Spark 实例

1. parquet

object testSparkReadParquet {
    def main(args: Array[String]): Unit = {
        var spark = SparkSession.builder().appName("TestSparkSession").master("local").getOrCreate()
        
        val df = spark.read.parquet("D:\tools\testSparkFile\users.parquet");
        
        df.printSchema()
        
        df.select("name","favorite_color","favorite_numbers").show()
        
         df.select("name","favorite_color").write.mode("overwrite").save("D:\tools\testSparkFile\namesAndFavColors.parquet")
         
         val df2 = spark.read.parquet("D:\tools\testSparkFile\namesAndFavColors.parquet");
        df2.printSchema()
    }
}

object TestParquet {
    def main(args: Array[String]): Unit = {
       val spark = SparkSession
                  .builder()
                  .appName("Java Spark SQL basic example")
                  .config("spark.some.config.option", "some-value")
                  .master("local")
                  .getOrCreate();
        import spark.implicits._
        val squaresDF = spark.sparkContext.makeRDD(1 to 5).map(i => (i, i * i)).toDF("value", "square")
        squaresDF.write.mode("append").parquet("D:\tools\testSparkFile\test\key=1")
        
        val cubesDF = spark.sparkContext.makeRDD(6 to 10).map(i => (i, i * i * i)).toDF("value", "cube")
        cubesDF.write.mode("append").parquet("D:\tools\testSparkFile\test\key=2")
        
        // Read the partitioned table
        val mergedDF = spark.read.option("mergeSchema", "true").parquet("D:\tools\testSparkFile\test\")
        mergedDF.select("value", "square","key").show()
        mergedDF.printSchema()
    }
}

2. DataFrame

object DFExample {
    case class Student(id:Int, name:String, phone:String, email:String, age:Int)
    def main(args: Array[String]): Unit = {
        val spark = SparkSession.builder().appName("DFExample").master("local").getOrCreate()
        import spark.implicits._
        val studentDF = spark.sparkContext.textFile("D:\tools\testSparkFile\dfTestFile.txt").map(_.split("\|"))
        .map(line =>Student(line(0).trim().toInt, line(1).trim(), line(2).trim(),line(3).trim(),line(4).trim().toInt)).toDF()
        
        val studentDF2 = studentDF
        
        studentDF.show()
        
        //过滤掉名字为空的数据
        //studentDF.filter("name !='' OR name != 'NULL' " ).show()  //没起作用
        studentDF.filter("name !='NULL'" ).filter("name !=''").show()  
        
        //找出名字以l开头的人
        studentDF.filter("substr(name,0,1)='l'").show()
        
//        spark.sql("show functions").show(2000)
        import org.apache.spark.sql.functions._
        import spark.implicits._
        
        //按照名字排序操作
        studentDF.sort("name").show()
        studentDF.sort(studentDF.col("name").desc).show()
        
        //队列进行重新命名
        studentDF.select(studentDF.col("name").as("student_name")).show()
        
        //join操作
        studentDF.join(studentDF2,studentDF.col("id") === studentDF2.col("id")).sort(studentDF.col("id")).show()
    }
}

object ScalaREPL {
    case class REPL(id_1:Int,id_2:Int,cmp_fname_c1:Double,cmp_fname_c2:String,cmp_lname_c1:Double,cmp_lname_c2:String,
        cmp_sex:Double,cmp_bd:Double,cmp_bm:Double,cmp_by:Double,cmp_plz:Double,is_match:Boolean)
    case class REPL2(summary:String,cmp_fname_c1:Double,cmp_fname_c2:Double)
    def main(args: Array[String]): Unit = {
        val spark = SparkSession.builder().appName("DFExample").master("local").getOrCreate()
        import spark.implicits._
        import org.apache.spark.sql.functions._
        val df = spark.sparkContext.textFile("D:\tools\testSparkFile\scalaREPL.txt").filter(_.length() != 0).map(_.split(","))
        .map(line =>REPL(line(0).trim().toInt,line(1).trim().toInt,line(2).trim().toDouble,line(3).trim(),
            line(4).trim().toDouble,line(5).trim(),line(6).trim().toDouble,line(7).trim().toDouble,
            line(8).trim().toDouble,line(9).trim().toDouble,line(10).trim().toDouble,line(11).trim().toBoolean)).toDF()
            
          df.show()
//        df.cache()    
//        df.rdd.map(_.getAs[Boolean]("is_match")).countByValue().foreach(println)
//        df.select(col("is_match")).groupBy(col("is_match")).count().show()
//          df.groupBy(col("is_match")).count().orderBy(col("is_match").desc).show()
          
//          df.agg(count(col("cmp_fname_c1")),mean(col("cmp_fname_c1")), stddev(col("cmp_fname_c1")), min(col("cmp_fname_c1")), max(col("cmp_fname_c1"))).show()
//          
//          df.createOrReplaceTempView("linkage")
          
//          spark.sql("""
//            select is_match, count(*) count 
//            from linkage 
//            group by is_match 
//            order by is_match desc
//            """).show()
          
          val matches = df.where("is_match = true")
          matches.createOrReplaceTempView("metch_desc")
          matches.show()
//          val matchSummary = matches.distinct()
          val misses = df.filter("is_match = false")
          misses.createOrReplaceTempView("miss_desc")
//          val missSummary = misses.distinct()
          
          
            //长表转宽表
        /*    val df2 = spark.sparkContext.textFile("D:\tools\testSparkFile\scalaREPL2.txt").filter(_.length() != 0).map(_.split(","))
            .map(line =>REPL2(line(0).trim(),line(1).trim().toDouble,line(2).trim().toDouble)).toDF()
            df2.show()
            val schema = df2.schema
            val ds = df2.flatMap(row =>{
              val metric = row.getString(0)
              (1 until row.size).map(i =>{
                (metric, schema(i).name, row.getDouble(i))
              })
            })
            val longDF = ds.toDF("metric","field","value")
            longDF.show()
            longDF.groupBy("field").pivot("metric",Seq("count","mean","suddev","max","min")).agg(first("value")).show()*/
            
            
            
    }
}

 3.  //dataSet 转 DataFrame  使用内置函数

object TestMachineLearning {
    def main(args: Array[String]): Unit = {
        val spark = SparkSession.builder().appName("DFExample").master("local").getOrCreate()
        import spark.implicits._
        val rawUserArtistData = spark.read.textFile("D:\tools\testSparkFile\profiledata\user_artist_data.txt")
        rawUserArtistData.take(5).foreach(println)
        
        //dataSet 转 DataFrame  使用内置函数
        val userArtistDF = rawUserArtistData.map(line =>{
          val Array(user,artist, _*) = line.split(" ")
          (user, artist.toInt)
        }).toDF("user","artist")
        import org.apache.spark.sql.functions._
        userArtistDF.agg(max("user"),min("user"),max("artist"),min("artist")).show()
    }
}