Spark Scala Approx Percentile over group

spark_approx_percentile.scala

val a_s = Seq.fill(9)("a" -> 1):+ ("a" -> 10)
// a_s: Seq[(String, Int)] = List((a,1), (a,1), (a,1), (a,1), (a,1), (a,1), (a,1), (a,1), (a,1), (a,10))

val b_s = Seq.fill(9)("b" -> 2):+ ("b" -> 10)
// b_s: Seq[(String, Int)] = List((b,2), (b,2), (b,2), (b,2), (b,2), (b,2), (b,2), (b,2), (b,2), (b,10))

val df = (a_s ++ b_s).toDF("kind", "value")
// df: org.apache.spark.sql.DataFrame = [kind: string, value: int]

df.groupBy("kind").agg(expr("approx_percentile(value, 0.90, 20)").as("x_percentile")).show
"""
+----+------------+
|kind|x_percentile|
+----+------------+
|   a|           1|
|   b|           2|
+----+------------+
"""

// https://spark.apache.org/docs/latest/api/sql/index.html#approx_percentile