Wsrtjtyk

I have a json file which has two types of data for store and online.But datatypes of few columns are different ,but column names and schema are same.
I would like to do some operations on json like sorting,re-arranging using spray joson library in Scala.But I am not able to apply the case class on josn becuase datatype of "storeId" and "amount" are different.
Is there any way we can re-write case class to handle bothe integer and string ?

Example : I will have them in a file, read it line by line and assign it to string


Given inputJosn String :
{
"purchaseType": "1",
"purchaseChannel": "store",
"storeId": 6167,
"paymentType": [{
"type": "CASH",
"category": "Cash",
"amount": 3.91
}]
},
{
"purchaseType": "2",
"purchaseChannel": "online",
"storeId": "6168",
"paymentType": [{
"type": "Card",
"category": "Card",
"amount": "5.04"
}]
}


Code :

import spray.json._

import DefaultJsonProtocol._



case class payType(`type`:String,category:String,amount:String)

case class Transactions(purchaseType:String,purchaseChannel:String,storeId:String,paymentType: payType )



object MyJsonpayType extends DefaultJsonProtocol {

implicit val payTypeFormat: JsonFormat[payType] = jsonFormat3(payType)

}

import MyJsonpayType._



object MyJsonTransactions extends DefaultJsonProtocol {

implicit val TransactionsFormat: JsonFormat[Transactions] = jsonFormat4(Transactions)

}

import MyJsonTransactions._







inputJosn.parseJson.convertTo[Transactions]

It throws errror saying expecing JsString but got 6167.

I know why is it throwing error, but is these any way to re-writw case class in bettwer way to handle both integer and string.

You can get away with using Either, like

case class payType(`type`:String,category:String,amount:Either[Double,String]) 

But then you are simply kicking the can down the road. If you have to do any operation on the amount field like- sort, aggregation you would need to normalise the data.

If that's case I would better write custom de-serialization for the object with having a single type.

In these situations I like to use the Jackson library, but in this situation you could just call the storeId:string. After that you can parse it to long by using a companion object.

Try this:

case class Transactions (

  purchaseType:String,

  purchaseChannel:String,

  storeId:String,

  paymentType: payType ) {



  def storeIdToLong(t: Transaction) = TransactionsAsLong(

    purchaseType = t.purchaseType,

    purchaseChannel = t.purchaseChannel,

    storeId = t.storeId.toLong,

    paymentType = t.paymentType

  )

}



case class TransactionsAsLong (

  purchaseType:String,

  purchaseChannel:String,

  storeId:Long,

  paymentType: payType )





object Transactions {

  implicit def transactions_to_long(t: Transactions): TransactionsAsLong = t.storeIdToLong()

}

Now you can write val test: Transaction = TransactionsAsLong(...) if you need that.

I would suggest to write a custom deserializer instead :-)