Setting parameters in scala classes

I am trying to get familiar with classes in Scala(spark) and I wrote a simple wrapper class over XGBoostClassifier to experiment with. Once that wrapper class was ready, I tried setting some parameters but I am getting errors and can't find much help online

This is the class I am using

class trialXGBClassifier{
  var xgb_cls_object: XGBoostClassifier = _
}

I also wrote a small utility function to convert JSON to a Map (string, string)

def jsonToMap(jsonString: String): Map[String, String] = {
  val jsonMap = JSON.parseFull(jsonString).get.asInstanceOf[Map[String, String]]
  return jsonMap
  }

And this is the code that's giving me the error

var params = """{"numRound": "10"}"""
var trial_xgb = new trialXGBClassifier
trial_xgb.xgb_cls_object.setSeed(jsonToMap(params)("numRound").toLong)

This is the error message I get

java.lang.NullPointerException
  ... 61 elided

I tried doing the following using the actual ml.dmlc class and it works

var xgb_orig = new XGBoostClassifier
xgb_orig.setSeed(jsonToMap(params)("numRound").toLong)
xgb_orig.getSeed #10

Whats the reason behind this error and how do I fix this?

In Scala, people try to use val instead of var in all places where such usage possible. Also, they try to use case class or sealed trait instead of just class.
So, there's a very nice way to avoid such kind of issues using case class:

case class TrialXGBClassifier(xgb_cls_object: XGBoostClassifier)

And then the best option to use your wrapper only when you have something to wrap:

val xgb_orig = new XGBoostClassifier
xgb_orig.setSeed(jsonToMap(params)("numRound").toLong)
val v = TrialXGBClassifier(xgb_orig)

And you even can omit xgb_orig declaration:

val v = TrialXGBClassifier(new XGBoostClassifier)
v.xgb_cls_object.setSeed(jsonToMap(params)("numRound").toLong)