B树——磁盘链式存储数据结构

B树详解

我们知道，内存的读写速度是远远高于磁盘的读写速度。磁盘的IO成本是极高的，所以，应尽量减少磁盘的访问次数。同样量级的数据，用二叉树和B树（多叉树）存储，树的高度是不同的。树的高度决定了访问磁盘的次数。

B树的性质

一颗M阶B树T，满足以下条件

1. 每个结点至多拥有M颗子树
2. 根结点至少拥有两颗子树
3. 除了根结点以外，其余每个分支结点至少拥有M/2课子树
4. 所有的叶结点都在同一层上
5. 有k课子树的分支结点则存在k-1个关键字，关键字按照递增顺序进行排序
6. 关键字数量满足ceil(M/2)-1 <= n <= M-1

B树添加节点（分裂）

void btree_split_child(btree *T, btree_node *x, int i) {
	int t = T->t;

	btree_node *y = x->childrens[i];
	btree_node *z = btree_create_node(t, y->leaf);

	z->num = t - 1;

	int j = 0;
	for (j = 0;j < t-1;j ++) {
		z->keys[j] = y->keys[j+t];
	}
	if (y->leaf == 0) {
		for (j = 0;j < t;j ++) {
			z->childrens[j] = y->childrens[j+t];
		}
	}

	y->num = t - 1;
	for (j = x->num;j >= i+1;j --) {
		x->childrens[j+1] = x->childrens[j];
	}

	x->childrens[i+1] = z;

	for (j = x->num-1;j >= i;j --) {
		x->keys[j+1] = x->keys[j];
	}
	x->keys[i] = y->keys[t-1];
	x->num += 1;
	
}

void btree_insert_nonfull(btree *T, btree_node *x, KEY_VALUE k) {

	int i = x->num - 1;

	if (x->leaf == 1) {
		
		while (i >= 0 && x->keys[i] > k) {
			x->keys[i+1] = x->keys[i];
			i --;
		}
		x->keys[i+1] = k;
		x->num += 1;
		
	} else {
		while (i >= 0 && x->keys[i] > k) i --;

		if (x->childrens[i+1]->num == (2*(T->t))-1) {
			btree_split_child(T, x, i+1);
			if (k > x->keys[i+1]) i++;
		}

		btree_insert_nonfull(T, x->childrens[i+1], k);
	}
}

void btree_insert(btree *T, KEY_VALUE key) {
	//int t = T->t;

	btree_node *r = T->root;
	if (r->num == 2 * T->t - 1) {
		
		btree_node *node = btree_create_node(T->t, 0);
		T->root = node;

		node->childrens[0] = r;

		btree_split_child(T, node, 0);

		int i = 0;
		if (node->keys[0] < key) i++;
		btree_insert_nonfull(T, node->childrens[i], key);
		
	} else {
		btree_insert_nonfull(T, r, key);
	}
}

B树删除节点（借位或合并）

void btree_merge(btree *T, btree_node *node, int idx) {

	btree_node *left = node->childrens[idx];
	btree_node *right = node->childrens[idx+1];

	int i = 0;

	/data merge
	left->keys[T->t-1] = node->keys[idx];
	for (i = 0;i < T->t-1;i ++) {
		left->keys[T->t+i] = right->keys[i];
	}
	if (!left->leaf) {
		for (i = 0;i < T->t;i ++) {
			left->childrens[T->t+i] = right->childrens[i];
		}
	}
	left->num += T->t;

	//destroy right
	btree_destroy_node(right);

	//node 
	for (i = idx+1;i < node->num;i ++) {
		node->keys[i-1] = node->keys[i];
		node->childrens[i] = node->childrens[i+1];
	}
	node->childrens[i+1] = NULL;
	node->num -= 1;

	if (node->num == 0) {
		T->root = left;
		btree_destroy_node(node);
	}
}

void btree_delete_key(btree *T, btree_node *node, KEY_VALUE key) {

	if (node == NULL) return ;

	int idx = 0, i;

	while (idx < node->num && key > node->keys[idx]) {
		idx ++;
	}

	if (idx < node->num && key == node->keys[idx]) {

		if (node->leaf) {
			
			for (i = idx;i < node->num-1;i ++) {
				node->keys[i] = node->keys[i+1];
			}

			node->keys[node->num - 1] = 0;
			node->num--;
			
			if (node->num == 0) { //root
				free(node);
				T->root = NULL;
			}

			return ;
		} else if (node->childrens[idx]->num >= T->t) {

			btree_node *left = node->childrens[idx];
			node->keys[idx] = left->keys[left->num - 1];

			btree_delete_key(T, left, left->keys[left->num - 1]);
			
		} else if (node->childrens[idx+1]->num >= T->t) {

			btree_node *right = node->childrens[idx+1];
			node->keys[idx] = right->keys[0];

			btree_delete_key(T, right, right->keys[0]);
			
		} else {

			btree_merge(T, node, idx);
			btree_delete_key(T, node->childrens[idx], key);
			
		}
		
	} else {

		btree_node *child = node->childrens[idx];
		if (child == NULL) {
			printf("Cannot del key = %d\n", key);
			return ;
		}

		if (child->num == T->t - 1) {

			btree_node *left = NULL;
			btree_node *right = NULL;
			if (idx - 1 >= 0)
				left = node->childrens[idx-1];
			if (idx + 1 <= node->num) 
				right = node->childrens[idx+1];

			if ((left && left->num >= T->t) ||
				(right && right->num >= T->t)) {

				int richR = 0;
				if (right) richR = 1;
				if (left && right) richR = (right->num > left->num) ? 1 : 0;

				if (right && right->num >= T->t && richR) { //borrow from next
					child->keys[child->num] = node->keys[idx];
					child->childrens[child->num+1] = right->childrens[0];
					child->num ++;

					node->keys[idx] = right->keys[0];
					for (i = 0;i < right->num - 1;i ++) {
						right->keys[i] = right->keys[i+1];
						right->childrens[i] = right->childrens[i+1];
					}

					right->keys[right->num-1] = 0;
					right->childrens[right->num-1] = right->childrens[right->num];
					right->childrens[right->num] = NULL;
					right->num --;
					
				} else { //borrow from prev

					for (i = child->num;i > 0;i --) {
						child->keys[i] = child->keys[i-1];
						child->childrens[i+1] = child->childrens[i];
					}

					child->childrens[1] = child->childrens[0];
					child->childrens[0] = left->childrens[left->num];
					child->keys[0] = node->keys[idx-1];
					
					child->num ++;

					node->key[idx-1] = left->keys[left->num-1];
					left->keys[left->num-1] = 0;
					left->childrens[left->num] = NULL;
					left->num --;
				}

			} else if ((!left || (left->num == T->t - 1))
				&& (!right || (right->num == T->t - 1))) {

				if (left && left->num == T->t - 1) {
					btree_merge(T, node, idx-1);					
					child = left;
				} else if (right && right->num == T->t - 1) {
					btree_merge(T, node, idx);
				}
			}
		}

		btree_delete_key(T, child, key);
	}
	
}


int btree_delete(btree *T, KEY_VALUE key) {
	if (!T->root) return -1;

	btree_delete_key(T, T->root, key);
	return 0;
}

文章参考与<零声教育>的C/C++linux服务期高级架构线上课学习。有兴趣的同学可以了解下哦。