Convert spark decision tree model debug string to nested JSON in scala - json

Question

Similar to the tree json parsing quoted here, I am trying to implement a simple visualization of decision trees in scala. It is exactly same as the display method available in databricks notebooks.
I am new to scala and struggling to get the logic right. I understand we have to make recursive calls to build the children and break when the final prediction values are shown. i have attempted a code here using the below mentioned input model debug string
def getStatmentType(x: String): (String, String) = {
val ifPattern = "If+".r
val ifelsePattern = "Else+".r
var t = ifPattern.findFirstIn(x.toString)
if(t != None){
("If", (x.toString).replace("If",""))
}else {
var ts = ifelsePattern.findFirstIn(x.toString)
if(ts != None) ("Else", (x.toString).replace("Else", ""))
else ("None", (x.toString).replace("(", "").replace(")",""))
}
}
def delete[A](test:List[A])(i: Int) = test.take(i) ++ test.drop((i+1))
def BuildJson(tree:List[String]):List[Map[String, Any]] = {
var block:List[Map[String, Any]] = List()
var lines:List[String] = tree
loop.breakable {
while (lines.length > 0) {
println("here")
var (cond, name) = getStatmentType(lines(0))
println("initial" + cond)
if (cond == "If") {
println("if" + cond)
// lines = lines.tail
lines = delete(lines)(0)
block = block :+ Map("if-name" -> name, "children" -> BuildJson(lines))
println("After pop Else State"+lines(0))
val (p_cond, p_name) = getStatmentType(lines(0))
// println(p_cond + " = "+ p_name+ "\n")
cond = p_cond
name = p_name
println(cond + " after="+ name+ "\n")
if (cond == "Else") {
println("else" + cond)
lines = lines.tail
block = block :+ Map("else-name" -> name, "children" -> BuildJson(lines))
}
}else if( cond == "None") {
println(cond + "NONE")
lines = delete(lines)(0)
block = block :+ Map("predict" -> name)
}else {
println("Finaly Break")
println("While loop--" +lines)
loop.break()
}
}
}
block
}
def treeJson1(str: String):JsValue = {
val str = "If (feature 0 in {1.0,2.0,3.0,4.0,5.0,6.0,7.0,8.0,10.0,11.0,12.0,13.0})\n If (feature 0 in {6.0})\n Predict: 17.0\n Else (feature 0 not in {6.0})\n Predict: 6.0\n Else (feature 0 not in {1.0,2.0,3.0,4.0,5.0,6.0,7.0,8.0,10.0,11.0,12.0,13.0})\n Predict: 20.0"
val x = str.replace(" ","")
val xs = x.split("\n").toList
var js = BuildJson(xs)
println(MapReader.mapToJson(js))
Json.toJson("")
}
Expected output:
[
{
'name': 'Root',
'children': [
{
'name': 'feature 0 in {1.0,2.0,3.0,4.0,5.0,6.0,7.0,8.0,10.0,11.0,12.0,13.0}',
'children': [
{
'name': 'feature 0 in {6.0}',
'children': [
{
'name': 'Predict: 17.0'
}
]
},
{
'name': 'feature 0 not in {6.0}',
'children': [
{
'name': 'Predict: 6.0'
}
]
}
]
},
{
'name': 'feature 0 not in {1.0,2.0,3.0,4.0,5.0,6.0,7.0,8.0,10.0,11.0,12.0,13.0}',
'children': [
{
'name': 'Predict: 20.0'
}
]
}
]

Answer 1

you don`t need to parse the debugstring, instead, you can parse from the rootnode of the model.
refer to enter link description here