Spark DataFrame入门学习笔记

Spark DataFrame入门学习笔记

1. 添加配置文件

1.1、 配置文件存放目录

resources 目录下，以 scala.properties 文件为例

1.2、 添加读取配置代码

import java.util.Properties

object LoadProperties {

  def get_instance: Properties = {
    val properties = new Properties()
    val in = Thread.currentThread().getContextClassLoader.
      getResourceAsStream("scala.properties")
    properties.load(in)
    properties
  }

}

2. 初始化Spark

// 初始化配置文件加载器
val pros: Properties = LoadProperties.get_instance
// 判断是否本地运行
val isLocal: Boolean = pros.get("is_load_local_data").toString == "true"
val conf: SparkConf =
	if (isLocal)
    	// 设置本地运行时 SparkConf 与 modelPath
    	new SparkConf().setAppName(getClass.getSimpleName).setMaster("local[2]")
    else
    	// 设置线上运行时 SparkConf 与 modelPath
        new SparkConf().setAppName(getClass.getSimpleName)
        
	// 设置 SparkContext
    val sc = new SparkContext(conf)
    // 设置 HiveContext
    val spark = new HiveContext(sc)
    spark.setConf("spark.sql.shuffle.partitions", "3000")
    import spark.implicits._

注意：

此处应该格外注意，对spark相关上下文初始化的过程中，应该放在main方法中，而不是放在静态变量加载区。由于Spark的运行是分布式的运行，静态变量(包括spark相关上下文及自定义对象的创建)在一个节点加载后，在其他节点运行时，静态变量及静态方法不会重新加载，从而导致该类初始化失败。

• SparkConf：Spark配置类，配置已键值对形式存储，封装了一个ConcurrentHashMap类实例settings用于存储Spark的配置信息。负责管理所有的Spark配置项。
• SparkContext：Spark的主要入口点，Spark的灵魂。
• HiveContext：Spark访问Hive表(或本地文件)的引擎

3. 读入数据

3.1、 本地文件导入

case class Data(phone: String, app_code: String, count: Double, master_flag: String, gender: String)
val dpi_path: Any = pros.get("dpi.data.path")   //可通过配置文件配置
data = = sc.textFile("" + dpi_path)
                .map(_.split("\t"))
                .map(x => Data(x(0), x(1), x(2).toDouble, x(3), x(4)))
                .toDF()

2.2 从Hive数据库中读取

val tableName1 = pros.get("tableName1").toString
val tableName2 = pros.get("tableName2").toString
val beginDay = pros.get("beginDay").toString
val endDay = pros.get("endDay").toString
val dpiSql: String =
                s" select a.phone phone, a.app_code app_code, a.count count," +                           s" b.master_flag master_flag " +
                s" from " +
                s" (select phone phone, appcode app_code, sum(usetimes) count" +
                s" from $tableName1 where day>=$beginDay and day <= $endDay" +
                s" group by phone, appcode) as a" +
                s" inner join " +
                s" (select phone, master_flag from $tableName2 where year > 2018)"
                s" as b " +
                s" on a.phone=b.phone"
dpiData = spark.sql(dpiSql)

2.3 从关系型数据库中读取(eg: Mysql)

def main(args: Array[String]): Unit = {
        val pros = LoadProperties.get_instance
        val conf: SparkConf = new 			     SparkConf().setAppName(getClass.getSimpleName).setMaster("local[2]")
        // 设置 SparkContext
        val sc = new SparkContext(conf)
        // 设置 HiveContext
        val spark = new HiveContext(sc)
        val jdbcDF = spark.read.format("jdbc")
            .option("url", "jdbc:mysql://" + pros.get("mysql.url").toString)
            .option("driver", pros.getProperty("mysql.driver"))
            .option("user", pros.getProperty("mysql.user"))
            .option("password", pros.getProperty("mysql.password"))
    		.option("dbtable", pros.getProperty("mysql.dbtable"))
            .load()
        jdbcDF.show()
    }

HiveContext 默认读入Mysql表中所有数据，若需要添加过滤条件，则需要通过 sql 语句构造中间表。修改 jdbcDF 如下：

val tableName = pros.getProperty("mysql.dbtable")
val table = s"(select id, username, email from $tableName where id < 3 ) as t1"
val jdbcDF = spark.read.format("jdbc")
	.option("url", "jdbc:mysql://" + pros.get("mysql.url").toString)
    .option("driver", pros.getProperty("mysql.driver"))
    .option("user", pros.getProperty("mysql.user"))
    .option("password", pros.getProperty("mysql.password"))
	.option("dbtable", table)
    .load()

4. 数据倾斜后的散列操作

4.1 添加随机数散列到不同节点

• 环境：40个节点
• 性别：男，女

问题描述：若不做散列，则容易出现数据倾斜到两个节点，导致两个节点崩掉

解决方案：将数据散列到不同节点

// 稀疏特征
def keyWithRandom(string: String): String = {
	string + Random.nextInt(200)
}

val w = Window.partitionBy("gender").orderBy(rand())
// 注册UDF函数
val udfKeyAddRandomValue = spark.udf.register("keyWithRandom", keyWithRandom _)
var tmpData = featureData.withColumn("gender", udfKeyAddRandomValue($"gender"))
	.withColumn("rk", row_number().over(w))
	.filter($"rk" <= 5000)
	.withColumn("gender", substring($"gender", 0, 1))
	.cache()

代码含义：

• udfKeyAddRandomValue : 将gender列特征的值加上一个随机数
• row_number.over(w) : over根据gender排序进行累计，row_number将产生一个新的编号，所以该代码含义是将gender对不同的partition(partition是性别)进行排序后对不同的partition添加新的编号
• w ：将render列排序。orderBy(rand())的作用是排序(每次排序的结果都不一样)【Sql】
• filter ：过滤rk(gender编号)小于5000的值，防止某个节点过载。

row_number.over(w) 函数功能示意图：

5. 数据分批次处理

spark.udf.register("hash_phone",(phone:String)=>phone.hashCode % 20)
hash = 10
val tableName1 = pros.get("tableName1").toString
val tableName2 = pros.get("tableName2").toString
val beginDay = pros.get("beginDay").toString
val endDay = pros.get("endDay").toString
val dpiSql: String =
                s" select a.phone phone, a.app_code app_code, a.count count," +                           s" b.master_flag master_flag " +
                s" from " +
                s" (select phone phone, appcode app_code, sum(usetimes) count" +
                s" from $tableName1" +
                s" where hash_phone(phoneno) = $hash" +
                s" group by phone, appcode) as a" +
                s" inner join " +
                s" (select phone, master_flag from $tableName2 where year > 2018)"
                s" as b " +
                s" on a.phone=b.phone"
dpiData = spark.sql(dpiSql)