Custom Data Structure for push, pop and finding minimum - algorithm

I was just asked an interview question with company A as follows:
Question : Design a data structure in which you have 3 operations, push, pop and find the minimum. You should be doing all the 3 operations in constant time.
My Answer : I would use a linked list in which i can do insertion and removal in constant time and i would use an extra memory to store the minimum.
He came up with a second question saying, if you pop the minimum, how do you find the second minimum? again, in constant time.
What would you tell him?

Minimum stack question - http://courses.csail.mit.edu/iap/interview/Hacking_a_Google_Interview_Practice_Questions_Person_A.pdf
From the PDF:
Question: Minimum Stack
Describe a stack data structure that supports "push", "pop", and "find minimum"
operations. "Find minimum" returns the smallest element in the stack.
Good Answer: Store two stacks, one of which contains all of the items in the stack
and one of which is a stack of minima. To push an element, push it onto the first
stack. Check whether it is smaller than the top item on the second stack; if so, push
it onto the second stack. To pop an item, pop it from the first stack. If it is the top
element of the second stack, pop it from the second stack. To find the minimum
element, simply return the element on the top of the second stack. Each operation
takes O(1) time.

What if you do a linked list, like you said, but also store the current minimum value. When you add a new number it looks at the previous min and changes the current min to the current value if the current value is lower.
E.g... Assume you have the data: 3, 6, 4, 2, 7, 1. Then this is what the list would look like
value|min
3|3 -> 6|3 -> 4|3 -> 2|2 -> 7|2 -> 1|1
That'll keep track of the mins as you add/remove items.
EDIT:
I mentioned going backwards, it would be something like this:
1|1 -> 7|2 -> 2|2 -> 4|3 -> 6|3 -> 3|3
Then you wouln't need the "footer".

Let every node keep another reference to the previously smallest item. So, when you pop the smallest item, you can restore the previously smallest. Because you can only push and pop it will be the correct node.

Here is the C code which implements the above algorithm given by Bryce Siedschlaw:
#include<stdio.h>
#include<stdlib.h>
#define minimumOf(a,b) (a<b) ? a : b;
struct node{
int data;
struct node* next;
int min;
};
void push(struct node **head_ref, int new_data){
struct node* new_node = (struct node *)malloc(sizeof(struct node));
new_node->data = new_data;
if(*head_ref == NULL)
new_node->min = new_data;
else
new_node->min = minimumOf(new_data,(*head_ref)->min);
new_node->next = *head_ref;
*head_ref = new_node;
}
int minimum(struct node *head){
return head->min;
}
int pop(struct node **head_ref){
int pop_data = (*head_ref)->data;
(*head_ref) = (*head_ref)->next;
return pop_data;
}
void printList(node *head){
while(head != NULL){
printf("%d->", head->data);
head = head->next;
}
printf("\b\n");
}
int main(){
struct node* a = NULL;
push(&a, 100);
push(&a, 24);
push(&a, 16);
push(&a, 19);
push(&a, 50);
printList(a);
printf("Minimum is:%d\n", minimum(a));
printf("Popped:%d\n",pop(&a));
printf("Minimum is:%d\n", minimum(a));
printf("Popped:%d\n",pop(&a));
printf("Minimum is:%d\n", minimum(a));
printf("Popped:%d\n",pop(&a));
printf("Minimum is:%d\n", minimum(a));
printf("Popped:%d\n",pop(&a));
printf("Minimum is:%d\n", minimum(a));
}

This will make you go wild - http://algods-cracker.blogspot.in/2011/09/design-data-structure-which-supports.html

Related

Singly linked list Tail

if you want create a singly linked list like this:
struct Node {
int data;
Node *next;
};
struct List{
Node *head;
// Node *tail; --> necessary?
Node *last;
};
And this list has the methods "append", "remove", "printList" and "findElement".
Is it necessary to have a tail? Because with "last" you can address the last node.
So when it is necessary to have all three Nodes "head", "tail" and "last"? When you want to insert the node sorted into the list for example?
No, it's not necessary. The tail is equal to head->next and thus it would be redundant and add bookkeeping overhead to keep this field updated.
Also note that the field last is kind of unusual. In most use cases, you add elements to the head of a singly linked list and use a different data structure when you really need to add to the end.
Actually, you can implement enqueue (append at tail), push (prepend at head), dequeue (remove from head), and of course find and print with with a one-pointer header. The trick is to make the list circular and have the header point to the tail. Then tail->next is the head.
#include <stdio.h>
#include <stdlib.h>
typedef struct node_s {
struct node_s *next;
int data;
} Node;
typedef struct list_s {
Node *tail;
} List;
Node *new_node(int data) {
Node *node = malloc(sizeof *node);
node->data = data;
node->next = node;
return node;
}
void init_list(List *list) {
list->tail = NULL;
}
int is_empty(List *list) {
return list->tail == NULL;
}
void enqueue(List *list, Node *node) {
if (list->tail) {
Node *head = list->tail->next;
node->next = head;
list->tail->next = node;
list->tail = node;
} else list->tail = node->next = node;
}
void push(List *list, Node *node) {
if (list->tail) {
Node *head = list->tail->next;
node->next = head;
list->tail->next = node;
} else list->tail = node->next = node;
}
Node *dequeue(List *list) {
Node *head = list->tail->next;
if (head == list->tail)
list->tail = NULL;
else
list->tail->next = head->next;
return head;
}
void print_list(List *list) {
printf("The list:\n");
if (list->tail) {
Node *head = list->tail->next;
Node *p = head;
do {
printf("%d\n", p->data);
p = p->next;
} while (p != head);
}
}
int main(int argc, char *argv[]) {
List list[1];
init_list(list);
// Build the list in order and print it.
for (int i = 0; i < 4; i++) enqueue(list, new_node(i));
print_list(list);
// Remove elements from head until empty.
printf("Dequeueing:\n");
while (!is_empty(list)) {
Node *node = dequeue(list);
printf("%d\n", node->data);
free(node);
}
// Build the list in reverse order and print it.
for (int i = 0; i < 4; i++) push(list, new_node(i));
print_list(list);
return 0;
}
I think it depends on what operations you want to use.
Assuming you want to insert and delete nodes at the tail of a list, it is certainly a wise choice to keep a last node in your list.
Otherwise, if you want to do operations at the beginning of the list, a last node is unnecessary.
It's not necessary but a tail can be useful if you're working with the linked list in a queue-like FIFO fashion rather than a stack-like LIFO fashion or want to be able to transfer entire lists of elements from one head to another's tail without disrupting the relative order of the elements.
Note that I'm referring to 'tail' as a reference to the last node in the list which I believe is safe to assume that the question is about.
A lot of very micro-optimized SLL implementations often are tail-less and work like a stack while backed by an efficient fixed allocator for locality of reference (cache-friendliness) and faster node allocations/deallocations. There the primary benefit of the SLL over a variable-sized array-based sequence is the ability to start moving things around by just changing the value of the next pointer/reference and the lack of invalidation on inserting/removing elements if you're working in native, lower-level languages that involve pointers. The lack of a tail can boost performance quite a bit by reducing the amount of branching instructions required in operations to push and pop from the stack.
For the needs you listed, whether the tail is going to help or just add unnecessary complexity and overhead is if your append and remove operations can work strictly from the front in a LIFO stack fashion or if you want to be able to append to the back but remove from the front in a FIFO fashion without any iteration involved, e.g. If you don't have a tail in the latter case, one of these operations are going to go from constant-time complexity to linear-time complexity, and you might improve your use cases by exchanging that linear-time algorithmic complexity for the relatively smaller micro-level overhead of maintaining a tail.

how to find lowest common ancestor of a nary tree?

Is there a way without using extra space to find LCA of nary tree.
I did it using a string saving the preorder of both the nodes and finding common prefix
If nodes "know" their depth - or you're willing to allow the space to compute the depth of your nodes, you can back up from the lower node to the same depth of the higher node, and then go up one level at a time until they meet.
Depends on what "extra space" means in this context. You can do it with one integer - the difference in depths of the two nodes. Is that too much space?
Another possibility is given you don't have a parent pointer, you can use pointer reversal - every time you traverse a pointer, remember the location from which you came, remember the pointer you will next traverse, and then just before the next pointer traversal, replace that pointer with the back pointer. You have to reverse this when going up the tree to restore it. This takes the space of one pointer as a temporary. And another integer to keep the depth as you work your way down and up. Do this synchronously for the two nodes you seek, so that you can work your way back up from the lower one until you're at the same height in both traversals, and then work back up from both until you're at the common node. This takes three extra pieces of memory - one for each of the current depths, one for the temporary used during a pointer reversal. Very space efficient. Is it worth it?
Go back and do it for a binary tree. If you can do it for a binary tree you can do it for an n-ary tree.
Here's a link to LCA in a binary tree:
And here's how it looks after converting it to a n-ary tree LCA:
public class LCA {
public static <V> Node<V>
lowestCommonAncestor(Node<V> argRoot, Node<V> a, Node<V> b) {
if (argRoot == null) {
return null;
}
if (argRoot.equals(a) || argRoot.equals(b)) {
// if at least one matched, no need to continue
// this is the LCA for this root
return argRoot;
}
Iterator<Node<V>> it = argRoot.childIterator();
// nr of branches that a or b are on,
// could be max 2 (considering unique nodes)
int i = 0;
Node<V> lastFoundLCA = null;
while (it.hasNext()) {
Node<V> node = lowestCommonAncestor(it.next(), a, b);
if (node != null) {
lastFoundLCA = node;
i++ ;
}
if (i >= 2) {
return argRoot;
}
}
return lastFoundLCA;
}
}
Do a synchronous walk to both the nodes.
Start with LCA=root;
loop:
find the step to take for A and the step for B
if these are equal { LCA= the step; decend A; descend B; goto loop; }
done: LCA now contains the lca for A and B
Pseudocode in C:
struct node {
struct node *offspring[1234];
int payload;
};
/* compare function returning the slot in which this should be found/placed */
int find_index (struct node *par, struct node *this);
struct node *lca(struct node *root, struct node *one, struct node *two)
{
struct node *lca;
int idx1,idx2;
for (lca=root; lca; lca=lca->offspring[idx1] ) {
idx1 = find_index(lca, one);
idx2 = find_index(lca, two);
if (idx1 != idx2 || idx1 < 0) break;
if (lca->offspring[idx1] == NULL) break;
}
return lca;
}

Binary tree level order traversal

Three types of tree traversals are inorder, preorder, and post order.
A fourth, less often used, traversal is level-order traversal. In a
level-order traveresal, all nodes at depth "d" are processed before
any node at depth d + 1. Level-order traversal differs from the other
traversals in that it is not done recursively; a queue is used,
instead of the implied stack of recursion.
My questions on above text snippet are
Why level order traversals are not done recursively?
How queue is used in level order traversal? Request clarification with Pseudo code will be helpful.
Thanks!
Level order traversal is actually a BFS, which is not recursive by nature. It uses Queue instead of Stack to hold the next vertices that should be opened. The reason for it is in this traversal, you want to open the nodes in a FIFO order, instead of a LIFO order, obtained by recursion
as I mentioned, the level order is actually a BFS, and its [BFS] pseudo code [taken from wikipedia] is:
1 procedure BFS(Graph,source):
2 create a queue Q
3 enqueue source onto Q
4 mark source
5 while Q is not empty:
6 dequeue an item from Q into v
7 for each edge e incident on v in Graph:
8 let w be the other end of e
9 if w is not marked:
10 mark w
11 enqueue w onto Q
(*) in a tree, marking the vertices is not needed, since you cannot get to the same node in 2 different paths.
void levelorder(Node *n)
{ queue < Node * >q;
q.push(n);
while(!q.empty())
{
Node *node = q.front();
cout<<node->value;
q.pop();
if(node->left != NULL)
q.push(node->left);
if (node->right != NULL)
q.push(node->right);
}
}
Instead of a queue, I used a map to solve this. Take a look, if you are interested. As I do a postorder traversal, I maintain the depth at which each node is positioned and use this depth as the key in a map to collect values in the same level
class Solution {
public:
map<int, vector<int> > levelValues;
void recursivePrint(TreeNode *root, int depth){
if(root == NULL)
return;
if(levelValues.count(root->val) == 0)
levelValues.insert(make_pair(depth, vector<int>()));
levelValues[depth].push_back(root->val);
recursivePrint(root->left, depth+1);
recursivePrint(root->right, depth+1);
}
vector<vector<int> > levelOrder(TreeNode *root) {
recursivePrint(root, 1);
vector<vector<int> > result;
for(map<int,vector<int> >::iterator it = levelValues.begin(); it!= levelValues.end(); ++it){
result.push_back(it->second);
}
return result;
}
};
The entire solution can be found here - http://ideone.com/zFMGKU
The solution returns a vector of vectors with each inner vector containing the elements in the tree in the correct order.
you can try solving it here - https://oj.leetcode.com/problems/binary-tree-level-order-traversal/
And, as you can see, we can also do this recursively in the same time and space complexity as the queue solution!
My questions on above text snippet are
Why level order traversals are not done recursively?
How queue is used in level order traversal? Request clarification with Pseudo code will be helpful.
I think it'd actually be easier to start with the second question. Once you understand the answer to the second question, you'll be better prepared to understand the answer to the first.
How level order traversal works
I think the best way to understand how level order traversal works is to go through the execution step by step, so let's do that.
We have a tree.
We want to traverse it level by level.
So, the order that we'd visit the nodes would be A B C D E F G.
To do this, we use a queue. Remember, queues are first in, first out (FIFO). I like to imagine that the nodes are waiting in line to be processed by an attendant.
Let's start by putting the first node A into the queue.
Ok. Buckle up. The setup is over. We're about to start diving in.
The first step is to take A out of the queue so it can be processed. But wait! Before we do so, let's put A's children, B and C, into the queue also.
Note: A isn't actually in the queue anymore at this point. I grayed it out to try to communicate this. If I removed it completely from the diagram, it'd make it harder to visualize what's happening later on in the story.
Note: A is being processed by the attendant at the desk in the diagram. In real life, processing a node can mean a lot of things. Using it to compute a sum, send an SMS, log to the console, etc, etc. Going off the metaphor in my diagram, you can tell the attendant how you want them to process the node.
Now we move on to the node that is next in line. In this case, B.
We do the same thing that we did with A: 1) add the children to the line, and 2) process the node.
Hey, check it out! It looks like what we're doing here is going to get us that level order traversal that we were looking for! Let's prove this to ourselves by continuing the step through.
Once we finish with B, C is next in line. We place C's children at the back of the line, and then process C.
Now let's see what happens next. D is next in line. D doesn't have any children, so we don't place anything at the back of the line. We just process D.
And then it's the same thing for E, F, and G.
Why it's not done recursively
Imagine what would happen if we used a stack instead of a queue. Let's rewind to the point where we had just visited A.
Here's how it'd look if we were using a stack.
Now, instead of going "in order", this new attendant likes to serve the most recent clients first, not the ones who have been waiting the longest. So C is who is up next, not B.
Here's where the key point is. Where the stack starts to cause a different processing order than we had with the queue.
Like before, we add C's children and then process C. We're just adding them to a stack instead of a queue this time.
Now, what's next? This new attendant likes to serve the most recent clients first (ie. we're using a stack), so G is up next.
I'll stop the execution here. The point is that something as simple as replacing the queue with a stack actually gives us a totally different execution order. I'd encourage you to finish the step through though.
You might be thinking: "Ok... but the question asked about recursion. What does this have to do with recursion?" Well, when you use recursion, something sneaky is going on. You never did anything with a stack data structure like s = new Stack(). However, the runtime uses the call stack. This ends up being conceptually similar to what I did above, and thus doesn't give us that A B C D E F G ordering we were looking for from level order traversal.
https://github.com/arun2pratap/data-structure/blob/master/src/main/java/com/ds/tree/binarytree/BinaryTree.java
for complete can look out for the above link.
public void levelOrderTreeTraversal(List<Node<T>> nodes){
if(nodes == null || nodes.isEmpty()){
return;
}
List<Node<T>> levelNodes = new ArrayList<>();
nodes.stream().forEach(node -> {
if(node != null) {
System.out.print(" " + node.value);
levelNodes.add(node.left);
levelNodes.add(node.right);
}
});
System.out.println("");
levelOrderTreeTraversal(levelNodes);
}
Also can check out
http://www.geeksforgeeks.org/
here you will find Almost all Data Structure related answers.
Level order traversal implemented by queue
# class TreeNode:
# def __init__(self, val=0, left=None, right=None):
# self.val = val
# self.left = left
# self.right = right
def levelOrder(root: TreeNode) -> List[List[int]]:
res = [] # store the node value
queue = [root]
while queue:
node = queue.pop()
# visit the node
res.append(node.val)
if node.left:
queue.insert(0, node.left)
if node.right:
queue.insert(0, node.right)
return res
Recursive implementation is also possible. However, it needs to know the max depth of the root in advance.
def levelOrder(root: TreeNode) -> List[int]:
res = []
max_depth = maxDepth(root)
for i in range(max_depth):
# level start from 0 to max_depth-1
visitLevel(root, i, action)
return res
def visitLevel(root:TreeNode, level:int, res: List):
if not root:
return
if level==0:
res.append(node.val)
else:
self.visitLevel(root.left, level-1, res)
self.visitLevel(root.right, level-1, res)
def maxDepth(root: TreeNode) -> int:
if not root:
return 0
if not root.left and not root.right:
return 1
return max([ maxDepth(root.left), maxDepth(root.right)]) + 1
For your point 1) we can use Java below code for level order traversal in recursive order, we have not used any library function for tree, all are user defined tree and tree specific functions -
class Node
{
int data;
Node left, right;
public Node(int item)
{
data = item;
left = right = null;
}
boolean isLeaf() { return left == null ? right == null : false; }
}
public class BinaryTree {
Node root;
Queue<Node> nodeQueue = new ConcurrentLinkedDeque<>();
public BinaryTree() {
root = null;
}
public static void main(String args[]) {
BinaryTree tree = new BinaryTree();
tree.root = new Node(1);
tree.root.left = new Node(2);
tree.root.right = new Node(3);
tree.root.left.left = new Node(4);
tree.root.left.right = new Node(5);
tree.root.right.left = new Node(6);
tree.root.right.right = new Node(7);
tree.root.right.left.left = new Node(8);
tree.root.right.left.right = new Node(9);
tree.printLevelOrder();
}
/*Level order traversal*/
void printLevelOrder() {
int h = height(root);
int i;
for (i = 1; i <= h; i++)
printGivenLevel(root, i);
System.out.println("\n");
}
void printGivenLevel(Node root, int level) {
if (root == null)
return;
if (level == 1)
System.out.print(root.data + " ");
else if (level > 1) {
printGivenLevel(root.left, level - 1);
printGivenLevel(root.right, level - 1);
}
}
/*Height of Binary tree*/
int height(Node root) {
if (root == null)
return 0;
else {
int lHeight = height(root.left);
int rHeight = height(root.right);
if (lHeight > rHeight)
return (lHeight + 1);
else return (rHeight + 1);
}
}
}
For your point 2) If you want to use non recursive function then you can use queue as below function-
public void levelOrder_traversal_nrec(Node node){
System.out.println("Level order traversal !!! ");
if(node == null){
System.out.println("Tree is empty");
return;
}
nodeQueue.add(node);
while (!nodeQueue.isEmpty()){
node = nodeQueue.remove();
System.out.printf("%s ",node.data);
if(node.left !=null)
nodeQueue.add(node.left);
if (node.right !=null)
nodeQueue.add(node.right);
}
System.out.println("\n");
}
Recursive Solution in C++
/**
* Definition for a binary tree node.
* struct TreeNode {
* int val;
* TreeNode *left;
* TreeNode *right;
* TreeNode() : val(0), left(nullptr), right(nullptr) {}
* TreeNode(int x) : val(x), left(nullptr), right(nullptr) {}
* TreeNode(int x, TreeNode *left, TreeNode *right) : val(x), left(left), right(right) {}
* };
*/
class Solution {
public:
vector<vector<int>> levels;
void helper(TreeNode* node,int level)
{
if(levels.size() == level) levels.push_back({});
levels[level].push_back(node->val);
if(node->left)
helper(node->left,level+1);
if(node->right)
helper(node->right,level+1);
}
vector<vector<int>> levelOrder(TreeNode* root) {
if(!root) return levels;
helper(root,0);
return levels;
}
};
We can use queue to solve this problem in less time complexity. Here is the solution of level order traversal suing Java.
class Solution {
public List<List<Integer>> levelOrder(TreeNode root) {
List<List<Integer>> levelOrderTraversal = new ArrayList<List<Integer>>();
List<Integer> currentLevel = new ArrayList<Integer>();
Queue<TreeNode> queue = new LinkedList<TreeNode>();
if(root != null)
{
queue.add(root);
queue.add(null);
}
while(!queue.isEmpty())
{
TreeNode queueRoot = queue.poll();
if(queueRoot != null)
{
currentLevel.add(queueRoot.val);
if(queueRoot.left != null)
{
queue.add(queueRoot.left);
}
if(queueRoot.right != null)
{
queue.add(queueRoot.right);
}
}
else
{
levelOrderTraversal.add(currentLevel);
if(!queue.isEmpty())
{
currentLevel = new ArrayList<Integer>();
queue.add(null);
}
}
}
return levelOrderTraversal;
}
}

how to find number of nodes in loop of linked list?

how to find the number of nodes in a loop of linked list?
for e.g
A ----> B ----> C -----> D -----> E
Λ |
| |
| V
H <----- G <----- F
Find the number of nodes in a loop from C to H
Fundamental problem is how to find point C. We can use traditional hare and tortoise algo but it does not meet every time at point C.
See here more solutions for how to find a loop in a linked list. Adding the nodes counting is pretty simple then. (Although The Tortoise and the Hare is probably the best one)
If you simply want to find the answer, do the tortoise-hare to determine at what point there is definitely a loop. Then start a counter, and count how many iterations you must make to reach the point that you first found. This may not be the most efficient possible, but it gives a correct answer.
Some C++ code:
#include <iostream>
struct node
{
node(node* next)
: next(next)
{ }
node* next;
};
int main(int argc, char* argv[])
{
node h(NULL), g(&h), f(&g), e(&f), d(&e), c(&d), b(&c), a(&b);
h.next = &c;
node* tortoise = &a;
node* hare = &b;
while(tortoise != hare)
{
tortoise = tortoise->next;
hare = hare->next->next;
}
int count = 1;
tortoise = tortoise->next;
while(tortoise != hare)
{
++count;
tortoise = tortoise->next;
}
std::cout << "Size of cycle: " << count << "\n";
return 0;
}
Note that you'll have to do some extra work to determine if you hit the end, rather than looping, in the case that you don't actually have a cycle. Traditional tortoise-hare should take care of that:
http://en.wikipedia.org/wiki/Cycle_detection
List visited;
List toVisit;
toVisit.add(A); // add the first Node
while(toVisit is not empty){
Node current = visited.remove();
Array <Node> links = current.getLinks();
for(int i=0; i<links.size(); i++){
if(!visited.contains(links[i])){ // if the link has NOT already been visited add it to the toVisit List
toVisit.add(links[i]);
}
visited.add(current); // mark current as visited
}
}
return visited.size(); // to get the number of nodes in the graph
I don't think that I would consider this a linkedList. LinkedLists usually end with a null pointer or a pointer pointing to an ending symbol. Ie: start -> A -> B -> C -> end. I think that this would be a specific kind of graph.
To find the total number of nodes in the graph I would do this:
List visited;
List toVisit;
toVisit.add(A); // add the first Node
while(toVisit is not empty){
Node current = visited.remove();
Array <Node> links = current.getLinks();
for(int i=0; i<links.size(); i++){
if(!visited.contains(links[i])){ // if the link has NOT already been visited add it to the toVisit List
toVisit.add(links[i]);
}
visited.add(current); // mark current as visited
}
}
return visited.size(); // to get the number of nodes in the graph
If you always know that there will one loop like (note the ...):
A ---> ... ---> C -----> D -----> E
Λ |
| |
| V
... <----- G <--- F
You could modify the code like this:
List visited;
Node current = firstNode;
while(!visited.contains(firstNode)){
Node next = current.getNext();
visited.add(current); // mark current as visited
current=next;
}
// our ending condition is when we have found the same node again.
int currentIndex = visited.indexOf(current);
int size = visited.size();
int sizeOfLoop = size - currentIndex;
return sizeOfLoop;
1) flyod alogo find the loop
2) when slow_ptr=fast_ptr , find the number of nodes in loop (k)
Additionally you can also go to C like this:-
3) start 2 ptr , one from head and another from head+k.
4) You will meet at starting of Loop (C)

Link Tree nodes at each level

Given a binary tree, how would you join the nodes at each level, left to right.
Say there are 5 nodes at level three, link all of them from left to right.
I don't need anybody to write code for this.. but just an efficient algorithm.
Thanks
Idea is:
1. Traverse tree with BFS.
2. When you do traversing, you're linking nodes on next level - if node has left and right node, you'll link left to right. If node has next node, then you link rightmost child of current node to leftmost child of next node.
public void BreadthFirstSearch(Action<Node> currentNodeAction)
{
Queue<Node> q = new Queue<Node>();
q.Enqueue(root);
while (q.Count != 0)
{
Node current = q.Dequeue();
if (currentNodeAction != null)
currentNodeAction(current);
if (current.left != null) q.Enqueue(current.left);
if (current.right != null) q.Enqueue(current.right);
}
}
private void Linker(Node node)
{
Link(node.left, node.right);
if (node.next != null)
Link(node.right ?? node.left, node.next.left ?? node.next.right);
}
private void Link(Node node1, Node node2)
{
if (node1 != null && node2 != null)
node1.next = node2;
}
public void LinkSameLevel()
{
BreadthFirstSearch(Linker);
}
Create a vector of linked lists.
Do a DFS keeping track of your level, and for each node you find, add it to the linked list of the level.
This will run in O(n) which is optimal.
Is this what you want to do?
This is not a direct answer to the question and may not be applicable based on your situation. But if you have control over the creation and maintenance of the binary tree, it would probably be more efficient to maintain the links while building/updating the tree.
If you kept both left and right pointers at each level, then it would be "simple" (always easy to say that word when someone else is doing the work) to maintain them. When inserting a new node at a given level, you know its direct siblings from the parent node information. You can adjust the left and right pointers for the three nodes involved (assuming not at the edge of the tree). Likewise, when removing a node, simply update the left and right pointers of the siblings of the node being removed. Change them to point to each other.
I agree with Thomas Ahle's answer if you want to make all of the row-lists at the same time. It seems that you are only interested in making the list for a one specific row.
Let's say you have a giant tree, but you only want to link the 5th row. There's clearly no point in accessing any node below the 5th row. So just do an early-terminated DFS. Unfortunately, you still have to run through all of the ancestors of every node in the list.
But here's the good news. If you have a perfect binary tree (where every single node branches exactly twice except for the last row) then the first row will have 1 one, the second 2, the third 4, the fourth 8 and the fifth 16. Thus there are more nodes on the last row (16) than all the previous put together (1 + 2 + 4 + 8 = 15), so searching through all of the ancestors is still just O(n), where n is the number of nodes in the row.
The worst case on the other hand would be to have the fifth row consist of a single node with a full binary tree above it. Then you still have to search through all 15 ancestors just to put that one node on the list.
So while this algorithm is really your only choice without modifying your data structure its efficiency relies entirely on how populated the row is compared to higher rows.
#include <queue>
struct Node {
Node *left;
Node *right;
Node *next;
};
/** Link all nodes of the same level in a binary tree. */
void link_level_nodes(Node *pRoot)
{
queue<Node*> q;
Node *prev; // Pointer to the revious node of the current level
Node *node;
int cnt; // Count of the nodes in the current level
int cntnext; // Count of the nodes in the next level
if(NULL == pRoot)
return;
q.push(pRoot);
cnt = 1;
cntnext = 0;
prev = NULL;
while (!q.empty()) {
node = q.front();
q.pop();
/* Add the left and the right nodes of the current node to the queue
and increment the counter of nodes at the next level.
*/
if (node->left){
q.push(node->left);
cntnext++;
}
if (node->right){
q.push(node->right);
cntnext++;
}
/* Link the previous node of the current level to this node */
if (prev)
prev->next = node;
/* Se the previous node to the current */
prev = node;
cnt--;
if (0 == cnt) { // if this is the last node of the current level
cnt = cntnext;
cntnext = 0;
prev = NULL;
}
}
}
What I usually do to solve this problem is that I do a simple inorder traversal.
I initialize my tree with a constructor that gives a level or column value to every node. Hence my head is at Level 0.
public Node(int d)
{
head=this;
data=d;
left=null;
right=null;
level=0;
}
Now, if in the traversal, I take a left or a right, I simply do the traversal with a level indicator. For each level identifier, I make a Linked List, possibly in a Vector of Nodes.
Different approaches can be used to solve this problem. Some of them that comes to mind are -
1) Using level order traversal or BFS.
We can modify queue entries to contain level of nodes.So queue node will contain a pointer to a tree node and an integer level. When we deque a node we can check the level of dequeued node if it is same we can set right pointer to point to it.
Time complexity for this method would be O(n).
2) If we have complete binary tree we can extend Pre-Order traversal. In this method we shall set right pointer of parent before the children.
Time complexity for this method would be O(n).
3) In case of incomplete binary tree we can modify method (2) by traversing first root then right pointer and then left so we can make sure that all nodes at level i have the right pointer set, before the level i+1 nodes.
Time complexity for this method would be O(n^2).
private class Node
{
public readonly Node Left;
public readonly Node Right;
public Node Link { get; private set; }
public void Run()
{
LinkNext = null;
}
private Node LinkNext
{
get
{
return Link == null ? null : (Link.Left ?? Link.Right ?? Link.LinkNext);
}
set
{
Link = value;
if (Right != null)
Right.LinkNext = LinkNext;
if (Left != null)
Left.LinkNext = Right ?? LinkNext;
}
}
}
Keep a depth array while breadth-first search.
vector<forward_list<index_t>> level_link(MAX_NODES);
index_t fringe_depth = 0;
static index_t depth[MAX_NODES];
memset(depth,0,sizeof(depth));
depth[0] = 0;
Now when the depth-changes while de-queuing, you get all linked !
explored[0] = true;
static deque<index_t> fringe;
fringe.clear();
fringe.push_back(0); // start bfs from node 0
while(!fringe.empty()) {
index_t xindex = fringe.front();
fringe.pop_front();
if(fringe_depth < depth[xindex]) {
// play with prev-level-data
fringe_depth = depth[xindex];
}
Now we have fringe-depth, so we can level-link.
level_link[fringe_depth].push_front(xindex);
for(auto yindex : nodes[xindex].connected) {
if(explored[yindex])
continue;
explored[yindex] = true;
depth[yindex] = depth[xindex] + 1;
fringe.push_back(yindex);
}
}

Resources