They said I can't have fast TopK, but I proved them wrong. And they stabbed me in the stomach.

top-k-magic.scala

  def topKEstaimateMin(arr: Array[Double], k: Int): Double = {
    import java.lang.Integer
    val upperK = Integer.highestOneBit(k) * 2 // must be higher than k, otherwise can produce wrong minimum
    val bits = Integer.numberOfTrailingZeros(upperK)
    val mask = (1 << bits) - 1
    var i = 0
    val maximums = new Array[Double](upperK)
    while (i < arr.length) {
      val pos = i & mask
      val m = maximums(pos)
      val a = arr(i)
      maximums(pos) = (if (m < a) a else m)
      i += 1
    }

    var min = Double.PositiveInfinity
    var j = 0
    while (j < upperK) {
      min = (if (maximums(j) < min) maximums(j) else min)
      j += 1
    }

    return min
  }

Kukam na to uz pekne dlho a fakt nechapem. Mozes hodit nejaky odkaz na to aku fintu pouzivas? Lebo tento zapis algoritmu nedokazem rozlustit.