Reputation: 2705
I have a dataset which is the output of a pipe in scalding that looks like this:
'Var1, 'Var2, 'Var3, 'Var4 =
a,x,1,2
a,y,3,4
b,x,1,2
b,y,3,4
I'm trying to turn it into something like:
'Var1, 'Var3x, 'Var4x, 'Var3y, 'Var4y =
a,1,2,3,4
b,1,2,3,4
First I thought using flatMap somehow would work but that didn't seem right. Seems like some use of pivot function should work, but I can't quite work out how to pivot multiple columns.
Any help is appreciated.
Upvotes: 0
Views: 529
Reputation: 40510
You need to combine your two value columns into one, and then you can use .pivot. Something like this:
case class v34(v3: Int, v4: Int)
pipe
.map(('Var3, 'Var4) -> ('V34)) { vars: (Int, Int) => v34(vars._1, vars._2) }
.groupBy('Var1) { _.pivot(('Var2, 'V34) => ('x, 'y)) }
.mapTo(('Var1, 'x, 'y) -> ('Var1, 'Var3x, 'Var4x, 'Var3y, 'Var4y) {
vars: (String,V34,V34) =>
val (key, xval, yval) = vars
(key, xval.v3, xval.v4, yval.v3, yval.v4)
}
.
Upvotes: 1