1、broadcast广播

在Spark中，broadcast是一种优化技术，它可以将一个只读变量缓存到每个节点上，以便在执行任务时使用。这样可以避免在每个任务中重复传输数据。

2、构建缓存

import org.apache.spark.sql.SparkSession
import org.apache.spark.broadcast.Broadcast
import com.alibaba.fastjson.JSONObject

// 定义全局缓存单例对象
object GlobalCache extends Serializable {

  // 广播变量，用于存储缓存数据
  private var cacheData: Broadcast[collection.mutable.Map[String, JSONObject]] = _

  // 设置 SparkSession 和广播变量
  def setSparkSession(spark: SparkSession): Unit = {
    cacheData = spark.sparkContext.broadcast(collection.mutable.Map.empty[String, JSONObject])
  }


  // 按订单ID和用户ID缓存JSONObject对象
  def cacheJSONObject(orderId: String, userId: String, jsonObject: JSONObject): Unit = {
    // 获取广播变量的值并进行修改
    val data = cacheData.value
    data.synchronized {
      data.put(generateKey(orderId, userId), jsonObject)
    }
  }

  // 根据订单ID和用户ID删除缓存的JSONObject对象
  def removeJSONObject(orderId: String, userId: String): Unit = {
    // 获取广播变量的值并进行修改
    val data = cacheData.value
    data.synchronized {
      data.remove(generateKey(orderId, userId))
    }
  }

  // 根据订单ID和用户ID获取缓存的JSONObject对象
  def getJSONObjet(orderId: String, userId: String): JSONObject = {
    // 获取广播变量的值并进行访问
    val data = cacheData.value
    data.synchronized {
      data.get(generateKey(orderId, userId)).orNull
    }
  }

  // 生成缓存键，使用订单ID和用户ID拼接
  private def generateKey(orderId: String, userId: String): String = s"$orderId|$userId"
}

3、缓存测试

import org.apache.spark.sql.SparkSession
import org.apache.spark.broadcast.Broadcast
import com.alibaba.fastjson.JSONObject
import org.apache.log4j.{Level, Logger}

object CacheTest {
  Logger.getLogger("org").setLevel(Level.ERROR)
  Logger.getRootLogger().setLevel(Level.ERROR) // 设置日志级别


  def addItem(orderId:String, userId:String, name:String): Unit = {
    val jsonObject = new JSONObject()
    jsonObject.put("name", name)

    // 缓存JSONObject对象
    GlobalCache.cacheJSONObject(orderId, userId, jsonObject)
  }


  def getCache(orderId: String, userId: String): JSONObject = {
    // 获取缓存的JSONObject对象
    GlobalCache.getJSONObjet(orderId, userId)
  }

  def delItem(orderId:String, userId:String): Unit = {
    // 删除缓存的JSONObject对象
    GlobalCache.removeJSONObject(orderId, userId)
  }


  def getSparkSession(appName: String, localType: Int): SparkSession = {
    val builder: SparkSession.Builder = SparkSession.builder().appName(appName)
    if (localType == 1) {
      builder.master("local[8]") // 本地模式，启用8个核心
    }

    val spark = builder.getOrCreate() // 获取或创建一个新的SparkSession
    spark.sparkContext.setLogLevel("ERROR") // Spark设置日志级别
    spark
  }

  def main(args: Array[String]): Unit = {
    println("Start CacheTest")
    val spark: SparkSession = getSparkSession("CacheTest", 1)

    GlobalCache.setSparkSession(spark)  // 构造全局缓存

    addItem("001", "456", "苹果")      // 添加元素
    addItem("002", "789", "香蕉")      // 添加元素
    var cachedObject = getCache("001", "456")
    println(s"Cached Object: $cachedObject")

    delItem("001", "456")      // 删除元素
    cachedObject = getCache("001", "456")
    println(s"Cached Object: $cachedObject")
    spark.stop()
  }
}

4、控制台输出

Start CacheTest
Using Spark's default log4j profile: org/apache/spark/log4j-defaults.properties
Cached Object: {"name":"苹果"}
Cached Object: null

Process finished with exit code 0