), Int)] = subjectTeacherAndOne.reduceByKey(partitioner, _+_)
val topN = args(2).toInt
val result = reduced.mapPartitions(it => {
//定义一个key排序的集合TreeSet
val sorter = new mutable.TreeSet[OrderingBean]()
//遍历出迭代器中的数据
it.foreach(t => {
sorter += new OrderingBean(t._1._1, t._1._2, t._2)
if (sorter.size > topN) {
val last = sorter.last
//移除最后一个
sorter -= last
}
})
sorter.iterator
})
val r = result.collect()
println(r.toBuffer)
sc.stop()
}
}
class SubjectPartitioner3(val subjects: Array[String]) extends Partitioner {
//初始化分器的分区规则
val rules = new mutable.HashMap[String, Int]()
var index = 0
for(sub <- subjects) {
rules(sub) = index
index += 1
}
override def numPartitions: Int = subjects.length
//该方法会在Executor中的Task中被调用
override def getPartition(key: Any): Int = {
val tuple = key.asInstanceOf[(String, String)]
val subject = tuple._1
//到实现初始化的规则中查找这个学科对应的分区编号
rules(subject)
}
}
OrderingBean(重新定义的排序规则)
package com._51doit.spark04
import scala.collection.mutable.ArrayBuffer
class OrderingBean(val subject: String, val name: String, val count: Int) extends Ordered[OrderingBean] with Serializable {
val equiv = new ArrayBuffer[(String, String, Int)]()
equiv += ((subject, name, count))
override def compare(that: OrderingBean): Int = {
if (this.count == that.count) {
equiv += ((that.subject, that.name, that.count))
0
} else {
-(this.count - that.count)
}
}
override def toString: String =
if (equiv.size > 1) {
equiv.toString()
} else
s"($subject, $name, $count)"
}
实现方式六（使用repartitionAndSortWithinPartitions）
repartitionAndSortWithinPartitions按照指定的分区器进行排序并且在每个分区内进行排序
FavoriteTeacher6
object FavoriteTeacher06 {
def main(args: Array[String]): Unit = {
val isLocal = args(0).toBoolean
//创建SparkConf，然后创建SparkContext
val conf = new SparkConf().setAppName(this.getClass.getSimpleName)
if (isLocal) {
conf.setMaster("local[1]")
}
val sc = new SparkContext(conf)
//指定以后从哪里读取数据创建RDD
val lines: RDD[String] = sc.textFile(args(1))
//对数据进行切分
val subjectTeacherAndOne = lines.map(line => {
val fields = line.split("/")
val subject = fields(2).split("[.]")(0)
val teacher = fields(3)
((subject, teacher), 1)
})
//聚合
val reduced = subjectTeacherAndOne.reduceByKey(_ + _)
//计算所有的学科，并收集到Driver端
val subjects: Array[String] = reduced.map(_._1._1).distinct().collect()
//paritioner是在Driver端被new出来的，但是他的方法是在Executor中被调用的
val partitioner = new SubjectPartitionerV2(subjects)
//对原来的数据进行整理
val keyByRDD: RDD[((String, String, Int), Null)] = reduced
.map(t => ((t._1._1, t._1._2, t._2), null))
//隐式转换
implicit val orderRules = new Ordering[(String, String, Int)] {
override def compare(x: (String, String, Int), y: (String, String, Int)): Int = {
-(x._3 - y._3)
}
}
val topN = args(2).toInt
//repartitionAndSortWithinPartitions按照指定的分区器进行排序并且在每个分区内进行排序
val result: RDD[((String, String, Int), Null)] = keyByRDD
.repartitionAndSortWithinPartitions(partitioner)
result.foreachPartition(it => {
var index = 1
while (it.hasNext && index <= topN) {
val tuple = it.next()
println(tuple)
index += 1
}
})
sc.stop()
}
}
SubjectPartitionerV2
class SubjectPartitionerV2(val subjects: Array[String]) extends Partitioner {
//初始化分器的分区规则
val rules = new mutable.HashMap[String, Int]()
var index = 0
for(sub <- subjects) {
rules(sub) = index
index += 1
}
override def numPartitions: Int = subjects.length
//该方法会在Executor中的Task中被调用
override def getPartition(key: Any): Int = {
val tuple = key.asInstanceOf[(String, String, Int)]
val subject = tuple._1
//到实现初始化的规则中查找这个学科对应的分区编号
rules(subject)
}
}
2. 自定义排序
数据形式：姓名，年龄，颜值
需求：首先按照颜值排序（颜值高的排前面），当颜值相同的情况下，年龄小的人排前面
2.1 第一种形式
思路，定义一个Boy类（case class），用来加载这些属性，利用隐式转换定义一个排序规则，具体如下
CustomSort1
object CustomSort1 {
def main(args: Array[String]): Unit = {
import com._51doit.spark04.MyPredef
val conf: SparkConf = new SparkConf()
.setAppName(this.getClass.getSimpleName)
.setMaster("local[*]")
// 创建SparkContext
val sc: SparkContext = new SparkContext(conf)
val lines: RDD[String] = sc.parallelize(List("jack,30,99.99", "sherry,18,9999.99", "Tom,29,99.99"))
// 处理数据,使用mapPartitions，减少Boy的创建
val BoyRDD: RDD[Boy] = lines.mapPartitions(it => {
it.map(t => {
val split: Array[String] = t.split(",")
Boy(split(0), split(1).toInt, split(2).toDouble)
})
})
import MyPredef.Boy2OrderingBoy
val res: RDD[Boy] = BoyRDD.sortBy(t => t)
print(res.collect().toBuffer)
}
}
MyPredef
两种写法都行
Ordered的形式
object MyPredef {
implicit val Boy2OrderingBoy: Boy => Ordered[Boy] = (boy:Boy) => new Ordered[Boy]{
override def compare(that: Boy): Int = {
if(boy.fv == that.fv){
boy.age - that.age
} else{
-(boy.fv - that.fv).toInt
}
}
}
}
Ordering的形式
object MyPredef {
implicit val Boy2OrderingBoy: Ordering[Boy] = new Ordering[Boy] {
override def compare(x: Boy, y:Boy): Int = {
if (x.fv == y.fv) {
x.age - y.age
} else {
-(x.fv - y.fv).toInt
}
}
}
}
Boy
case class Boy(name:String, age:Int, fv: Double)
2.2 第二种形式（借助元组）
object CustomSort2 {
def main(args: Array[String]): Unit = {
val conf: SparkConf = new SparkConf()
.setAppName(this.getClass.getSimpleName)
.setMaster("local[*]")
// 创建SparkContext
val sc: SparkContext = new SparkContext(conf)
val lines: RDD[String] = sc.parallelize(List("jack,30,99.99", "sherry,18,9999.99", "Tom,29,99.99"))
// 处理数据,使用mapPartitions，减少Boy的创建
val tpRDD: RDD[(String, Int, Double)] = lines.mapPartitions(it => {
it.map(t => {
val split: Array[String] = t.split(",")
(split(0), split(1).toInt, split(2).toDouble)
})
})
// 使用元组的默认规则进行排序
val sorted: RDD[(String, Int, Double)] = tpRDD.sortBy(t => (-t._3, t._2))
println(sorted.collect().toBuffer)
}
}
来源：https://www.cnblogs.com/jj1106/p/12014214.html