2nd Largest Item in a Binary Search Tree (Practice Interview Question)

You only have 3 free questions left (including this one).

But it doesn't have to end here! Sign up for the 7-day coding interview crash course and you'll get a free Interview Cake problem every week.

You're in!

Write a function to find the 2nd largest element in a binary search tree. ↴

Costs

Balanced Unbalanced (Worst Case)

space $O(n)$ $O(n)$

insert $O(lg(n))$ $O(n)$

lookup $O(lg(n))$ $O(n)$

delete $O(lg(n))$ $O(n)$

Quick reference

A binary search tree is a binary tree where the nodes are ordered in a specific way. For every node:

The nodes to the left are smaller than the current node.

The nodes to the right are larger than the current node.

Checking if a binary tree is a binary search tree is a favorite question from interviews.

Strengths:

Good performance across the board. Assuming they're balanced, binary search trees are good at lots of operations, even if they're not constant time for anything.

Compared to a sorted array, lookups take the same amount of time ( $O(lg(n))$ ), but inserts and deletes are faster ( $O(lg(n))$ for BSTs, $O(n)$ for arrays).

Compared to arrays, BSTs have better worst case performance— $O(lg(n))$ instead of $O(n)$ . But, on average arrays perform better than BSTs (meaning $O(1)$ time complexity).

BSTs are sorted. Taking a binary search tree and pulling out all of the elements in sorted order can be done in $O(n)$ using an in-order traversal. Finding the element closest to a value can be done in $O(lg(n))$ (again, if the BST is balanced!).

Weaknesses:

Poor performance if unbalanced. Some types of binary search trees balance automatically, but not all. If a BST is not balanced, then operations become $O(n)$ .

No $O(1)$ operations. BSTs aren't the fastest for anything. On average, an array or an array will be faster.

Balanced Binary Search Trees

Two binary search trees can store the same values in different ways:

Some trees (like AVL trees or Red-Black trees) rearrange nodes as they're inserted to ensure the tree is always balanced. With these, the worst case complexity for searching, inserting, or deleting is always $O(lg(n))$ , not $O(n)$ .

	Balanced	Unbalanced (Worst Case)
space	$O(n)$	$O(n)$
insert	$O(lg(n))$	$O(n)$
lookup	$O(lg(n))$	$O(n)$
delete	$O(lg(n))$	$O(n)$

Here's a sample binary tree node class:

  class BinaryTreeNode
{
    public $value;
    public $left;
    public $right;

    public function __construct($value)
    {
        $this->value = $value;
    }

    public function insertLeft($leftValue)
    {
        $this->left = new BinaryTreeNode($leftValue);
        return $this->left;
    }

    public function insertRight($rightValue)
    {
        $this->right = new BinaryTreeNode($rightValue);
        return $this->right;
    }
}

Gotchas

Our first thought might be to do an in-order traversal of the BST ↴

Sometimes we have a BST and we want to go through the items in order from smallest to largest. This is called an "in-order traversal."

We can write a recursive algorithm for this:

Everything in the left subtree is smaller, so print that first, then,
print the current node, then
print the right subtree.

  function inOrderPrint($node)
{
    if ($node) {
        inOrderPrint($node->left);
        echo $node->value . PHP_EOL;
        inOrderPrint($node->right);
    }
}

This takes $O(n)$ time (we're traversing the whole tree, so we're looking at all $n$ items) and $O(h)$ space, where $h$ is the height of the tree (this is the max depth of the call stack during our recursion).

If the tree is balanced its height is $lg(n)$ , so we have $O(lg(n))$ space. If we can't make any assumptions about the "shape" of the tree, in the worst case it's just one continuous line, giving us a height of $n$ , and a space complexity of $O(n)$ for our recursive in-order traversal.

and return the second-to-last item. This means looking at every node in the BST. That would take

O(n)

time and

O(h)

space, where

h

is the max height of the tree (which is

lg(n)

if the tree is balanced, ↴

Formally, a tree is said to be balanced if the difference between the depths of any node's left tree and right tree is no greater than 1.

Thus a 'balanced' tree 'looks full', without any apparent chunks missing or any branches that end much earlier than other branches.

but could be as much as

n

if not).

We can do better than $O(n)$ time and $O(h)$ space.

We can do this in one walk from top to bottom of our BST. This means $O(h)$ time (again, that's $O(\lg{n})$ if the tree is balanced, $O(n)$ otherwise).

A clean recursive implementation will take $O(h)$ space in the call stack, ↴

Overview

The call stack is what a program uses to keep track of function calls. The call stack is made up of stack frames—one for each function call.

For instance, say we called a function that rolled two dice and printed the sum.

  def roll_die():
    return random.randint(1, 6)

def roll_two_and_sum():
    total = 0
    total += roll_die()
    total += roll_die()
    print total

roll_two_and_sum()

We're still translating this code to PHP. Here it is in Python 2.7:

First, our program calls rollTwoAndSum(). It goes on the call stack:

rollTwoAndSum()

That function calls rollDie(), which gets pushed on to the top of the call stack:

rollDie()

rollTwoAndSum()

Inside of rollDie(), we call random.randint(). Here's what our call stack looks like then:

random.randint()

rollDie()

rollTwoAndSum()

When random.randint() finishes, we return back to rollDie() by removing ("popping") random.randint()'s stack frame.

rollDie()

rollTwoAndSum()

Same thing when rollDie() returns:

rollTwoAndSum()

We're not done yet! rollTwoAndSum() calls rollDie() again:

rollDie()

rollTwoAndSum()

Which calls random.randint() again:

random.randint()

rollDie()

rollTwoAndSum()

random.randint() returns, then rollDie() returns, putting us back in rollTwoAndSum():

rollTwoAndSum()

Which calls print():

print()

rollTwoAndSum()

What's stored in a stack frame?

What actually goes in a function's stack frame?

A stack frame usually stores:

Local variables
Arguments passed into the function
Information about the caller's stack frame
The return address—what the program should do after the function returns (i.e.: where it should "return to"). This is usually somewhere in the middle of the caller's code.

Some of the specifics vary between processor architectures. For instance, AMD64 (64-bit x86) processors pass some arguments in registers and some on the call stack. And, ARM processors (common in phones) store the return address in a special register instead of putting it on the call stack.

The Space Cost of Stack Frames

Each function call creates its own stack frame, taking up space on the call stack. That's important because it can impact the space complexity of an algorithm. Especially when we use recursion.

For example, if we wanted to multiply all the numbers between $1$ and $n$ , we could use this recursive approach:

  function product1ToN($n)
{
    // we assume $n >= 1
    return ($n > 1) ? ($n * product1ToN($n - 1)) : 1;
}

What would the call stack look like when n = 10?

First, product1ToN() gets called with n = 10:

    product1ToN()    n = 10

This calls product1ToN() with n = 9.

    product1ToN()    n = 9

    product1ToN()    n = 10

Which calls product1ToN() with n = 8.

    product1ToN()    n = 8

    product1ToN()    n = 9

    product1ToN()    n = 10

And so on until we get to n = 1.

    product1ToN()    n = 1

    product1ToN()    n = 2

    product1ToN()    n = 3

    product1ToN()    n = 4

    product1ToN()    n = 5

    product1ToN()    n = 6

    product1ToN()    n = 7

    product1ToN()    n = 8

    product1ToN()    n = 9

    product1ToN()    n = 10

Look at the size of all those stack frames! The entire call stack takes up $O(n)$ space. That's right—we have an $O(n)$ space cost even though our function itself doesn't create any data structures!

What if we'd used an iterative approach instead of a recursive one?

  function product1ToN($n)
{
    // we assume $n >= 1
    $result = 1;
    for ($num = 1; $num <= $n; $num++) {
        $result *= $num;
    }

    return $result;
}

This version takes a constant amount of space. At the beginning of the loop, the call stack looks like this:

    product1ToN()    n = 10, result = 1, num = 1

As we iterate through the loop, the local variables change, but we stay in the same stack frame because we don't call any other functions.

    product1ToN()    n = 10, result = 2, num = 2

    product1ToN()    n = 10, result = 6, num = 3

    product1ToN()    n = 10, result = 24, num = 4

In general, even though the compiler or interpreter will take care of managing the call stack for you, it's important to consider the depth of the call stack when analyzing the space complexity of an algorithm.

Be especially careful with recursive functions! They can end up building huge call stacks.

What happens if we run out of space? It's a stack overflow! In PHP, you'll get a Fatal Error when it can't allocate additional memory for a new entry on the stack.

If the very last thing a function does is call another function, then its stack frame might not be needed any more. The function could free up its stack frame before doing its final call, saving space.

This is called tail call optimization (TCO). If a recursive function is optimized with TCO, then it may not end up with a big call stack.

In general, most languages don't provide TCO. Scheme is one of the few languages that guarantee tail call optimization. Some Ruby, C, and Javascript implementations may do it. Python and Java decidedly don't.

but we can bring our algorithm down to

O(1)

space overall.

Breakdown

Let's start by solving a simplified version of the problem and see if we can adapt our approach from there. How would we find the largest element in a BST?

A reasonable guess is to say the largest element is simply the "rightmost" element.

So maybe we can start from the root and just step down right child pointers until we can't anymore (until the right child is null). At that point the current node is the largest in the whole tree.

Is this sufficient? We can prove it is by contradiction:

If the largest element were not the "rightmost," then the largest element would either:

be in some ancestor node's left subtree, or
have a right child.

But each of these leads to a contradiction:

If the node is in some ancestor node's left subtree it's smaller than that ancestor node, so it's not the largest.
If the node has a right child that child is larger than it, so it's not the largest.

So the "rightmost" element must be the largest.

How would we formalize getting the "rightmost" element in code?

We can use a simple recursive approach. At each step:

If there is a right child, that node and the subtree below it are all greater than the current node. So step down to this child and recurse.
Else there is no right child and we're already at the "rightmost" element, so we return its value.

  function findLargest($rootNode)
{
    if (!$rootNode) {
        throw new Exception('Tree must have at least 1 node');
    }
    if ($rootNode->right) {
        return findLargest($rootNode->right);
    }
    return $rootNode->value;
}

Okay, so we can find the largest element. How can we adapt this approach to find the second largest element?

Our first thought might be, "it's simply the parent of the largest element!" That seems obviously true when we imagine a nicely balanced tree like this one:

  .        ( 5 )
        /     \
      (3)     (8)
     /  \     /  \
   (1)  (4) (7)  (9)

But what if the largest element itself has a left subtree?

  .        ( 5 )
        /     \
      (3)     (8)
     /  \     /  \
   (1)  (4) (7)  (12)
                 /
               (10)
               /  \
             (9)  (11)

Here the parent of our largest is 8, but the second largest is 11.

Drat, okay so the second largest isn't necessarily the parent of the largest...back to the drawing board...

Wait. No. The second largest is the parent of the largest if the largest does not have a left subtree. If we can handle the case where the largest does have a left subtree, we can handle all cases, and we have a solution.

So let's try sticking with this. How do we find the second largest when the largest has a left subtree?

It's the largest item in that left subtree! Whoa, we freaking just wrote a function for finding the largest element in a tree. We could use that here!

How would we code this up?

  function findLargest($rootNode)
{
    if (!$rootNode) {
        throw new Exception('Tree must have at least 1 node');
    }
    if ($rootNode->right) {
        return findLargest($rootNode->right);
    }
    return $rootNode->value;
}

function findSecondLargest($rootNode)
{
    if (!$rootNode || (!$rootNode->left && !$rootNode->right)) {
        throw new Exception('Tree must have at least 2 nodes');
    }

    // case: we're currently at largest, and largest has a left subtree,
    // so 2nd largest is largest in said subtree
    if ($rootNode->left && !$rootNode->right) {
        return findLargest($rootNode->left);
    }

    // case: we're at parent of largest, and largest has no children,
    // so 2nd largest must be current node
    if ($rootNode->right &&
            !$rootNode->right->left &&
            !$rootNode->right->right) {
        return $rootNode->value;
    }

    // otherwise: step right
    return findSecondLargest($rootNode->right);
}

Okay awesome. This'll work. It'll take $O(h)$ time (where $h$ is the height of the tree) and $O(h)$ space.

But that $h$ space in the call stack ↴

Overview

The call stack is what a program uses to keep track of function calls. The call stack is made up of stack frames—one for each function call.

For instance, say we called a function that rolled two dice and printed the sum.

  def roll_die():
    return random.randint(1, 6)

def roll_two_and_sum():
    total = 0
    total += roll_die()
    total += roll_die()
    print total

roll_two_and_sum()

We're still translating this code to PHP. Here it is in Python 2.7:

First, our program calls rollTwoAndSum(). It goes on the call stack:

rollTwoAndSum()

That function calls rollDie(), which gets pushed on to the top of the call stack:

rollDie()

rollTwoAndSum()

Inside of rollDie(), we call random.randint(). Here's what our call stack looks like then:

random.randint()

rollDie()

rollTwoAndSum()

When random.randint() finishes, we return back to rollDie() by removing ("popping") random.randint()'s stack frame.

rollDie()

rollTwoAndSum()

Same thing when rollDie() returns:

rollTwoAndSum()

We're not done yet! rollTwoAndSum() calls rollDie() again:

rollDie()

rollTwoAndSum()

Which calls random.randint() again:

random.randint()

rollDie()

rollTwoAndSum()

random.randint() returns, then rollDie() returns, putting us back in rollTwoAndSum():

rollTwoAndSum()

Which calls print():

print()

rollTwoAndSum()

What's stored in a stack frame?

What actually goes in a function's stack frame?

A stack frame usually stores:

Local variables
Arguments passed into the function
Information about the caller's stack frame
The return address—what the program should do after the function returns (i.e.: where it should "return to"). This is usually somewhere in the middle of the caller's code.

The Space Cost of Stack Frames

For example, if we wanted to multiply all the numbers between $1$ and $n$ , we could use this recursive approach:

  function product1ToN($n)
{
    // we assume $n >= 1
    return ($n > 1) ? ($n * product1ToN($n - 1)) : 1;
}

What would the call stack look like when n = 10?

First, product1ToN() gets called with n = 10:

    product1ToN()    n = 10

This calls product1ToN() with n = 9.

    product1ToN()    n = 9

    product1ToN()    n = 10

Which calls product1ToN() with n = 8.

    product1ToN()    n = 8

    product1ToN()    n = 9

    product1ToN()    n = 10

And so on until we get to n = 1.

    product1ToN()    n = 1

    product1ToN()    n = 2

    product1ToN()    n = 3

    product1ToN()    n = 4

    product1ToN()    n = 5

    product1ToN()    n = 6

    product1ToN()    n = 7

    product1ToN()    n = 8

    product1ToN()    n = 9

    product1ToN()    n = 10

What if we'd used an iterative approach instead of a recursive one?

  function product1ToN($n)
{
    // we assume $n >= 1
    $result = 1;
    for ($num = 1; $num <= $n; $num++) {
        $result *= $num;
    }

    return $result;
}

This version takes a constant amount of space. At the beginning of the loop, the call stack looks like this:

    product1ToN()    n = 10, result = 1, num = 1

As we iterate through the loop, the local variables change, but we stay in the same stack frame because we don't call any other functions.

    product1ToN()    n = 10, result = 2, num = 2

    product1ToN()    n = 10, result = 6, num = 3

    product1ToN()    n = 10, result = 24, num = 4

Be especially careful with recursive functions! They can end up building huge call stacks.

What happens if we run out of space? It's a stack overflow! In PHP, you'll get a Fatal Error when it can't allocate additional memory for a new entry on the stack.

This is called tail call optimization (TCO). If a recursive function is optimized with TCO, then it may not end up with a big call stack.

is avoidable. How can we get this down to constant space?

Solution

We start with a function for getting the largest value. The largest value is simply the "rightmost" one, so we can get it in one walk down the tree by traversing rightward until we don't have a right child anymore:

  function findLargest($rootNode)
{
    $current = $rootNode;
    while ($current) {
        if (!$current->right) {
            return $current->value;
        }
        $current = $current->right;
    }
}

With this in mind, we can also find the second largest in one walk down the tree. At each step, we have a few cases:

If we have a left subtree but not a right subtree, then the current node is the largest overall (the "rightmost") node. The second largest element must be the largest element in the left subtree. We use our findLargest() function above to find the largest in that left subtree!
If we have a right child, but that right child node doesn't have any children, then the right child must be the largest element and our current node must be the second largest element!
Else, we have a right subtree with more than one element, so the largest and second largest are somewhere in that subtree. So we step right.

  function findLargest($rootNode)
{
    $current = $rootNode;
    while ($current) {
        if (!$current->right) {
            return $current->value;
        }
        $current = $current->right;
    }
}

function findSecondLargest($rootNode)
{
    if (!$rootNode || (!$rootNode->left && !$rootNode->right)) {
        throw new Exception('Tree must have at least 2 nodes');
    }

    $current = $rootNode;

    while ($current) {

        // case: current is largest and has a left subtree
        // 2nd largest is the largest in that subtree
        if ($current->left && !$current->right) {
            return findLargest($current->left);
        }

        // case: current is parent of largest, and largest has no children,
        // so current is 2nd largest
        if ($current->right &&
                !$current->right->left &&
                !$current->right->right) {
            return $current->value;
        }

        $current = $current->right;
    }
}

Complexity

We're doing one walk down our BST, which means $O(h)$ time, where $h$ is the height of the tree (again, that's $O(\lg{n})$ if the tree is balanced, $O(n)$ otherwise). $O(1)$ space.

What We Learned

Here we used a "simplify, solve, and adapt" strategy.

The question asks for a function to find the second largest element in a BST, so we started off by simplifying the problem: we thought about how to find the first largest element.

Once we had a strategy for that, we adapted that strategy to work for finding the second largest element.

It may seem counter-intuitive to start off by solving the wrong question. But starting off with a simpler version of the problem is often much faster, because it's easier to wrap our heads around right away.

One more note about this one:

Breaking things down into cases is another strategy that really helped us here.

Notice how simple finding the second largest node got when we divided it into two cases:

The largest node has a left subtree.
The largest node does not have a left subtree.

Whenever a problem is starting to feel complicated, try breaking it down into cases.

It can be really helpful to actually draw out sample inputs for those cases. This'll keep your thinking organized and also help get your interviewer on the same page.

Do you have an answer?

Show answer

I have an answer Tell me more

Ready for more?

Check out our full course

You only have 3 free questions left (including this one).

Costs

Quick reference

Strengths:

Weaknesses:

Balanced Binary Search Trees

Gotchas

Overview

What's stored in a stack frame?

The Space Cost of Stack Frames

Breakdown

Overview

What's stored in a stack frame?

The Space Cost of Stack Frames

Solution

Complexity

What We Learned

Ready for more?

Programming interview questions by company:

Programming interview questions by topic: