第四章 Spark SQL多数据源交互

Spark SQL可以与多种数据源交互，如普通文本、json、parquet、csv、MySQL等

1.写入不同数据源

2.读取不同数据源

 写数据

package cn.itcast.sql

  

  import java.util.Properties

  

  import org.apache.spark.SparkContext

  import org.apache.spark.rdd.RDD

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

  

  

  object WriterDataSourceDemo {
case class Person(id:Int,name:String,age:Int)

  

  def main(args: Array[String]): Unit = {

    //1.创建SparkSession

    val spark: SparkSession = SparkSession.builder().master("local[*]").appName("SparkSQL")
.getOrCreate()

    val sc: SparkContext = spark.sparkContext

    sc.setLogLevel("WARN")

    //2.读取文件

    val fileRDD: RDD[String] = sc.textFile("D:\\data\\person.txt")

    val linesRDD: RDD[Array[String]] = fileRDD.map(_.split(" "))

    val rowRDD: RDD[Person] = linesRDD.map(line =>Person(line(0).toInt,line(1),line(2).toInt))

    //3.将RDD转成DF

    //注意:RDD中原本没有toDF方法,新版本中要给它增加一个方法,可以使用隐式转换

    import spark.implicits._

    //注意:上面的rowRDD的泛型是Person,里面包含了Schema信息

    //所以SparkSQL可以通过反射自动获取到并添加给DF

    val personDF: DataFrame = rowRDD.toDF

    //==================将DF写入到不同数据源===================

    //Text data source supports only a single column, and you have 3 columns.;

    //personDF.write.text("D:\\data\\output\\text")

    personDF.write.json("D:\\data\\output\\json")

    personDF.write.csv("D:\\data\\output\\csv")

    personDF.write.parquet("D:\\data\\output\\parquet")

    val prop = new Properties()

    prop.setProperty("user","root")

    prop.setProperty("password","root")

    personDF.write.mode(SaveMode.Overwrite).jdbc(
"jdbc:mysql://localhost:3306/bigdata?characterEncoding=UTF-8","person",prop)

    println("写入成功")

    sc.stop()

    spark.stop()

  }

 }

 

读数据

package cn.itcast.sql

  

  import java.util.Properties

  

  import org.apache.spark.SparkContext

  import org.apache.spark.sql.SparkSession

  

  

  object ReadDataSourceDemo {

  def main(args: Array[String]): Unit = {

    //1.创建SparkSession

    val spark: SparkSession = SparkSession.builder().master("local[*]").appName("SparkSQL")
.getOrCreate()

    val sc: SparkContext = spark.sparkContext

    sc.setLogLevel("WARN")

    //2.读取文件

    spark.read.json("D:\\data\\output\\json").show()

    spark.read.csv("D:\\data\\output\\csv").toDF("id","name","age").show()

    spark.read.parquet("D:\\data\\output\\parquet").show()

    val prop = new Properties()

    prop.setProperty("user","root")

    prop.setProperty("password","root")

    spark.read.jdbc(
"jdbc:mysql://localhost:3306/bigdata?characterEncoding=UTF-8","person",prop).show()

    sc.stop()

    spark.stop()

  }

}

 

 总结

1.SparkSQL写数据:

DataFrame/DataSet.write.json/csv/jdbc

2.SparkSQL读数据:

SparkSession.read.json/csv/text/jdbc/format