Spark中的爆炸结构答案

【问题标题】：Exploded Struct in SparkSpark中的爆炸结构
【发布时间】：2018-04-27 09:45:50
【问题描述】：

我有以下架构的 DataFrame：

 |-- data: struct (nullable = true)
 |    |-- asin: string (nullable = true)
 |    |-- customerId: long (nullable = true)
 |    |-- eventTime: long (nullable = true)
 |    |-- marketplaceId: long (nullable = true)
 |    |-- rating: long (nullable = true)
 |    |-- region: string (nullable = true)
 |    |-- type: string (nullable = true)
 |-- uploadedDate: long (nullable = true)

我想分解结构，使 asin、customerId、eventTime 等所有元素成为 DataFrame 中的列。我尝试了explode 函数，但它适用于Array 而不是struct 类型。是否可以将能够的数据框转换为以下数据框：

     |-- asin: string (nullable = true)
     |-- customerId: long (nullable = true)
     |-- eventTime: long (nullable = true)
     |-- marketplaceId: long (nullable = true)
     |-- rating: long (nullable = true)
     |-- region: string (nullable = true)
     |-- type: string (nullable = true)
     |-- uploadedDate: long (nullable = true)

【问题讨论】：

标签： hadoop apache-spark apache-spark-sql

【解决方案1】：

很简单：

val newDF = df.select("uploadedDate", "data.*");

您告诉选择uploadedDate，然后选择字段数据的所有子元素

例子：

scala> case class A(a: Int, b: Double)
scala> val df = Seq((A(1, 1.0), "1"), (A(2, 2.0), "2")).toDF("data", "uploadedDate")
scala> val newDF = df.select("uploadedDate", "data.*")
scala> newDF.show()
+------------+---+---+
|uploadedDate|  a|  b|
+------------+---+---+
|           1|  1|1.0|
|           2|  2|2.0|
+------------+---+---+

scala> newDF.printSchema()
root
 |-- uploadedDate: string (nullable = true)
 |-- a: integer (nullable = true)
 |-- b: double (nullable = true)

【讨论】：