sparksql实现单词计数

1、创建SparkSession

val sparkSession = SparkSession.builder()
	.appName("SparkWordCount")
	.master("local[2]")
	.getOrCreate()

2、加载数据，使用dataset处理数据集
read来读取可以直接返回DataSet[String],这是个比RDD更高级的数据集
它返回一个列名为value的视图

+------------+
|       value|
+------------+
|hello tunter|
|  hello tony|
| tony hunter|
+------------+

val datas:DataSet[String] = 
sparkSession.read.textFile("hdfs://192.168.252.121:9000/words.txt")

3、切分数据

//导入隐式
import sparkSession.implicits._
val word:DataSet[String]=datas.flatMap(_.split(" "))

4、注册视图

word.createTempView("wc_t")

5、执行sql

val df:DataFrame = sparkSession.sql("select value, count(*) sum from wc_t 
group by value order by sum desc")

df.show()
sparkSession.stop()

执行结果：

+------+---+
| value|sum|
+------+---+
| hello|  2|
|  tony|  2|
|hunter|  1|
|tunter|  1|
+------+---+

完整代码：

import org.apache.spark.sql.{DataFrame, Dataset, SparkSession}

object SparkSqlWordCount {
  def main(args: Array[String]): Unit = {
    //1.创建sparksession
    val sparkSession = SparkSession.builder()
      .appName("SparkSqlWordCount")
      .master("local[2]")
      .getOrCreate()

    //2.加载数据 使用dataset处理数据集
    val datas: Dataset[String] = sparkSession.read
      .textFile("hdfs://192.168.252.121:9000/words.txt")

    //3.sparksql 注册表/注册视图
    import sparkSession.implicits._
    val word: Dataset[String] = datas.flatMap(_.split(" "))

    //4.注册视图
    word.createTempView("wc_t")

    //5.执行sql
    val frame: DataFrame = sparkSession.sql("select value,count(*) sum from wc_t group by value order by sum desc")

    frame.show()
    sparkSession.stop()
  }
}