DATA STRUCTURES AND ALGORITHMS

LECTURE 12

Lect. PhD. Oneţ-Marian Zsuzsanna

Babeş - Bolyai University

Computer Science and Mathematics Faculty

2024 - 2025

Lect. PhD. Oneţ-Marian Zsuzsanna DATA STRUCTURES AND ALGORITHMS

In Lecture 11...

Tree

Binary tree

Lect. PhD. Oneţ-Marian Zsuzsanna DATA STRUCTURES AND ALGORITHMS

Today

Huffman encoding

Binary search tree

Lect. PhD. Oneţ-Marian Zsuzsanna DATA STRUCTURES AND ALGORITHMS

Huffman coding

When building the Huffman encoding, we need to have a

message that we want to encode.

First we have to compute the frequency of every character

from the message, because we are going to define the codes
based on the frequencies.

Assume that we have the following message: WE WERE

HERE.

It contains the following characters and frequencies:

w e r h space .
2 5 2 1 2 1

Lect. PhD. Oneţ-Marian Zsuzsanna DATA STRUCTURES AND ALGORITHMS

Huffman coding

For defining the Huffman code a binary tree is build in the

following way:

Start with trees containing only a root node, one for every
character. Each tree has a weight, which is the frequency of
the character.

Get the two trees with the least weight (if there is a tie,
choose randomly), combine them into one tree which has as
weight the sum of the two weights.

Repeat until we have only one tree.

The implementation can simply be done with a Priority Queue

which stores these partially built trees (and considers the
weight as priority).

Lect. PhD. Oneţ-Marian Zsuzsanna DATA STRUCTURES AND ALGORITHMS

Huffman coding

Lect. PhD. Oneţ-Marian Zsuzsanna DATA STRUCTURES AND ALGORITHMS

Huffman coding
Code for each character can be read from the tree in the
following way: start from the root and go towards the
corresponding leaf node. Every time we go left add the bit 0
to encoding and when we go right add bit 1.
Code for the characters:
w - 000
r - 001
h - 0100
. - 0101
space - 011
e-1
We can see that the most frequent character (e), indeed has
the shortest code, and the least frequent ones have the
longest. Also, no code is the prefix of another.
In order to encode a message, just replace each character with
the corresponding code.
Lect. PhD. Oneţ-Marian Zsuzsanna DATA STRUCTURES AND ALGORITHMS
Huffman coding

Assume we have the following code and we want to decode it:

0110110001000100111001000000

We do not know where the code of each character ends, but

we can use the previously built tree to decode it.

Start parsing the code and iterate through the tree in the
following way:
Start from the root
If the current bit from the code is 0 go to the left child,
otherwise go to the right child
If we are at a leaf node we have decoded a character and have
to start over from the root
The decoded message (quotation marks added to see the
spaces at the beginning): ” wewereerww”

Lect. PhD. Oneţ-Marian Zsuzsanna DATA STRUCTURES AND ALGORITHMS

Binary search trees

A Binary Search Tree is a binary tree that satisfies the

following property:
if x is a node of the binary search tree then:

For every node y from the left subtree of x, the information

from y is less than or equal to the information from x

For every node y from the right subtree of x, the information

from y is greater than the information from x

Obviously, the relation used to order the nodes can be

considered in an abstract way (instead of having ”≤” as in
the definition).

Lect. PhD. Oneţ-Marian Zsuzsanna DATA STRUCTURES AND ALGORITHMS

Binary Search Tree Example

If we do an inorder traversal of a binary search tree, we will

get the elements in increasing order (according to the relation
used).

Lect. PhD. Oneţ-Marian Zsuzsanna DATA STRUCTURES AND ALGORITHMS

Binary Search Tree

The terminology and many properties discussed for binary tree

is valid for binary search trees as well:
We can have a binary search tree that is full, complete, almost
complete, degenerate or balanced
The maximum number of nodes in a binary search tree of
height N is 2N+1 − 1 - if the tree is complete.
The minimum number of nodes in a binary search tree of
height N is N - if the tree is degenerate.
A binary search tree with N nodes has a height between log2 N
and N (we will denote the height of the tree by h).

Lect. PhD. Oneţ-Marian Zsuzsanna DATA STRUCTURES AND ALGORITHMS

Binary Search Tree

Binary search trees can be used as representation for sorted

containers: sorted maps, sorted multimaps, priority queues,
sorted sets, etc.

In order to implement these containers on a binary search

tree, we need to define the following basic operations:
search for an element
insert an element
remove an element
Other operations that can be implemented for binary search
trees (and can be used by the containers): get
minimum/maximum element, find the successor/predecessor
of an element.

Lect. PhD. Oneţ-Marian Zsuzsanna DATA STRUCTURES AND ALGORITHMS

Binary Search Tree - Representation

We will use a linked representation with dynamic allocation

(similar to what we used for binary trees)

BSTNode:
info: TElem
left: ↑ BSTNode
right: ↑ BSTNode

BinarySearchTree:
root: ↑ BSTNode

Normally, BST would contain a relation as well. In our

examples we will use the ≤ relation directly.

Lect. PhD. Oneţ-Marian Zsuzsanna DATA STRUCTURES AND ALGORITHMS

Binary Search Tree - search operation

How can we search for an element in a binary search tree?

How can we search for element 15? And for element 14?

Lect. PhD. Oneţ-Marian Zsuzsanna DATA STRUCTURES AND ALGORITHMS

BST - search operation - recursive implementation

How can we implement the search algorithm recursively?

Lect. PhD. Oneţ-Marian Zsuzsanna DATA STRUCTURES AND ALGORITHMS

BST - search operation - recursive implementation

How can we implement the search algorithm recursively?

function search rec (node, elem) is:

//pre: node is a BSTNode and elem is the TElem we are searching for
if node = NIL then
search rec ← false
else
if [node].info = elem then
search rec ← true
else if [node].info < elem then
search rec ← search rec([node].right, elem)
else
search rec ← search rec([node].left, elem)
end-if
end-function

Lect. PhD. Oneţ-Marian Zsuzsanna DATA STRUCTURES AND ALGORITHMS

BST - search operation - recursive implementation

Since the search algorithm takes as parameter a node, we

need a wrapper function to call it with the root of the tree.

function search (tree, e) is:

//pre: tree is a BinarySearchTree, e is the elem we are looking for
search ← search rec(tree.root, e)
end-function

Lect. PhD. Oneţ-Marian Zsuzsanna DATA STRUCTURES AND ALGORITHMS

BST - search operation - non-recursive implementation

How can we define the search operation non-recursively?

Lect. PhD. Oneţ-Marian Zsuzsanna DATA STRUCTURES AND ALGORITHMS

BST - search operation - non-recursive implementation

How can we define the search operation non-recursively?

function search (tree, elem) is:

//pre: tree is a BinarySearchTree and elem is the TElem we are searching for
currentNode ← tree.root
found ← false
while currentNode ̸= NIL and not found execute
if [currentNode].info = elem then
found ← true
else if [currentNode].info < elem then
currentNode ← [currentNode].right
else
currentNode ← [currentNode].left
end-if
end-while
search ← found
end-function

Lect. PhD. Oneţ-Marian Zsuzsanna DATA STRUCTURES AND ALGORITHMS

BST operation complexities

Most BST operations will follow one single path from the root
to a leaf node (or maybe stop somewhere at an internal node,
depending on the problem), which means that their
complexity depends on the height of the tree, h.

As discussed, height of the tree can be between log2 n and n,

so the worst case complexities will in general be Θ(n) for the
operations, which means that total complexity will be in many
cases O(n).

Nevertheless, average complexity is Θ(log2 n) (assuming that

the elements were inserted in a random order).

Regarding the search algorithm, best case complexity is Θ(1),

average case is Θ(log2 n) and worst case is Θ(n).

Lect. PhD. Oneţ-Marian Zsuzsanna DATA STRUCTURES AND ALGORITHMS

BST - insert operation

How/Where can we insert element 14?

Lect. PhD. Oneţ-Marian Zsuzsanna DATA STRUCTURES AND ALGORITHMS

BST - insert operation

Lect. PhD. Oneţ-Marian Zsuzsanna DATA STRUCTURES AND ALGORITHMS

BST - insert operation - recursive implementation

How can we implement the insert operation?

Lect. PhD. Oneţ-Marian Zsuzsanna DATA STRUCTURES AND ALGORITHMS

BST - insert operation - recursive implementation

How can we implement the insert operation?

We will start with a function that creates a new node given
the information to be stored in it.

function initNode(e) is:

//pre: e is a TComp
//post: initNode ← a node with e as information
allocate(node)
[node].info ← e
[node].left ← NIL
[node].right ← NIL
initNode ← node
end-function

Lect. PhD. Oneţ-Marian Zsuzsanna DATA STRUCTURES AND ALGORITHMS

BST - insert operation - recursive implementation

function insert rec(node, e) is:

//pre: node is a BSTNode, e is TComp
//post: a node containing e was added in the tree starting from node
if node = NIL then
node ← initNode(e)
else if [node].info ≥ e then
[node].left ← insert rec([node].left, e)
else
[node].right ← insert rec([node].right, e)
end-if
insert rec ← node
end-function

Complexity:

Lect. PhD. Oneţ-Marian Zsuzsanna DATA STRUCTURES AND ALGORITHMS

BST - insert operation - recursive implementation

function insert rec(node, e) is:

//pre: node is a BSTNode, e is TComp
//post: a node containing e was added in the tree starting from node
if node = NIL then
node ← initNode(e)
else if [node].info ≥ e then
[node].left ← insert rec([node].left, e)
else
[node].right ← insert rec([node].right, e)
end-if
insert rec ← node
end-function

Complexity: O(n) (Θ(n) in worst case, but Θ(log2 n) on

average)

Like in case of the search operation, we need a wrapper

function to call insert rec with the root of the tree.

Lect. PhD. Oneţ-Marian Zsuzsanna DATA STRUCTURES AND ALGORITHMS

BST - Finding the minimum element

How can we find the minimum element of the binary search

tree?

Lect. PhD. Oneţ-Marian Zsuzsanna DATA STRUCTURES AND ALGORITHMS

BST - Finding the minimum element

function minimum(tree) is:

//pre: tree is a BinarySearchTree
//post: minimum ← the minimum value from the tree
currentNode ← tree.root
if currentNode = NIL then
@empty tree, no minimum
else
while [currentNode].left ̸= NIL execute
currentNode ← [currentNode].left
end-while
minimum ← [currentNode].info
end-if
end-function

Lect. PhD. Oneţ-Marian Zsuzsanna DATA STRUCTURES AND ALGORITHMS

BST - Finding the minimum element

Complexity of the minimum operation:

Lect. PhD. Oneţ-Marian Zsuzsanna DATA STRUCTURES AND ALGORITHMS

BST - Finding the minimum element

Complexity of the minimum operation: O(n)

We can have an implementation for the minimum, when we

want to find the minimum element of a subtree, in this case
the parameter to the function would be a node, not a tree.

We can have an implementation where we return the node

containing the minimum element, instead of just the value
(depends on what we want to do with the operation)

Maximum element of the tree can be found similarly.

Lect. PhD. Oneţ-Marian Zsuzsanna DATA STRUCTURES AND ALGORITHMS

Finding the parent of a node

Given a node, how can we find the parent of the node?

(assume a representation where the node has no parent field).

Lect. PhD. Oneţ-Marian Zsuzsanna DATA STRUCTURES AND ALGORITHMS

Finding the parent of a node

function parent(tree, node) is:

//pre: tree is a BinarySearchTree, node is a pointer to a BSTNode, node ̸= NIL
//post: returns the parent of node, or NIL if node is the root
c ← tree.root
if c = node then //node is the root
parent ← NIL
else
while c ̸= NIL and [c].left ̸= node and [c].right ̸= node execute
if [c].info ≥ [node].info then
c ← [c].left
else
c ← [c].right
end-if
end-while
parent ← c
end-if
end-function

Complexity:

Lect. PhD. Oneţ-Marian Zsuzsanna DATA STRUCTURES AND ALGORITHMS

Finding the parent of a node

function parent(tree, node) is:

//pre: tree is a BinarySearchTree, node is a pointer to a BSTNode, node ̸= NIL
//post: returns the parent of node, or NIL if node is the root
c ← tree.root
if c = node then //node is the root
parent ← NIL
else
while c ̸= NIL and [c].left ̸= node and [c].right ̸= node execute
if [c].info ≥ [node].info then
c ← [c].left
else
c ← [c].right
end-if
end-while
parent ← c
end-if
end-function

Complexity: O(n)

Lect. PhD. Oneţ-Marian Zsuzsanna DATA STRUCTURES AND ALGORITHMS

BST - Finding the successor of a node

Given a node, how can we find the node containing the next
value (considering the relation used for ordering the
elements)?

How can we find the next after 11?

Lect. PhD. Oneţ-Marian Zsuzsanna DATA STRUCTURES AND ALGORITHMS
BST - Finding the successor of a node

Given a node, how can we find the node containing the next
value (considering the relation used for ordering the
elements)?

How can we find the next after 11? After 13?

Lect. PhD. Oneţ-Marian Zsuzsanna DATA STRUCTURES AND ALGORITHMS
BST - Finding the successor of a node

Given a node, how can we find the node containing the next
value (considering the relation used for ordering the
elements)?

How can we find the next after 11? After 13? After 17?
Lect. PhD. Oneţ-Marian Zsuzsanna DATA STRUCTURES AND ALGORITHMS
BST - Finding the successor of a node

function successor(tree, node) is:

//pre: tree is a BinarySearchTree, node is a pointer to a BSTNode, node ̸= NIL
//post: returns the node with the next value after the value from node
//or NIL if node is the maximum
if [node].right ̸= NIL then
c ← [node].right
while [c].left ̸= NIL execute
c ← [c].left
end-while
successor ← c
else
p ← parent(tree, c)
while p ̸= NIL and [p].left ̸= c execute
c←p
p ← parent(tree, p)
end-while
successor ← p
end-if
end-function

Lect. PhD. Oneţ-Marian Zsuzsanna DATA STRUCTURES AND ALGORITHMS

BST - Finding the successor of a node

Complexity of successor:

Lect. PhD. Oneţ-Marian Zsuzsanna DATA STRUCTURES AND ALGORITHMS

BST - Finding the successor of a node

Complexity of successor: depends on parent function:

If parent is Θ(1), complexity of successor is O(n)

If parent is O(n), complexity of successor is O(n2 )

What if, instead of receiving a node, successor algorithm

receives as parameter a value from a node (assume unique
values in the nodes)? How can we find the successor then?

Similar to successor, we can define a predecessor function as

well.

Lect. PhD. Oneţ-Marian Zsuzsanna DATA STRUCTURES AND ALGORITHMS

BST - Finding successor of a node

If we do not have direct access to the parent, finding the

successor of a node is O(n2 ).

Can we reduce this complexity?

Lect. PhD. Oneţ-Marian Zsuzsanna DATA STRUCTURES AND ALGORITHMS

BST - Finding successor of a node

If we do not have direct access to the parent, finding the

successor of a node is O(n2 ).

Can we reduce this complexity?

O(n2 ) is given by the potentially repeated calls for the parent
function. But we only need the last parent, where we went
left. We can do one single traversal from the root to our
node, and every time we continue left (i.e. current node is
greater than the one we are looking for) we memorize that
node in a variable (and change the variable when we find a
new such node). When the current node is at the one we are
looking for, this variable contains its successor.

Lect. PhD. Oneţ-Marian Zsuzsanna DATA STRUCTURES AND ALGORITHMS

BST - Remove a node

How can we remove the value 25?

Lect. PhD. Oneţ-Marian Zsuzsanna DATA STRUCTURES AND ALGORITHMS

BST - Remove a node

How can we remove the value 25? And value 2?

Lect. PhD. Oneţ-Marian Zsuzsanna DATA STRUCTURES AND ALGORITHMS

BST - Remove a node

How can we remove the value 25? And value 2? And value
11?

Lect. PhD. Oneţ-Marian Zsuzsanna DATA STRUCTURES AND ALGORITHMS

BST - Remove a node

When we want to remove a value (a node containing the

value) from a binary search tree we have three cases:
The node to be removed has no descendant
Set the corresponding child of the parent to NIL

The node to be removed has one descendant

Set the corresponding child of the parent to the descendant

The node to be removed has two descendants

Find the maximum of the left subtree, move it to the node to
be deleted, and delete the maximum
OR
Find the minimum of the right subtree, move it to the node to
be deleted, and delete the minimum

Lect. PhD. Oneţ-Marian Zsuzsanna DATA STRUCTURES AND ALGORITHMS

Binary Search Tree

Think about it:

Can we define a Sorted List on a Binary Search Tree? If not,

why not? If yes, how exactly? What would be the most
complicated part?

Lect. PhD. Oneţ-Marian Zsuzsanna DATA STRUCTURES AND ALGORITHMS

Binary Search Tree

Think about it:

Can we define a Sorted List on a Binary Search Tree? If not,

why not? If yes, how exactly? What would be the most
complicated part?
Lists have positions and operations based on positions. In case
of a SortedList we have an operation to remove an element
from a position and to return an element from a position (but
no insert to position). How can we keep track of positions in a
binary search tree?

Lect. PhD. Oneţ-Marian Zsuzsanna DATA STRUCTURES AND ALGORITHMS

Binary Search Tree

Think about it:

Can we define a Sorted List on a Binary Search Tree? If not,

why not? If yes, how exactly? What would be the most
complicated part?
Lists have positions and operations based on positions. In case
of a SortedList we have an operation to remove an element
from a position and to return an element from a position (but
no insert to position). How can we keep track of positions in a
binary search tree?
We can keep in every node, besides the information, the
number of nodes in the left subtree. This gives us
automatically the ”position” of the root in the SortedList.
When we have operations that are based on positions, we use
these values to decide if we go left or right.

Lect. PhD. Oneţ-Marian Zsuzsanna DATA STRUCTURES AND ALGORITHMS

Binary Search Tree

Lect. PhD. Oneţ-Marian Zsuzsanna DATA STRUCTURES AND ALGORITHMS

Binary Search Tree with duplicate values

Starting from an initially empty Binary Search Tree and the

relation ≤, insert into it, in the given order, the following
values: 10, 20, 5, 7, 15, 5, 30, 3, 5, 5, 1, 9, 29, 2.

Lect. PhD. Oneţ-Marian Zsuzsanna DATA STRUCTURES AND ALGORITHMS

Binary Search Tree with duplicate values

Starting from an initially empty Binary Search Tree and the

relation ≤, insert into it, in the given order, the following
values: 10, 20, 5, 7, 15, 5, 30, 3, 5, 5, 1, 9, 29, 2.

How would you count how many times the value 5 is in the
tree?

Lect. PhD. Oneţ-Marian Zsuzsanna DATA STRUCTURES AND ALGORITHMS

Binary Search Tree with duplicate values

Starting from an initially empty Binary Search Tree and the

relation ≤, insert into it, in the given order, the following
values: 10, 20, 5, 7, 15, 5, 30, 3, 5, 5, 1, 9, 29, 2.

How would you count how many times the value 5 is in the
tree?

Remove 3 (show both options)

Lect. PhD. Oneţ-Marian Zsuzsanna DATA STRUCTURES AND ALGORITHMS

Binary Search Tree with duplicate values

Starting from an initially empty Binary Search Tree and the

relation ≤, insert into it, in the given order, the following
values: 10, 20, 5, 7, 15, 5, 30, 3, 5, 5, 1, 9, 29, 2.

How would you count how many times the value 5 is in the
tree?

Remove 3 (show both options)

How would you count now how many times the value 5 is in
the tree now?

Lect. PhD. Oneţ-Marian Zsuzsanna DATA STRUCTURES AND ALGORITHMS

Today

Huffman encoding

Binary search tree

Lect. PhD. Oneţ-Marian Zsuzsanna DATA STRUCTURES AND ALGORITHMS