在 C++ 中查詢 BST 的中位數,時間複雜度為 O(n),空間複雜度為 O(1)
概念
對於給定的二叉搜尋樹 (BST),我們的任務是確定它的中位數。
對於偶數個節點,中位數 = ((n/2 個節點 + (n+1)/2 個節點) /2 對於奇數個節點,中位數 = (n+1)/2 個節點。
對於給定的 BST(具有奇數個節點)為 -
7 / \ 4 9 / \ / \ 2 5 8 10
給定 BST 的中序遍歷將是:2、4、5、7、8、9、10 因此,這裡的中位數將是 7。
對於給定的 BST(具有偶數個節點)為 -
7 / \ 4 9 / \ / 2 5 8
給定 BST 的中序遍歷將是 - 2、4、5、7、8、9
因此,這裡的中位數將是 (5+7)/2 = 6。
方法
為了確定中位數,我們需要確定 BST 的中序遍歷,因為它的中序遍歷將按排序順序排列,然後確定中位數。在這裡,這個概念基於使用 O(1) 額外空間的 BST 中第 K 個最小元素。現在,如果允許我們實現額外的空間,那麼任務非常簡單,但中序遍歷實現遞迴和棧都使用空間,這裡不允許。
因此,解決方案是執行 Morris 中序遍歷,因為它不需要任何額外空間。
Morris 中序遍歷的解釋如下 -
- 我們將 current 初始化為根節點
- 當 current 不為 NULL 時
如果 current 沒有左子節點
- 列印 current 的資料
- 移到右側,即 current = current->right
否則
- 將 current 構造為 current 的左子樹中最右側節點的右子節點
- 移到這個左子節點,即 current = current->left
最終實現以以下方式討論 -
我們使用 Morris 中序遍歷計算給定 BST 中的節點數。
之後,再次執行 Morris 中序遍歷,同時計算節點數並驗證計數是否等於中位數點。
為了考慮偶數個節點,實現了一個額外的指標指向前一個節點。
示例
/* C++ program to find the median of BST in O(n) time and O(1) space*/ #include<bits/stdc++.h> using namespace std; /* Implements a binary search tree Node1 which has data, pointer to left child and a pointer to right child */ struct Node1{ int data1; struct Node1* left1, *right1; }; //Shows a utility function to create a new BST node struct Node1 *newNode(int item1){ struct Node1 *temp1 = new Node1; temp1->data1 = item1; temp1->left1 = temp1->right1 = NULL; return temp1; } /* Shows a utility function to insert a new node with given key in BST */ struct Node1* insert(struct Node1* node1, int key1){ /* It has been seen that if the tree is empty, return a new node */ if (node1 == NULL) return newNode(key1); /* Else, recur down the tree */ if (key1 < node1->data1) node1->left1 = insert(node1->left1, key1); else if (key1 > node1->data1) node1->right1 = insert(node1->right1, key1); /* return the (unchanged) node pointer */ return node1; } /* Shows function to count nodes in a binary search tree using Morris Inorder traversal*/ int counNodes(struct Node1 *root1){ struct Node1 *current1, *pre1; // Used to initialise count of nodes as 0 int count1 = 0; if (root1 == NULL) return count1; current1 = root1; while (current1 != NULL){ if (current1->left1 == NULL){ // Now count node if its left is NULL count1++; // Go to its right current1 = current1->right1; } else { /* Determine the inorder predecessor of current */ pre1 = current1->left1; while (pre1->right1 != NULL && pre1->right1 != current1) pre1 = pre1->right1; /* Construct current1 as right child of its inorder predecessor */ if(pre1->right1 == NULL){ pre1->right1 = current1; current1 = current1->left1; } /* we have to revert the changes made in if part to restore the original tree i.e., fix the right child of predecssor */ else { pre1->right1 = NULL; // Now increment count if the current // node is to be visited count1++; current1 = current1->right1; } /* End of if condition pre1->right1 == NULL */ } /* End of if condition current1->left1 == NULL*/ } /* End of while */ return count1; } /* Shows function to find median in O(n) time and O(1) space using Morris Inorder traversal*/ int findMedian(struct Node1 *root1){ if (root1 == NULL) return 0; int count1 = counNodes(root1); int currCount1 = 0; struct Node1 *current1 = root1, *pre1, *prev1; while (current1 != NULL){ if (current1->left1 == NULL){ // Now count current node currCount1++; // Verify if current node is the median // Odd case if (count1 % 2 != 0 && currCount1 == (count1+1)/2) return prev1->data1; // Even case else if (count1 % 2 == 0 && currCount1 == (count1/2)+1) return (prev1->data1 + current1->data1)/2; // Now update prev1 for even no. of nodes prev1 = current1; //Go to the right current1 = current1->right1; } else { /* determine the inorder predecessor of current1 */ pre1 = current1->left1; while (pre1->right1 != NULL && pre1->right1 != current1) pre1 = pre1->right1; /* Construct current1 as right child of its inorder predecessor */ if (pre1->right1 == NULL){ pre1->right1 = current1; current1 = current1->left1; } /* We have to revert the changes made in if part to restore the original tree i.e., fix the right child of predecssor */ else { pre1->right1 = NULL; prev1 = pre1; // Now count current node currCount1++; // Verify if the current node is the median if (count1 % 2 != 0 && currCount1 == (count1+1)/2 ) return current1->data1; else if (count1%2==0 && currCount1 == (count1/2)+1) return (prev1->data1+current1->data1)/2; // Now update prev1 node for the case of even // no. of nodes prev1 = current1; current1 = current1->right1; } /* End of if condition pre1->right1 == NULL */ } /* End of if condition current1->left1 == NULL*/ } /* End of while */ } /* Driver program to test above functions*/ int main(){ /* Let us create following BST 7 / \ 4 9 / \ / \ 2 5 8 10 */ struct Node1 *root1 = NULL; root1 = insert(root1, 7); insert(root1, 4); insert(root1, 2); insert(root1, 5); insert(root1, 9); insert(root1, 8); insert(root1, 10); cout << "\nMedian of BST is(for odd no. of nodes) "<< findMedian(root1) <<endl; /* Let us create following BST 7 / \ 4 9 / \ / 2 5 8 */ struct Node1 *root2 = NULL; root2 = insert(root2, 7); insert(root2, 4); insert(root2, 2); insert(root2, 5); insert(root2, 9); insert(root2, 8); cout << "\nMedian of BST is(for even no. of nodes) " << findMedian(root2); return 0; }
輸出
Median of BST is(for odd no. of nodes) 7 Median of BST is(for even no. of nodes) 6
廣告