scala - 使用 Spark 窗口函数计算移动平均值时丢弃前几个值

Question

我正在尝试计算按名称分组的列的季度移动平均值，并且我已将 Spark 窗口函数规范定义为

val wSpec1 = Window.partitionBy("name").orderBy("date").rowsBetween(-2, 0)

我的 DataFrame 如下所示:

+-----+----------+-----------+------------------+
| name|      date|amountSpent|         movingAvg|
+-----+----------+-----------+------------------+
|  Bob|2016-01-01|       25.0|              25.0|
|  Bob|2016-02-02|       25.0|              25.0|
|  Bob|2016-03-03|       25.0|              25.0|
|  Bob|2016-04-04|       29.0|26.333333333333332|
|  Bob|2016-05-06|       27.0|              27.0|
|Alice|2016-01-01|       50.0|              50.0|
|Alice|2016-02-03|       45.0|              47.5|
|Alice|2016-03-04|       55.0|              50.0|
|Alice|2016-04-05|       60.0|53.333333333333336|
|Alice|2016-05-06|       65.0|              60.0|
+-----+----------+-----------+------------------+

为每个名称组突出显示第一个准确计算的值。我想用一些字符串替换前两个值，比如 NULL。由于我对 Spark/Scala 的了解有限，我考虑过从 DataFrame 中提取此列并在 Scala 中使用 patch 函数。但是，我无法弄清楚如何像第二个名称组的开头那样每隔一段时间替换这些值。这是我的代码:

import com.datastax.spark.connector._
import org.apache.spark.SparkContext
import org.apache.spark.SparkContext._
import org.apache.spark.SparkConf
import org.apache.spark.sql._
import org.apache.spark.sql.types.StructType
import org.apache.spark.sql.types.StructField
import org.apache.spark.sql.types.StringType
import org.apache.spark.sql.types.IntegerType
import org.apache.spark.sql.types.DoubleType
import org.apache.spark.sql.expressions.Window
import org.apache.spark.sql.functions._
import org.apache.spark.sql.Row
import org.apache.spark.sql.types._
object Test {

  def main(args: Array[String]) {
    //val sparkSession = SparkSession.builder.master("local").appName("Test").config("spark.cassandra.connection.host", "localhost").config("spark.driver.host", "localhost").getOrCreate()
    val sparkSession = SparkSession.builder.master("local").appName("Test").config("spark.cassandra.connection.host", "localhost").config("spark.driver.host", "localhost").getOrCreate()
    val sc = sparkSession.sparkContext

    val sqlContext = new org.apache.spark.sql.SQLContext(sc)
    import sparkSession.implicits._

    val customers = sc.parallelize(List(("Alice", "2016-01-01", 50.00),
      ("Alice", "2016-02-03", 45.00),
      ("Alice", "2016-03-04", 55.00),
      ("Alice", "2016-04-05", 60.00),
      ("Alice", "2016-05-06", 65.00),
      ("Bob", "2016-01-01", 25.00),
      ("Bob", "2016-02-02", 25.00),
      ("Bob", "2016-03-03", 25.00),
      ("Bob", "2016-04-04", 29.00),
      ("Bob", "2016-05-06", 27.00))).toDF("name", "date", "amountSpent")

    import org.apache.spark.sql.expressions.Window
    import org.apache.spark.sql.functions._

    // Create a window spec.
    val wSpec1 = Window.partitionBy("name").orderBy("date").rowsBetween(-2, 0)

    val ls=customers.withColumn("movingAvg",avg(customers("amountSpent")).over(wSpec1))
    ls.show()

  }
}

Answer 1

如果窗口正好包含 3 行(即跨越整个范围 -2 到 0)，我建议只计算平均值

val ls=customers
.withColumn("count",count(($"amountSpent")).over(wSpec1))
.withColumn("movingAvg",when($"count"===3,avg(customers("amountSpent")).over(wSpec1)))

ls.show()


+-----+----------+-----------+-----+------------------+
| name|      date|amountSpent|count|         movingAvg|
+-----+----------+-----------+-----+------------------+
|  Bob|2016-01-01|       25.0|    1|              null|
|  Bob|2016-02-02|       25.0|    2|              null|
|  Bob|2016-03-03|       25.0|    3|              25.0|
|  Bob|2016-04-04|       29.0|    3|26.333333333333332|
|  Bob|2016-05-06|       27.0|    3|              27.0|
|Alice|2016-01-01|       50.0|    1|              null|
|Alice|2016-02-03|       45.0|    2|              null|
|Alice|2016-03-04|       55.0|    3|              50.0|
|Alice|2016-04-05|       60.0|    3|53.333333333333336|
|Alice|2016-05-06|       65.0|    3|              60.0|
+-----+----------+-----------+-----+------------------+