用户画像：最喜爱的明星前3名

package com.profile.main

import java.math.BigDecimal
import com.profile.comment.Comments
import com.profile.tools._
import org.apache.commons.lang3.StringUtils
import scala.collection.mutable.ListBuffer
import com.profile.main
 /**
  * listBuffer(userid,;;;;)-->,userId+"|"+actor+"|"+playtime
  *  用户画像：7）最喜爱的明星前3名
  * @param userId
  * @param actor
  * @param playtime
  */
case class most_love_actors(userId:String,actor:String,playtime:Int){
  override def toString: String = {
    userId+"|"+actor+"|"+playtime
  }
}
 /**
  * 7）最喜爱的明星前3名,用户画像最喜欢的演员
  * @author denghd
  * date 2017-11-08 16:16
  */
object UserLoveActors {

   def main(args: Array[String]): Unit = {
     val date=if(args.length==1) args(0) else DateTools.getYestodayDate //yyyy-MM-dd
     val sc=SparkTools.getSparkContext
     val vodProgramMap = sc.broadcast(JdbcTools.getLoveVodProgramMap)  //点播节目
     val sqlContext = new org.apache.spark.sql.SQLContext(sc)
     import sqlContext.implicits._
     val rdd=if(args.length==1) ReadData.readDataFromHdfs(sc,Comments.ilogslave_log_hdfs_path+date) else ReadData.readDataFromLocal(sc,"E:\logs\2017-11-07")
     val listRDD = rdd.filter(x=> x.contains("vod") && x.contains("VideoPlay")) //x=>"vod".equals(LogTools.getKeywords(x,"ProgramMethod")
       .map(x=>{
            val userId = LogTools.getKeywords(x,Comments.UserId)
            val ProgramID = if(LogTools.getKeywords(x,Comments.ProgramID) !="") new BigDecimal(LogTools.getKeywords(x,Comments.ProgramID)) else new BigDecimal(0)
            val playTime = LogTools.getKeywords(x,Comments.PlayS)
            val actors = vodProgramMap.value.get(ProgramID)
            val actor = if(null == actors || StringUtils.isBlank(actors)) "" else actors
         (userId,actor,playTime)
       }).filter(x=>StringUtils.isNoneBlank(x._2) && null !=x._2)
         .map(x=>{
              val listBuffer = ListBuffer[(String,String,Int)]()
              if(x._2.contains("|")){
                 for (actor <- x._2.split("\|")){
                    listBuffer +=((x._1,actor,x._3.toInt))
                 }
              }else{
                  listBuffer +=((x._1,x._2,x._3.toInt))
              }
             listBuffer.toList
         })
        listRDD.flatMap(x=>x).map(x=>{
              most_love_actors(x._1,x._2,x._3)
         }).toDF().registerTempTable("most_love_actors") // userId+"|"+actor+"|"+playtime
         val most_love_actors_df = sqlContext.sql("select '"+date+"' as date,userId,actor,sum(playtime) as playtime from most_love_actors group by userId,actor")
         most_love_actors_df.toDF("date","userId","actor","playtime").show(10)
         most_love_actors_df.map(r=>{
           (r.getAs[String]("userId"),(r.getAs[String]("actor"),r.getAs[Long]("playtime")))
         }).groupByKey().map(x=>{
           val i2 = x._2.toBuffer
           val i2_2 = i2.sortBy(_._2) //按时长从大到小排序
           if (i2_2.length > Comments.top_N_love_actor) i2_2.remove(0, (i2_2.length - Comments.top_N_love_actor))
           (x._1, i2_2.toIterable)
         }).flatMap(x => {
           val y = x._2
           for (w <- y) yield (x._1, w._1, w._2)
         }).toDF("userId","actor","playtime").registerTempTable("most_love_actors")
          val most_love_actors_df2=sqlContext.sql("select '"+date+"' as date,userId,actor,playtime from most_love_actors")
          SparkTools.writeDataframeToPhoenixHbase(most_love_actors_df2,Comments.hbase_t_user_most_love_actor)
   }

}