mirror of
https://github.com/krahets/hello-algo.git
synced 2024-12-28 05:46:29 +08:00
302 lines
9.3 KiB
Markdown
302 lines
9.3 KiB
Markdown
|
---
|
|||
|
comments: true
|
|||
|
---
|
|||
|
|
|||
|
# 9.3. 图的遍历
|
|||
|
|
|||
|
!!! note "图与树的关系"
|
|||
|
|
|||
|
树代表的是“一对多”的关系,而图则自由度更高,可以代表任意“多对多”关系。本质上,**可以把树看作是图的一类特例**。那么显然,树遍历操作也是图遍历操作的一个特例,两者的方法是非常类似的,建议你在学习本章节的过程中将两者融会贯通。
|
|||
|
|
|||
|
「图」与「树」都是非线性数据结构,都需要使用「搜索算法」来实现遍历操作。
|
|||
|
|
|||
|
类似地,图的遍历方式也分为两种,即「广度优先遍历 Breadth-First Traversal」和「深度优先遍历 Depth-First Travsersal」,也称「广度优先搜索 Breadth-First Search」和「深度优先搜索 Depth-First Search」,简称为 BFS 和 DFS 。
|
|||
|
|
|||
|
## 9.3.1. 广度优先遍历
|
|||
|
|
|||
|
**广度优先遍历优是一种由近及远的遍历方式,从距离最近的顶点开始访问,并一层层向外扩张**。具体地,从某个顶点出发,先遍历该顶点的所有邻接顶点,随后遍历下个顶点的所有邻接顶点,以此类推……
|
|||
|
|
|||
|
![graph_bfs](graph_traversal.assets/graph_bfs.png)
|
|||
|
|
|||
|
### 算法实现
|
|||
|
|
|||
|
BFS 常借助「队列」来实现。队列具有“先入先出”的性质,这与 BFS “由近及远”的思想是异曲同工的。
|
|||
|
|
|||
|
1. 将遍历起始顶点 `startVet` 加入队列,并开启循环;
|
|||
|
2. 在循环的每轮迭代中,弹出队首顶点弹出并记录访问,并将该顶点的所有邻接顶点加入到队列尾部;
|
|||
|
3. 循环 `2.` ,直到所有顶点访问完成后结束。
|
|||
|
|
|||
|
为了防止重复遍历顶点,我们需要借助一个哈希表 `visited` 来记录哪些结点已被访问。
|
|||
|
|
|||
|
=== "Java"
|
|||
|
|
|||
|
```java title="graph_bfs.java"
|
|||
|
/* 广度优先遍历 BFS */
|
|||
|
// 使用邻接表来表示图,以便获取指定顶点的所有邻接顶点
|
|||
|
List<Vertex> graphBFS(GraphAdjList graph, Vertex startVet) {
|
|||
|
// 顶点遍历序列
|
|||
|
List<Vertex> res = new ArrayList<>();
|
|||
|
// 哈希表,用于记录已被访问过的顶点
|
|||
|
Set<Vertex> visited = new HashSet<>() {{ add(startVet); }};
|
|||
|
// 队列用于实现 BFS
|
|||
|
Queue<Vertex> que = new LinkedList<>() {{ offer(startVet); }};
|
|||
|
// 以顶点 vet 为起点,循环直至访问完所有顶点
|
|||
|
while (!que.isEmpty()) {
|
|||
|
Vertex vet = que.poll(); // 队首顶点出队
|
|||
|
res.add(vet); // 记录访问顶点
|
|||
|
// 遍历该顶点的所有邻接顶点
|
|||
|
for (Vertex adjVet : graph.adjList.get(vet)) {
|
|||
|
if (visited.contains(adjVet))
|
|||
|
continue; // 跳过已被访问过的顶点
|
|||
|
que.offer(adjVet); // 只入队未访问的顶点
|
|||
|
visited.add(adjVet); // 标记该顶点已被访问
|
|||
|
}
|
|||
|
}
|
|||
|
// 返回顶点遍历序列
|
|||
|
return res;
|
|||
|
}
|
|||
|
```
|
|||
|
|
|||
|
=== "C++"
|
|||
|
|
|||
|
```cpp title="graph_bfs.cpp"
|
|||
|
|
|||
|
```
|
|||
|
|
|||
|
=== "Python"
|
|||
|
|
|||
|
```python title="graph_bfs.py"
|
|||
|
|
|||
|
```
|
|||
|
|
|||
|
=== "Go"
|
|||
|
|
|||
|
```go title="graph_bfs.go"
|
|||
|
|
|||
|
```
|
|||
|
|
|||
|
=== "JavaScript"
|
|||
|
|
|||
|
```javascript title="graph_bfs.js"
|
|||
|
|
|||
|
```
|
|||
|
|
|||
|
=== "TypeScript"
|
|||
|
|
|||
|
```typescript title="graph_bfs.ts"
|
|||
|
|
|||
|
```
|
|||
|
|
|||
|
=== "C"
|
|||
|
|
|||
|
```c title="graph_bfs.c"
|
|||
|
|
|||
|
```
|
|||
|
|
|||
|
=== "C#"
|
|||
|
|
|||
|
```csharp title="graph_bfs.cs"
|
|||
|
|
|||
|
```
|
|||
|
|
|||
|
=== "Swift"
|
|||
|
|
|||
|
```swift title="graph_bfs.swift"
|
|||
|
|
|||
|
```
|
|||
|
|
|||
|
=== "Zig"
|
|||
|
|
|||
|
```zig title="graph_bfs.zig"
|
|||
|
|
|||
|
```
|
|||
|
|
|||
|
代码相对抽象,建议对照以下动画图示来加深理解。
|
|||
|
|
|||
|
=== "Step 1"
|
|||
|
![graph_bfs_step1](graph_traversal.assets/graph_bfs_step1.png)
|
|||
|
|
|||
|
=== "Step 2"
|
|||
|
![graph_bfs_step2](graph_traversal.assets/graph_bfs_step2.png)
|
|||
|
|
|||
|
=== "Step 3"
|
|||
|
![graph_bfs_step3](graph_traversal.assets/graph_bfs_step3.png)
|
|||
|
|
|||
|
=== "Step 4"
|
|||
|
![graph_bfs_step4](graph_traversal.assets/graph_bfs_step4.png)
|
|||
|
|
|||
|
=== "Step 5"
|
|||
|
![graph_bfs_step5](graph_traversal.assets/graph_bfs_step5.png)
|
|||
|
|
|||
|
=== "Step 6"
|
|||
|
![graph_bfs_step6](graph_traversal.assets/graph_bfs_step6.png)
|
|||
|
|
|||
|
=== "Step 7"
|
|||
|
![graph_bfs_step7](graph_traversal.assets/graph_bfs_step7.png)
|
|||
|
|
|||
|
=== "Step 8"
|
|||
|
![graph_bfs_step8](graph_traversal.assets/graph_bfs_step8.png)
|
|||
|
|
|||
|
=== "Step 9"
|
|||
|
![graph_bfs_step9](graph_traversal.assets/graph_bfs_step9.png)
|
|||
|
|
|||
|
=== "Step 10"
|
|||
|
![graph_bfs_step10](graph_traversal.assets/graph_bfs_step10.png)
|
|||
|
|
|||
|
=== "Step 11"
|
|||
|
![graph_bfs_step11](graph_traversal.assets/graph_bfs_step11.png)
|
|||
|
|
|||
|
!!! question "广度优先遍历的序列是否唯一?"
|
|||
|
|
|||
|
不唯一。广度优先遍历只要求“由近及远”,而相同距离的多个顶点的遍历顺序允许任意被打乱。以上图为例,顶点 $1$ , $3$ 的访问顺序可以交换、顶点 $2$ , $4$ , $6$ 的访问顺序也可以任意交换、以此类推……
|
|||
|
|
|||
|
### 复杂度分析
|
|||
|
|
|||
|
**时间复杂度:** 所有顶点都会入队、出队一次,使用 $O(|V|)$ 时间;在遍历邻接顶点的过程中,由于是无向图,因此所有边都会被访问 $2$ 次,使用 $O(2|E|)$ 时间;总体使用 $O(|V| + |E|)$ 时间。
|
|||
|
|
|||
|
**空间复杂度:** 列表 `res` ,哈希表 `visited` ,队列 `que` 中的顶点数量最多为 $|V|$ ,使用 $O(|V|)$ 空间。
|
|||
|
|
|||
|
## 9.3.2. 深度优先遍历
|
|||
|
|
|||
|
**深度优先遍历是一种优先走到底、无路可走再回头的遍历方式**。具体地,从某个顶点出发,不断地访问当前结点的某个邻接顶点,直到走到尽头时回溯,再继续走到底 + 回溯,以此类推……直至所有顶点遍历完成时结束。
|
|||
|
|
|||
|
![graph_dfs](graph_traversal.assets/graph_dfs.png)
|
|||
|
|
|||
|
### 算法实现
|
|||
|
|
|||
|
这种“走到头 + 回溯”的算法形式一般基于递归来实现。与 BFS 类似,在 DFS 中我们也需要借助一个哈希表 `visited` 来记录已被访问的顶点,以避免重复访问顶点。
|
|||
|
|
|||
|
=== "Java"
|
|||
|
|
|||
|
```java title="graph_dfs.java"
|
|||
|
/* 深度优先遍历 DFS 辅助函数 */
|
|||
|
void dfs(GraphAdjList graph, Set<Vertex> visited, List<Vertex> res, Vertex vet) {
|
|||
|
res.add(vet); // 记录访问顶点
|
|||
|
visited.add(vet); // 标记该顶点已被访问
|
|||
|
// 遍历该顶点的所有邻接顶点
|
|||
|
for (Vertex adjVet : graph.adjList.get(vet)) {
|
|||
|
if (visited.contains(adjVet))
|
|||
|
continue; // 跳过已被访问过的顶点
|
|||
|
// 递归访问邻接顶点
|
|||
|
dfs(graph, visited, res, adjVet);
|
|||
|
}
|
|||
|
}
|
|||
|
|
|||
|
/* 深度优先遍历 DFS */
|
|||
|
// 使用邻接表来表示图,以便获取指定顶点的所有邻接顶点
|
|||
|
List<Vertex> graphDFS(GraphAdjList graph, Vertex startVet) {
|
|||
|
// 顶点遍历序列
|
|||
|
List<Vertex> res = new ArrayList<>();
|
|||
|
// 哈希表,用于记录已被访问过的顶点
|
|||
|
Set<Vertex> visited = new HashSet<>();
|
|||
|
dfs(graph, visited, res, startVet);
|
|||
|
return res;
|
|||
|
}
|
|||
|
```
|
|||
|
|
|||
|
=== "C++"
|
|||
|
|
|||
|
```cpp title="graph_dfs.cpp"
|
|||
|
|
|||
|
```
|
|||
|
|
|||
|
=== "Python"
|
|||
|
|
|||
|
```python title="graph_dfs.py"
|
|||
|
|
|||
|
```
|
|||
|
|
|||
|
=== "Go"
|
|||
|
|
|||
|
```go title="graph_dfs.go"
|
|||
|
|
|||
|
```
|
|||
|
|
|||
|
=== "JavaScript"
|
|||
|
|
|||
|
```javascript title="graph_dfs.js"
|
|||
|
|
|||
|
```
|
|||
|
|
|||
|
=== "TypeScript"
|
|||
|
|
|||
|
```typescript title="graph_dfs.ts"
|
|||
|
|
|||
|
```
|
|||
|
|
|||
|
=== "C"
|
|||
|
|
|||
|
```c title="graph_dfs.c"
|
|||
|
|
|||
|
```
|
|||
|
|
|||
|
=== "C#"
|
|||
|
|
|||
|
```csharp title="graph_dfs.cs"
|
|||
|
|
|||
|
```
|
|||
|
|
|||
|
=== "Swift"
|
|||
|
|
|||
|
```swift title="graph_dfs.swift"
|
|||
|
|
|||
|
```
|
|||
|
|
|||
|
=== "Zig"
|
|||
|
|
|||
|
```zig title="graph_dfs.zig"
|
|||
|
|
|||
|
```
|
|||
|
|
|||
|
深度优先遍历的算法流程如下图所示,其中
|
|||
|
|
|||
|
- **直虚线代表向下递推**,代表开启了一个新的递归方法来访问新顶点;
|
|||
|
- **曲虚线代表向上回溯**,代表此递归方法已经返回,回溯到了开启此递归方法的位置;
|
|||
|
|
|||
|
为了加深理解,请你将图示与代码结合起来,在脑中(或者用笔画下来)模拟整个 DFS 过程,包括每个递归方法何时开启、何时返回。
|
|||
|
|
|||
|
=== "Step 1"
|
|||
|
![graph_dfs_step1](graph_traversal.assets/graph_dfs_step1.png)
|
|||
|
|
|||
|
=== "Step 2"
|
|||
|
![graph_dfs_step2](graph_traversal.assets/graph_dfs_step2.png)
|
|||
|
|
|||
|
=== "Step 3"
|
|||
|
![graph_dfs_step3](graph_traversal.assets/graph_dfs_step3.png)
|
|||
|
|
|||
|
=== "Step 4"
|
|||
|
![graph_dfs_step4](graph_traversal.assets/graph_dfs_step4.png)
|
|||
|
|
|||
|
=== "Step 5"
|
|||
|
![graph_dfs_step5](graph_traversal.assets/graph_dfs_step5.png)
|
|||
|
|
|||
|
=== "Step 6"
|
|||
|
![graph_dfs_step6](graph_traversal.assets/graph_dfs_step6.png)
|
|||
|
|
|||
|
=== "Step 7"
|
|||
|
![graph_dfs_step7](graph_traversal.assets/graph_dfs_step7.png)
|
|||
|
|
|||
|
=== "Step 8"
|
|||
|
![graph_dfs_step8](graph_traversal.assets/graph_dfs_step8.png)
|
|||
|
|
|||
|
=== "Step 9"
|
|||
|
![graph_dfs_step9](graph_traversal.assets/graph_dfs_step9.png)
|
|||
|
|
|||
|
=== "Step 10"
|
|||
|
![graph_dfs_step10](graph_traversal.assets/graph_dfs_step10.png)
|
|||
|
|
|||
|
=== "Step 11"
|
|||
|
![graph_dfs_step11](graph_traversal.assets/graph_dfs_step11.png)
|
|||
|
|
|||
|
!!! question "深度优先遍历的序列是否唯一?"
|
|||
|
|
|||
|
与广度优先遍历类似,深度优先遍历序列的顺序也不是唯一的。给定某顶点,先往哪个方向探索都行,都是深度优先遍历。
|
|||
|
|
|||
|
以树的遍历为例,“根 $\rightarrow$ 左 $\rightarrow$ 右”、“左 $\rightarrow$ 根 $\rightarrow$ 右”、“左 $\rightarrow$ 右 $\rightarrow$ 根”分别对应前序、中序、后序遍历,体现三种不同的遍历优先级,而三者都属于深度优先遍历。
|
|||
|
|
|||
|
### 复杂度分析
|
|||
|
|
|||
|
**时间复杂度:** 所有顶点都被访问一次;所有边都被访问了 $2$ 次,使用 $O(2|E|)$ 时间;总体使用 $O(|V| + |E|)$ 时间。
|
|||
|
|
|||
|
**空间复杂度:** 列表 `res` ,哈希表 `visited` 顶点数量最多为 $|V|$ ,递归深度最大为 $|V|$ ,因此使用 $O(|V|)$ 空间。
|