SparkSQL架构、运行原理、Dataset、dataframe

Spark SQL架构

Spark SQL是Spark的核心组件之一（2014.4 Spark1.0）
能够直接访问现存的Hive数据
提供JDBC/ODBC接口供第三方工具借助Spark进行数据处理
提供了更高层级的接口方便地处理数据
支持多种操作方式：SQL、API编程
支持多种外部数据源：Parquet、JSON、RDBMS等

Spark SQL运行原理

Catalyst优化器是Spark SQL的核心
Catalyst Optimizer：Catalyst优化器，将逻辑计划转为物理计划

SparkContext
SQLContext
Spark SQL的编程入口
HiveContext
SQLContext的子集，包含更多功能
SparkSession（Spark 2.x推荐）
SparkSession：合并了SQLContext与HiveContext
提供与Spark功能交互单一入口点，并允许使用DataFrame和Dataset API对Spark进行编程

Dataset

特定域对象中的强类型集合
1、createDataset()的参数可以是：Seq、Array、RDD
2、上面三行代码生成的Dataset分别是：
Dataset[Int]、Dataset[(String,Int)]、Dataset[(String,Int,Int)]
3、Dataset=RDD+Schema，所以Dataset与RDD有大部共同的函数，如map、filter等

Dataset创建

   val spark: SparkSession = SparkSession.builder().master("local[*]").appName("demo").getOrCreate()import spark.implicits._//Dataset创建方法val ds1: Dataset[Int] = spark.createDataset(1 to 6)ds1.show()val ds2: Dataset[(String, Int)] = spark.createDataset(List(("a",1),("b",2),("c",3)))ds2.show()val ds3: Dataset[(String, Int, Int)] = spark.createDataset(sc.parallelize(List(("gree",17,175),("tom",22,180))))ds3.show()

使用样例类创建dataset

  //样例类case class Point(label:String,x:Double,y:Double)case class Category(i: Int, str: String)val conf: SparkConf = new SparkConf().setMaster("local[*]").setAppName("demo1")val sc = SparkContext.getOrCreate(conf)val spark: SparkSession = SparkSession.builder().master("local[*]").appName("demo").getOrCreate()import spark.implicits._val points = Seq(Point("njzb",23.23,48.71),Point("njnz",26.12,48.33))val pointsDS: Dataset[Point] = points.toDS()pointsDS.printSchema()pointsDS.select("label").show()val categories = Seq(Category(1,"njzb"),Category(2,"njnz"))val cateDS: Dataset[Category] = categories.toDS()cateDS.printSchema()//rdd转Datasetval rdd: RDD[Int] = sc.parallelize(1 to 6)val ds: Dataset[Int] = rdd.toDS()

DataFrame

DataFrame=Dataset[Row]
类似传统数据的二维表格
在RDD基础上加入了Schema（数据结构信息）
DataFrame Schema支持嵌套数据类型
struct
map
array
提供更多类似SQL操作的AP

RDD与DataFrame对比

通过加载文件 创建dataframe

val spark = SparkSession.builder().master("local[*]").appName("jsonsession").getOrCreate()val frame = spark.read.format("json").option("header",true).load("in/user.json")frame.printSchema()frame.show()val nameColumn: Column = frame("name")frame.select(nameColumn).show()frame.select("name").show()

rdd -> dataframe

val conf: SparkConf = new SparkConf().setMaster("local[*]").setAppName("demo1")val sc = SparkContext.getOrCreate(conf)val spark: SparkSession = SparkSession.builder().master("local[*]").appName("demo").getOrCreate()import spark.implicits._val people: RDD[String] = sc.textFile("in/people.txt")val schemaString = "id name age"
//    val schema: StructType = StructType(schemaString.split(" ").map(x => StructField(x, StringType, true)))
//    val fields: Array[StructField] = schemaString.split(" ").map(x => StructField(x, StringType, true))val fields1 = Array(StructField("id", IntegerType, true),StructField("name", StringType, true),StructField("age", IntegerType, true))val schema: StructType = StructType(fields1)val row: RDD[Row] = people.map(x => x.split(" ")).map(x => Row(x(0).toInt, x(1), x(2).toInt))val peopleDF: DataFrame = spark.createDataFrame(row, schema)peopleDF.printSchema()peopleDF.show()peopleDF.createOrReplaceTempView("people")val resultDF: DataFrame = spark.sql("select name from people where age>30")resultDF.show()//读写parquet文件 
//    peopleDF.write.parquet("out/parquettest")val frame1: DataFrame = spark.read.parquet("out/parquettest")frame1.printSchema()frame1.show()

dataframe ->rdd

/** people.json内容如下* {"name":"Michael"}* {"name":"Andy", "age":30}* {"name":"Justin", "age":19}*/
val df = spark.read.json("file:///home/hadoop/data/people.json")
//将DF转为RDD
df.rdd.collect