Merging Scala Spark sql schemas

Question

I'm trying to merge two dataframes and create a dataframe with a new column containing the other dataframe as an array. Does anyone know how this can be implemented in scala?

//Schema 1

PRIM_KEY: decimal(20,0) (nullable = true)
|-- SOME_DECIMAL: decimal(20,0) (nullable = true)
|-- SOME_INTEGER: integer (nullable = true)

//Schema 2

PRIM_KEY: decimal(20,0) (nullable = true)
|-- COLUMN1: string (nullable = false)
|-- COLUMN2: string (nullable = false)

//Resulting schema

RIM_KEY: decimal(20,0) (nullable = true)
|-- SOME_DECIMAL: decimal(20,0) (nullable = true)
|-- SOME_INTEGER: integer (nullable = true)
|-- an_array: array (nullable = true)
|    |-- element: String (containsNull = false)

Answer 1

One approach would be to create an UDF that combines two lists into one, perform a groupBy on the joined dataframes, and apply the UDF like in the following:

val df1 = Seq(
  (1, 100.1, 10),
  (2, 200.2, 20)
).toDF("pk", "col1", "col2")

val df2 = Seq(
  (1, "a1", "b1"),
  (1, "c1", "d1"),
  (2, "a2", "b2")
).toDF("pk", "str_col1", "str_col2")

def combineLists = udf(
  (a: Seq[String], b: Seq[String]) => a ++ b
)

val df3 = df1.join(df2, Seq("pk")).
  groupBy(df1("pk"), df1("col1"), df1("col2")).agg(
    combineLists(collect_list(df2("str_col1")), collect_list(df2("str_col2"))).alias("arr_col")
  ).
  select(df1("pk"), df1("col1"), df1("col2"), col("arr_col"))

df3.show
+---+-----+----+----------------+
| pk| col1|col2|         arr_col|
+---+-----+----+----------------+
|  1|100.1|  10|[c1, a1, d1, b1]|
|  2|200.2|  20|        [a2, b2]|
+---+-----+----+----------------+

Answer 2

The result you are seeking:

RIM_KEY: decimal(20,0) (nullable = true)
|-- SOME_DECIMAL: decimal(20,0) (nullable = true)
|-- SOME_INTEGER: integer (nullable = true)
|-- an_array: array (nullable = true)
|    |-- element: String (containsNull = false)

Let me tell you first:

array (nullable = true) is not a datatype but its a data structure. So you can't at all define schema as DataType Array.

One way is to concatenate with string using concat_ws and perform a withcolumn operation on the 2nd dataset.

EG:

val tmpDf = test2Df.select(concat_ws(",", col("NAME"), col("CLASS")).as("ARRAY_COLUMN"))
val mergedDf = test1Df.withColumn("ARRAY_COLUMN",tmpDf.col("ARRAY_COLUMN"))

I don't understand what is your use case to use an array type as schema but instead you can use the concatenated result and convert into Array.

Hope this helps you, i know am bit late here to answer but still I'll be happy if it helps you even now.