迭代深度优先树遍历每个节点的访问前和访问后

时间:2011-01-12 00:00:53

标签: algorithm tree-traversal

任何人都可以指出我在伪代码中进行迭代深度优先树遍历,可以在订单前后对每个节点执行操作吗?

也就是说,在进入节点的孩子之前的行动,然后是从孩子们上升后的行动?

此外,我的树不是二进制 - 每个节点都有0..n个孩子。

基本上,我的情况是转换递归遍历,我在当前节点上执行前后操作,在递归到子节点的任一侧。

7 个答案:

答案 0 :(得分:10)

我的计划是使用两个堆栈。一个用于预订遍历,另一个用于后序遍历。 现在,我运行标准的迭代DFS(深度优先遍历),并且只要我从{pre“堆栈pop我将其推入”post“堆栈。 最后,我的“post”堆栈将在顶部具有子节点,在底部具有根节点。

treeSearch(Tree root) {
    Stack pre;
    Stack post;
    pre.add(root);
    while (pre.isNotEmpty()) {
        int x = pre.pop();
        // do pre-order visit here on x
        post.add(x);
        pre.addAll(getChildren(x));
    }
    while (post.isNotEmpty()) {
        int y = post.pop();
        // do post-order visit here on y
    }
}

root将始终从post堆栈遍历,因为它将保持在底部。

这是简单的java代码:

public void treeSearch(Tree tree) {
    Stack<Integer> preStack = new Stack<Integer>();
    Stack<Integer> postStack = new Stack<Integer>();
    preStack.add(tree.root);
    while (!preStack.isEmpty()) {
        int currentNode = preStack.pop();
        // do pre-order visit on current node
        postStack.add(currentNode);
        int[] children = tree.getNeighbours(currentNode);
        for (int child : children) {
            preStack.add(child);
        }
    }

    while (!postStack.isEmpty()) {
        int currentNode = postStack.pop();
        // do post-order visit on current node
    }
}

我假设这是一棵树,所以:没有循环并且再没有重新访问同一个节点。但是,如果我们想要,我们总是可以拥有一个访问过的数组并检查它。

答案 1 :(得分:5)

我意识到这篇文章已经有好几年了,但没有一个答案似乎能直接回答这个问题,所以我想我会写一些简单的东西。

这假设一个整数索引图;但你必须根据需要调整它。迭代地执行DFS并且仍然具有预订/后序操作的关键是不要一次追加每个子项,而是完全像递归DFS一样,这只是添加一个孩子-node一次到堆栈,只有在完成后才从堆栈中删除它们。我通过创建一个包含邻接列表作为堆栈的包装器节点来完成此示例。如果您希望允许循环,则省略循环检查(无论如何它都不会遍历被访问的节点,因此它仍然可以工作)

class Stack(object):
    def __init__(self, l=None):
        if l is None:
            self._l = []
        else:
            self._l = l
        return

    def pop(self):
        return self._l.pop()

    def peek(self):
        return self._l[-1]

    def push(self, value):
        self._l.append(value)
        return

    def __len__(self):
        return len(self._l)

class NodeWrapper(object):
    def __init__(self, graph, v):
        self.v = v
        self.children = Stack(graph[v])
        return

def iterative_postorder(G, s):
    onstack = [False] * len(G)
    edgeto = [None] * len(G)
    visited = [False] * len(G)

    st = Stack()
    st.push(NodeWrapper(G, s))

    while len(st) > 0:
        vnode = st.peek()
        v = vnode.v
        if not onstack[v]:
            print "Starting %d" % (v)
        visited[v] = True
        onstack[v] = True
        if len(vnode.children) > 0:
            e = vnode.children.pop()
            if onstack[e]:
                cycle = [e]
                e = v
                while e != cycle[0]:
                    cycle.append(e)
                    e = edgeto[e]
                raise StandardError("cycle detected: %s, graph not acyclic" % (cycle))
            if not visited[e]:
                edgeto[e] = v
                st.push(NodeWrapper(G, e))
        else:
            vnode = st.pop()
            onstack[vnode.v] = False
            print 'Completed %d' % (vnode.v)

答案 2 :(得分:3)

class Node:
  def __init__( self, value ):
    self.value    = value
    self.children = []

def preprocess( node ):
  print( node.value )

def postprocess( node ):
  print( node.value )

def preorder( root ):
  # Always a flat, homogeneous list of Node instances.
  queue = [ root ]
  while len( queue ) > 0:
    a_node = queue.pop( 0 )
    preprocess( a_node )
    queue = a_node.children + queue

def postorder( root ):
  # Always a flat, homogeneous list of Node instances:
  queue   = [ root ]
  visited = set()
  while len( queue ) > 0:
    a_node = queue.pop( 0 )
    if a_node not in visited:
      visited.add( a_node )
      queue = a_node.children + [ a_node ] + queue
    else:
      # this is either a leaf or a parent whose children have all been processed
      postprocess( a_node )

答案 3 :(得分:1)

我认为通过在El Mariachi提供的后序功能中插入preProcess,我可以得到我所需要的:

def postorder( root ):
 # Always a flat, homogeneous list of Node instances:
 queue   = [ root ]
 visited = set()
 while len( queue ) > 0:
   a_node = queue.pop( 0 )
   if a_node not in visited:
     preprocess( a_node )                  # <<<<<<<< Inserted
     visited.add( a_node )
     queue = a_node.children + [ a_node ] + queue
   else:
     # this is either a leaf or a parent whose children have all been processed
     postprocess( a_node )

答案 4 :(得分:1)

我希望你觉得它很有用。

http://www.vvlasov.com/2013/07/post-order-iterative-dfs-traversal.html

代码:

public void dfsPostOrderIterative(AdjGraph graph, AdjGraph.Node vertex, Callback callback) {
    Stack<Level> toVisit = new Stack<Level>();
    toVisit.push(new Level(Collections.singletonList(vertex)));

    while (!toVisit.isEmpty()) {
        Level level = toVisit.peek();

        if (level.index >= level.nodes.size()) {
            toVisit.pop();
            continue;
        }

        AdjGraph.Node node = level.nodes.get(level.index);

        if (!node.isVisited()) {
            if (node.isChildrenExplored()) {
                node.markVisited();
                callback.nodeVisited(graph, node);
                level.index++;
            } else {
                List<AdjGraph.Node> edges = graph.edges(node);
                List<AdjGraph.Node> outgoing = Lists.newArrayList(Collections2.filter(edges, new Predicate<AdjGraph.Node>() {
                    @Override
                    public boolean apply(AdjGraph.Node input) {
                        return !input.isChildrenExplored();
                    }
                }));

                if (outgoing.size() > 0)
                    toVisit.add(new Level(outgoing));
                node.markChildrenExplored();
            }
        } else {
            level.index++;
        }
    }
}

答案 5 :(得分:1)

一个简单的python实现,有两个不同的访问者

class Print_visitor(object):
    def __init__(self):
        pass
    def pre_visit(self, node):
        print "pre: ", node.value
    def post_visit(self, node):
        print "post:", node.value

class Prettyprint_visitor(object):
    def __init__(self):
        self.level=0
    def pre_visit(self, node):
        print "{}<{}>".format("    "*self.level, node.value)
        self.level += 1
    def post_visit(self, node):
        self.level -= 1
        print "{}</{}>".format("    "*self.level, node.value)

class Node(object):
    def __init__(self, value):
        self.value = value
        self.children = []
    def traverse(self, visitor):
        visitor.pre_visit(self)
        for child in self.children:
            child.traverse(visitor)
        visitor.post_visit(self)

if __name__ == '__main__':
    #test
    tree = Node("root")
    tree.children = [Node("c1"), Node("c2"), Node("c3")]
    tree.children[0].children = [Node("c11"), Node("c12"), Node("c13")]
    tree.children[1].children = [Node("c21"), Node("c22"), Node("c23")]
    tree.children[2].children = [Node("c31"), Node("c32"), Node("c33")]
    tree.traverse(Print_visitor())
    tree.traverse(Prettyprint_visitor())

答案 6 :(得分:1)

非树图的迭代解决方案

在尝试不同的解决方案失败之后,让我为迭代解决方案添加伪代码,在该迭代解决方案中,您实际上在堆栈变量中重新创建了函数调用堆栈空间(在递归版本中可能会溢出)。

您需要存储的所有状态是顶点编号已处理的邻居数量。无论您的语言允许什么,它都可以是一个元组,一个列表,一个对象。

此解决方案的优点是您不会出现堆栈溢出,它也适用于具有周期的图,并且非常健壮。如果使用邻接表或矩阵,则获取下一个邻居很容易。

这是伪代码,因为它更易于理解,您不仅会从SO复制代码,对吗?

globals: isProcessed, preOrder, postOrder

depthFirstSearch()
    set isProcessed to all false
    for each vertex
        if !isProcessed(vertex)
            explore(vertex)

explore(root)
    create stack
    add (root, 0) to stack
    visited = empty list 
        // a list of visited vertices e.g. for finding connected components

    while stack is not empty
        (vertex, processedNeighbors) ← pop from stack

        if !isProcessed(vertex)
            // previsit
            add vertex to preOrder list
            isProcessed(vertex) ← true

        if processedNeighbors < number of vertex's neighbors
            nextNeighborNumber ← processedNeighbors + 1
            push (vertex, nextNeighborNumber) to stack
            nextNeighbor ← 'nextNeighborNumber'th neighbor of vertex

            if !isProcessed(nextNeighbor)
                push (nextNeighbor, 0) to stack
        else
            // postvisit
            add vertex to postOrder list