dataframe-aggs.scala

val df = spark.createDataFrame(Seq(("col1", "col2", 4, 5, 7, 5), 
                                   ("col1", "col2", 2, 0, 2, 2), 
                                   ("col1", "col2", 2, 0, 2, 2), 
                                   ("col1", "col1", 2, 0, 2, 2), 
                                   ("col1", "col1", 5, 10, 3, 4)))
.toDF("first_group", "second_group", "col1", "col2", "col3", "col4")

df.groupBy("first_group", "second_group").min()
  .groupBy("first_group").avg()

/*
first_group:string
avg(min(col1)):double
avg(min(col2)):double
avg(min(col3)):double
avg(min(col4)):double
*/