[Leetcode]274. H-Index

好吧，原来辛苦打的被csdn吃了，跟以往一样，被吃掉的文章我就只暂时放出简要的描述和代码

解法一：基于排序的O(nlogn)，就是先排序，然后找到第一个citations[i]，它比citations.length - i要大（也就是剩下的论文）。给出代码如下：
 

    public int hIndex(int[] citations) {
        Arrays.sort(citations);
        for (int i = 0; i < citations.length; i++) {
            int citationsLeft = citations.length - i;
            if (citations[i] >= citationsLeft) return citationsLeft;
        }
        
        return 0;
    }

解法二： 基于selection rank的最坏O(n^2)，平均O(n)，就是不停通过partition找，找到最小的citations[mid], 它大于等于citations.length - mid，其实和排序的最终完结条件是一样的，但我们并没有做完全的排序。
 

    public int hIndex(int[] citations) {
        int head = 0, tail = citations.length - 1, maxCandidate = 0;
        while (head <= tail) {
            int mid = partition(head, tail, citations);
            int candidate = citations.length - mid;
            if (citations[mid] == candidate) {
                return candidate;
            } else if (citations[mid] > candidate) {
                maxCandidate = Math.max(candidate, maxCandidate);
                tail = mid - 1;
            } else {
                head = mid + 1;
            }
        }
        
        return maxCandidate;
    }
    
    public void swap(int[] arr, int a, int b) {
        if (a != b) {
            arr[a] ^= arr[b];
            arr[b] ^= arr[a];
            arr[a] ^= arr[b];
        }
    }
    
    public int partition(int head, int tail, int[] arr) {
        Random rand = new Random();
        int pivotIndx = head + rand.nextInt(tail - head + 1);
        swap(arr, pivotIndx, tail);
        int pivot = arr[tail];
        int j = head;
        for (int i = head; i < tail; i++) {
            if (pivot >= arr[i]) {
                swap(arr, i, j);
                j++;
            }
        }
        swap(arr, tail, j);
        return j;
    }