Apache Spark 将多行连接成单行列表

Question

我需要从源表创建一个表（hive 表/spark 数据框），该表将多行用户的数据存储到单行列表中。

User table:
Schema:  userid: string | transactiondate:string | charges: string |events:array<struct<name:string,value:string>> 
----|------------|-------| ---------------------------------------
123 | 2017-09-01 | 20.00 | [{"name":"chargeperiod","value":"this"}]
123 | 2017-09-01 | 30.00 | [{"name":"chargeperiod","value":"last"}]
123 | 2017-09-01 | 20.00 | [{"name":"chargeperiod","value":"recent"}]
123 | 2017-09-01 | 30.00 | [{"name":"chargeperiod","value":"0"}]
456 | 2017-09-01 | 20.00 | [{"name":"chargeperiod","value":"this"}]
456 | 2017-09-01 | 30.00 | [{"name":"chargeperiod","value":"last"}]
456 | 2017-09-01 | 20.00 | [{"name":"chargeperiod","value":"recent"}]
456 | 2017-09-01 | 30.00 | [{"name":"chargeperiod","value":"0"}]

输出表应该是

userid:String | concatenatedlist :List[Row]
-------|-----------------
123    | [[2017-09-01,20.00,[{"name":"chargeperiod","value":"this"}]],[2017-09-01,30.00,[{"name":"chargeperiod","value":"last"}]],[2017-09-01,20.00,[{"name":"chargeperiod","value":"recent"}]], [2017-09-01,30.00, [{"name":"chargeperiod","value":"0"}]]]
456    | [[2017-09-01,20.00,[{"name":"chargeperiod","value":"this"}]],[2017-09-01,30.00,[{"name":"chargeperiod","value":"last"}]],[2017-09-01,20.00,[{"name":"chargeperiod","value":"recent"}]], [2017-09-01,30.00, [{"name":"chargeperiod","value":"0"}]]]

星火版本：1.6.2

Answer 1

Seq(("1", "2017-02-01", "20.00", "abc"),
  ("1", "2017-02-01", "30.00", "abc2"),
  ("2", "2017-02-01", "20.00", "abc"),
  ("2", "2017-02-01", "30.00", "abc"))
.toDF("id", "date", "amt", "array")

df.withColumn("new", concat_ws(",", $"date", $"amt", $"array"))
  .select("id", "new")
  .groupBy("id")
  .agg(concat_ws(",", collect_list("new")))