|
|
|
|
# 二叉搜索树
|
|
|
|
|
|
|
|
|
|
「二叉搜索树 Binary Search Tree」满足以下条件:
|
|
|
|
|
|
|
|
|
|
1. 对于根结点,左子树中所有结点的值 $<$ 根结点的值 $<$ 右子树中所有结点的值;
|
|
|
|
|
2. 任意结点的左子树和右子树也是二叉搜索树,即也满足条件 `1.` ;
|
|
|
|
|
|
|
|
|
|
![二叉搜索树](binary_search_tree.assets/binary_search_tree.png)
|
|
|
|
|
|
|
|
|
|
## 二叉搜索树的操作
|
|
|
|
|
|
|
|
|
|
### 查找结点
|
|
|
|
|
|
|
|
|
|
给定目标结点值 `num` ,可以根据二叉搜索树的性质来查找。我们声明一个结点 `cur` ,从二叉树的根结点 `root` 出发,循环比较结点值 `cur.val` 和 `num` 之间的大小关系
|
|
|
|
|
|
|
|
|
|
- 若 `cur.val < num` ,说明目标结点在 `cur` 的右子树中,因此执行 `cur = cur.right` ;
|
|
|
|
|
- 若 `cur.val > num` ,说明目标结点在 `cur` 的左子树中,因此执行 `cur = cur.left` ;
|
|
|
|
|
- 若 `cur.val = num` ,说明找到目标结点,跳出循环并返回该结点即可;
|
|
|
|
|
|
|
|
|
|
=== "<1>"
|
|
|
|
|
![查找结点步骤](binary_search_tree.assets/bst_search_step1.png)
|
|
|
|
|
|
|
|
|
|
=== "<2>"
|
|
|
|
|
![bst_search_step2](binary_search_tree.assets/bst_search_step2.png)
|
|
|
|
|
|
|
|
|
|
=== "<3>"
|
|
|
|
|
![bst_search_step3](binary_search_tree.assets/bst_search_step3.png)
|
|
|
|
|
|
|
|
|
|
=== "<4>"
|
|
|
|
|
![bst_search_step4](binary_search_tree.assets/bst_search_step4.png)
|
|
|
|
|
|
|
|
|
|
二叉搜索树的查找操作和二分查找算法如出一辙,也是在每轮排除一半情况。循环次数最多为二叉树的高度,当二叉树平衡时,使用 $O(\log n)$ 时间。
|
|
|
|
|
|
|
|
|
|
=== "Java"
|
|
|
|
|
|
|
|
|
|
```java title="binary_search_tree.java"
|
|
|
|
|
[class]{BinarySearchTree}-[func]{search}
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
=== "C++"
|
|
|
|
|
|
|
|
|
|
```cpp title="binary_search_tree.cpp"
|
|
|
|
|
[class]{BinarySearchTree}-[func]{search}
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
=== "Python"
|
|
|
|
|
|
|
|
|
|
```python title="binary_search_tree.py"
|
|
|
|
|
[class]{BinarySearchTree}-[func]{search}
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
=== "Go"
|
|
|
|
|
|
|
|
|
|
```go title="binary_search_tree.go"
|
|
|
|
|
[class]{binarySearchTree}-[func]{search}
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
=== "JavaScript"
|
|
|
|
|
|
|
|
|
|
```javascript title="binary_search_tree.js"
|
|
|
|
|
[class]{}-[func]{search}
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
=== "TypeScript"
|
|
|
|
|
|
|
|
|
|
```typescript title="binary_search_tree.ts"
|
|
|
|
|
[class]{}-[func]{search}
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
=== "C"
|
|
|
|
|
|
|
|
|
|
```c title="binary_search_tree.c"
|
|
|
|
|
[class]{binarySearchTree}-[func]{search}
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
=== "C#"
|
|
|
|
|
|
|
|
|
|
```csharp title="binary_search_tree.cs"
|
|
|
|
|
[class]{BinarySearchTree}-[func]{search}
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
=== "Swift"
|
|
|
|
|
|
|
|
|
|
```swift title="binary_search_tree.swift"
|
|
|
|
|
[class]{BinarySearchTree}-[func]{search}
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
=== "Zig"
|
|
|
|
|
|
|
|
|
|
```zig title="binary_search_tree.zig"
|
|
|
|
|
[class]{BinarySearchTree}-[func]{search}
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
### 插入结点
|
|
|
|
|
|
|
|
|
|
给定一个待插入元素 `num` ,为了保持二叉搜索树“左子树 < 根结点 < 右子树”的性质,插入操作分为两步:
|
|
|
|
|
|
|
|
|
|
1. **查找插入位置**:与查找操作类似,我们从根结点出发,根据当前结点值和 `num` 的大小关系循环向下搜索,直到越过叶结点(遍历到 $\text{null}$ )时跳出循环;
|
|
|
|
|
2. **在该位置插入结点**:初始化结点 `num` ,将该结点放到 $\text{null}$ 的位置 ;
|
|
|
|
|
|
|
|
|
|
二叉搜索树不允许存在重复结点,否则将会违背其定义。因此若待插入结点在树中已经存在,则不执行插入,直接返回即可。
|
|
|
|
|
|
|
|
|
|
![在二叉搜索树中插入结点](binary_search_tree.assets/bst_insert.png)
|
|
|
|
|
|
|
|
|
|
=== "Java"
|
|
|
|
|
|
|
|
|
|
```java title="binary_search_tree.java"
|
|
|
|
|
[class]{BinarySearchTree}-[func]{insert}
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
=== "C++"
|
|
|
|
|
|
|
|
|
|
```cpp title="binary_search_tree.cpp"
|
|
|
|
|
[class]{BinarySearchTree}-[func]{insert}
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
=== "Python"
|
|
|
|
|
|
|
|
|
|
```python title="binary_search_tree.py"
|
|
|
|
|
[class]{BinarySearchTree}-[func]{insert}
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
=== "Go"
|
|
|
|
|
|
|
|
|
|
```go title="binary_search_tree.go"
|
|
|
|
|
[class]{binarySearchTree}-[func]{insert}
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
=== "JavaScript"
|
|
|
|
|
|
|
|
|
|
```javascript title="binary_search_tree.js"
|
|
|
|
|
[class]{}-[func]{insert}
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
=== "TypeScript"
|
|
|
|
|
|
|
|
|
|
```typescript title="binary_search_tree.ts"
|
|
|
|
|
[class]{}-[func]{insert}
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
=== "C"
|
|
|
|
|
|
|
|
|
|
```c title="binary_search_tree.c"
|
|
|
|
|
[class]{binarySearchTree}-[func]{insert}
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
=== "C#"
|
|
|
|
|
|
|
|
|
|
```csharp title="binary_search_tree.cs"
|
|
|
|
|
[class]{BinarySearchTree}-[func]{insert}
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
=== "Swift"
|
|
|
|
|
|
|
|
|
|
```swift title="binary_search_tree.swift"
|
|
|
|
|
[class]{BinarySearchTree}-[func]{insert}
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
=== "Zig"
|
|
|
|
|
|
|
|
|
|
```zig title="binary_search_tree.zig"
|
|
|
|
|
[class]{BinarySearchTree}-[func]{insert}
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
为了插入结点,需要借助 **辅助结点 `pre`** 保存上一轮循环的结点,这样在遍历到 $\text{null}$ 时,我们也可以获取到其父结点,从而完成结点插入操作。
|
|
|
|
|
|
|
|
|
|
与查找结点相同,插入结点使用 $O(\log n)$ 时间。
|
|
|
|
|
|
|
|
|
|
### 删除结点
|
|
|
|
|
|
|
|
|
|
与插入结点一样,我们需要在删除操作后维持二叉搜索树的“左子树 < 根结点 < 右子树”的性质。首先,我们需要在二叉树中执行查找操作,获取待删除结点。接下来,根据待删除结点的子结点数量,删除操作需要分为三种情况:
|
|
|
|
|
|
|
|
|
|
**当待删除结点的子结点数量 $= 0$ 时**,表明待删除结点是叶结点,直接删除即可。
|
|
|
|
|
|
|
|
|
|
![在二叉搜索树中删除结点(度为 0)](binary_search_tree.assets/bst_remove_case1.png)
|
|
|
|
|
|
|
|
|
|
**当待删除结点的子结点数量 $= 1$ 时**,将待删除结点替换为其子结点即可。
|
|
|
|
|
|
|
|
|
|
![在二叉搜索树中删除结点(度为 1)](binary_search_tree.assets/bst_remove_case2.png)
|
|
|
|
|
|
|
|
|
|
**当待删除结点的子结点数量 $= 2$ 时**,删除操作分为三步:
|
|
|
|
|
|
|
|
|
|
1. 找到待删除结点在 **中序遍历序列** 中的下一个结点,记为 `nex` ;
|
|
|
|
|
2. 在树中递归删除结点 `nex` ;
|
|
|
|
|
3. 使用 `nex` 替换待删除结点;
|
|
|
|
|
|
|
|
|
|
=== "<1>"
|
|
|
|
|
![删除结点(度为 2)步骤](binary_search_tree.assets/bst_remove_case3_step1.png)
|
|
|
|
|
|
|
|
|
|
=== "<2>"
|
|
|
|
|
![bst_remove_case3_step2](binary_search_tree.assets/bst_remove_case3_step2.png)
|
|
|
|
|
|
|
|
|
|
=== "<3>"
|
|
|
|
|
![bst_remove_case3_step3](binary_search_tree.assets/bst_remove_case3_step3.png)
|
|
|
|
|
|
|
|
|
|
=== "<4>"
|
|
|
|
|
![bst_remove_case3_step4](binary_search_tree.assets/bst_remove_case3_step4.png)
|
|
|
|
|
|
|
|
|
|
删除结点操作也使用 $O(\log n)$ 时间,其中查找待删除结点 $O(\log n)$ ,获取中序遍历后继结点 $O(\log n)$ 。
|
|
|
|
|
|
|
|
|
|
=== "Java"
|
|
|
|
|
|
|
|
|
|
```java title="binary_search_tree.java"
|
|
|
|
|
[class]{BinarySearchTree}-[func]{remove}
|
|
|
|
|
|
|
|
|
|
[class]{BinarySearchTree}-[func]{getInOrderNext}
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
=== "C++"
|
|
|
|
|
|
|
|
|
|
```cpp title="binary_search_tree.cpp"
|
|
|
|
|
[class]{BinarySearchTree}-[func]{remove}
|
|
|
|
|
|
|
|
|
|
[class]{BinarySearchTree}-[func]{getInOrderNext}
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
=== "Python"
|
|
|
|
|
|
|
|
|
|
```python title="binary_search_tree.py"
|
|
|
|
|
[class]{BinarySearchTree}-[func]{remove}
|
|
|
|
|
|
|
|
|
|
[class]{BinarySearchTree}-[func]{get_inorder_next}
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
=== "Go"
|
|
|
|
|
|
|
|
|
|
```go title="binary_search_tree.go"
|
|
|
|
|
[class]{binarySearchTree}-[func]{remove}
|
|
|
|
|
|
|
|
|
|
[class]{binarySearchTree}-[func]{getInOrderNext}
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
=== "JavaScript"
|
|
|
|
|
|
|
|
|
|
```javascript title="binary_search_tree.js"
|
|
|
|
|
[class]{}-[func]{remove}
|
|
|
|
|
|
|
|
|
|
[class]{}-[func]{getInOrderNext}
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
=== "TypeScript"
|
|
|
|
|
|
|
|
|
|
```typescript title="binary_search_tree.ts"
|
|
|
|
|
[class]{}-[func]{remove}
|
|
|
|
|
|
|
|
|
|
[class]{}-[func]{getInOrderNext}
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
=== "C"
|
|
|
|
|
|
|
|
|
|
```c title="binary_search_tree.c"
|
|
|
|
|
[class]{binarySearchTree}-[func]{remove}
|
|
|
|
|
|
|
|
|
|
[class]{binarySearchTree}-[func]{getInOrderNext}
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
=== "C#"
|
|
|
|
|
|
|
|
|
|
```csharp title="binary_search_tree.cs"
|
|
|
|
|
[class]{BinarySearchTree}-[func]{remove}
|
|
|
|
|
|
|
|
|
|
[class]{BinarySearchTree}-[func]{getInOrderNext}
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
=== "Swift"
|
|
|
|
|
|
|
|
|
|
```swift title="binary_search_tree.swift"
|
|
|
|
|
[class]{BinarySearchTree}-[func]{remove}
|
|
|
|
|
|
|
|
|
|
[class]{BinarySearchTree}-[func]{getInOrderNext}
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
=== "Zig"
|
|
|
|
|
|
|
|
|
|
```zig title="binary_search_tree.zig"
|
|
|
|
|
[class]{BinarySearchTree}-[func]{remove}
|
|
|
|
|
|
|
|
|
|
[class]{BinarySearchTree}-[func]{getInOrderNext}
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
### 排序
|
|
|
|
|
|
|
|
|
|
我们知道,「中序遍历」遵循“左 $\rightarrow$ 根 $\rightarrow$ 右”的遍历优先级,而二叉搜索树遵循“左子结点 $<$ 根结点 $<$ 右子结点”的大小关系。因此,在二叉搜索树中进行中序遍历时,总是会优先遍历下一个最小结点,从而得出一条重要性质:**二叉搜索树的中序遍历序列是升序的**。
|
|
|
|
|
|
|
|
|
|
借助中序遍历升序的性质,我们在二叉搜索树中获取有序数据仅需 $O(n)$ 时间,而无需额外排序,非常高效。
|
|
|
|
|
|
|
|
|
|
![二叉搜索树的中序遍历序列](binary_search_tree.assets/bst_inorder_traversal.png)
|
|
|
|
|
|
|
|
|
|
## 二叉搜索树的效率
|
|
|
|
|
|
|
|
|
|
假设给定 $n$ 个数字,最常用的存储方式是「数组」,那么对于这串乱序的数字,常见操作的效率为:
|
|
|
|
|
|
|
|
|
|
- **查找元素**:由于数组是无序的,因此需要遍历数组来确定,使用 $O(n)$ 时间;
|
|
|
|
|
- **插入元素**:只需将元素添加至数组尾部即可,使用 $O(1)$ 时间;
|
|
|
|
|
- **删除元素**:先查找元素,使用 $O(n)$ 时间,再在数组中删除该元素,使用 $O(n)$ 时间;
|
|
|
|
|
- **获取最小 / 最大元素**:需要遍历数组来确定,使用 $O(n)$ 时间;
|
|
|
|
|
|
|
|
|
|
为了得到先验信息,我们也可以预先将数组元素进行排序,得到一个「排序数组」,此时操作效率为:
|
|
|
|
|
|
|
|
|
|
- **查找元素**:由于数组已排序,可以使用二分查找,平均使用 $O(\log n)$ 时间;
|
|
|
|
|
- **插入元素**:先查找插入位置,使用 $O(\log n)$ 时间,再插入到指定位置,使用 $O(n)$ 时间;
|
|
|
|
|
- **删除元素**:先查找元素,使用 $O(\log n)$ 时间,再在数组中删除该元素,使用 $O(n)$ 时间;
|
|
|
|
|
- **获取最小 / 最大元素**:数组头部和尾部元素即是最小和最大元素,使用 $O(1)$ 时间;
|
|
|
|
|
|
|
|
|
|
观察发现,无序数组和有序数组中的各项操作的时间复杂度是“偏科”的,即有的快有的慢;**而二叉搜索树的各项操作的时间复杂度都是对数阶,在数据量 $n$ 很大时有巨大优势**。
|
|
|
|
|
|
|
|
|
|
<div class="center-table" markdown>
|
|
|
|
|
|
|
|
|
|
| | 无序数组 | 有序数组 | 二叉搜索树 |
|
|
|
|
|
| ------------------- | -------- | ----------- | ----------- |
|
|
|
|
|
| 查找指定元素 | $O(n)$ | $O(\log n)$ | $O(\log n)$ |
|
|
|
|
|
| 插入元素 | $O(1)$ | $O(n)$ | $O(\log n)$ |
|
|
|
|
|
| 删除元素 | $O(n)$ | $O(n)$ | $O(\log n)$ |
|
|
|
|
|
| 获取最小 / 最大元素 | $O(n)$ | $O(1)$ | $O(\log n)$ |
|
|
|
|
|
|
|
|
|
|
</div>
|
|
|
|
|
|
|
|
|
|
## 二叉搜索树的退化
|
|
|
|
|
|
|
|
|
|
理想情况下,我们希望二叉搜索树的是“左右平衡”的(详见「平衡二叉树」章节),此时可以在 $\log n$ 轮循环内查找任意结点。
|
|
|
|
|
|
|
|
|
|
如果我们动态地在二叉搜索树中插入与删除结点,**则可能导致二叉树退化为链表**,此时各种操作的时间复杂度也退化之 $O(n)$ 。
|
|
|
|
|
|
|
|
|
|
!!! note
|
|
|
|
|
|
|
|
|
|
在实际应用中,如何保持二叉搜索树的平衡,也是一个需要重要考虑的问题。
|
|
|
|
|
|
|
|
|
|
![二叉搜索树的平衡与退化](binary_search_tree.assets/bst_degradation.png)
|
|
|
|
|
|
|
|
|
|
## 二叉搜索树常见应用
|
|
|
|
|
|
|
|
|
|
- 系统中的多级索引,高效查找、插入、删除操作。
|
|
|
|
|
- 各种搜索算法的底层数据结构。
|
|
|
|
|
- 存储数据流,保持其已排序。
|