CODEWITHSUNDEEP
scalacurrying
xiaojia zhang
xiaojia zhang

Reputation: 1058

Scala: How to design the high order function?

I want to design a high order function in scala which may looks like flowing:

def process(data: Seq[Double], costFun: **): Double

costFun is a function which can be used to calculate the cost of a method, since I have serval cost functions, which may have different signatures, like:

def costGauss(data: Seq[Double], scalaShift: Boolean): Double

def costKernal(data: Seq[Double], theta: Int): Double

how should I design the process function to enable that cost functions which have different signatures can be passed to it as parameter costFun?

Upvotes: 0

Views: 63

Answers (1)

Andrey Tyukin
Andrey Tyukin

Reputation: 44918

Seems that you need just Seq[Double] => Double there:

def processData(data: Seq[Double], lossFunc: Seq[Double] => Double): Double = ???

def lossGauss(data: Seq[Double], scalaShift: Boolean): Double = ???
def lossKernel(data: Seq[Double], theta: Int): Double = ???

val data: Seq[Double] = Seq(1.0, 2.0, 3.0)
processData(data, lossGauss(_, true))
processData(data, lossKernel(_, 1234))

Even better, use multiple argument lists with currying:

def processData(data: Seq[Double], lossFunc: Seq[Double] => Double): Double = ???

def lossGauss(scalaShift: Boolean)(data: Seq[Double]): Double = ???
def lossKernel(theta: Int)(data: Seq[Double]): Double = ???

val data: Seq[Double] = Seq(1.0, 2.0, 3.0)
processData(data, lossGauss(true))
processData(data, lossKernel(1234))

And by the way: don't use Floats, especially not for tiny results that take O(1) memory. Seq[Float] => Double would make some sense, but not the other way round.

Upvotes: 3

Related Questions