i2i样本制作，第3步。做枚举值

src/main/scala/com/aliyun/odps/spark/examples/makedata_dssm/makedata_i2i_03_onehotFile_20241128.scala

@@ -0,0 +1,119 @@
+package com.aliyun.odps.spark.examples.makedata_dssm
+
+import com.alibaba.fastjson.JSON
+import com.aliyun.odps.TableSchema
+import com.aliyun.odps.data.Record
+import com.aliyun.odps.spark.examples.myUtils.{MyDateUtils, MyHdfsUtils, ParamUtils, env}
+import org.apache.hadoop.io.compress.GzipCodec
+import org.apache.spark.sql.SparkSession
+import scala.collection.mutable
+import scala.collection.JavaConversions._
+import scala.collection.mutable.ArrayBuffer
+
+object makedata_i2i_03_onehotFile_20241128 {
+  def func(record: Record, schema: TableSchema): Record = {
+    record
+  }
+  def main(args: Array[String]): Unit = {
+    val spark = SparkSession
+      .builder()
+      .appName(this.getClass.getName)
+      .getOrCreate()
+    val sc = spark.sparkContext
+
+    // 1 读取参数
+    val param = ParamUtils.parseArgs(args)
+    val tablePart = param.getOrElse("tablePart", "64").toInt
+    val dt = param.getOrElse("dt", "20240620")
+    val readPath = param.getOrElse("readPath", "/dw/recommend/model/53_dssm_i2i_onehot/20240101")
+    val savePath = param.getOrElse("savePath", "/dw/recommend/model/53_dssm_i2i_onehot/20250101")
+    val project = param.getOrElse("project", "loghubods")
+    val repartition = param.getOrElse("repartition", "100").toInt
+    val ifDebug = param.getOrElse("ifDebug", "false").toBoolean
+    // 2 读取odps+表信息
+    val odpsOps = env.getODPS(sc)
+    // 3 vid中的sparse特征
+    if (MyHdfsUtils.hdfs_exits(readPath)){
+      val onehot = sc.textFile(readPath).map(r => {
+        val rList = r.split("\t")
+        (rList(0), rList(1))
+      })
+      val onehotMapOldStatic = onehot.map(_._1.split(":")(0)).map(r => (r, 1)).reduceByKey(_ + _).collect().sortBy(-_._2)
+      print(s"读入路径:$readPath \n数据量:" + onehot.count())
+      print("打印各个特征多少枚举值：")
+      onehotMapOldStatic.foreach(r => println(r.productIterator.mkString("\t")))
+    }
+    val onehotMap = if (MyHdfsUtils.hdfs_exits(readPath)){
+      mutable.Map(sc.textFile(readPath).map(r => {
+        val rList = r.split("\t")
+        (rList(0), rList(1))
+      }).collectAsMap().toSeq: _*)
+    }else{
+      mutable.Map[String, String]()
+    }
+
+    val vidStaticFeature = odpsOps.readTable(project = project,
+        table = "t_vid_tag_feature",
+        partition = s"dt=$dt",
+        transfer = func,
+        numPartition = tablePart)
+      .map(record => {
+        val vid = record.getString("vid")
+        val feature = record.getString("feature")
+        (vid, feature)
+      }).flatMap{
+        case (vid, feature) =>
+          val result = new ArrayBuffer[String]()
+          result.add("vid:" + vid)
+          JSON.parseObject(feature).foreach(r =>{
+            val value = if (r._2 == null) "无" else r._2.toString
+            r._1 match {
+              case "category1" => result += "cate1:" + value
+              case "category2_1" => result += "cate2:" + value
+              case "category2_2" => result += "cate2:" + value
+              case "category2_3" => result += "cate2:" + value
+              case "valid_time" => result += "valid_time:" + value
+              case " timeliness" => result += "timeliness:" + value
+              case "sentiment_tendency" => result += "sentiment_tendency:" + value
+              case "has_end_credit_guide" => result += "has_end_credit_guide:" + value
+              case "background_music_type" => result += "background_music_type:" + value
+              case "cover_persons_num" => result += "cover_persons_num:" + value
+              case "captions" => result += "captions:" + value
+              case "captions_color" => result += "captions_color:" + value
+              case "audience_value_type" => result += "audience_value_type:" + value
+              case "audience_gender" => result += "audience_gender:" + value
+              case "audience_age_group" => result += "audience_age_group:" + value
+              case "font_size" => result += "font_size:" + value
+              case "video_type" => result += "video_type:" + value
+              case "video_style" => result += "video_style:" + value
+              case _ =>
+            }
+          })
+          result.distinct
+      }.distinct().collect()
+    val maxEnumMap = mutable.Map[String, Int]()
+    onehotMap.foreach { case (key, value) =>
+      val prefix = key.split(":")(0)
+      val currentMax = maxEnumMap.getOrElse(prefix, 0)
+      maxEnumMap(prefix) = Math.max(currentMax, value.toInt)
+    }
+    vidStaticFeature.foreach { feature =>
+      val prefix = feature.split(":")(0)
+      if (!onehotMap.contains(feature)) {
+        val newEnumValue = maxEnumMap.getOrElse(prefix, 0) + 1
+        maxEnumMap(prefix) = newEnumValue
+        onehotMap(feature) = newEnumValue.toString
+      }
+    }
+
+    val dataRdd = sc.parallelize(onehotMap.toSeq.map(_.productIterator.mkString("\t")))
+    val hdfsPath = savePath
+    if (hdfsPath.nonEmpty && hdfsPath.startsWith("/dw/recommend/model/")) {
+      println("删除路径并开始数据写入:" + hdfsPath)
+      MyHdfsUtils.delete_hdfs_path(hdfsPath)
+      dataRdd.repartition(10).saveAsTextFile(hdfsPath, classOf[GzipCodec])
+    } else {
+      println("路径不合法，无法写入:" + hdfsPath)
+    }
+  }
+}