最近又写到有向图的拓扑排序,这段排序代码在几年前也写过。做个简单记录,本次用到了google guava的Graph类。
主要是用方法可以参考:https://www.jianshu.com/p/78786a4f2cf1
相关排序可以参考: https://juejin.cn/post/7251844608774209592
图1拓扑排序:
Visited node: A
Visited node: D
Visited node: E
Visited node: B
Visited node: C
Visited node: F
但在实际的工作中,并不是单纯的对节点进行排序或者遍历,比如说常用的就是A有三条下游节点时,则选择其中一条路径执行,其余路径均不可执行。
图2
在拓扑排序的基础上,我们增加不执行节点的判断与处理。具体处理逻辑如下:
public class UnReachableTopologyTraversal {
public static void topologyTraverse(MutableGraph<String> graph, Function<String, Boolean> function) {
// 1.1 拓扑排序需要的辅助队列。
Queue<String> queue = Queues.newLinkedBlockingQueue();
//1.2 初始化本次不执行节点集合
List<String> unReachNodeList = Lists.newArrayList();
// 2. 将root节点(入度为0的节点)入队列。
for (String node : graph.nodes()) {
if (graph.inDegree(node) == 0) {
queue.add(node);
}
}
// 3. 非root节点(入度非0的节点)对应的入度记录表。
Map<String, Integer> nodeIdInDegreeMap = graph.nodes().stream()
.filter(task -> graph.inDegree(task) != 0)
.collect(Collectors.toMap(node -> node, node -> graph.inDegree(node)));
// 4. 拓扑遍历
while (!queue.isEmpty()) {
//5. 随机选取其中一个可执行节点
List<String> allNodes = new ArrayList<>(queue);
Collections.shuffle(allNodes);
queue = new LinkedList<>(allNodes);
String current = queue.poll();
// 6. 判断是不是不可执行节点,如果是不可执行节点,则跳过
if (unReachNodeList.contains(current)) {
continue;
}
//7.1 执行本次节点函数
function.apply(current);
//7.2 后续节点入度减一
subSuccessorInDegree(current, graph, nodeIdInDegreeMap);
//8.1 标记不可执行节点
for (String node : allNodes) {
if (!unReachNodeList.contains(node) && !StringUtils.equals(node,
current)) {
unReachNodeList.add(node);
subSuccessorInDegree(node, graph, nodeIdInDegreeMap);
markSuccessorUnReachable(graph, node, unReachNodeList, nodeIdInDegreeMap);
}
}
//9. 拓扑排序, 将当前节点的后继节点入度减一, 如果入度为0, 则加入队列, 等待调度。
for (String node : graph.successors(current)) {
if (nodeIdInDegreeMap.get(node) == 0) {
queue.add(node);
}
}
}
}
/**
* 将当前节点的后续节点入度-1
* @param currentNode
* @param graph
* @param nodeIdInDegreeMap
*/
private static void subSuccessorInDegree(String currentNode,
Graph<String> graph, Map<String, Integer> nodeIdInDegreeMap) {
Set<String> successorNodes = graph.successors(currentNode);
if (successorNodes == null || successorNodes.size() == 0) {
return;
}
for (String successor : successorNodes) {
Integer inDegree = nodeIdInDegreeMap.get(successor);
nodeIdInDegreeMap.put(successor, inDegree - 1);
}
}
/**
* 标记当前节点的后续节点为不可执行节点
* @param graph
* @param currentNode
* @param unReachNodeList
* @param nodeIdInDegreeMap
*/
private static void markSuccessorUnReachable(Graph<String> graph, String currentNode,
List<String> unReachNodeList, Map<String, Integer> nodeIdInDegreeMap) {
Set<String> afterNodeList = graph.successors(currentNode);
for (String afterNode : afterNodeList) {
boolean unReachable = true;
//当前节点的前置节点如果存在不可执行节点,则不需要递归不可执行
Set<String> preAfterNodeList = graph.predecessors(afterNode);
for (String preAfterNode : preAfterNodeList) {
if (!unReachNodeList.contains(preAfterNode)) {
unReachable = false;
}
}
//后续节点为不可执行节点,继续递归不可执行
if (unReachable) {
unReachNodeList.add(afterNode);
subSuccessorInDegree(afterNode, graph, nodeIdInDegreeMap);
markSuccessorUnReachable(graph, afterNode, unReachNodeList, nodeIdInDegreeMap);
}
}
}
public static void main(String[] args) {
MutableGraph<String> graph = GraphBuilder.directed() //指定为有向图
.nodeOrder(ElementOrder.<String>insertion()) //节点按插入顺序输出
//(还可以取值无序unordered()、节点类型的自然顺序natural())
.expectedNodeCount(6) //预期节点数
.allowsSelfLoops(false) //不允许自环
.build();
graph.putEdge("A", "B");
graph.putEdge("A", "D");
graph.putEdge("A", "E");
graph.putEdge("B", "C");
graph.putEdge("C", "F");
graph.putEdge("D", "F");
graph.addNode("A");
graph.addNode("B");
graph.addNode("C");
graph.addNode("D");
graph.addNode("E");
System.out.println("选择分支执行:");
UnReachableTopologyTraversal.topologyTraverse(graph, node -> {
System.out.println("Visited node: " + node);
});
}
}
结果1
结果2
结果3
网友评论