你可以用 groupBy 以及 collect_list 将相关列“合并”到单个嵌套结构中后：

val input: DataFrame = Seq(
  (1, "JAMES", "SMITH"),
  (2, "MARY", "BROWN"),
  (2, "DAVID", "WILLIAMS"),
  (1, "ROBERT", "DAVIS")
).toDF("id", "first_name", "last_name")

import org.apache.spark.sql.functions._
val result = input
  .withColumn("entity", struct($"first_name", $"last_name"))
  .groupBy("id").agg(collect_list($"entity"))

result.show(false)
// +---+--------------------------------+
// |id |entities                        |
// +---+--------------------------------+
// |1  |[[JAMES,SMITH], [ROBERT,DAVIS]] |
// |2  |[[MARY,BROWN], [DAVID,WILLIAMS]]|
// +---+--------------------------------+

result.printSchema()
// root
//  |-- id: integer (nullable = false)
//  |-- entities: array (nullable = true)
//  |    |-- element: struct (containsNull = true)
//  |    |    |-- first_name: string (nullable = true)
//  |    |    |-- last_name: string (nullable = true)