Efficiently, get the count of numbers in a sorted array that is less than a given number using binary search

Question

My problem statement is this - Find the count of numbers in a sorted array that are less than a given number, and this should be done efficiently with respect to time. I wrote a program using binary search that gets the count but time complexity wise it's failing. Need help in achieving this.

import java.util.Arrays;

public class SortedSearch {
    public static int countNumbers(int[] sortedArray, int lessThan) {
        if(sortedArray.length ==1 || sortedArray.length == 0) {
            return singleElement(sortedArray, lessThan);
        }
        else {
            return binarySearch(sortedArray, lessThan);
        }
    }

    public static int singleElement(int[] sortedArray, int searchVal) {
        if(sortedArray.length == 0) {
            return 0;
        }
        if(sortedArray[0] < searchVal) {
            return 1;
        }
        return 0;
    }

    private static int binarySearch(int[] sortedArray, int searchVal) {
        int low = 0;
        int high = (sortedArray.length)-1;
        int mid = (low + high)/2;
        if((sortedArray.length == 0) || (sortedArray[0] > searchVal)) {
            return 0;
        }
        if(sortedArray[high] < searchVal) {
            return sortedArray.length;
        }
        if(sortedArray[high] == searchVal) {
            return sortedArray.length-1;
        }
        if(sortedArray[mid] < searchVal) {
            int newLow = low;
            int newHigh = calculateNewHigh(sortedArray, newLow, 0, searchVal);
            int[] newArray = Arrays.copyOfRange(sortedArray, newLow, newHigh+1);
            return newArray.length;
        }
        else {
            int newLow = low;
            int newHigh = mid;
            int[] newArray = Arrays.copyOfRange(sortedArray, newLow, newHigh+1);
            return binarySearch(newArray, searchVal);
        }
    }

    private static int calculateNewHigh(int[] sortedArray, int low, int previousHigh, int searchVal) {
        int newHigh =  previousHigh + (sortedArray.length-low)/2;
        if(sortedArray[newHigh] < searchVal) {
            newHigh = calculateNewHigh(sortedArray, newHigh, newHigh, searchVal);
        }
        if(sortedArray[newHigh] == searchVal) {
            newHigh--;
        }
        if(sortedArray[newHigh] > searchVal) {
            newHigh--;
        }
        return newHigh;
    }

    public static void main(String[] args) {
        System.out.println(SortedSearch.countNumbers(new int[] { 1, 3, 5, 7 }, 4));
    }
}

Don't use `copyOfRange`. There is no need to copy a slice of the array, or create a new array at all, and this degrades the performance to O(n) instead of O(log n). — kaya3, Dec 22 '19 at 03:45

Andreas · Answer 1 · 2019-12-22T07:33:51.807

Since you're using Arrays anyway, way not use the Arrays.binarySearch(int[] a, int key) method, instead of attempting to write your own?

public static int countNumbers(int[] sortedArray, int lessThan) {
    int idx = Arrays.binarySearch(sortedArray, lessThan);
    if (idx < 0)
        return -idx - 1; // insertion point
    while (idx > 0 && sortedArray[idx - 1] == lessThan)
        idx--;
    return idx; // index of first element with given value
}

The while loop¹ is necessary because the javadoc says:

If the array contains multiple elements with the specified value, there is no guarantee which one will be found.

^{1) This loop is not optimal if e.g. all values are the same, see e.g. Finding multiple entries with binary search}

Efficiently, get the count of numbers in a sorted array that is less than a given number using binary search

1 Answers1