Hadoop+Spark大数据技术（微课版）曾国荪、曹洁版 第七章 Spark RDD编程实验

Apache Spark中，`sc` 和 `reduce` 分别代表 SparkContext 和 reduce 函数。

使用文本文件创建RDD

注意：需要先在hadoop分布式文件系统中创建文件

1.先在本地文件系统创建data.txt文件

cd /usr/local/hadoop/input

gedit data.txt

2.启动hadoop分布式文件系统

./sbin/start-dfs.sh

3.上传本地文件data.txt到hadoop分布式文件系统

hdfs dfs -put /usr/local/hadoop/input/data.txt input

查看分布式文件系统中是否存在data.txt
hdfs dfs -ls input

textFile 函数用于读取文本文件中的内容，并将其转换为RDD。

rdd.map(line => line.length)：这一行代码将每一行文本映射为其长度，即字符数。

.map()函数对RDD中的每个元素应用给定的函数，并返回结果组成的新RDD。

.reduce(_ + _)：这一行代码对RDD中的所有元素进行累加，得到它们的总和。.reduce()函数将RDD中的元素两两合并，直到只剩下一个元素，这里使用的匿名函数 _ + _ 表示将两个元素相加。

val wordCount = rdd.map(line => line.length).reduce(_ + _)：这行代码将文本文件中所有行的字符数求和，并将结果保存在wordCount变量中。

wordCount：最后，wordCount 变量中保存着文本文件中所有行的字符总数。

val rdd = sc.textFile("/user/hadoop/input/data.txt")
rdd: org.apache.spark.rdd.RDD[String] = /user/hadoop/input/data.txt MapPartitionsRDD[31] at textFile at <console>:25

val wordCount = rdd.map(line => line.length).reduce(_ + _)
wordCount

wordCount: Int = 274
res11: Int = 274

一些JSON数据

{
  "name": "中国",
  "province": [
    {
      "name": "河南",
      "cities": [
        {
          "city": ["郑州","洛阳"]
      ]
    }
  ]
}
```

{
  "code": 0,
  "msg": "",
  "count": 2,
  "data": [
    {
      "id": 101,
      "username": "zhangsan",
      "city": "xiamen"
    },
    {
      "id": 102,
      "username": "liming",
      "city": "zhengzhou"
    }
  ]
}
{"学号":"106","姓名":"李明","数据结构":"92"}

{"学号":"242","姓名":"李乐","数据结构":"93"}

{"学号":"107","姓名":"冯涛","数据结构":"99"}

完整程序练习

seq是一个包含元组的列表，每个元组都有一个字符串键和一个字符串列表值。

sc.parallelize方法将seq转换为一个RDD。

rddP.partitions.size是用来获取RDD的分区数量的。RDD的分区是数据的逻辑划分，决定了数据在集群中的分布方式。每个分区都可以在集群的不同节点上进行并行处理。

//使用程序中的数据创建RDD

val arr = Array(1,2,3,4,5,6)

val rdd = Array(1,2,3,4,5,6)

arr: Array[Int] = Array(1, 2, 3, 4, 5, 6)
rdd: Array[Int] = Array(1, 2, 3, 4, 5, 6)

val sum = rdd.reduce(_ + _)

sum: Int = 21

//parallelize() 创建RDD

val seq = List( ("num", List("one","two","three") ), ("study",List("Scala","Python","Hadoop")) , ("color", List("blue","white","black")) )

val rddP = sc.parallelize(seq)   

rddp.partitions.size

seq: List[(String, List[String])] = List((num,List(one, two, three)), (study,List(Scala, Python, Hadoop)), (color,List(blue, white, black)))
rddP: org.apache.spark.rdd.RDD[(String, List[String])] = ParallelCollectionRDD[7] at parallelize at <console>:31
res5: Int = 2

//makeRDD() 创建RDD

val rddM = sc.makeRDD(seq)

rddM.partitions.size

rddM: org.apache.spark.rdd.RDD[String] = ParallelCollectionRDD[8] at makeRDD at <console>:30
res6: Int = 3

//hdfs的文件创建RDD

val rdd = sc.textFile("/user/hadoop/input/data.txt")

rdd: org.apache.spark.rdd.RDD[String] = /user/hadoop/input/data.txt MapPartitionsRDD[31] at textFile at <console>:25

val wordCount = rdd.map(line => line.length).reduce(_ + _)

wordCount

wordCount: Int = 274
res11: Int = 274

//本地文件创建RDD

val rdd = sc.textFile("file:/home/hadoop/data.txt")

val wordCount = rdd.map(line => line.length).reduce(_ + _)

wordCount

rdd: org.apache.spark.rdd.RDD[String] = file:/home/hadoop/data.txt MapPartitionsRDD[34] at textFile at <console>:29
wordCount: Int = 274
res12: Int = 274

val rddw1 = sc.textFile("file:/home/hadoop/input")

rddw1.collect()

rddw1: org.apache.spark.rdd.RDD[String] = file:/home/hadoop/input MapPartitionsRDD[37] at textFile at <console>:25
res13: Array[String] = Array(Hello Spark!, Hello Scala!)

val rddw1 = sc.wholeTextFiles("file:/home/hadoop/input")

rddw1.collect()

rddw1: org.apache.spark.rdd.RDD[(String, String)] = file:/home/hadoop/input MapPartitionsRDD[39] at wholeTextFiles at <console>:27
res14: Array[(String, String)] =
Array((file:/home/hadoop/input/text1.txt,"Hello Spark!
"), (file:/home/hadoop/input/text2.txt,"Hello Scala!
"))

val jsonStr = sc.textFile("file:/home/hadoop/student.json")

jsonStr.collect()

jsonStr: org.apache.spark.rdd.RDD[String] = file:/home/hadoop/student.json MapPartitionsRDD[48] at textFile at <console>:28
res18: Array[String] = Array({"学号":"106","姓名":"李明","数据结构":"92"}, {"学号":"242","姓名":"李乐","数据结构":"93"}, {"学号":"107","姓名":"冯涛","数据结构":"99"})

import scala.util.parsing.json.JSON

val jsonStr = sc.textFile("file:/home/hadoop/student.json")

val result  = jsonStr.map(s => JSON.parseFull(s))

result.foreach(println)

Some(Map(学号 -> 107, 姓名 -> 冯涛, 数据结构 -> 99))
Some(Map(学号 -> 106, 姓名 -> 李明, 数据结构 -> 92))
Some(Map(学号 -> 242, 姓名 -> 李乐, 数据结构 -> 93))

import scala.util.parsing.json.JSON
jsonStr: org.apache.spark.rdd.RDD[String] = file:/home/hadoop/student.json MapPartitionsRDD[50] at textFile at <console>:31
result: org.apache.spark.rdd.RDD[Option[Any]] = MapPartitionsRDD[51] at map at <console>:32

import java.io.StringReader

import au.com.bytecode.opencsv.CSVReader

val gradeRDD = sc.textFile("file:/home/hadoop/sparkdata/grade.csv")

val result = gradeRDD.map{line => val reader = new CSVReader(new StringReader(line));reader.readNext()}

result.collect().foreach(x => println(x(0),x(1),x(2)))

(101,LiNing,95)
(102,LiuTao,90)
(103,WangFei,96)

import java.io.StringReader
import au.com.bytecode.opencsv.CSVReader
gradeRDD: org.apache.spark.rdd.RDD[String] = file:/home/hadoop/sparkdata/grade.csv MapPartitionsRDD[53] at textFile at <console>:31
result: org.apache.spark.rdd.RDD[Array[String]] = MapPartitionsRDD[54] at map at <console>:32