比如原始表的schema如下：

image.png

現(xiàn)在想將該DataFrame 的schema轉(zhuǎn)換成:
id:String,
goods_name:String
price: Array<String>

1. sql 轉(zhuǎn)換
   spark.sql("create table speedup_tmp_test_spark_schema_parquet12 using parquet as select cast(id as string),cast(goods_name as string),cast(price as array<string>) from tmp_test_spark_schema_parquet")

2. case class 變換
   case class newSchemaClass(id: String, goods_name: String, price: Array[String])

// 原dataframe
val df = spark.sql("select * from tmp_test_spark_schema_parquet")

// 新dataframe
val newDF = df .rdd.map { r =>
newSchemaClass(r(0).toString, r(1).toString, r.getSeqInt.map(_.toString).toArray)
}.toDF()

// 獲取具體數(shù)據(jù)
newDF.collect()(2).getListString