如何在 Spark SQL 中合并两行？

Question

我需要根据关键列“id”合并同一数据框中的行。在示例数据框中，1 行包含 id、name 和 age 的数据。另一行有id、name 和salary。具有相同键“id”的行必须在最终数据框中合并为一条记录。如果只有一个记录，也应该用空值 [Smith, and Jake] 显示它们，如下例所示。

计算需要在实时数据上进行，基于火花本机函数的解决方案将是理想的。我曾尝试根据年龄和城市列过滤记录以分隔数据框，并对 ID 执行左连接。但它的效率不是很高。寻找任何替代建议。提前致谢！

示例数据框

val inputDF= Seq(("100","John", Some(35),None)
,("100","John", None,Some("Georgia")),
("101","Mike", Some(25),None),
("101","Mike", None,Some("New York")),
("103","Mary", Some(22),None),
("103","Mary", None,Some("Texas")),
("104","Smith", Some(25),None),
("105","Jake", None,Some("Florida")))
.toDF("id","name","age","city")

输入数据框

+---+-----+----+--------+
|id |name |age |city    |
+---+-----+----+--------+
|100|John |35  |null    |
|100|John |null|Georgia |
|101|Mike |25  |null    |
|101|Mike |null|New York|
|103|Mary |22  |null    |
|103|Mary |null|Texas   |
|104|Smith|25  |null    |
|105|Jake |null|Florida |
+---+-----+----+--------+ 

预期输出数据帧

+---+-----+----+---------+
| id| name| age|     city|
+---+-----+----+---------+
|100| John|  35|  Georgia|
|101| Mike|  25| New York|
|103| Mary|  22|    Texas|
|104|Smith|  25|     null|
|105| Jake|null|  Florida|
+---+-----+----+---------+

Answer 1

使用first或last带有ignoreNulls标志的标准函数。

第一标准函数

val q = inputDF
  .groupBy("id", "name")
  .agg(first("age", ignoreNulls = true) as "age", first("city", ignoreNulls = true) as "city")
  .orderBy("id")

最后一个标准函数

val q = inputDF
  .groupBy("id","name")
  .agg(last("age", true) as "age", last("city") as "city")
  .orderBy("id")