今天我接受了关于Scala的采访,请求是:
实现具有固定大小N的数据结构,具有以下功能:
复杂性应为O(1)
example:
val ds = DS(100)
ds.set(4,5)
ds.get(4) //would return 5
ds.set(1,4)
ds.setall(10)
ds.get(4) //would return 10
ds.get(7) //would return 10
ds.set(1,7)
ds.get(1) //would return 7
请查找我在下面发送的代码 我的问题是这是正确的解决方案吗?如果有更好的方法吗?
import scala.collection.mutable.HashMap
trait Operations {
def get(index: Int): Int
def set(index: Int, value: Int)
def setall(value: Int)
}
class DS(N: Int) extends Operations {
var myMutableHashMap = new HashMap[Int, Int]().withDefaultValue(0)
def get(index: Int) = myMutableHashMap(index)
def set(index: Int, value: Int) = {
if (index <= N) myMutableHashMap += (index -> value)
}
def setall(value: Int) = {
myMutableHashMap = new HashMap[Int, Int]().withDefaultValue(value)
}
}
object Task {
def main(args: Array[String]): Unit = {
val ds = new DS(100)
ds.set(4, 5)
ds.get(4)
println(ds.get(4)) // -> 5
ds.setall(10)
println(ds.get(4)) //-> 10
println(ds.get(7)) //-> 10
ds.set(1, 7)
println(ds.get(1)) //-> 7
}
}
答案 0 :(得分:1)
我不确定它是否更好,但我认为HashMap可能有点矫枉过正。 以下解决方案可能占用空间更小,代码更少。 虽然,我可能宁愿实现更通用的东西,但它应该满足你提到的要求。
trait Operations {
def get(index: Int): Int
def set(index: Int, value: Int): Unit
def setall(fill: Int): Unit
}
class DS(size: Int) extends Operations {
val underlying: Array[Int] = new Array(size)
def get(index: Int): Int = underlying(index)
def set(index: Int, value: Int): Unit = underlying(index) = value
def setall(fill: Int): Unit = (0 until size).foreach(underlying(_) = fill)
}
另类,这可能会给我们一个更好的&#39; setall&#39;复杂但需要付出代价...
trait Operations {
def get(index: Int): Int
def set(index: Int, value: Int): Unit
def setall(fill: Int): Unit
}
class DS(size: Int) extends Operations {
var underlying: Array[Integer] = new Array(size)
var default: Integer = new Integer(0)
def get(index: Int): Int = {
val result = underlying(index)
if (result == null) default else result
}
def set(index: Int, value: Int): Unit = underlying(index) = value
def setall(fill: Int): Unit = {
default = fill
underlying = new Array(size)
}
}