算法系列笔记5(扩展数据结构-动态顺序统计和区间树)

╰半橙微兮° 2022-08-06 15:09 137阅读 0赞

在编程中，我们往往使用已有的数据结构无法解决问题，这是不必要急着创建新的数据结构，而是在已有数据结构的基础上添加新的字段。本节在上一次笔记红黑树这一基础数据结构上进行扩展，得出两个重要的应用—动态顺序统计和区间树。

**动态顺序统计**

在[算法系列笔记2][2]中我们在线性时间内完成了静态表的顺序统计，而这里我们在红黑树上进行扩展，在O(lgn)时间内完成该操作，主要包括返回第i 排名的元素os\_select(i)和给定一个元素x，返回其排名(os\_rank(x)).

**思想：**添加新项：在红黑树的结点上记录下该结点的子树个数。size\[x\] = size\[left\[x\]\] + size\[right\[x\]\] +1。 若结点为空，则为0。

此外当你对该扩展的数据结构进行插入和删除操作时，需随时更新子树的大小，与插入和删除操作同步进行，但是需要重新使其回到平衡。主要在于case2和case3这两种情况的旋转。<可以与算法系列笔记4>红黑树的插入代码进行对比，看修改情况。

代码：

返回第i 排名的元素os\_select(i)

BSTNode* OSRBTree::os_select(BSTNode *p, const int &ith){
    	if(p == NULL) return p;
    	int k = 1;
    	if(p->left != NULL){
    		k = p->left->size + 1;           // 当前该结点所对应的rank
    		
    	}
    	if(ith == k) return p;
    	if(ith < k) return os_select(p->left, ith);
    	else return os_select(p->right, ith - k);
    }

给定一个元素x，返回其排名(os\_rank(x))

//  return the rank of value
    int OSRBTree::os_rank(BSTNode *p, const int &value){
    	if(p == NULL) return 0;
    	int k = 1;
    	if(p->left != NULL)
    		k = p->left->size + 1;
    	if(p->val == value)
    		return k;
    	else if(p->val > value) return os_rank(p->left, value);
    	else return os_rank(p->right, value)+k;
    }

完整代码：

OSTree.h

#ifndef OSRBTREE
    #define OSRBTREE
    #include <iostream>
    #include <string>
    using namespace std;
    
    class BSTNode{
    public:
    	BSTNode *left, *right;
    	BSTNode *parent;
    	int val;
    	string color;
    	int size;
    };
    
    class OSRBTree{
    public:
    	OSRBTree(const int &rootVal){
    		root = new BSTNode();
    		root->val = rootVal;
    		root->left = NULL;
    		root->right = NULL;
    		root->color = "black";
    		root->size = 1;
    		root->parent = NULL;
    	}
    	BSTNode* insertBST(BSTNode *p, const int &value);
    	void insertOSRBTree(BSTNode *root1, const int &value);
    	void inorderOSRBTree(BSTNode *p);
    	BSTNode* os_select(BSTNode *p, const int &ith);
    	int os_rank(BSTNode *p, const int &value);
    public:
    	BSTNode *root;
    };
    
    #endif

OSTree.cpp

#include "OSRBTree.h"
    
    // 二叉查找树的插入
    BSTNode* OSRBTree::insertBST(BSTNode *p, const int &value){
    	BSTNode *y = NULL;
    	BSTNode *in = new BSTNode();
    	in->left = NULL;
    	in->right = NULL;
    	in->val = value;
    	in->parent = NULL;
    	in->size = 1;
    	while(p != NULL){
    		y = p;
    		p->size += 1;
    		if(p->val > in->val)
    			p = p->left;
    		else p = p->right;
    	}
    	
    	if(y == NULL)
    		p = in;
    	else{
    		in->parent = y;
    		if(y->val > in->val) y->left = in;
    		else y->right = in;
    	}
    	return in;
    }
    
    // 插入红黑树
    void OSRBTree::insertOSRBTree(BSTNode *root1, const int &value){
    	BSTNode * in = insertBST(root1, value);
    	in->color = "red";
    	while(in != root1 && in->color == "red"){            // 对红黑特性进行调整
    		if(in->parent->color == "black") return;   //  也就保证了必须
    		if(in->parent == in->parent->parent->left){
    			BSTNode *y = in->parent->parent->right;
    			if(y != NULL && y->color == "red"){   //  case 1
    				y->color = "black";
    				y->parent->color = "red";
    				in->parent->color ="black";
    				in = in->parent->parent;
    			}
    			else{
    				if(in == in->parent->right){      // case 2   in->parent 左旋
    					BSTNode *pa = in->parent;
    					in->size = pa->size;    //  修改该结点所包含子树结点个数
    					in->parent = pa->parent;
    					pa->parent->left = in;
    					pa->parent = in;
    					if(pa->left != NULL)
    						pa->size = pa->left->size + 1;        // 修改结点子树结点大小 
    					else pa->size = 1;
    					if(in->left != NULL){
    						in->left->parent = pa;
    						pa->size += in->left->size;
    					}
    					pa->right = in->left;
    					in->left = pa;
    
    					
    					in = pa;
    				}
    				// case 3    in->parent->parent 右旋
    				BSTNode *pa = in->parent;
    				BSTNode *gpa = in->parent->parent;
    				pa->size = gpa->size;
    				if(gpa->parent != NULL){
    					if(gpa == gpa->parent->left){
    						gpa->parent->left = pa;
    					}else
    						gpa->parent->right = pa;
    				}
    				pa->parent = gpa->parent;
    				if(gpa->right != NULL)gpa->size = gpa->right->size + 1;
    				else gpa->size = 1;
    				if(pa->right != NULL){
    					gpa->size += pa->right->size;
    					pa->right->parent = gpa;
    				}
    				gpa->left = pa->right;
    				pa->right = gpa;
    				gpa->parent = pa;
    				pa->color = "black";
    				gpa->color = "red";
    			}
    		}
    		else{
    			BSTNode *y = in->parent->parent->left;
    			if(y != NULL && y->color == "red"){   //  case 1
    				y->color = "black";
    				y->parent->color = "red";
    				in->parent->color ="black";
    				in = in->parent->parent;
    			}else{                               // do the same as A  but left与right对换
    				if(in == in->parent->left){      // case 2   in->parent 右旋
    					BSTNode *pa = in->parent;
    
    					in->size = pa->size;    //  修改该结点所包含子树结点个数
    
    					in->parent = pa->parent;
    					pa->parent->right = in;
    					pa->parent = in;
    					if(pa->right != NULL)
    						pa->size = pa->right->size + 1;
    					else pa->size = 1;
    					if(in->right != NULL){
    						in->right->parent = pa;
    						pa->size += in->right->size;
    					}
    					pa->left = in->right;
    					in->right = pa;
    					in = pa;
    				}
    				// case 3    in->parent->parent 左旋
    				BSTNode *pa = in->parent;
    				BSTNode *gpa = in->parent->parent;
    				pa->size = gpa->size;
    				if(gpa->parent != NULL){
    					if(gpa == gpa->parent->left){
    						gpa->parent->left = pa;
    					}else
    						gpa->parent->right = pa;
    				}
    				pa->parent = gpa->parent;
    				if(gpa->left != NULL)gpa->size = gpa->left->size+1;
    				else gpa->size = 1;
    				if(pa->left != NULL){
    					pa->left->parent = gpa;
    					gpa->size += pa->left->size;
    				}
    				gpa->right = pa->left;
    				pa->left = gpa;
    				gpa->parent = pa;
    				pa->color = "black";
    				gpa->color = "red";
    			}
    		}
    	}
    	root1->color = "black";
    }
    
    // 中序遍历输出
    void OSRBTree::inorderOSRBTree(BSTNode *p){
    	if(p == NULL) return;
    	if(p->left != NULL) inorderOSRBTree(p->left);
    	cout << p->val << p->color << p->size << " ";
    	if(p->right != NULL) inorderOSRBTree(p->right);
    }
    
    //  give ith smallest value
    BSTNode* OSRBTree::os_select(BSTNode *p, const int &ith){
    	if(p == NULL) return p;
    	int k = 1;
    	if(p->left != NULL){
    		k = p->left->size + 1;           // 当前该结点所对应的rank
    		
    	}
    	if(ith == k) return p;
    	if(ith < k) return os_select(p->left, ith);
    	else return os_select(p->right, ith - k);
    }
    
    //  return the rank of value
    int OSRBTree::os_rank(BSTNode *p, const int &value){
    	if(p == NULL) return 0;
    	int k = 1;
    	if(p->left != NULL)
    		k = p->left->size + 1;
    	if(p->val == value)
    		return k;
    	else if(p->val > value) return os_rank(p->left, value);
    	else return os_rank(p->right, value)+k;
    }

Main.cpp

int a[10] = {5,4,6, 7,2,4, 1, 8, 5, 10};
    	OSRBTree osbrt(a[0]);
    	for(int i = 1; i < 10; i++)
    		osbrt.insertOSRBTree(osbrt.root, a[i]);
    	cout << "中序遍历的结果: " << endl;
    	osbrt.inorderOSRBTree(osbrt.root);
    	cout << endl;
    
    	int ith = 6;
    	BSTNode *rank = osbrt.os_select(osbrt.root, ith);
    	if(rank == NULL) cout << "排名" << ith << "不存在！！" << endl;
    	cout << "排名" << ith << ": " << rank->val << endl;
    
    	int x = 6;
    	cout << x << "排名为: ";
    	cout << osbrt.os_rank(osbrt.root, x) << endl;

Result：

![Center][]

它们的时间复杂度都为O(lgn)，因为红黑树的高度为O(lgn)。

问题：为什么不直接使用这些结点排名作为新添加的项呢？原因在于当你此时对树进行修改时，维护这个树就变得很费劲。

**方法论：**如<OSTree—顺序统计树>

1：选择一个基础的数据结构(red-black tree)

2：在数据统计中维护一些附加信息(子树大小)

3：验证这个数据结构上的信息不会受修改操作的影响(insert, delete---rotations)

4：建立新的运算。假设新的数据已经存好了，然后开始使用这些信息(os\_select, os\_rank).

**区间树(Interval Tree)**

**问题：**保存一系列的区间，比如说时间区间。需要查询集合中的所有区间，与给定区间发生重叠的有哪些？

我们按照上面提到的方法论来进行：

1：选择红黑树作为基本的数据结构，并将区间的较低值(low)作为键值

2：将结点子树的最大值作为新添加的项(m\[x\] = max\{high\[int\[x\]\],m\[left\[x\]\], m\[right\[x\]\]\}).

3：是否受插入删除等操作的影响？受，但是在O(1)时间内就能调整过来，见代码。

4：新的操作，查询集合中与给定区间重叠的一个区间。

代码：

IntervalTree.h

#ifndef INTERVALTREE
    #define INTERVALTREE
    #include <iostream>
    #include <string>
    using namespace std;
    struct dataNode{
    	int low;
    	int high;
    };
    class BSTNode{
    public:
    	BSTNode *left, *right;
    	BSTNode *parent;
    	int val;
    	dataNode d;
    	string color;
    	int m;         // 最大值
    };
    
    class IntervalTree{
    public:
    	IntervalTree(const dataNode &d)
    	{
    		root = new BSTNode();
    		root->d = d;
    		root->color = "black";
    		root->left = NULL;
    		root->right = NULL;
    		root->m = d.high;
    		root->parent = NULL;
    		root->val = d.low;
    	}
    
    	BSTNode* insertBST(BSTNode *p, const dataNode &d);
    	void insertIntervalTree(BSTNode *root1, const dataNode &d);
    	void inorderOSRBTree(BSTNode *p);
    	BSTNode* intervalSearch(BSTNode *p, const dataNode &d);
    public:
    	BSTNode *root;
    	void destroyBST(BSTNode *p);
    };
    
    #endif

IntervalTree.cpp

#include "IntervalTree.h"
    using namespace std;
    
    BSTNode* IntervalTree::insertBST(BSTNode *p, const dataNode &d){
    	BSTNode *y = NULL;
    	BSTNode *in = new BSTNode();
    	in->left = NULL;
    	in->right = NULL;
    	in->val = d.low;
    	in->parent = NULL;
    	in->m = d.high;
    	in->d = d;
    	while(p != NULL){
    		y = p;
    		if(p->m < in->m) p->m = in->m;  // 为子树结点的最大值
    		if(p->val > in->val)
    			p = p->left;
    		else p = p->right;
    	}
    	
    	if(y == NULL)
    		p = in;
    	else{
    		in->parent = y;
    		if(y->val > in->val) y->left = in;
    		else y->right = in;
    	}
    	return in;
    }
    
    void IntervalTree::insertIntervalTree(BSTNode *root1, const dataNode &d){
    	BSTNode * in = insertBST(root1, d);
    	in->color = "red";
    	while(in != root1 && in->color == "red"){            // 对红黑特性进行调整
    		if(in->parent->color == "black") return;   //  也就保证了必须
    		if(in->parent == in->parent->parent->left){
    			BSTNode *y = in->parent->parent->right;
    			if(y != NULL && y->color == "red"){   //  case 1
    				y->color = "black";
    				y->parent->color = "red";
    				in->parent->color ="black";
    				in = in->parent->parent;
    			}
    			else{
    				if(in == in->parent->right){      // case 2   in->parent 左旋
    					BSTNode *pa = in->parent;
    					in->m = pa->m;    //  修改该结点所包含子树结点个数
    					in->parent = pa->parent;
    					pa->parent->left = in;
    					pa->parent = in;
    					if(pa->left != NULL)
    						pa->m = pa->left->m > pa->m ? pa->left->m : pa->m;
    
    					if(in->left != NULL){
    						in->left->parent = pa;
    						pa->m = in->left->m > pa->m ? pa->left->m : pa->m;
    					}
    					pa->right = in->left;
    					in->left = pa;
    
    					
    					in = pa;
    				}
    				// case 3    in->parent->parent 右旋
    				BSTNode *pa = in->parent;
    				BSTNode *gpa = in->parent->parent;
    				pa->m = gpa->m;
    				if(gpa->parent != NULL){
    					if(gpa == gpa->parent->left){
    						gpa->parent->left = pa;
    					}else
    						gpa->parent->right = pa;
    				}
    				pa->parent = gpa->parent;
    				if(gpa->right != NULL)gpa->m = gpa->right->m > gpa->m ? gpa->right->m : gpa->m;
    				if(pa->right != NULL){
    					gpa->m = pa->right->m > gpa->m ? pa->right->m : gpa->m;
    					pa->right->parent = gpa;
    				}
    				gpa->left = pa->right;
    				pa->right = gpa;
    				gpa->parent = pa;
    				pa->color = "black";
    				gpa->color = "red";
    			}
    		}
    		else{
    			BSTNode *y = in->parent->parent->left;
    			if(y != NULL && y->color == "red"){   //  case 1
    				y->color = "black";
    				y->parent->color = "red";
    				in->parent->color ="black";
    				in = in->parent->parent;
    			}else{                               // do the same as A  but left与right对换
    				if(in == in->parent->left){      // case 2   in->parent 右旋
    					BSTNode *pa = in->parent;
    
    					in->m = pa->m;    //  修改该结点所包含子树结点个数
    
    					in->parent = pa->parent;
    					pa->parent->right = in;
    					pa->parent = in;
    					if(pa->right != NULL)
    						pa->m = pa->right->m > pa->m ? pa->right->m : pa->m;
    					if(in->right != NULL){
    						in->right->parent = pa;
    						pa->m = in->right->m > pa->m ? in->right->m : pa->m;
    					}
    					pa->left = in->right;
    					in->right = pa;
    					in = pa;
    				}
    				// case 3    in->parent->parent 左旋
    				BSTNode *pa = in->parent;
    				BSTNode *gpa = in->parent->parent;
    				pa->m = gpa->m;
    				if(gpa->parent != NULL){
    					if(gpa == gpa->parent->left){
    						gpa->parent->left = pa;
    					}else
    						gpa->parent->right = pa;
    				}
    				pa->parent = gpa->parent;
    				if(gpa->left != NULL)gpa->m = gpa->left->m > gpa->m ? gpa->left->m : gpa->m;
    				if(pa->left != NULL){
    					pa->left->parent = gpa;
    					gpa->m = pa->left->m > gpa->m ? pa->left->m : gpa->m;
    				}
    				gpa->right = pa->left;
    				pa->left = gpa;
    				gpa->parent = pa;
    				pa->color = "black";
    				gpa->color = "red";
    			}
    		}
    	}
    	root1->color = "black";
    }
    
    void IntervalTree::inorderOSRBTree(BSTNode *p){
    	if(p == NULL) return;
    	if(p->left != NULL) inorderOSRBTree(p->left);
    	cout << p->val << p->color << p->m << " ";
    
    	//cout << p->d.low << p->color << p->d.high << " ";
    	if(p->right != NULL) inorderOSRBTree(p->right);
    }
    
    BSTNode* IntervalTree::intervalSearch(BSTNode *p, const dataNode &d){
    	while(p != NULL && (d.low > p->d.high || d.high < p->d.low)){
    		if(p->left != NULL && d.low < p->m)
    			p = p->left;
    		else p = p->right;
    	}
    	return p;
    }
    
    void IntervalTree::destroyBST(BSTNode *p){
    	if(p == NULL) return;
    	if(p->left != NULL){
    		destroyBST(p->left);
    	}
    	if(p->right != NULL){
    		destroyBST(p->right);
    	}
    	delete p;
    }

Main.cpp

int a[6] = {17, 5, 21, 4, 15, 7};
    	int b[6] = {19, 11, 23, 8, 18, 10};
    	vector<dataNode> data;
    	for(int i = 0; i < 6; i++)
    	{
    		dataNode d;
    		d.low = a[i];
    		d.high = b[i];
    		data.push_back(d);
    	}
    	IntervalTree interval(data[0]);
    	for(int i = 1; i < data.size(); i++){
    		interval.insertIntervalTree(interval.root, data[i]);
    	}
    	cout << "中序遍历的结果: " << endl;
    	interval.inorderOSRBTree(interval.root);
    	cout << endl;
    
    	dataNode sd;
    	sd.low = 18;
    	sd.high = 25;
    	BSTNode * bst = interval.intervalSearch(interval.root, sd);
    	cout << "[" << bst->d.low << "," << bst->d.high << "]" << endl;

Result:

![Center 1][]

时间复杂度都为O(lgn)，因为红黑树的高度为O(lgn)。

[2]: http://blog.csdn.net/lu597203933/article/details/43418179
[Center]: /images/20220805/a3ff68c4e89f441b9911b0faa215db9e.png
[Center 1]: /images/20220805/9d06134fea2546769935b039ee763841.png