How to implement generic average function in scala

问题:

It seems easy problem for any specific kind of Number i.e. Double/Integer but it is hard to write in general case.

implicit def iterebleWithAvg(data:Iterable[Double]) = new {
    def avg:Double = data.sum / data.size
}

How to implement this for any kind of Number(Int,Float,Double,BigDecemial)?

回答1:

You have to pass an implicit Numeric which will allow summation and conversion to Double:

def average[T]( ts: Iterable[T] )( implicit num: Numeric[T] ) = {
  num.toDouble( ts.sum ) / ts.size
}

The compiler will provide the correct instance for you:

scala> average( List( 1,2,3,4) )
res8: Double = 2.5

scala> average( 0.1 to 1.1 by 0.05 )
res9: Double = 0.6000000000000001

scala> average( Set( BigInt(120), BigInt(1200) ) )
res10: Double = 660.0

You can the use the function to define an implicit view (provided you propagate the implicit numeric dependency):

implicit def iterebleWithAvg[T:Numeric](data:Iterable[T]) = new {
  def avg = average(data)
}

scala> List(1,2,3,4).avg
res13: Double = 2.5

回答2:

Here's the way I define it in my code.

Instead of using Numeric, I use Fractional, since Fractional defines a division operation (Numeric doesn't necessarily have division). This means that when you call .avg, you will get back the same type you put in, instead of always getting Double.

I also define it over all GenTraversableOnce collections so that it works on, for example, Iterator.

class EnrichedAvgFractional[A](self: GenTraversableOnce[A]) {
  def avg(implicit num: Fractional[A]) = {
    val (total, count) = self.toIterator.foldLeft((num.zero, num.zero)) {
      case ((total, count), x) => (num.plus(total, x), num.plus(count, num.one))
    }
    num.div(total, count)
  }
}
implicit def enrichAvgFractional[A: Fractional](self: GenTraversableOnce[A]) = new EnrichedAvgFractional(self)

Notice how if we give it a collection of Double, we get back Double and if we give it BigDecimal, we get back BigDecimal. We could even define our own Fractional number type (which I do occasionally), and it will work for that.

scala> Iterator(1.0, 2.0, 3.0, 4.0, 5.0).avg
res0: Double = 3.0

scala> Iterator(1.0, 2.0, 3.0, 4.0, 5.0).map(BigDecimal(_)).avg
res1: scala.math.BigDecimal = 3.0

However, Int is not a kind of Fractional, meaning that it doesn't make sense to get an Int and the result of averaging Ints, so we have to have a special case for Int that converts to a Double.

class EnrichedAvgInt(self: GenTraversableOnce[Int]) {
  def avg = {
    val (total, count) = self.toIterator.foldLeft(0, 0) {
      case ((total, count), x) => (total + x, count + 1)
    }
    total.toDouble / count
  }
}
implicit def enrichAvgInt(self: GenTraversableOnce[Int]) = new EnrichedAvgInt(self)

So averaging Ints gives us a Double:

scala> Iterator(1, 2, 3, 4, 5).avg
res2: Double = 3