How to create schema for Spark SQL for Array of struct?

Question

How to create a schema for the below json to read schema. I am using hiveContext.read.schema().json("input.json"), and I want to ignore the first two "ErrorMessage" and "IsError" read only Report. Below is the JSON:

 {  
  "ErrorMessage": null,
  "IsError": false,
   "Report":{  
      "tl":[  
         {  
            "TlID":"F6",
            "CID":"mo"
         },
         {  
            "TlID":"Fk",
            "CID":"mo"
         }
      ]
   }
}

I created the below schema :

val schema = StructType(
            Array(
                   StructField("Report", StructType(
                     Array(
                                                     StructField
                                                     ("tl",ArrayType(StructType(Array(
                                                                    StructField("TlID", StringType),
                                                                    StructField("CID", IntegerType)
                                                                  )))))))))

Below is my json.printSchema() :
root
 |-- Report: struct (nullable = true)
 |    |-- tl: array (nullable = true)
 |    |    |-- element: struct (containsNull = true)
 |    |    |    |-- TlID: string (nullable = true)
 |    |    |    |-- CID: integer (nullable = true)

score 3 · Accepted Answer · answered Nov 15 '17 at 16:09

The schema is incorrect. CID in your data is clearly not String ("mo"). Use

val schema = StructType(Array(
  StructField("Report", StructType(
    Array(
        StructField
        ("tl",ArrayType(StructType(Array(
                       StructField("CID", StringType),
                       StructField("TlID", StringType)
                          )))))))))

and:

val df = Seq("""{  
  "ErrorMessage": null,
  "IsError": false,
   "Report":{  
      "tl":[  
         {  
            "TlID":"F6",
            "CID":"mo"
         },
         {  
            "TlID":"Fk",
            "CID":"mo"
         }
      ]
   }
}""").toDS

spark.read.schema(schema).json(df).show(false)
+--------------------------------+
|Report                          |
+--------------------------------+
|[WrappedArray([mo,F6], [mo,Fk])]| 
+--------------------------------+

I changed the CID to StringType, but when I do show(), it is returning "null" — Divya, Nov 15 '17 at 17:55

score -1 · Answer 2 · answered Feb 10 '20 at 16:06

Datatype: array<struct<metrics_name:string,metrics_value:string>>


  import org.apache.spark.sql.types.{ArrayType}

  StructField("usage_metrics", ArrayType(StructType(
    Array(
      StructField("metric_name", StringType, true),
      StructField("metric_value", StringType, true)
    )
  ))))

How to create schema for Spark SQL for Array of struct?

2 Answers2