Scala - Как преобразовать JSON ключи и значения в столбцы

Question

Как разобрать ниже Ввод Json в столбцы ключ и значение. Любая помощь приветствуется.

Ввод:

{
"name" : "srini",
"value": {
"1" : "val1",
"2" : "val2",
"3" : "val3"
}
}

    Output DataFrame Column:

    name      key        value
    -----------------------------
    srini      1         val1
    srini      2         val2
    srini      3         val3



        //++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++Input DataFrame :
        +--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
        |json_file                                                                                                                                                                                                                                                                                                     |
        +--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
        |{"file_path":"AAA/BBB.CCC.zip","file_name":"AAA_20200202122754.json","received_time":"2020-03-31","obj_cls":"Monitor","obj_cls_inst":"Monitor","relation_tree":"Source~>HD_Info~>Monitor","s_tag":"ABC1234","Monitor":{"Index":"0","Vendor_Data":"58F5Y","Monitor_Type":"Lenovo Monitor","HnfoID":"650FEC74"}}| 
        +--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+


        How to convert this above json file in a DataFrame like below :

        +----------------+-----------------------+--------------+--------+-------------+-------------------------+----------+----------------+----------------+
        |file_path       |file_name              |received_time |obj_cls |obj_cls_inst |relation_tree            |s_tag     |attribute_name  |attribute_value |
        +----------------+-----------------------+--------------+--------+-------------+-------------------------+----------+----------------+----------------+
        |AAA/BBB.CCC.zip |AAA_20200202122754.json|2020-03-31    |Monitor |Monitor      |Source~>HD_Info~>Monitor |ABC1234   |Index           |0               |
        +----------------+-----------------------+--------------+--------+-------------+-------------------------+----------+----------------+----------------+
        |AAA/BBB.CCC.zip |AAA_20200202122754.json|2020-03-31    |Monitor |Monitor      |Source~>HD_Info~>Monitor |ABC1234   |Vendor_Data     |58F5Y           |
        +----------------+-----------------------+--------------+--------+-------------+-------------------------+----------+----------------+----------------+
        |AAA/BBB.CCC.zip |AAA_20200202122754.json|2020-03-31    |Monitor |Monitor      |Source~>HD_Info~>Monitor |ABC1234   |Monitor_Type    |Lenovo Monitor  |
        +----------------+-----------------------+--------------+--------+-------------+-------------------------+----------+----------------+----------------+
        |AAA/BBB.CCC.zip |AAA_20200202122754.json|2020-03-31    |Monitor |Monitor      |Source~>HD_Info~>Monitor |ABC1234   |HnfoID          |650FEC74        |
        +----------------+-----------------------+--------------+--------+-------------+-------------------------+----------+----------------+----------------+

//**********************************************
val rawData = sparkSession.sql("select 1").withColumn("obj_cls", lit("First")).withColumn("s_tag", lit("S_12345")).withColumn("jsonString", lit("""{"id":""1,"First":{"Info":"ABCD123","Res":"5.2"}}"""))

Answer 1

Как только вы загрузите json в DF следующим образом:

+-----+------------------+
| name|             value|
+-----+------------------+
|srini|[val1, val2, val3]|
+-----+------------------+

Сначала вы выбираете целые значения элементов:

df.select($"name", $"value.*")

Это даст вам следующее:

+-----+----+----+----+
| name|   1|   2|   3|
+-----+----+----+----+
|srini|val1|val2|val3|
+-----+----+----+----+

Затем вам нужно повернуть столбцы, чтобы они стали строками, для этого я обычно определяю вспомогательную функцию kv:

 def kv (columnsToTranspose: Array[String]) = explode(array(columnsToTranspose.map {
    c => struct(lit(c).alias("k"), col(c).alias("v"))
  }: _*))

Затем вы создаете массив для нужных столбцов:

val pivotCols = Array("1", "2", "3")

И, наконец, примените функцию к предыдущему DF:

df.select($"name", $"value.*")
.withColumn("kv", kv(pivotCols))
.select($"name", $"kv.k" as "key", $"kv.v" as "value")

Результат:

+-----+---+-----+
| name|key|value|
+-----+---+-----+
|srini|  1| val1|
|srini|  2| val2|
|srini|  3| val3|
+-----+---+-----+

РЕДАКТИРОВАТЬ

Если вы не хотите вручную указывать столбцы для поворота, вы можете использовать промежуточный df следующим образом:

val dfIntermediate = df.select($"name", $"value.*")

dfIntermediate.withColumn("kv", kv(dfIntermediate.columns.tail))
.select($"name", $"kv.k" as "key", $"kv.v" as "value")

И вы получите тот же результат:

+-----+---+-----+
| name|key|value|
+-----+---+-----+
|srini|  1| val1|
|srini|  2| val2|
|srini|  3| val3|
+-----+---+-----+

EDIT2

С новым примером то же самое, вам просто нужно изменить, какие столбцы вы читаете / сводите

val pivotColumns = Array("HnfoId", "Index", "Monitor_Type", "Vendor_Data")

df.select("file_path", "file_name", "received_time", "obj_cls", "obj_cls_inst", "relation_tree", "s_Tag", "Monitor.*").withColumn("kv", kv(pivotColumns)).select($"file_path", $"file_name", $"received_time", $"obj_cls", $"obj_cls_inst", $"relation_tree", $"s_Tag", $"kv.k" as "attribute_name", $"kv.v" as "attribute_value").show
+---------------+--------------------+-------------+-------+------------+--------------------+-------+--------------+---------------+
|      file_path|           file_name|received_time|obj_cls|obj_cls_inst|       relation_tree|  s_Tag|attribute_name|attribute_value|
+---------------+--------------------+-------------+-------+------------+--------------------+-------+--------------+---------------+
|AAA/BBB.CCC.zip|AAA_2020020212275...|   2020-03-31|Monitor|     Monitor|Source~>HD_Info~>...|ABC1234|        HnfoId|       650FEC74|
|AAA/BBB.CCC.zip|AAA_2020020212275...|   2020-03-31|Monitor|     Monitor|Source~>HD_Info~>...|ABC1234|         Index|              0|
|AAA/BBB.CCC.zip|AAA_2020020212275...|   2020-03-31|Monitor|     Monitor|Source~>HD_Info~>...|ABC1234|  Monitor_Type| Lenovo Monitor|
|AAA/BBB.CCC.zip|AAA_2020020212275...|   2020-03-31|Monitor|     Monitor|Source~>HD_Info~>...|ABC1234|   Vendor_Data|          58F5Y|
+---------------+--------------------+-------------+-------+------------+--------------------+-------+--------------+---------------+