flink 实现 dau 去重

flink 实现 dau 去重（状态实现）

//主程序
object DauFirstInfo{
	def main(args: Array[String]):Unit = {
		// 定义环境变量，创建消费环境配置
		val testDauGroupId = "test_daufirstinfo_20240303"
		val source_kafka_host = ConfigHelper.getBootStrapServers(Config.TEST_SOURCE_KAFKA,Region.TEST_CLUSTER.getName())
		val testTopicProprties:Properties = KafkaConnector.createConsumerProperties(source_kafka_host,testDauGroupId)
		val testConsumer = KafkaConnector.createStringSource(Config.TEST_SOURCE_KAFKA,testTopicProprties)
		val env:StreamExecutionEnvironment = StreamingUtils.getStreamExecutionment 
		
		env.getCheckpointConfig.setMaxConcurrentCheckpoints(1)
		val dataStream = env
		.addSource(testConsumer).uid("xxx").name("xxx")
		// 筛选出有设备 id + 启动时间的日志（有效日志）
		.flatMap(new RichFlatMapFunction[String,String]){
			override def flatMap(in:String,collector:Collector[String]): Unit = {
				val sourceObject: JsonObject = new JsonParser().parse(in).getJsonObject()
				if(StringUnits.isNotBlank(sourceObject.get("shebei_id").getAsString) && sourceObject.get("qidong_timestamp").getAsLong){
				collector.collect(in)
				}
			}
		}).uid("map_valid_info").name("map_valid_info")
		//找出标识 dau 的key：设备id+当天日期
		.keyBy(new KeySelector[String,String]{
			override def getKey(value:String):String = {
				val sourceObject: JsonObject = new JsonParser().parse(in).getJsonObject()
				sourceObject.get("sheibei_id").getAsString + DateFormatUtils.format(sourceObject.get("qidong_timestamp").getAsLong,"yyyy-MM-dd")
			}
		})
		//将 key 存在状态里面，进行判断
		.flatMap(new FirstNotBlankDauProcess()).uid("first_dau_state").name("first_dau_state")
		//设置处理的并发度
		.setMaxParallelism(100)
		//设置产出的环境变量
		val sink_kafka_host:String = Confighelper.getBootstrapServers(CommonConfig.TEST_SINK_KAFKA,Region.TEST_CLUSTER.getName())
		val props: Properties = new Properties()
		props.put("bootstrap.servers",sink_kafka_host)
		props.put("request.timeout.ms",200000)
		val kafkaProducer: FlinkKafkaProducer[String] = kafkaConnector.createStringSink(Commonfig.TEST_SINK_KAFKA,props)
		dataStream.addSink(kafkaProducer).uid("sink_id").name("sink_name").setParallelism(30)
		env.execute("ValidDauJob")
	}
}

// 状态变量去重逻辑
class FirstNotBlankDauProcess extends RichFlatMapFunction[String,String]{
	var clientState : ValueState[Integer] = _
	override def open(parameters: Configuration): Unit = {
		super.open(parameters)
		val stateTtlConfig: StateTtlConfig = StateTtlConfig
		.newBulider(Time.minutes(1560))
		.useProcessingTime()
		.setUpdateType(StateTtlConfig.UpdateType.OnCreateAndWrite)
		.setStateVisibility(StateTtlConfig.StateVisibility.ReturnExpiredIfNotCleanedUp)
		.bulid
		
		val roaringBitmapDescriptor = new ValueStateDescriptor[Integer]("FirstNotBlankDauState",Types.GENERIC(classOf[Integer]))
		roaringBitmapDescriptor.enableTimeToLive(stateTtlConfig)

		clientState = getRuntimeContext.getSate[Integer](roaringBitmapDescriptor)
	}

	override def flatMap(in:String,out:Collector[String]): Unit = {
		val existFlag = clientState.value()
		if(existFlag != 1){
			clientstate.update(1)
			out.collect(in)
		}
	}

	override def close(): Unit = {
		super.close()
	}
}