package mobvista.dmp.datasource.adn import mobvista.dmp.common.{CommonSparkJob, MobvistaConstant} import mobvista.dmp.util.MRUtils import org.apache.commons.cli.{BasicParser, Options} import org.apache.hadoop.fs.{FileSystem, Path} import org.apache.hadoop.io.compress.GzipCodec import java.net.URI import scala.collection.mutable.ArrayBuffer /** * @package: mobvista.dmp.datasource.adn * @author: wangjf * @date: 2021/2/23 * @time: 5:12 下午 * @email: jinfeng.wang@mobvista.com */ class AdnClickJoinInstallDaily extends CommonSparkJob with Serializable { def commandOptions(): Options = { val options = new Options() options.addOption("input_click", true, "input_click") options.addOption("input_install", true, "input_install") options.addOption("output", true, "output") options.addOption("coalesce", true, "coalesce") options } override protected def run(args: Array[String]): Int = { val parser = new BasicParser() val options = commandOptions() val commandLine = parser.parse(options, args) val input_click = commandLine.getOptionValue("input_click") val input_install = commandLine.getOptionValue("input_install") val output = commandLine.getOptionValue("output") val coalesce = commandLine.getOptionValue("coalesce") val spark = MobvistaConstant.createSparkSession(s"AdnClickJoinInstallDaily") try { val sc = spark.sparkContext // deviceId, deviceType, requestId, campaignId, platform val clickRdd = sc.textFile(input_click).map(_.split("\t")).map(r => { val deviceId = r(0) val deviceType = r(1) val requestId = r(2) val campaignId = r(3) val platform = r(4) (MRUtils.JOINER.join(requestId, campaignId), (deviceId, deviceType, platform)) }).combineByKey( (v: (String, String, String)) => Iterable(v), (c: Iterable[(String, String, String)], v: (String, String, String)) => c ++ Seq(v), (c1: Iterable[(String, String, String)], c2: Iterable[(String, String, String)]) => c1 ++ c2 ) // deviceId, deviceType, platform, requestId, campaignId, packageName val installRdd = sc.textFile(input_install).map(_.split("\t")).map(r => { val deviceId = r(0) val deviceType = r(1) val platform = r(2) val requestId = r(3) val campaignId = r(4) val packageName = r(5) (MRUtils.JOINER.join(requestId, campaignId), (deviceId, deviceType, platform, packageName)) }).combineByKey( (v: (String, String, String, String)) => Iterable(v), (c: Iterable[(String, String, String, String)], v: (String, String, String, String)) => c ++ Seq(v), (c1: Iterable[(String, String, String, String)], c2: Iterable[(String, String, String, String)]) => c1 ++ c2 ) val rdd = installRdd.join(clickRdd).map(rs => { val arrayBuffer = new ArrayBuffer[String] val keys = MRUtils.SPLITTER.split(rs._1, -1) val campaignId = keys(1) val install = rs._2._1 // val click = rs._2._2 install.foreach(e => { val deviceId = e._1 val deviceType = e._2 val platform = e._3 val packageName = e._4 arrayBuffer += MRUtils.JOINER.join(deviceId, deviceType, platform, campaignId, packageName) }) arrayBuffer.toIterator }).flatMap(l => l) FileSystem.get(new URI(s"s3://mob-emr-test"), spark.sparkContext.hadoopConfiguration).delete(new Path(output), true) rdd.repartition(coalesce.toInt).saveAsTextFile(output, classOf[GzipCodec]) } finally { if (spark != null) { spark.stop() } } 0 } } object AdnClickJoinInstallDaily { def main(args: Array[String]): Unit = { new AdnClickJoinInstallDaily().run(args) } }