The Problem Solver’s Guide To Coding

Nhut Nguyen, Ph. D.

March, 2024
The Problem Solver’s Guide To Coding
First edition. March, 2024.
ISBN 978-87-975174-0-6 (paperback)
ISBN 978-87-975174-1-3 (PDF)
ISBN 978-87-430-5777-2 (hardcover)
Copyright © 2024 Nhut Nguyen.
All rights reserved.
www.nhutnguyen.com
To my dearest mother, Nguyen Thi Kim Sa.

iii
iv
 PREFACE

Welcome to The Problem Solver’s Guide To Coding, a comprehensive journey to mas-

ter problem-solving, data structures, algorithms, and programming techniques using
C++. This book results from my experiences, challenges, failures and successes in
my programming career.
One of the most critical phases in software developer hiring process is the coding
interview, a rigorous process where candidates are tested on their problem-solving
skills and technical proficiency. Companies like FAANG (Facebook, Amazon, Apple,
Netflix, and Google) often subject candidates to three to four rounds of interviews,
making preparation essential for success.
My journey with coding interviews began during a pivotal career change in the sum-
mer of 2021. During this transitional period, I discovered LeetCode, a platform that
soon became my daily companion in honing my programming skills. What started
as a means to practice new languages (Golang and C#) quickly evolved into a deep
exploration of my strongest language, C++.
One day, I decided to write an article for each challenge and share it on my blog
leetsolve.com. Over time, my daily practice transformed into more than 70 chal-
lenges, each accompanied by a detailed article encompassing problem statements,
various approaches, C++ code implementations, thorough explanations, and com-
plexity analyses.
As I delved into these coding challenges, I realized their potential to empower aspiring
programmers, students, and junior developers to excel in interviews and master
problem-solving and algorithmic thinking.

v
The book’s overview

The Problem Solver’s Guide to Coding presents challenges covering fundamental data
structures, algorithms, and mathematical problems. Challenges are grouped in top-
ics, starting with the simplest data structure - Array. Most are arranged in order of
increasing difficulty, but you can pick any chapter or any challenge to start since I
write each independently to the other.
Challenges in this book are curated from LeetCode.com, focusing on those that are
not difficult but provide valuable learning experiences. You might encounter some
simple challenges I go directly to the code without saying much about the idea (intu-
ition) since their solution is straightforward.
I also keep the problems’ original constraints (inputs’ size, limits, etc.) as the code in
this book is the ones I submitted on Leetcode.com. It explains why I usually focus on
the core algorithm and do not consider/handle corner cases or invalid inputs.
The problems in each chapter comes with a detailed solution, explaining the logic
behind the solution and how to implement it in C++, my strongest programming
language.
At the end of some problems, I also provide similar problems on leetcode.com for
you to solve on your own, as practicing is essential for reinforcing understanding
and mastery of the concepts presented in the book. By engaging in problem-solving
exercises, you can apply what you have learned, develop your problem-solving skills,
and gain confidence in your ability to tackle real-world challenges.
In this book, I focus on readable code rather than optimal one, as most of you are
at the beginner level. Some of my solutions might need to be in better runtime or
memory. But I keep my code in my style or coding convention, where readability is
vital.
Moreover, my weekly sharing of articles with various developer communities has re-
fined the content and established a connection with a diverse group of programming
enthusiasts.

Who is this book for?

This book is tailored to benefit a wide audience, from students beginning their pro-
gramming journey to experienced developers looking to enhance their skills. Re-
gardless of your experience level, whether you’re preparing for coding interviews
or simply seeking to improve your problem-solving abilities, this book is designed
to meet your needs.

vi
As a minimum requirement, you are supposed to have some basic background in
C++ programming language, data structures and algorithms like a second-year un-
dergraduate in Computer Science.
What sets this book apart is its focus on practicality. The challenges presented here
are not just exercises; they mirror real coding interviews from top companies like
FAANG.
As you work through the coding challenges in this book, you’ll learn new skills, im-
prove your problem-solving abilities, and develop your confidence as a programmer.

Acknowledgement

I am deeply grateful to my invaluable previewers, especially Alexander Volkodav,

Phung Phu Tai, Le Nhat Tung, Tran Anh Tuan A, Bui Nguyen Cong Duy, Cao Minh
Thinh, Nguyen Trung Tuan, Nguyen Tuan Hung, Nguyen Hoang Nhat Minh, Nhan
Nguyen, Phan Dinh Thai and Nguyen Quang Trung, whose dedication and insights
have shaped this book into its final form. Your contributions and unwavering support
are truly appreciated.
I would like to express my heartfelt gratitude to Ninh Pham, Hoang Thanh Lam, Dinh
Thai Minh Tam and Tran Anh Tuan B, whose invaluable feedback contributed to the
refinement of this book in future versions. Your insightful comments and constructive
criticism have played a pivotal role in shaping its content and enhancing its quality.
Thank you for your dedication and generosity in sharing your expertise. Your input
will undoubtedly pave the way for future improvements and iterations of this work.

vii
Students and developers! By immersing yourself in the challenges and insights shared
in this book, you will not only prepare for coding interviews but also cultivate a
mindset beyond the scope of a job interview. You will become a problem solver, a
strategic thinker, and a proficient C++ programmer.
As you embark on this journey, remember that every challenge you encounter is an
opportunity for growth. Embrace the complexities, learn from each solution, and let
the knowledge you gain propel you to new heights in your programming career.
Thank you for joining me on this expedition.
May your code be elegant, your algorithms efficient, and your programming
journey genuinely transformative.
Happy coding!
Copenhagen, March 2024.
Nhut Nguyen, Ph.D.

viii
 CONTENTS

1 Introduction 1
1.1 Why LeetCode? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 A brief about algorithm complexity . . . . . . . . . . . . . . . . . . . . 2
1.3 Why readable code? . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

2 Array 7
2.1 Transpose Matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.2 Valid Mountain Array . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.3 Shift 2D Grid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.4 Find All Numbers Disappeared in an Array . . . . . . . . . . . . . . . . 18
2.5 Rotate Image . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.6 Spiral Matrix II . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.7 Daily Temperatures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

3 Linked List 37
3.1 Merge Two Sorted Lists . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.2 Remove Linked List Elements . . . . . . . . . . . . . . . . . . . . . . . 42
3.3 Intersection of Two Linked Lists . . . . . . . . . . . . . . . . . . . . . . 48
3.4 Swap Nodes in Pairs . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
3.5 Add Two Numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62

4 Hash Table 67
4.1 Roman to Integer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
4.2 Maximum Erasure Value . . . . . . . . . . . . . . . . . . . . . . . . . . 71
4.3 Find and Replace Pattern . . . . . . . . . . . . . . . . . . . . . . . . . . 74

5 String 79
5.1 Valid Anagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80

ix
 5.2 Detect Capital . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
5.3 Unique Morse Code Words . . . . . . . . . . . . . . . . . . . . . . . . . 87
5.4 Unique Email Addresses . . . . . . . . . . . . . . . . . . . . . . . . . . 90
5.5 Longest Substring Without Repeating Characters . . . . . . . . . . . . 96
5.6 Compare Version Numbers . . . . . . . . . . . . . . . . . . . . . . . . . 100

6 Stack 105
6.1 Baseball Game . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
6.2 Valid Parentheses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
6.3 Backspace String Compare . . . . . . . . . . . . . . . . . . . . . . . . . 112
6.4 Remove All Adjacent Duplicates in String II . . . . . . . . . . . . . . . 115

7 Priority Queue (Heap) 121

7.1 Last Stone Weight . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
7.2 Kth Largest Element in a Stream . . . . . . . . . . . . . . . . . . . . . 124
7.3 Kth Smallest Element in a Sorted Matrix . . . . . . . . . . . . . . . . . 130
7.4 Construct Target Array With Multiple Sums . . . . . . . . . . . . . . . 134

8 Bit Manipulation 141

8.1 Hamming Distance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
8.2 Power of Four . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
8.3 Find the Duplicate Number . . . . . . . . . . . . . . . . . . . . . . . . 148
8.4 Maximum Product of Word Lengths . . . . . . . . . . . . . . . . . . . . 153

9 Sorting 159
9.1 Majority Element . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
9.2 Merge Sorted Array . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
9.3 Remove Covered Intervals . . . . . . . . . . . . . . . . . . . . . . . . . 169
9.4 My Calendar I . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
9.5 Remove Duplicates from Sorted Array II . . . . . . . . . . . . . . . . . 178

10 Greedy Algorithm 185

10.1 Can Place Flowers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186
10.2 Minimum Deletions to Make Character Frequencies Unique . . . . . . 188
10.3 Wiggle Subsequence . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192
10.4 Partitioning Into Minimum Number Of Deci-Binary Numbers . . . . . . 196
10.5 Maximum Units on a Truck . . . . . . . . . . . . . . . . . . . . . . . . 198

11 Dynamic Programming 203

11.1 Fibonacci Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204
11.2 Unique Paths . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208
11.3 Largest Divisible Subset . . . . . . . . . . . . . . . . . . . . . . . . . . 215
11.4 Triangle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222

x
 11.5 Unique Paths II . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226

12 Counting 231
12.1 Single Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 232
12.2 First Unique Character in a String . . . . . . . . . . . . . . . . . . . . . 235
12.3 Max Number of K-Sum Pairs . . . . . . . . . . . . . . . . . . . . . . . . 239

13 Prefix Sums 243

13.1 Running Sum of 1d Array . . . . . . . . . . . . . . . . . . . . . . . . . 244
13.2 Maximum Subarray . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247
13.3 Product of Array Except Self . . . . . . . . . . . . . . . . . . . . . . . . 250
13.4 Subarray Sum Equals K . . . . . . . . . . . . . . . . . . . . . . . . . . 255

14 Two Pointers 263

14.1 Middle of the Linked List . . . . . . . . . . . . . . . . . . . . . . . . . . 264
14.2 Linked List Cycle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 269
14.3 Sort Array By Parity II . . . . . . . . . . . . . . . . . . . . . . . . . . . 275
14.4 Container With Most Water . . . . . . . . . . . . . . . . . . . . . . . . 280
14.5 Remove Nth Node From End of List . . . . . . . . . . . . . . . . . . . . 285
14.6 Shortest Unsorted Continuous Subarray . . . . . . . . . . . . . . . . . 291

15 Mathematics 297
15.1 Excel Sheet Column Number . . . . . . . . . . . . . . . . . . . . . . . 298
15.2 Power of Three . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 301
15.3 Best Time to Buy and Sell Stock . . . . . . . . . . . . . . . . . . . . . . 304
15.4 Subsets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 310
15.5 Minimum Moves to Equal Array Elements II . . . . . . . . . . . . . . . 313
15.6 Array Nesting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 317
15.7 Count Sorted Vowel Strings . . . . . . . . . . . . . . . . . . . . . . . . 321
15.8 Concatenation of Consecutive Binary Numbers . . . . . . . . . . . . . . 326
15.9 Perfect Squares . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 328

16 Conclusion 337

A Coding challenge best practices 339

A.1 Read the problem carefully . . . . . . . . . . . . . . . . . . . . . . . . 339
A.2 Plan and pseudocode . . . . . . . . . . . . . . . . . . . . . . . . . . . . 339
A.3 Test your code . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 339
A.4 Optimize for time and space complexity . . . . . . . . . . . . . . . . . 340
A.5 Write clean, readable code . . . . . . . . . . . . . . . . . . . . . . . . . 340
A.6 Submit your code and learn from feedback . . . . . . . . . . . . . . . . 340
A.7 Keep practicing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 340

CONTENTS xi
xii CONTENTS
 CHAPTER

ONE

INTRODUCTION

1.1 Why LeetCode?

Coding challenges are a great way to practice problem-solving, algorithm develop-

ment, and logical thinking. They showcase your creativity and innovation while im-
proving your coding techniques. This book offers diverse coding challenges to help
you develop your skills.
Coding challenges could be programming puzzles or mathematical problems that re-
quire coding solutions. Each challenge requires different coding skills and is designed
to challenge and develop a particular set of skills.
The coding challenges in this book are picked from LeetCode1 . It is a popular online
platform for programmers and software engineers that provides many coding chal-
lenges and problems. The website was launched in 2015 and has since grown to
become one of the go-to resources for coding practice, technical interview prepara-
tion, and skills enhancement.
LeetCode offers diverse coding challenges, ranging from easy to hard, covering a
wide range of topics such as algorithms, data structures, databases, system design,
and more. The problems are created by industry experts and are designed to simulate
real-world scenarios, allowing you to gain practical experience in problem-solving.
One feature that makes LeetCode stand out is its extensive discussion forum, where
you can interact, share your solutions, and learn from one another. This fosters com-
munity and collaboration, as you can receive feedback on their solutions and ask for
clarification on difficult problems.
LeetCode also provides premium services like mock interviews with real-world com-
panies, career coaching, and job postings. These premium services are designed to
1 https://leetcode.com

1
help you prepare for technical interviews, sharpen your skills, and advance your ca-
reers.
LeetCode has become a popular resource for technical interview preparation, as many
companies use similar problems to screen and evaluate potential candidates. The plat-
form has helped many users to secure job offers from top companies in the technology
industry, including Google, Microsoft, and Facebook.
In summary, LeetCode is a valuable resource for programmers and software engineers
looking to improve their coding skills, prepare for technical interviews, and advance
their careers. Its extensive collection of coding challenges, community discussion
forums, and premium services make it an all-in-one platform for coding practice and
skills enhancement.

1.2 A brief about algorithm complexity

Algorithm complexity, also known as runtime complexity, is a measure of how the

running time of an algorithm increases as the input size grows. It is an essential
concept in computer science, as it helps programmers evaluate and optimize their
algorithms’ performance.
The complexity of an algorithm is usually measured in terms of its Big O notation,
which describes the upper bound of the algorithm’s running time as a function of
the input size. For example, an algorithm with a time complexity of O(n) will have
a running time proportional to the input size. In contrast, an algorithm with a time
complexity of O(n^2) will have a running time proportional to the square of the input
size.
Algorithm complexity is important because it helps programmers determine their al-
gorithms’ efficiency and scalability. In general, algorithms with lower complexity are
more efficient, as they require less time and resources to process larger inputs. By
analyzing the time complexity of an algorithm, programmers can identify potential
performance bottlenecks and optimize their code accordingly.
In addition to time complexity, algorithms may also have space complexity, which
measures the memory required to execute the algorithm. Space complexity is also
measured in Big O notation and is important for optimizing the memory usage of an
algorithm.

2 Chapter 1. Introduction
While it is important to optimize the performance of algorithms, it is also important to
balance this with readability and maintainability. A highly optimized algorithm may
be difficult to understand and maintain, which can lead to problems in the long run.
Therefore, it is important to balance performance and readability when designing and
implementing algorithms.
In summary, algorithm complexity is an essential concept in computer science that
helps programmers evaluate and optimize their algorithms’ performance. By ana-
lyzing an algorithm’s time and space complexity, programmers can identify potential
performance bottlenecks and optimize their code to improve efficiency and scalability.

1.2. A brief about algorithm complexity 3

1.3 Why readable code?

Readable code is code that is easy to understand, maintain, and modify. It is an essen-
tial aspect of programming, as it ensures that code is accessible to other programmers
and helps to prevent errors and bugs. Readable code is important for several reasons.
Firstly, readable code makes it easier for other programmers to understand and mod-
ify it. This is particularly important in collaborative projects where multiple program-
mers work on the same codebase. If the code is not readable, it can lead to confusion
and errors, making it difficult for others to work on it.
Secondly, readable code helps to prevent bugs and errors. When code is easy to un-
derstand, it is easier to identify and fix potential issues before they become problems.
This is important for ensuring the code is reliable and performs as expected.
Thirdly, readable code can improve the overall quality of the codebase. When code is
easy to understand, it is easier to identify areas for improvement and make changes
to improve the code. This can help improve the codebase’s efficiency and maintain-
ability, leading to a better overall product.
Finally, readable code can save time and money. When code is easy to understand, it
is easier to maintain and modify. This can help reduce the time and resources required
to make changes to the codebase, leading to cost savings in the long run.
In conclusion, readable code is an essential aspect of programming that ensures that
code is accessible, error-free, and efficient. By focusing on readability when designing
and implementing code, programmers can improve the quality and reliability of their
code, leading to a better overall product.

4 Chapter 1. Introduction
 I hope this book is an enjoyable and educational experience that will chal-
lenge and inspire you. Whether you want to enhance your skills, prepare for
a technical interview, or just have fun, this book has something for you. So,
get ready to put your coding skills to the test and embark on a challenging
and rewarding journey through the world of coding challenges!

1.3. Why readable code? 5

6 Chapter 1. Introduction
 CHAPTER

TWO

ARRAY

This chapter will explore the basics of arrays - collections of elements organized in a
sequence. While they may seem simple, you can learn many concepts and techniques
from arrays to improve your coding skills. We’ll cover topics like indexing, itera-
tion, and manipulation, as well as dynamic arrays (std::vector) and time/space
complexity.
Along the way, we’ll tackle challenging problems like searching, sorting, and subar-
ray problems, using a structured approach to break down complex tasks into man-
ageable steps.
What this chapter covers:
1. Fundamentals of Arrays: Gain a solid understanding of arrays, their proper-
ties, and how to access and manipulate elements efficiently.
2. Array Operations: Learn essential array operations like insertion, deletion, and
updating elements, and understand their trade-offs.
3. Dynamic Arrays: Explore dynamic arrays, their advantages over static arrays,
and the mechanics of resizing.
4. Time and Space Complexity: Grasp the importance of analyzing the efficiency
of algorithms and how to evaluate the time and space complexity of array-
related operations.
5. Common Array Algorithms: Discover classic algorithms such as searching,
sorting, and various techniques for tackling subarray problems.
6. Problem-Solving Strategies: Develop systematic strategies to approach array-
related challenges, including how to break down problems, devise algorithms,
and validate solutions.

7
2.1 Transpose Matrix

2.1.1 Problem statement

1
You are given a 2D integer array matrix, and your objective is to find the transpose
of the given matrix.
The transpose of a matrix involves flipping the matrix over its main diagonal, effec-
tively swapping its row and column indices.
⎡ ⎤ ⎡ ⎤
1 2 3 1 4 7
⎣4 5 6⎦ −→ ⎣2 5 8⎦
7 8 9 3 6 9

Example 1

Input: matrix = [[1,2,3],[4,5,6],[7,8,9]]

Output: [[1,4,7],[2,5,8],[3,6,9]]

Example 2

Input: matrix = [[1,2,3],[4,5,6]]

Output: [[1,4],[2,5],[3,6]]

Constraints

• m == matrix.length.
• n == matrix[i].length.
• 1 <= m, n <= 1000.
• 1 <= m * n <= 10^5.
• -10^9 <= matrix[i][j] <= 10^9.
1 https://leetcode.com/problems/transpose-matrix/

8 Chapter 2. Array
2.1.2 Solution

Code

#include <iostream>
#include <vector>
using namespace std;
vector<vector<int>> transpose(const vector<vector<int>>& matrix) {
// declare the transposed matrix mt of desired size, i.e.
// mt's number of rows = matrix's number of columns
// mt's number of columns = matrix's number of rows
vector<vector<int>> mt(matrix[0].size(),
vector<int>(matrix.size()));
for (int i = 0; i < mt.size(); i++) {
for (int j = 0; j < mt[i].size(); j++) {
mt[i][j] = matrix[j][i];
}
}
return mt;
}
void printResult(const vector<vector<int>>& matrix) {
cout << "[";
for (auto& row : matrix) {
cout << "[";
for (int m : row) {
cout << m << ",";
}
cout << "]";
}
cout << "]\n";
}
int main() {
vector<vector<int>> matrix = {{1,2,3},{4,5,6},{7,8,9}};
auto result = transpose(matrix);
printResult(result);
matrix = {{1,2,3},{4,5,6}};
result = transpose(matrix);
printResult(result);
}

Output:
(continues on next page)

2.1. Transpose Matrix 9

 (continued from previous page)
[[1,4,7,][2,5,8,][3,6,9,]]
[[1,4,][2,5,][3,6,]]

Complexity

• Runtime: O(m*n), where m = matrix.length and n = matrix[i].length.

• Extra space: O(1).

2.1.3 Implementation note

Note that the matrix might not be square, you cannot just swap the elements using
for example the function std::swap.

2.2 Valid Mountain Array

2.2.1 Problem statement

1
You are given an array of integers arr, and your task is to determine whether it is a
valid mountain array.
A valid mountain array must meet the following conditions:
1. The length of arr should be greater than or equal to 3.
2. There should exist an index i such that 0 < i < arr.length - 1, and the
elements up to i (arr[0] to arr[i]) should be in strictly ascending order, while
the elements starting from i (arr[i] to arr[arr.length-1]) should be in strictly
descending order.
1 https://leetcode.com/problems/valid-mountain-array/

10 Chapter 2. Array
Example 1

Input: arr = [2,1]

Output: false

Example 2

Input: arr = [3,5,5]

Output: false

2.2. Valid Mountain Array 11

Example 3

Input: arr = [0,3,2,1]

Output: true

Constraints

• 1 <= arr.length <= 10^4.

• 0 <= arr[i] <= 10^4.

2.2.2 Solution

Following the conditions, we have the following implementation.

Code

#include <vector>
#include <iostream>
using namespace std;
bool validMountainArray(const vector<int>& arr) {
if (arr.size() < 3) {
return false;
}
const int N = arr.size() - 1;
int i = 0;
// find the top of the mountain
while (i < N && arr[i] < arr[i + 1]) {
i++;
}
// condition: 0 < i < N - 1
if (i == 0 || i == N) {
return false;
}
// going from the top to the bottom
while (i < N && arr[i] > arr[i + 1]) {
i++;
}
return i == N;
(continues on next page)

12 Chapter 2. Array
 (continued from previous page)
}
int main() {
vector<int> arr{2,1};
cout << validMountainArray(arr) << endl;
arr = {3,5,5};
cout << validMountainArray(arr) << endl;
arr = {0,3,2,1};
cout << validMountainArray(arr) << endl;
arr = {9,8,7,6,5,4,3,2,1,0};
cout << validMountainArray(arr) << endl;
}

Output:
0
0
1
0

This solution iteratively checks for the two slopes of a mountain array, ensuring that
the elements to the left are strictly increasing and the elements to the right are strictly
decreasing. If both conditions are met, the function returns true, indicating that the
input array is a valid mountain array; otherwise, it returns false.

Complexity

• Runtime: O(N), where N = arr.length.

• Extra space: O(1).

2.2.3 Coding best practices

Breaking down the problem into distinct stages, like finding the peak of the mountain
and then traversing down from there, can simplify the logic and improve code read-
ability. This approach facilitates a clear understanding of the algorithm’s progression
and helps in handling complex conditions effectively.

2.2. Valid Mountain Array 13

2.2.4 Exercise

• Beautiful Towers I2 .

2.3 Shift 2D Grid

2.3.1 Problem statement

1
You are given a 2D grid with dimension mxn and an integer k. Your task is to perform
k shift operations on the grid.
In each shift operation:
• The element at grid[i][j] moves to grid[i][j+1].
• The element at grid[i][n-1] moves to grid[i+1][0].
• The element at grid[m-1][n-1] moves to grid[0][0].
After performing k shift operations, return the updated 2D grid.

Example 1
⎡ ⎤ ⎡ ⎤
1 2 3 9 1 2
⎣4 5 6⎦ −→ ⎣3 4 5⎦
7 8 9 6 7 8

Input: grid = [[1,2,3],[4,5,6],[7,8,9]], k = 1

Output: [[9,1,2],[3,4,5],[6,7,8]]

Example 2
⎡ ⎤ ⎡ ⎤ ⎡ ⎤
3 8 1 9 13 3 8 1 21 13 3 8
⎢19 7 2 5⎥ ⎢9 19 7 2 ⎥ ⎢1 9 19 7⎥
⎢ ⎥→⎢ ⎥→⎢ ⎥
⎣4 6 11 10⎦ ⎣5 4 6 11⎦ ⎣2 5 4 6⎦
12 0 21 13 10 12 0 21 11 10 12 0
2 https://leetcode.com/problems/beautiful-towers-i/
1 https://leetcode.com/problems/shift-2d-grid/

14 Chapter 2. Array
 ⎡ ⎤ ⎡ ⎤
0 21 13 3 12 0 21 13
⎢8 1 9 19⎥ ⎢3 8 1 9⎥
→⎢
⎣7
⎥→⎢ ⎥
2 5 4⎦ ⎣19 7 2 5⎦
6 11 10 12 4 6 11 10

Input: grid = [[3,8,1,9],[19,7,2,5],[4,6,11,10],[12,0,21,13]], k = 4

Output: [[12,0,21,13],[3,8,1,9],[19,7,2,5],[4,6,11,10]]

Example 3

Input: grid = [[1,2,3],[4,5,6],[7,8,9]], k = 9

Output: [[1,2,3],[4,5,6],[7,8,9]]

Constraints

• 1 <= grid.length <= 50.

• 1 <= grid[i].length <= 50.
• -1000 <= grid[i][j] <= 1000.
• 0 <= k <= 100.

2.3.2 Solution: Convert a 2D array into a 1D one

You can convert the 2D grid into a 1D vector v to perform the shifting easier. One
way of doing this is concatenating the rows of the matrix.
• If you shift the grid k = i*N times where N = v.size() and i is any non-negative
integer, you go back to the original grid; i.e. you did not shift it.
• If you shift the grid k times with 0 < k < N, the first element of the result starts
from v[N-k].
• In general, the first element of the result starts from v[N - k%N].

2.3. Shift 2D Grid 15

Example 1

For grid = [[1,2,3],[4,5,6],[7,8,9]]:

• It can be converted into a 1D vector v = [1,2,3,4,5,6,7,8,9] of size m*n = 9.
• With k = 1 the shifted grid now starts from v[9-1] = 9.
• The final result is grid = [[9,1,2][3,4,5][6,7,8]].

Code

#include <vector>
#include <iostream>
using namespace std;
vector<vector<int>> shiftGrid(vector<vector<int>>& grid, int k) {
vector<int> v;
// store the 2D grid values into a 1D vector v
for (auto& r : grid) {
v.insert(v.end(), r.begin(), r.end());
}
const int N = v.size();

// perform the shifting

int p = N - k % N;

// number of rows
const int m = grid.size();

// number of columns
const int n = grid[0].size();

for (int i = 0; i < m; i++) {

for (int j = 0; j < n; j++) {
if (p == N) {
p = 0;
}
// reconstruct the grid
grid[i][j] = v[p++];
}
}
return grid;
(continues on next page)

16 Chapter 2. Array
 (continued from previous page)
}
void printResult(const vector<vector<int>>& grid) {
cout << "[";
for (auto& r : grid) {
cout << "[";
for (int a: r) {
cout << a << ",";
}
cout << "]";
}
cout << "]\n";
}
int main() {
vector<vector<int>> grid{{1,2,3},{4,5,6},{7,8,9}};
auto result = shiftGrid(grid, 1);
printResult(result);
grid = {{3,8,1,9},{19,7,2,5},{4,6,11,10},{12,0,21,13}};
result = shiftGrid(grid, 4);
printResult(result);
grid = {{1,2,3},{4,5,6},{7,8,9}};
result = shiftGrid(grid, 9);
printResult(result);
}

Output:
[[9,1,2,][3,4,5,][6,7,8,]]
[[12,0,21,13,][3,8,1,9,][19,7,2,5,][4,6,11,10,]]
[[1,2,3,][4,5,6,][7,8,9,]]

This solution flattens the 2D grid into a 1D vector v, representing the grid’s elements
in a linear sequence. Then, by calculating the new position for each element after
the shift operation, it reconstructs the grid by placing the elements back into their re-
spective positions based on the calculated indices. This approach avoids unnecessary
copying or shifting of elements within the grid, optimizing both memory and time
complexity.

2.3. Shift 2D Grid 17

Complexity

• Runtime: O(m*n) (the nested for loops), where m = grid.length and n =

grid[i].length.
• Extra space: O(m*n) (the vector v).

2.3.3 Key takeaway

1. To convert a 2D matrix into a 1D vector, you can use the std::vector’s function
insert()2 .
2. The modulo operator % is usually used to ensure the index is inbound.

2.4 Find All Numbers Disappeared in an Array

2.4.1 Problem statement

1
You are given an array nums of n integers, where each integer nums[i] falls within
the range [1, n]. Your task is to find and return an array containing all the integers
in the range [1, n] that are not present in the given array nums.

Example 1

Input: nums = [4,3,2,7,8,2,3,1]

Output: [5,6]

Example 2

Input: nums = [1,1]

Output: [2]

2 https://en.cppreference.com/w/cpp/container/vector/insert
1 https://leetcode.com/problems/find-all-numbers-disappeared-in-an-array/

18 Chapter 2. Array
Constraints

• n == nums.length.
• 1 <= n <= 10^5.
• 1 <= nums[i] <= n.

Follow up

Can you solve the problem without using additional memory and achieve a linear
runtime complexity? You can assume that the list you return does not count as extra
space.

2.4.2 Solution 1: Marking the appearances by bool

You can use a vector of bool to mark which value appeared in the array.

Code

#include <vector>
#include <iostream>
using namespace std;
vector<int> findDisappearedNumbers(const vector<int>& nums) {

const int n = nums.size();

vector<bool> exist(n + 1, false);
for (auto& a : nums) {
exist[a] = true;
}
vector<int> result;
for (int i = 1; i <= n; i++) {
if (!exist.at(i)) {
result.push_back(i);
}
}
return result;
}
void print(const vector<int>& nums) {
cout << "[";
(continues on next page)

2.4. Find All Numbers Disappeared in an Array 19

 (continued from previous page)
for (auto& a : nums) {
cout << a << ",";
}
cout << "]\n";
}
int main() {
vector<int> nums = {4,3,2,7,8,2,3,1};
auto result = findDisappearedNumbers(nums);
print(result);
nums = {1,1};
result = findDisappearedNumbers(nums);
print(result);
}

Output:
[5,6,]
[2,]

This code declares a vector named exist of type bool and initializes all of its values
to false. Its size is declared as n + 1 where n = nums.size() so it can mark the
values ranged from 1 to n.
Then it performs the marking of all nums’s elements to true. The ones that are false
will belong to the result.

Complexity

• Runtime: O(n), where n = nums.length.

• Extra space: much less than O(n). vector<bool> is optimized for space effi-
ciency; it stores single bits.

2.4.3 Solution 2: Marking the appearances by sign

You could use the indices of the array nums to mark the appearances of its elements
because they are just a shift ([1, n] vs. [0, n-1]).
One way of marking the appearance of a value j (1 <= j <= n) is making the element
nums[j-1] to be negative. Then the indices j’s whose nums[j-1] are still positive are
the ones that do not appear in nums.

20 Chapter 2. Array
Example 1

With nums = [4,3,2,7,8,2,3,1]:

• To indicate 4 is present, make nums[4-1] is negative, i.e. changing nums[4-1] =
nums[3] to -7.
• To indicate 3 is present, make nums[3-1] is negative, i.e. changing nums[3-1] =
nums[2] to -2.
• And so on.
• nums becomes [-4,-3,-2,-7,8,2,-3,-1].
• The positive values 8 corresponds to nums[4] = nums[5-1], indicates 5 was not
present in nums.
• Similarly, the positive values 2 corresponds to nums[5] = nums[6-1], indicates
6 was not present in nums.

Code

#include <vector>
#include <iostream>
using namespace std;
vector<int> findDisappearedNumbers(vector<int>& nums) {
const int n = nums.size();
int j;
for (int i{0}; i < n; i++) {
// make sure j is positive since nums[i] might be
// changed to be negative in previous steps
j = abs(nums.at(i));

// Mark nums[j - 1] as negative to indicate its presence

nums[j - 1] = -abs(nums.at(j - 1));
}
vector<int> result;
for (int j{1}; j <= n; j++) {
// If nums[j - 1] is positive, it means j is missing
if (nums.at(j - 1) > 0) {
result.push_back(j);
}
}
return result;
(continues on next page)

2.4. Find All Numbers Disappeared in an Array 21

 (continued from previous page)
}
void print(const vector<int>& nums) {
cout << "[";
for (auto& a : nums) {
cout << a << ",";
}
cout << "]\n";
}
int main() {
vector<int> nums = {4,3,2,7,8,2,3,1};
auto result = findDisappearedNumbers(nums);
print(result);
nums = {1,1};
result = findDisappearedNumbers(nums);
print(result);
}

Output:
[5,6,]
[2,]

The key to this solution is that it utilizes the array to mark the presence of numbers.
Negating the value at the index corresponding to each number found in the input ar-
ray effectively marks that number as present. Then, by iterating through the modified
array, it identifies the missing numbers by checking which indices still hold positive
values.

Complexity

• Runtime: O(n), where n = nums.length.

• Extra space: O(1) (the returned list does not count as extra space).

22 Chapter 2. Array
2.4.4 Readable code

• Solution 2 helps to avoid allocating extra memory but it is not straightforward

to understand.
• Though Solution 1 requires some extra space, that memory is not much since
std::vector<bool>2 is optimized for space efficiency. Moreover, it is easier to
understand than Solution 2.

2.4.5 Exercise

• Find All Duplicates in an Array3 .

2.5 Rotate Image

2.5.1 Problem statement

1
Given an n x n 2D matrix representing an image, your task is to rotate the image by
90 degrees clockwise. The rotation must be performed in-place, meaning you need
to modify the original input 2D matrix directly. It is not allowed to create another 2D
matrix for the rotation.

Example 1
⎡ ⎤ ⎡ ⎤
1 2 3 7 4 1
⎣4 5 6⎦ −→ ⎣8 5 2⎦
7 8 9 9 6 3

Input: matrix = [[1,2,3],[4,5,6],[7,8,9]]

Output: [[7,4,1],[8,5,2],[9,6,3]]

2 https://en.cppreference.com/w/cpp/container/vector_bool
3 https://leetcode.com/problems/find-all-duplicates-in-an-array
1 https://leetcode.com/problems/rotate-image/

2.5. Rotate Image 23

Example 2
⎡ ⎤ ⎡ ⎤
5 1 9 11 15 13 2 5
⎢2 4 8 10⎥⎥ −→ ⎢14 3 4 1⎥
⎢
⎢ ⎥
⎣13 3 6 7⎦ ⎣12 6 8 9⎦
15 14 12 16 16 7 10 11

Input: matrix = [[5,1,9,11],[2,4,8,10],[13,3,6,7],[15,14,12,16]]

Output: [[15,13,2,5],[14,3,4,1],[12,6,8,9],[16,7,10,11]]

Constraints

• n == matrix.length == matrix[i].length.
• 1 <= n <= 20.
• -1000 <= matrix[i][j] <= 1000.

2.5.2 Solution: The math behind

For any square matrix, the rotation 90 degrees clockwise can be performed in two
steps:
1. Transpose the matrix.
2. Mirror the matrix vertically.

Code

#include <iostream>
#include <vector>
using namespace std;
void rotate(vector<vector<int>>& matrix) {
const int n = matrix.size();
// transpose
for (int i = 0; i < n; i++) {
for (int j = i + 1; j < n; j++) {
swap(matrix[i][j], matrix[j][i]);
}
}
(continues on next page)

24 Chapter 2. Array
 (continued from previous page)
// vertical mirror
for (int i = 0; i < n; i++) {
for (int j = 0; j < n / 2; j++ ) {
swap(matrix[i][j], matrix[i][n - 1 - j]);
}
}
}
void printMatrix(const vector<vector<int>>& matrix) {
cout << "[";
for (auto& row: matrix) {
cout << "[";
for (auto& a: row) {
cout << a << ",";
}
cout << "],";
}
cout << "]\n";
}
int main() {
vector<vector<int>> matrix{{1,2,3},{4,5,6},{7,8,9}};
rotate(matrix);
printMatrix(matrix);
matrix = {{5,1,9,11},{2,4,8,10},{13,3,6,7},{15,14,12,16}};
rotate(matrix);
printMatrix(matrix);
}

Output:
[[7,4,1,],[8,5,2,],[9,6,3,],]
[[15,13,2,5,],[14,3,4,1,],[12,6,8,9,],[16,7,10,11,],]

Complexity

• Runtime: O(n^2), where n = matrix.length.

• Extra space: O(1).

2.5. Rotate Image 25

2.5.3 Implementation tips

1. The function std::swap2 can be used to exchange two values.

2. When doing the transpose or mirroring, you could visit over only one-half of
the matrix.

2.5.4 Exercise

• Determine Whether Matrix Can Be Obtained By Rotation3 .

2.6 Spiral Matrix II

2.6.1 Problem statement

1
Given a positive integer n, generate an n x n matrix filled with elements from 1 to
n^2 in spiral order.

Example 1

Input: n = 3
Output: [[1,2,3],[8,9,4],[7,6,5]]

2 https://en.cppreference.com/w/cpp/algorithm/swap
3 https://leetcode.com/problems/determine-whether-matrix-can-be-obtained-by-rotation/
1 https://leetcode.com/problems/spiral-matrix-ii/

26 Chapter 2. Array
Example 2

Input: n = 1
Output: [[1]]

Constraints

• 1 <= n <= 20.

2.6.2 Solution

1. Starting from the top left of the matrix.

2. Going along the spiral direction.
3. Put the value to the matrix, starting from 1.

Code

#include <vector>
#include <iostream>
using namespace std;
enum Direction {RIGHT, DOWN, LEFT, UP};
vector<vector<int>> generateMatrix(int n) {
vector<vector<int>> m(n, vector<int>(n));
int bottom = n - 1;
int right = n - 1;
int top = 0;
int left = 0;
int row = 0;
int col = 0;
Direction d = RIGHT;
int a = 1;
while (top <= bottom && left <= right) {
m[row][col] = a++;
switch (d) {
case RIGHT: if (col == right) {
top++;
d = DOWN;
(continues on next page)

2.6. Spiral Matrix II 27

 (continued from previous page)
row++;
} else {
col++;
}
break;
case DOWN: if (row == bottom) {
right--;
d = LEFT;
col--;
} else {
row++;
}
break;
case LEFT: if (col == left) {
bottom--;
d = UP;
row--;
} else {
col--;
}
break;
case UP: if (row == top) {
left++;
d = RIGHT;
col++;
} else {
row--;
}
break;
}
}
return m;
}
void printResult(const vector<vector<int>>& m) {
cout << "[";
for (auto& r : m) {
cout << "[";
for (int a : r) {
cout << a << ",";
}
(continues on next page)

28 Chapter 2. Array
 (continued from previous page)
cout << "]";
}
cout << "]\n";
}
int main() {
auto m = generateMatrix(3);
printResult(m);
m = generateMatrix(1);
printResult(m);
}

Output:
[[1,2,3,][8,9,4,][7,6,5,]]
[[1,]]

This solution uses a Direction enum and boundary variables to iteratively fill the
matrix in a spiral pattern. Updating the direction of movement based on the cur-
rent position and boundaries efficiently populates the matrix with sequential values,
traversing in a clockwise direction from the outer layer to the inner layer.

Complexity

• Runtime: O(n^2), where n x n is the size of the matrix.

• Extra space: O(1).

2.6. Spiral Matrix II 29

2.6.3 Key Takeaway

Enumerating directions with an enum (like Direction) can enhance code readability
and maintainability, especially in algorithms involving traversal or movement. It aids
in clearly defining and referencing the possible directions within the problem domain.

2.6.4 Exercise

• Spiral Matrix2 .

2.7 Daily Temperatures

2.7.1 Problem statement

1
You are given an array of integers temperatures, which represents the daily tem-
peratures. Your task is to create an array answer such that answer[i] represents the
number of days you need to wait after the i-th day to experience a warmer tempera-
ture. If there is no future day with a warmer temperature, then answer[i] should be
set to 0.

Example 1

Input: temperatures = [73,74,75,71,69,72,76,73]

Output: [1,1,4,2,1,1,0,0]

Example 2

Input: temperatures = [30,40,50,60]

Output: [1,1,1,0]

2 https://leetcode.com/problems/spiral-matrix/
1 https://leetcode.com/problems/daily-temperatures/

30 Chapter 2. Array
Example 3

Input: temperatures = [30,60,90]

Output: [1,1,0]

Constraints

• 1 <= temperatures.length <= 10^5.

• 30 <= temperatures[i] <= 100.

2.7.2 Solution 1: Starting from the first day

For each temperatures[i], find the closest temperatures[j] with j > i such that
temperatures[j] > temperatures[i], then answer[i] = j - i. If not found,
answer[i] = 0.

Example 1

For temperatures = [73,74,75,71,69,72,76,73]:

• answer[0] = 1 since the next day is warmer (74 > 73).
• answer[1] = 1 since the next day is warmer (75 > 74).
• answer[2] = 4 since only after 4 days it is warmer (76 > 75).
• And so on.

Code

#include <vector>
#include <iostream>
using namespace std;
vector<int> dailyTemperatures(const vector<int>& temperatures) {
vector<int> answer(temperatures.size());
for (int i = 0; i < temperatures.size(); i++) {
answer[i] = 0;
for (int j = i + 1; j < temperatures.size(); j++) {
if (temperatures[j] > temperatures[i]) {
(continues on next page)

2.7. Daily Temperatures 31

 (continued from previous page)
answer[i] = j - i;
break;
}
}
}
return answer;
}
void print(const vector<int>& answer) {
cout << "[";
for (auto& v : answer ) {
cout << v << ",";
}
cout << "]\n";
}
int main() {
vector<int> temperatures{73,74,75,71,69,72,76,73};
auto answer = dailyTemperatures(temperatures);
print(answer);
temperatures = {30,40,50,60};
answer = dailyTemperatures(temperatures);
print(answer);
temperatures = {30,60,90};
answer = dailyTemperatures(temperatures);
print(answer);
}

Output:
[1,1,4,2,1,1,0,0,]
[1,1,1,0,]
[1,1,0,]

This solution iterates through the temperatures array and, for each temperature, it-
erates through the remaining temperatures to find the next higher temperature. Stor-
ing the time difference between the current day and the next higher temperature day
constructs the resulting array representing the number of days until warmer temper-
atures.

32 Chapter 2. Array
Complexity

• Runtime: O(N^2), where N = temperatures.length.

• Extra space: O(1).

2.7.3 Solution 2: Starting from the last day

The straightforward solution above is easy to understand, but the complexity is

O(N^2).
The way starting from the first day to the last day does not make use of the knowledge
of the answer[i] values.
• The value answer[i] > 0 tells you that temperatures[i + answer[i]] is the
next temperature that is warmer than temperatures[i].
• The value answer[i] = 0 tells you that there is no warmer temperature than
temperatures[i].
When computing answer[i] in the reversed order, you can use that knowledge more
efficiently.
Suppose you already know the future values answer[j]. To compute an older
value answer[i] with i < j, you need only to compare temperatures[i] with
temperatures[i + 1] and its chain of warmer temperatures.

Example 1

For temperatures = [73,74,75,71,69,72,76,73].

Suppose you have computed all answer[j] with j > 2, answer = [?,?,?,2,1,1,0,0].
To compute answer[i = 2] for temperatures[2] = 75, you need to compare it with
• temperatures[3] = 71 (< 75). Go to the next warmer temperature than
temperatures[3], which is temperatures[3 + answer[3]] = temperatures[3
+ 2].
• temperatures[5] = 72 (< 75). Go to the next warmer temperature than
temperatures[5], which is temperatures[5 + answer[5]] = temperatures[5
+ 1].
• temperatures[6] = 76 (> 75). Stop.
• answer[i = 2] = j - i = 6 - 2 = 4.

2.7. Daily Temperatures 33

Code

#include <vector>
#include <iostream>
using namespace std;
vector<int> dailyTemperatures(const vector<int>& temperatures) {
vector<int> answer(temperatures.size(), 0);
for (int i = temperatures.size() - 2; i >= 0 ; i--) {
int j = i + 1;
while (j < temperatures.size() &&
temperatures[j] <= temperatures[i]) {
// some temperature is bigger than temperatures[j],
// go to that value
if (answer[j] > 0) {
j += answer[j];
} else {
j = temperatures.size();
}
}
if (j < temperatures.size()) {
answer[i] = j - i;
}
}
return answer;
}
void print(const vector<int>& answer) {
cout << "[";
for (auto& v : answer ) {
cout << v << ",";
}
cout << "]\n";
}
int main() {
vector<int> temperatures{73,74,75,71,69,72,76,73};
auto answer = dailyTemperatures(temperatures);
print(answer);
temperatures = {30,40,50,60};
answer = dailyTemperatures(temperatures);
print(answer);
temperatures = {30,60,90};
answer = dailyTemperatures(temperatures);
(continues on next page)

34 Chapter 2. Array
 (continued from previous page)
print(answer);
}

Output:
[1,1,4,2,1,1,0,0,]
[1,1,1,0,]
[1,1,0,]

The key to this solution lies in its optimized approach to finding the next higher
temperature. It utilizes a while loop to traverse the temperatures array efficiently,
skipping elements if they are not potential candidates for a higher temperature. Up-
dating the index based on previously calculated values stored in the answer array
avoids unnecessary iterations, resulting in improved performance compared to the
straightforward nested loop approach.
This improved solution reduces the time complexity to O(N) as it iterates through the
temperatures vector only once, resulting in a more efficient algorithm for finding the
waiting periods for each day.

Complexity

Worse cases for the while loop are when most temperatures[j] in their chain are
cooler than temperatures[i].
In these cases, the resulting answer[i] will be either 0 or a big value j - i. Those
extreme values give you a huge knowledge when computing answer[i] for other
older days i.
The value 0 would help the while loop terminates very soon. On the other hand, the
big value j - i would help the while loop skips the days j very quickly.
• Runtime: O(N), where N = temperatures.length.
• Extra space: O(1).

2.7. Daily Temperatures 35

2.7.4 Tips

In some computations, you could improve the performance by using the knowledge
of the results you have computed.
In this particular problem, it can be achieved by doing it in the reversed order.

2.7.5 Exercise

• Next Greater Element I2 .

2 https://leetcode.com/problems/next-greater-element-i/description/

36 Chapter 2. Array
 CHAPTER

THREE

LINKED LIST

In this chapter, we’ll learn about linked list - a unique and dynamic data structure
that challenges our understanding of sequential data.
Unlike arrays, linked lists do not impose a fixed size or continuous memory block.
Rather, they consist of nodes that contain data and a reference to the next node. This
seemingly simple concept unlocks many possibilities, from creating efficient insertions
and deletions to creatively solving problems that may seem specifically designed for
linked list manipulation.
Our exploration of linked lists will encompass a variety of variations and intricacies,
including singly linked lists. By delving into these lists, you’ll discover how they
empower us to tackle problems that may initially appear complicated.
What this chapter covers:
1. Introduction to Linked Lists: Gain a comprehensive understanding of linked
lists, their advantages, and their role in problem-solving.
2. Singly Linked Lists: Explore the mechanics of singly linked lists, mastering the
art of traversal, insertion, and deletion.
3. Advanced Linked List Concepts: Learn about sentinel nodes, dummy nodes,
and techniques to handle common challenges like detecting cycles and reversing
lists.
4. Problem-Solving Strategies: Develop strategies to approach linked list prob-
lems systematically, including strategies for merging lists, detecting intersec-
tions, and more.

37
3.1 Merge Two Sorted Lists

3.1.1 Problem statement

1
Given the starting nodes of two sorted linked lists, list1 and list2, your task is to
combine these lists into a single sorted linked list.
This merged list should be created by connecting the nodes from both list1 and
list2. Finally, you should return the starting node of the resulting merged linked list.

Example 1

Input: list1 = [1,2,4], list2 = [1,3,4]

Output: [1,1,2,3,4,4]

Example 2

Input: list1 = [], list2 = []

Output: []

1 https://leetcode.com/problems/merge-two-sorted-lists/

38 Chapter 3. Linked List

Example 3

Input: list1 = [], list2 = [0]

Output: [0]

Constraints

• The number of nodes in both lists is in the range [0, 50].

• -100 <= Node.val <= 100.
• Both list1 and list2 are sorted in non-decreasing order.

3.1.2 Solution: Constructing a new list

For each pair of nodes between the two lists, pick the node having smaller value to
append to the new list.

Code

#include <iostream>
struct ListNode {
int val;
ListNode *next;
ListNode() : val(0), next(nullptr) {}
ListNode(int x) : val(x), next(nullptr) {}
ListNode(int x, ListNode *next) : val(x), next(next) {}
};

ListNode* mergeTwoLists(ListNode* list1, ListNode* list2) {

if (list1 == nullptr) {
return list2;
} else if (list2 == nullptr) {
return list1;
}
// identify which list is head of the merged one
ListNode* head = list1;
if (list2->val < head->val) {
head = list2;
(continues on next page)

3.1. Merge Two Sorted Lists 39

 (continued from previous page)
list2 = list2->next;
} else {
list1 = list1->next;
}
ListNode* node = head;
while (list1 && list2) {
// pick the smaller node to append to the new list.
if (list1->val < list2->val) {
node->next = list1;
list1 = list1->next;
} else {
node->next = list2;
list2 = list2->next;
}
node = node->next;
}
if (list1 == nullptr) {
node->next = list2;
} else {
node->next = list1;
}
return head;
}
void printResult(ListNode* head) {
std::cout << "[";
while (head) {
std::cout << head->val << ",";
head = head->next;
}
std::cout << "]\n";
}
int main() {
ListNode four1(4);
ListNode two1(2, &four1);
ListNode one1(1, &two1);
ListNode four2(4);
ListNode three2(3, &four2);
ListNode one2(1, &three2);
auto newOne = mergeTwoLists(&one1, &one2);
printResult(newOne);
(continues on next page)

40 Chapter 3. Linked List

 (continued from previous page)

auto empty = mergeTwoLists(nullptr, nullptr);

printResult(empty);

ListNode zero(0);
auto z = mergeTwoLists(nullptr, &zero);
printResult(z);
}

Output:
[1,1,2,3,4,4,]
[]
[0,]

Complexity

• Runtime: O(N), where N = list1.length + list2.length.

• Extra space: O(1).

3.1.3 Conclusion

This solution merges two sorted linked lists efficiently without using extra space.
It identifies the head of the merged list by comparing the values of the first nodes of
the input lists. Then, it iterates through both lists, linking nodes in ascending order
until one list is exhausted.
Finally, it appends the remaining nodes from the non-empty list to the merged list,
ensuring the resulting list remains sorted.

3.1. Merge Two Sorted Lists 41

3.2 Remove Linked List Elements

3.2.1 Problem statement

1
You are given the starting node, head, of a linked list, and an integer val. Your task
is to eliminate all nodes from the linked list that have a value equal to val. After
removing these nodes, return the new starting node of the modified linked list.

Example 1

Input: head = [1,2,6,3,4,5,6], val = 6

Output: [1,2,3,4,5]

Example 2

Input: head = [], val = 1

Output: []

Example 3

Input: head = [7,7,7,7], val = 7

Output: []

1 https://leetcode.com/problems/remove-linked-list-elements/

42 Chapter 3. Linked List

Constraints

• The number of nodes in the list is in the range [0, 10^4].

• 1 <= Node.val <= 50.
• 0 <= val <= 50.

Linked list data structure

struct ListNode {
int val;
ListNode *next;
ListNode() : val(0), next(nullptr) {}
ListNode(int x) : val(x), next(nullptr) {}
ListNode(int x, ListNode *next) : val(x), next(next) {}
};

3.2.2 Solution 1: Consider the special case for head

Removing a node A in a linked list means instead of connecting the previous node
A.pre to A, you connect A.pre to A.next.

Code

#include <iostream>
struct ListNode {
int val;
ListNode *next;
ListNode() : val(0), next(nullptr) {}
ListNode(int x) : val(x), next(nullptr) {}
ListNode(int x, ListNode *next) : val(x), next(next) {}
};
ListNode* removeElements(ListNode* head, int val) {
// remove head if its value matches val
while (head && head->val == val) {
head = head->next;
}
if (head == nullptr) return nullptr;
(continues on next page)

3.2. Remove Linked List Elements 43

 (continued from previous page)
ListNode* pre = head;
while (pre->next) {
if (pre->next->val == val) {
// remove pre->next
pre->next = pre->next->next;
} else {
pre = pre->next;
}
}
return head;
}
void print(const ListNode* head) {
ListNode* node = head;
std::cout << "[";
while (node) {
std::cout << node->val << ",";
node = node->next;
}
std::cout << "]\n";
}
int main() {
ListNode sixb(6);
ListNode five(5, &sixb);
ListNode four(4, &five);
ListNode three(3, &four);
ListNode sixa(6, &three);
ListNode two(2, &sixa);
ListNode head(1, &two);
ListNode* newHead = removeElements(&head, 6);
print(newHead);
newHead = removeElements(nullptr, 1);
print(newHead);
ListNode seven4(7);
ListNode seven3(7, &seven4);
ListNode seven2(7, &seven3);
ListNode seven1(7, &seven2);
newHead = removeElements(&seven1, 7);
print(newHead);
}

44 Chapter 3. Linked List

Output:
[1,2,3,4,5,]
[]
[]

This solution efficiently removes nodes with a specified value val from a linked list
by using two pointers (head and pre) to traverse the list and update the next pointers
to bypass nodes with the specified value.

Complexity

• Runtime: O(N), where N is the number of nodes.

• Memory: O(1).

3.2.3 Solution 2: Create a previous dummy node for head

head has no pre. You can create a dummy node for head.pre whose values is out of
the contraints.

Code

#include <iostream>
struct ListNode {
int val;
ListNode *next;
ListNode() : val(0), next(nullptr) {}
ListNode(int x) : val(x), next(nullptr) {}
ListNode(int x, ListNode *next) : val(x), next(next) {}
};
ListNode* removeElements(ListNode* head, int val) {

// create a new head of value 2023 > 50

ListNode preHead(2023, head);
ListNode* pre = &preHead;
while (pre->next) {
if (pre->next->val == val) {
pre->next = pre->next->next;
} else {
(continues on next page)

3.2. Remove Linked List Elements 45

 (continued from previous page)
pre = pre->next;
}
}
return preHead.next;
}
void print(const ListNode* head) {
ListNode* node = head;
std::cout << "[";
while (node) {
std::cout << node->val << ",";
node = node->next;
}
std::cout << "]\n";
}
int main() {
ListNode sixb(6);
ListNode five(5, &sixb);
ListNode four(4, &five);
ListNode three(3, &four);
ListNode sixa(6, &three);
ListNode two(2, &sixa);
ListNode head(1, &two);
ListNode* newHead = removeElements(&head, 6);
print(newHead);

newHead = removeElements(nullptr, 1);

print(newHead);

ListNode seven4(7);
ListNode seven3(7, &seven4);
ListNode seven2(7, &seven3);
ListNode seven1(7, &seven2);
newHead = removeElements(&seven1, 7);
print(newHead);
}

Output:
[1,2,3,4,5,]
[]
[]

46 Chapter 3. Linked List

This solution creates a preHead node with a value of 2023 (an arbitrary value larger
than 50) and sets its next pointer to point to the original head of the linked list.
The purpose of this preHead node is to serve as the dummy or sentinel node at the
beginning of the linked list. Having a preHead node simplifies the code because it
eliminates the need to handle the special case of removing nodes from the beginning
of the list separately.
The remaining code is the same.

Complexity

• Runtime: O(N), where N is the number of nodes.

• Memory: O(1).

Attention!

Depending on your real situation, in practice, you might need to deallocate memory
for the removed nodes; especially when they were allocated by the new operator.

ListNode* removeElements(ListNode* head, int val) {

ListNode preHead(2022, head);
ListNode* pre = &preHead;
while (pre->next) {
if (pre->next->val == val) {
ListNode* node = pre->next;
pre->next = node->next;
delete node;
} else {
pre = pre->next;
}
}
return preHead.next;
}

3.2. Remove Linked List Elements 47

3.2.4 Key takeaway

• In some linked list problems where head needs to be treated as a special case,
you can create a previous dummy node for it to adapt the general algorithm.
• Be careful with memory leak when removing nodes of the linked list containing
pointers.

3.2.5 Exercise

• Delete Node in a Linked List2 .

3.3 Intersection of Two Linked Lists

3.3.1 Problem statement

1
You are provided with the starting nodes of two singly linked lists, headA and headB.
Your task is to find the node where these two lists intersect. If there is no point of
intersection, return null.
For example, the following two linked lists begin to intersect at node c1:

Note that the linked lists do not have any cycles, and you must ensure that the original
structure of the linked lists remains unchanged after solving this problem.
2 https://leetcode.com/problems/delete-node-in-a-linked-list/
1 https://leetcode.com/problems/intersection-of-two-linked-lists/

48 Chapter 3. Linked List

Example 1

Input: listA = [4,1,8,4,5], listB = [5,6,1,8,4,5].

Output: Intersected at '8'

Example 2

Input: listA = [1,9,1,2,4], listB = [3,2,4]

Output: Intersected at '2'

3.3. Intersection of Two Linked Lists 49

Example 3

Input: listA = [2,6,4], listB = [1,5]

Output: No intersection.

Constraints

• The number of nodes of listA is in the m.

• The number of nodes of listB is in the n.
• 1 <= m, n <= 3 * 10^4.
• 1 <= Node.val <= 10^5.

Follow up

• Could you write a solution that runs in O(m + n) time and use only O(1) mem-
ory?

3.3.2 Solution 1: Store the nodes

You can store all nodes of listA then iterate listB to determine which node is the
intersection. If none is found, the two lists have no intersection.

50 Chapter 3. Linked List

Example 1

• Store all nodes of listA = [4,1,8,4,5] in a map.

• Iterate listB and found node '8' was stored.
• Return '8'.

Code

#include <iostream>
#include <unordered_map>
using namespace std;
struct ListNode {
int val;
ListNode *next;
ListNode(int x) : val(x), next(NULL) {}
};

ListNode *getIntersectionNode(ListNode *headA, ListNode *headB) {

unordered_map<ListNode*, bool> m;
ListNode *node = headA;
while (node != nullptr) {
m[node] = true;
node = node->next;
}
node = headB;
while (node != nullptr) {
if (m.find(node) != m.end()) {
return node;
}
node = node->next;
}
return nullptr;
}
int main() {
{ // Example 1
ListNode five(5);
ListNode four(4);
four.next = &five;
ListNode eight(8);
eight.next = &four;
(continues on next page)

3.3. Intersection of Two Linked Lists 51

 (continued from previous page)

ListNode one1(1);
one1.next = &eight;
ListNode four1(4);
four1.next = &one1;

ListNode one2(1);
one2.next = &eight;
ListNode six2(6);
six2.next = &one2;
ListNode five2(5);
five2.next = &six2;
cout << (getIntersectionNode(&four1, &five2) == &eight) << endl;
}
{ // Example 2
ListNode four(4);
ListNode two(2);
two.next = &four;

ListNode one12(1);
one12.next = &two;
ListNode nine1(9);
nine1.next = &one12;
ListNode one11(1);
one11.next = &nine1;

ListNode three2(3);
three2.next = &two;
cout << (getIntersectionNode(&one11, &three2) == &two) << endl;
}
{ // Example 3
ListNode four(4);
ListNode six(6);
six.next = &four;
ListNode two(2);
two.next = &six;

ListNode five(5);
ListNode one(1);
one.next = &five;
(continues on next page)

52 Chapter 3. Linked List

 (continued from previous page)
cout << (getIntersectionNode(&two, &one) == nullptr) << endl;
}
}

Output:
1
1
1

This code uses an unordered map to store the nodes of headA while traversing it.
Then, it traverses headB and checks if each node in headB exists in the map of nodes
from headA. If a common node is found, it returns that node as the intersection point;
otherwise, it returns nullptr to indicate no intersection.

Complexity

• Runtime: O(m + n), where m, n are the number of nodes of listA and listB.
• Extra space: O(m).

3.3.3 Solution 2: Reiterating the two lists at the same time

If the two lists do not share the same tail, they have no intersection. Otherwise, they
must intersect at some node.
After iterating to find the tail node, you know the length of the two lists. That infor-
mation gives you a hint of how to reiterate to find the intersection node.

Example 1

• After iterating listA = [4,1,8,4,5], you find the tail node is '5' and listA.
length = 5.
• After iterating listB = [5,6,1,8,4,5], you find the tail node is the last '5' and
listB.length = 6.
• The two lists share the same tail. They must intersect at some node.
• To find that intersection node, you have to reiterate the two lists.

3.3. Intersection of Two Linked Lists 53

 • Since listB.length = 6 > 5 = listA.length, you can start iterating listB first
until the number of its remaining nodes is the same as listA. In this case, it is
the node '6' of listB.
• Now you can iterate them at the same time to find which node is shared.
• Found and return the intersection node '8'.

Code

#include <iostream>
#include <unordered_map>
using namespace std;
struct ListNode {
int val;
ListNode *next;
ListNode(int x) : val(x), next(NULL) {}
};
ListNode *getIntersectionNode(ListNode *headA, ListNode *headB)
{
int lengthA = 0;
ListNode *nodeA = headA;
while (nodeA->next != nullptr) {
lengthA++;
nodeA = nodeA->next;
}
int lengthB = 0;
ListNode *nodeB = headB;
while (nodeB->next != nullptr) {
lengthB++;
nodeB = nodeB->next;
}
// not the same tail -> no intersection
if (nodeA != nodeB) {
return nullptr;
}
nodeA = headA;
nodeB = headB;
// find the nodeA in listA and nodeB in listB
// that make two lists have the same length
while (lengthA > lengthB) {
nodeA = nodeA->next;
(continues on next page)

54 Chapter 3. Linked List

 (continued from previous page)
lengthA--;
}
while (lengthB > lengthA) {
nodeB = nodeB->next;
lengthB--;
}
// find the intersection
while (nodeA != nodeB) {
nodeA = nodeA->next;
nodeB = nodeB->next;
}
return nodeA;
}
int main() {
{ // Example 1
ListNode five(5);
ListNode four(4);
four.next = &five;
ListNode eight(8);
eight.next = &four;

ListNode one1(1);
one1.next = &eight;
ListNode four1(4);
four1.next = &one1;

ListNode one2(1);
one2.next = &eight;
ListNode six2(6);
six2.next = &one2;
ListNode five2(5);
five2.next = &six2;
cout << (getIntersectionNode(&four1, &five2) == &eight) << endl;
}
{ // Example 2
ListNode four(4);
ListNode two(2);
two.next = &four;

ListNode one12(1);
(continues on next page)

3.3. Intersection of Two Linked Lists 55

 (continued from previous page)
one12.next = &two;
ListNode nine1(9);
nine1.next = &one12;
ListNode one11(1);
one11.next = &nine1;

ListNode three2(3);
three2.next = &two;
cout << (getIntersectionNode(&one11, &three2) == &two) << endl;
}
{ // Example 3
ListNode four(4);
ListNode six(6);
six.next = &four;
ListNode two(2);
two.next = &six;

ListNode five(5);
ListNode one(1);
one.next = &five;
cout << (getIntersectionNode(&two, &one) == nullptr) << endl;
}
}

Output:
1
1
1

This improved solution finds the intersection of two linked lists by first determining
their lengths and adjusting the pointers so that they start from the same relative
position to the intersection point. Then, it iterates through both linked lists until it
finds the common intersection node.

56 Chapter 3. Linked List

Complexity

• Runtime: O(m + n), where m, n are the number of nodes of listA and listB.
• Extra space: O(1).

3.3.4 Implementation tip

• The technique used in Solution 2 is known as the Two-pointer technique since

you use two pointers to iterate the list at the same time.

3.3.5 Exercise

• Minimum Index Sum of Two Lists2 .

3.4 Swap Nodes in Pairs

3.4.1 Problem statement

1
You are provided with a linked list. Your goal is to exchange every two adjacent
nodes in the list and then return the head of the modified list.
You must solve this problem without altering the values within the nodes; you should
only modify the arrangement of the nodes themselves.

2 https://leetcode.com/problems/minimum-index-sum-of-two-lists/
1 https://leetcode.com/problems/swap-nodes-in-pairs/

3.4. Swap Nodes in Pairs 57

Example 1

Input: head = [1,2,3,4]

Output: [2,1,4,3]

Example 2

Input: head = []
Output: []

Example 3

Input: head = [1]

Output: [1]

Constraints

• The number of nodes in the list is in the range [0, 100].

• 0 <= Node.val <= 100.

58 Chapter 3. Linked List

3.4.2 Solution

Draw a picture of the swapping to identify the correct order of the update.

Denote (cur, next) the pair of nodes you want to swap and prev be the previous
node that links to cur. Here are the steps you need to perform for the swapping.
1. Update the links between nodes.
2. Go to the next pair.

Code

#include <iostream>
struct ListNode {
int val;
ListNode *next;
ListNode() : val(0), next(nullptr) {}
ListNode(int x) : val(x), next(nullptr) {}
ListNode(int x, ListNode *next) : val(x), next(next) {}
};
ListNode* swapPairs(ListNode* head) {
// the list does not have enough nodes to swap
(continues on next page)

3.4. Swap Nodes in Pairs 59

 (continued from previous page)
if (head == nullptr || head->next == nullptr) {
return head;
}
ListNode* preNode = nullptr;
ListNode* curNode = head;
ListNode* nextNode = head->next;
head = nextNode;
while (curNode != nullptr && nextNode != nullptr) {

// swap curNode and nextNode

curNode->next = nextNode->next;
nextNode->next = curNode;

// update links/pointers after swap

if (preNode) {
preNode->next = nextNode;
}

// update nodes for next step

preNode = curNode;
curNode = curNode->next;
if (curNode) {
nextNode = curNode->next;
}
}
return head;
}
void print(const ListNode* head) {
ListNode* node = head;
std::cout << "[";
while (node != nullptr) {
std::cout << node->val << ",";
node = node->next;
}
std::cout << "]" << std::endl;
}
int main() {
ListNode four(4);
ListNode three(3, &four);
ListNode two(2, &three);
(continues on next page)

60 Chapter 3. Linked List

 (continued from previous page)
ListNode one(1, &two);
print(swapPairs(&one));
ListNode five(5);
print(swapPairs(nullptr));
print(swapPairs(&five));
}

Output:
[2,1,4,3,]
[]
[5,]

Complexity

• Runtime: O(N), where N is the number of nodes.

• Extra space: O(1).

3.4.3 Conclusion

This solution swaps pairs of nodes in a linked list by adjusting the pointers accord-
ingly.
It initializes pointers to the current node (curNode), its next node (nextNode), and the
previous node (preNode). Then, it iterates through the list, swapping pairs of nodes
by adjusting their next pointers and updating the preNode pointer.
This approach efficiently swaps adjacent nodes in the list without requiring additional
space, effectively transforming the list by rearranging pointers.

3.4.4 Exercise

• Swapping Nodes in a Linked List2 .

2 https://leetcode.com/problems/swapping-nodes-in-a-linked-list/

3.4. Swap Nodes in Pairs 61

3.5 Add Two Numbers

3.5.1 Problem statement

1
You have two linked lists that represent non-negative integers. The digits of these
numbers are stored in reverse order, with each node containing a single digit.
Your task is to add the two numbers represented by these linked lists and return the
result as a new linked list.
You can assume that the two numbers don’t have leading zeros, except for the number
0 itself.

Example 1

Input: l1 = [2,4,3], l2 = [5,6,4]

Output: [7,0,8]
Explanation: 342 + 465 = 807.

Example 2

Input: l1 = [0], l2 = [0]

Output: [0]

Example 3

Input: l1 = [9,9,9,9,9,9,9], l2 = [9,9,9,9]

Output: [8,9,9,9,0,0,0,1]

1 https://leetcode.com/problems/add-two-numbers/

62 Chapter 3. Linked List

Constraints

• The number of nodes in each linked list is in the range [1, 100].
• 0 <= Node.val <= 9.
• It is guaranteed that the list represents a number that does not have leading
zeros.

3.5.2 Solution: Addition With Remember

Perform the school addition calculation and store the result in one of the lists.
Without loss of generality, let us store the result in l1. Then you might need to
extend it when l2 is longer than l1 and when the result requires one additional node
(Example 3).

Code

#include <iostream>
struct ListNode {
int val;
ListNode *next;
ListNode() : val(0), next(nullptr) {}
ListNode(int x) : val(x), next(nullptr) {}
ListNode(int x, ListNode *next) : val(x), next(next) {}
};

ListNode* addTwoNumbers(ListNode* l1, ListNode* l2) {

// dummy node to hook the head of the list
ListNode prehead;

// let's use l1's nodes to store result

ListNode* node = l1;
prehead.next = node;
int sum = 0;
while (node) {

// perform the addition

if (l1) {
sum += l1->val;
(continues on next page)

3.5. Add Two Numbers 63

 (continued from previous page)
l1 = l1->next;
}
if (l2) {
sum += l2->val;
l2 = l2->next;
}
node->val = sum % 10;

// keep track the carry

sum /= 10;
if (!l1) { // l1 ends
if (l2) { // l1 is shorter than l2
node->next = l2;
} else if (sum == 1) {
// both l1 and l2 end but the remember is not zero
ListNode* newNode = new ListNode(sum);
node->next = newNode;
}
}
node = node->next;
}
return prehead.next;
}
void printResult(ListNode* l) {
std::cout << "[";
while (l) {
std::cout << l->val << ",";
l = l->next;
}
std::cout << "]\n";
}
int main() {
{
ListNode three(3);
ListNode four1(4, &three);
ListNode two(2, &four1);
ListNode four2(4);
ListNode six(6, &four2);
ListNode five(5, &six);
printResult(addTwoNumbers(&two, &five));
(continues on next page)

64 Chapter 3. Linked List

 (continued from previous page)
}
{
ListNode zero1(0);
ListNode zero2(0);
printResult(addTwoNumbers(&zero1, &zero2));
}
{
ListNode nine0(9);
ListNode nine1(9, &nine0);
ListNode nine2(9, &nine1);
ListNode nine3(9, &nine2);
ListNode nine4(9, &nine3);
ListNode nine5(9, &nine4);
ListNode nine6(9, &nine5);
ListNode nine7(9);
ListNode nine8(9, &nine7);
ListNode nine9(9, &nine8);
ListNode nine10(9, &nine9);
printResult(addTwoNumbers(&nine6, &nine10));
}
}

Output:
[7,0,8,]
[0,]
[8,9,9,9,0,0,0,1,]

Complexity

• Runtime: O(N), where N = max(l1.length, l2.length).

• Extra space: O(1).

3.5. Add Two Numbers 65

3.5.3 Conclusion

This solution leverages a dummy node (prehead) to simplify the handling of edge
cases and to hook the head of the resulting list.
By iterating through both input lists simultaneously and performing addition digit by
digit while keeping track of carry, it efficiently computes the sum without the need
for additional checks for the head of the resulting list.
This approach streamlines the addition process, resulting in a concise and straightfor-
ward implementation.

3.5.4 Exercise

• Double a Number Represented as a Linked List2 .

2 https://leetcode.com/problems/double-a-number-represented-as-a-linked-list/

66 Chapter 3. Linked List

 CHAPTER

FOUR

HASH TABLE

This chapter is about the C++ Standard Template Library’s std::unordered_map and
how it can help your programming.
With hash-based data structures, you can store and retrieve information quickly,
like a well-organized library. Hash tables allow you to efficiently manage data
by inserting, locating, and removing elements, even from large datasets. C++’s
std::unordered_map makes it easy to use hash tables without manual implementa-
tion.
What this chapter covers:
1. Exploring std::unordered_map: Dive into the C++ Standard Template Li-
brary’s std::unordered_map container, learning how to use it effectively for
mapping keys to values.
2. Problem-Solving with Hash Tables: Learn strategies for solving many prob-
lems using hash tables, including frequency counting, anagram detection, and
more.

4.1 Roman to Integer

4.1.1 Problem statement

1
Roman numerals utilize seven symbols: I, V, X, L, C, D, and M to represent numbers.

Symbol Value
I 1
(continues on next page)
1 https://leetcode.com/problems/roman-to-integer/

67
 (continued from previous page)
V 5
X 10
L 50
C 100
D 500
M 1000

For example, 2 is denoted as II, which is essentially two ones added together. Simi-
larly, 12 is represented as XII, indicating X + II. The number 27 is written as XXVII,
which stands for XX + V + II.
Roman numerals are generally written from the largest value to the smallest value,
moving from left to right. However, there are exceptions to this pattern. For instance,
the numeral for 4 is IV instead of IIII, where I is placed before V to subtract 1 from
5. Similarly, 9 is IX, representing the subtraction of 1 from 10. There are six such
subtraction instances:
• I before V (5) or X (10) forms 4 and 9.
• X before L (50) or C (100) forms 40 and 90.
• C before D (500) or M (1000) forms 400 and 900.
Your task is to convert a given Roman numeral into its equivalent integer value.

Example 1

Input: s = "III"
Output: 3
Explanation: III = 3.

Example 2

Input: s = "LVIII"
Output: 58
Explanation: L = 50, V= 5, III = 3.

68 Chapter 4. Hash Table

Example 3

Input: s = "MCMXCIV"
Output: 1994
Explanation: M = 1000, CM = 900, XC = 90 and IV = 4.

Constraints

• 1 <= s.length <= 15.

• s contains only the characters 'I', 'V', 'X', 'L', 'C', 'D', 'M'.
• It is guaranteed that s is a valid Roman numeral in the range [1, 3999].

4.1.2 Solution: Mapping and summing the values

To treat the subtraction cases easier you can iterate the string s backward.

Code

#include <iostream>
#include <unordered_map>
using namespace std;
const unordered_map<char, int> value = {
{'I', 1}, {'V', 5},
{'X', 10}, {'L', 50},
{'C', 100}, {'D', 500},
{'M', 1000}
};
int romanToInt(const string& s) {

// starting from the end character of the string s

int i = s.length() - 1;
int result = value.at(s[i--]);
while (i >= 0) {
// In cases of subtraction
if (value.at(s[i]) < value.at(s[i+1])) {
result -= value.at(s[i--]);
} else {
(continues on next page)

4.1. Roman to Integer 69

 (continued from previous page)
result += value.at(s[i--]);
}
}
return result;
}
int main() {
cout << romanToInt("III") << endl;
cout << romanToInt("LVIII") << endl;
cout << romanToInt("MCMXCIV") << endl;
}

Output:
3
58
1994

Complexity

• Runtime: O(N) where N = s.length.

• Extra space: O(1).

4.1.3 Conclusion

This problem can be solved using a map to store the values of each Roman numeral
character. This solution iterates through the string from right to left, accumulating
the integer value based on the corresponding Roman numeral characters.
By comparing the current character’s value with the previous one, the solution han-
dles cases of subtraction (e.g., IV, IX, etc.) by subtracting the value if it’s smaller and
adding it otherwise.

70 Chapter 4. Hash Table

4.1.4 Exercise

• Integer to Roman2 .

4.2 Maximum Erasure Value

4.2.1 Problem statement

1
You have an array of positive integers called nums, and you wish to remove a subarray
from it that consists of distinct elements. The score you achieve by removing this
subarray is the sum of its elements.
Your goal is to determine the highest possible score attainable by erasing exactly one
subarray from the provided array.
A subarray, denoted as b, is considered part of another array, a, if it appears consec-
utively within a, i.e., if it is equivalent to a[l], a[l+1], ..., a[r] for some indices
(l, r).

Example 1

Input: nums = [4,2,4,5,6]

Output: 17
Explanation: The optimal subarray here is [2,4,5,6].

Example 2

Input: nums = [5,2,1,2,5,2,1,2,5]

Output: 8
Explanation: The optimal subarray here is [5,2,1] or [1,2,5].

2 https://leetcode.com/problems/integer-to-roman/
1 https://leetcode.com/problems/maximum-erasure-value/

4.2. Maximum Erasure Value 71

Constraints

• 1 <= nums.length <= 10^5.

• 1 <= nums[i] <= 10^4.

4.2.2 Solution: Store the position of the visited elements

You can use a map to store the position of the elements of nums. Then when iterating
nums you can identify if an element has been visited before. That helps you to decide
if a subarray contains unique elements.

Code

#include <iostream>
#include <unordered_map>
#include <vector>
using namespace std;
int maximumUniqueSubarray(const vector<int>& nums) {
// sum stores the running sum of nums
// i.e., sum[i] = nums[0] + ... + nums[i]
vector<int> sum(nums.size(), 0);
sum[0] = nums.at(0);

// store the maximum sum of the maximum subarray

int maxSum = sum.at(0);

// position[a] keeps track of latest index i whose nums[i] = a

unordered_map<int, int> position;
position[nums.at(0)] = 0;

// the starting index of the current subarray

int start = -1;
for (int i = 1; i < nums.size(); i++) {
sum[i] = sum.at(i - 1) + nums.at(i);

// check if the current subarray's elements are still distinct

auto it = position.find(nums.at(i));

// found the value nums[i] in the map position

(continues on next page)

72 Chapter 4. Hash Table

 (continued from previous page)
if (it != position.end()) {
// a new subarray now starts from i
start = max(start, it->second);

// clear the latest index of the value nums[i]

// by updating it to i
it->second = i;
} else {
// current subarray continues with nums[i]
position.insert({nums.at(i), i});
}
if (start == -1) {
// still on the first subarray nums[0]..nums[i]
maxSum = sum.at(i);
} else {
// update maxSum with the sum of subarray nums[start]..
˓→nums[i]

maxSum = max(maxSum, sum.at(i) - sum.at(start));

}
}
return maxSum;
}
int main() {
vector<int> nums{4,2,4,5,6};
cout << maximumUniqueSubarray(nums) << endl;
nums = {5,2,1,2,5,2,1,2,5};
cout << maximumUniqueSubarray(nums) << endl;
}

Output:
17
8

4.2. Maximum Erasure Value 73

Complexity

• Runtime: O(N), where N = nums.length.

• Extra space: O(N).

4.2.3 Conclusion

This solution computes the maximum sum of a subarray containing unique elements.
It uses a sliding window approach to maintain a running sum of the elements en-
countered so far and a hashmap to keep track of the positions of previously seen
elements. By updating the starting index of the window when a repeated element is
encountered, it ensures that the current subarray contains only unique elements.
This approach optimizes the computation of the maximum sum by handling the slid-
ing window and updating the sum accordingly, resulting in an overall efficient solu-
tion.

4.3 Find and Replace Pattern

4.3.1 Problem statement

1
You are provided with a list of strings named words and a string named pattern.
Your task is to find the strings from words that match the given pattern. The order in
which you return the answers does not matter.
A word is considered to match the pattern if there is a mapping p of the letters such
that, when each letter x in the pattern is replaced with p(x), the word is formed.
Keep in mind that a permutation of letters is a one-to-one correspondence from letters
to letters, where each letter is mapped to a distinct letter, and no two letters are
mapped to the same letter.
1 https://leetcode.com/problems/find-and-replace-pattern/

74 Chapter 4. Hash Table

Example 1

Input: words = ["abc","deq","mee","aqq","dkd","ccc"], pattern = "abb"

Output: ["mee","aqq"]
Explanation: "mee" matches the pattern because there is a permutation {a␣
˓→-> m, b -> e, ...}.

"ccc" does not match the pattern because {a -> c, b -> c, ...} is not a␣
˓→permutation, since a and b map to the same letter.

Example 2

Input: words = ["a","b","c"], pattern = "a"

Output: ["a","b","c"]

Constraints

• 1 <= pattern.length <= 20.

• 1 <= words.length <= 50.
• words[i].length == pattern.length.
• pattern and words[i] are lowercase English letters.

4.3.2 Solution: Construct the bijection and check the condition

Code

#include <iostream>
#include <vector>
#include <unordered_map>
using namespace std;
vector<string> findAndReplacePattern(const vector<string>& words, const␣
˓→string& pattern) {

vector<string> result;
// need two maps for the bijection
unordered_map<char,char> w_to_p, p_to_w;
int i;
for (auto& w : words) {
(continues on next page)

4.3. Find and Replace Pattern 75

 (continued from previous page)
w_to_p.clear();
p_to_w.clear();
i = 0;
while (i < w.length()) {
if (w_to_p.find(w[i]) != w_to_p.end()) {
// w[i] was mapped to some letter x
// but x != pattern[i]
if (w_to_p[w[i]] != pattern[i]) {
break;
}
} else {
if (p_to_w.find(pattern[i]) != p_to_w.end()) {
// w[i] was not mapped to any letter yet
// but pattern[i] was already mapped to some letter
break;
}
// build the bijection w[i] <-> pattern[i]
w_to_p[w[i]] = pattern[i];
p_to_w[pattern[i]] = w[i];
}
i++;
}
if (i == w.length()) {
result.push_back(w);
}
}
return result;
}
void printResult(const vector<string>& result) {
cout << "[";
for (auto& s : result) {
cout << s << ",";
}
cout << "]\n";
}
int main() {
vector<string> words{"abc","deq","mee","aqq","dkd","ccc"};
auto result = findAndReplacePattern(words, "abb");
printResult(result);
words = {"a", "b", "c"};
(continues on next page)

76 Chapter 4. Hash Table

 (continued from previous page)
result = findAndReplacePattern(words, "abb");
printResult(result);
}

Output:
[mee,aqq,]
[a,b,c,]

Complexity

• Runtime: O(NL), where N = words.length and L = pattern.length.

• Extra space: O(1). The maps w_to_p and p_to_w just map between 26 lowercase
English letters.

4.3.3 Conclusion

This solution efficiently finds and returns words from a vector of strings that match a
given pattern in terms of character bijection. It uses two unordered maps to establish
and maintain the bijection while iterating through the characters of the words and
the pattern.

4.3. Find and Replace Pattern 77

78 Chapter 4. Hash Table
 CHAPTER

FIVE

STRING

In this chapter, we’ll learn about the importance of strings in programming. Strings
help us work with text and are essential for many tasks, from processing data to cre-
ating better communication between programs and people. By understanding strings,
you’ll be better equipped to solve problems and make things easier for users.
What this chapter covers:
1. Understanding Strings: Lay the groundwork by comprehending the nature of
strings, character encoding schemes, and the basics of representing and storing
textual data.
2. String Manipulation: Explore the art of string manipulation, covering opera-
tions like concatenation, slicing, reversing, and converting cases.
3. String Searching and Pattern Matching: Delve into strategies for finding sub-
strings, detecting patterns, and performing advanced search operations within
strings.
4. Anagrams and Palindromes: Tackle challenges related to anagrams and palin-
dromes, honing your ability to discern permutations and symmetric constructs.
5. Problem-Solving with Strings: Learn how to approach coding problems that
involve string manipulation, from simple tasks to intricate algorithms.

79
5.1 Valid Anagram

5.1.1 Problem statement

1
You are given two strings, s and t. Your task is to determine whether t is an anagram
of s. If t is an anagram of s, return true; otherwise, return false.
An anagram is a term that describes a word or phrase formed by rearranging the
letters of another word or phrase, typically using all the original letters exactly once.

Example 1

Input: s = "anagram", t = "nagaram"

Output: true

Example 2

Input: s = "rat", t = "car"

Output: false

Constraints

• 1 <= s.length, t.length <= 5 * 10^4.

• s and t consist of lowercase English letters.

Follow up

• What if the inputs contain Unicode characters? How would you adapt your
solution to such a case?
1 https://leetcode.com/problems/valid-anagram/

80 Chapter 5. String
5.1.2 Solution 1: Rearrange both s and t into a sorted string

Code

#include <iostream>
#include <algorithm>
using namespace std;
bool isAnagram(string& s, string& t) {
// anagrams must have the same length
if (s.length() != t.length()) {
return false;
}
sort(s.begin(), s.end());
sort(t.begin(), t.end());
return s == t;
}
int main() {
cout << isAnagram("anagram", "nagaram") << endl;
cout << isAnagram("rat", "car") << endl;
}

Output:
1
0

This solution determines if two strings are anagrams by comparing their sorted ver-
sions. If the sorted versions are equal, the original strings are anagrams, and the
function returns true. Otherwise, it returns false.

5.1. Valid Anagram 81

Complexity

• Runtime: O(NlogN), where N = s.length.

• Extra space: O(1).

5.1.3 Solution 2: Count the appearances of each letter

Code

#include <iostream>
using namespace std;
bool isAnagram(const string& s, const string& t) {
if (s.length() != t.length()) {
return false;
}
// s and t consist of only lowercase English letters
// you can encode 0: 'a', 1: 'b', .., 25: 'z'.
int alphabet[26];
for (int i = 0; i < 26; i++) {
alphabet[i] = 0;
}
// count the frequency of each letter in s
for (auto& c : s) {
alphabet[c - 'a']++;
}
for (auto& c : t) {
alphabet[c - 'a']--;
// if s and t have the same length but are not anagrams,
// there must be some letter in t having higher frequency than s
if (alphabet[c - 'a'] < 0) {
return false;
}
}
return true;
}
int main() {
cout << isAnagram("anagram", "nagaram") << endl;
cout << isAnagram("rat", "car") << endl;
}

82 Chapter 5. String
Output:
1
0

This solution efficiently determines if two strings are anagrams by counting the fre-
quency of each character in both strings using an array. If the character frequencies
match for both strings, they are anagrams.

Complexity

• Runtime: O(N), where N = s.length.

• Extra space: O(1).

5.1.4 Solution 3: If the inputs contain Unicode characters

Replace the array alphabet in Solution 2 with a map.

Code

#include <iostream>
#include <unordered_map>
using namespace std;
bool isAnagram(const string& s, const string& t) {
if (s.length() != t.length()) {
return false;
}
// this alphabet can store all UTF-8 characters
unordered_map<char, int> alphabet;
for (auto& c : s) {
alphabet[c]++;
}
for (auto& c : t) {
alphabet[c]--;
if (alphabet[c] < 0) {
return false;
}
}
return true;
(continues on next page)

5.1. Valid Anagram 83

 (continued from previous page)
}
int main() {
cout << isAnagram("anagram", "nagaram") << endl;
cout << isAnagram("rat", "car") << endl;
}

Output:
1
0

Complexity

• Runtime: O(N), where N = s.length.

• Extra space: O(c) where c represents the number of unique characters present
in both strings s and t.

5.1.5 Key Takeaway

Instead of relying on a fixed-size array like the ASCII-based solutions, Solution 3 uses
an unordered_map to store character frequencies. Each character is used as a key in
the map, and the count of occurrences is stored as the associated value.
Unicode characters values are not restricted to a specific range. The unordered_map
approach accommodates this variability by allowing any character to be a key.

5.1.6 Exercise

• Find Resultant Array After Removing Anagrams2 .

2 https://leetcode.com/problems/find-resultant-array-after-removing-anagrams/

84 Chapter 5. String
5.2 Detect Capital

5.2.1 Problem statement

1
The task is to determine if the usage of capital letters in a given string, word, is
correct according to the following rules:
1. All letters in the word are capital, like “USA”.
2. All letters in the word are not capital, like “leetcode”.
3. Only the first letter in the word is capital, like “Google”.
If the capitalization in the given word adheres to these rules, the function should
return true; otherwise, it should return false.

Example 1

Input: word = "USA"

Output: true

Example 2

Input: word = "FlaG"

Output: false

Constraints

• 1 <= word.length <= 100,

• word consists of lowercase and uppercase English letters.
1 https://leetcode.com/problems/detect-capital/

5.2. Detect Capital 85

5.2.2 Solution

Only when the first two characters of the word are uppercase, the rest must be the
same. Otherwise, the rest is always lowercase.

Code

#include <string>
#include <iostream>
using namespace std;
//! @return true if (c is lowercase and isLower is true)
//! or (c is uppercase and isLower is false).
//! false, otherwise.
bool isValidCase(const char& c, const bool isLower) {
if (isLower) {
return 'a' <= c && c <= 'z';
}
return 'A' <= c && c <= 'Z';
}
bool detectCapitalUse(const string& word) {
if (word.length() == 1) {
return true;
}
bool isLower = true;

// if the first two characters are uppercase,

// the rest must be uppercase, too.
if (isValidCase(word[0], false) && isValidCase(word[1], false)) {
isLower = false;
}
for (int i = 1; i < word.length(); i++) {
if (!isValidCase(word[i], isLower)) {
return false;
}
}
return true;
}
int main() {
cout << detectCapitalUse("USA") << endl;
cout << detectCapitalUse("FlaG") << endl;
cout << detectCapitalUse("leetcode") << endl;
(continues on next page)

86 Chapter 5. String
 (continued from previous page)
cout << detectCapitalUse("Google") << endl;
}

Output:
1
0
1
1

Complexity

• Runtime: O(N), where N = word.length.

• Extra space: O(1).

5.2.3 Conclusion

This solution efficiently checks whether a given word follows one of the specified
capitalization rules by iterating through the characters of the word and using the
isValidCase function to validate each character’s capitalization based on the current
capitalization type (isLower). If no violations are found, the word is considered valid,
and the function returns true.

5.2.4 Exercise

• Capitalize the Title2 .

5.3 Unique Morse Code Words

5.3.1 Problem statement

1
The problem involves the International Morse Code, which defines a standard way
to encode letters with dots and dashes. Each English letter corresponds to a specific
2 https://leetcode.com/problems/capitalize-the-title/
1 https://leetcode.com/problems/unique-morse-code-words/

5.3. Unique Morse Code Words 87

sequence in Morse Code, and a full table mapping each letter is provided.
For instance, 'a' is encoded as ".-", 'b' as "-...", and so on.
The full table for the 26 letters of the English alphabet is given below:

[".-", "-...", "-.-.", "-..", ".", "..-.", "--.",

"....", "..", ".---", "-.-", ".-..", "--", "-.",
"---", ".--.", "--.-", ".-.", "...", "-", "..-",
"...-", ".--", "-..-", "-.--", "--.."]

You are given an array of strings named words, where each word can be represented
as a concatenation of the Morse code for each of its letters. For example, the word
"cab" can be represented as "-.-..--...", which is the concatenation of "-.-.", ".
-", and "-...". This concatenated Morse code representation is referred to as the
“transformation” of a word.
Your task is to count the number of different transformations that can be obtained
from all the words in the given array.

Example 1

Input: words = ["gin","zen","gig","msg"]

Output: 2
Explanation: The transformation of each word is:
"gin" -> "--...-."
"zen" -> "--...-."
"gig" -> "--...--."
"msg" -> "--...--."
There are 2 different transformations: "--...-." and "--...--.".

Example 2

Input: words = ["a"]

Output: 1

88 Chapter 5. String
Constraints

• 1 <= words.length <= 100.

• 1 <= words[i].length <= 12.
• words[i] consists of lowercase English letters.

5.3.2 Solution: Store the transformations in a set

Code

#include <iostream>
#include <vector>
#include <unordered_set>
using namespace std;
const vector<string> morse{
".-", "-...", "-.-.", "-..", ".", "..-.", "--.",
"....", "..", ".---", "-.-", ".-..", "--", "-.",
"---", ".--.", "--.-", ".-.", "...", "-", "..-",
"...-", ".--", "-..-", "-.--", "--.."
};

int uniqueMorseRepresentations(const vector<string>& words) {

unordered_set<string> transformations;
for (auto& w : words) {
string s{""};
for (auto& c : w) {
// concatnate the letter c's Morse code
s += morse[c - 'a'];
}
// only insert the transformation s to the set
// if the set did not consist s yet.
transformations.insert(s);
}
return transformations.size();
}
int main() {
vector<string> words{"gin","zen","gig","msg"};
cout << uniqueMorseRepresentations(words) << endl;
words = {"a"};
(continues on next page)

5.3. Unique Morse Code Words 89

 (continued from previous page)
cout << uniqueMorseRepresentations(words) << endl;
}

Output:
2
1

Complexity

• Runtime: O(N*M), where N = words.length and M = words[i].length.

• Extra space: O(N).

5.3.3 Conclusion

This solution converts each word into Morse code based on a predefined mapping
and uses an unordered set to keep track of unique representations. By inserting each
representation into the set, it automatically filters out duplicates. The final result is
the size of the set, which represents the number of unique Morse code representations
among the input words.

5.4 Unique Email Addresses

5.4.1 Problem statement

1
Each valid email address is composed of a local name and a domain name, separated
by the '@' sign. The local name may contain lowercase letters, one or more '.'
characters, and a plus '+' sign. However, the rules for dots and the plus sign do not
apply to the domain name.
For example, in the email "[email protected]", "alice" is the local name, and
"leetcode.com" is the domain name.
If you insert periods '.' between certain characters in the local name, the email will
still be forwarded to the same address without the dots in the local name. This rule
does not apply to the domain name.
1 https://leetcode.com/problems/unique-email-addresses/

90 Chapter 5. String
For example, "[email protected]" and "[email protected]" both forward to
the same email address.
If you include a plus '+' sign in the local name, everything after the first plus sign
is ignored, allowing for email filtering. This rule also does not apply to the domain
name.
For example, "[email protected]" will be forwarded to "[email protected]".
It is possible to use both of these rules at the same time.
Given an array of strings emails, where each element is an email address to which
an email is sent, your task is to determine the number of different addresses that will
actually receive the emails after applying the rules described above.

Example 1

Input: emails = ["[email protected]","test.e.mail+bob.

˓→[email protected]","[email protected]"]

Output: 2
Explanation: "[email protected]" and "[email protected]"␣
˓→actually receive mails.

Example 2

Input: emails = ["[email protected]","[email protected]","[email protected]"]

Output: 3

Constraints

• 1 <= emails.length <= 100.

• 1 <= emails[i].length <= 100.
• emails[i] consist of lowercase English letters, '+', '.' and '@'.
• Each emails[i] contains exactly one '@' character.
• All local and domain names are non-empty.
• Local names do not start with a '+' character.
• Domain names end with the ".com" suffix.

5.4. Unique Email Addresses 91

5.4.2 Solution 1: Removing the ignored characters

Do exactly the steps the problem describes:

1. Extract the local name.
2. Ignore all characters after '+' in it.
3. Ignore all '.' in it.
4. Combine the local name with the domain one to form the clean email address.

Code

#include<string>
#include<iostream>
#include<vector>
#include <unordered_set>
using namespace std;
int numUniqueEmails(const vector<string>& emails) {
unordered_set<string> s;
for (auto& e: emails) {
auto apos = e.find('@');

// extract the local name

string local = e.substr(0, apos);

// ignore all characters after '+'

local = local.substr(0, local.find('+'));
auto it = local.find('.');
while (it != string::npos) {
// remove each '.' found in local
local.erase(it, 1);
it = local.find('.');
}
// combine local name with domain one
s.insert(local + e.substr(apos));
}
return s.size();
}
int main() {
vector<string> emails{"[email protected]",
"[email protected]",
(continues on next page)

92 Chapter 5. String
 (continued from previous page)
"[email protected]"};
cout << numUniqueEmails(emails) << endl;
emails = {"[email protected]","[email protected]","[email protected]"};
cout << numUniqueEmails(emails) << endl;
emails = {"[email protected]","test.email.leet+alex@code.
˓→com"};

cout << numUniqueEmails(emails) << endl;

}

Output:
2
3
2

This solution parses a list of email addresses, normalizes each email address by re-
moving periods and ignoring characters after the plus sign in the local name, and then
counts the number of unique email addresses. The use of an unordered set ensures
that only unique email addresses are counted.

Complexity

• Runtime: O(N*M^2), where N = emails.length and M = max(emails[i].length).

Explanation: you loop over N emails. Then you might loop over the length of
each email, O(M), to remove the character '.'. The removal might cost O(M).
• Extra space: O(N*M) (the set of emails).

5.4. Unique Email Addresses 93

5.4.3 Solution 2: Building the clean email addresses from scratch

The runtime of removing characters in std::string is not constant. To avoid that

complexity you can build up the clean email addresses from scratch.

Code

#include<string>
#include<iostream>
#include<vector>
#include <unordered_set>
using namespace std;
int numUniqueEmails(const vector<string>& emails) {
unordered_set<string> s;
for (auto& e: emails) {
string address;
int i = 0;
// the local name ends here
while (e[i] != '@' && e[i] != '+') {
// ignore each '.' found
if (e[i++] == '.') {
continue;
}
// add valid characters to local name
address += e[i++];
}
// combine local name with domain one
address += e.substr(e.find('@', i));
s.insert(address);
}
return s.size();
}
int main() {
vector<string> emails{"[email protected]",
"[email protected]",
"[email protected]"};
cout << numUniqueEmails(emails) << endl;
emails = {"[email protected]","[email protected]","[email protected]"};
cout << numUniqueEmails(emails) << endl;
emails = {"[email protected]","test.email.leet+alex@code.
˓→com"};

(continues on next page)

94 Chapter 5. String
 (continued from previous page)
cout << numUniqueEmails(emails) << endl;
}

Output:
2
3
2

Complexity

• Runtime: O(N*M), where N = emails.length and M = max(emails[i].length).

• Extra space: O(N*M).

5.4.4 C++ Notes

• A string can be concatenated with a char and another string by + operator.

std::string address = "name";

address += '@'; // "name@"
address += "domain.com"; // "[email protected]"

• string::substr(pos = 0, count = npos)2 returns the substring of length count

starting from the position pos of the string string.

std::string address = "[email protected]";

cout << address.substr(address.find('.')); // ".com"
cout << address.substr(0, address.find('@')); // "name"

• string::find(char, pos=0)3 returns the position of the first char which ap-
pears in the string string starting from pos.
2 https://en.cppreference.com/w/cpp/string/basic_string/substr
3 https://en.cppreference.com/w/cpp/string/basic_string/find

5.4. Unique Email Addresses 95

5.4.5 High-performance C++

• Do not use std::set or std::map unless you want the keys to be in

order (sorted). Use unordered containers like std::unordered_set4 or
std::unordered_map5 instead. They use hashed keys for faster lookup.
• Do not blindly/lazily use string.find(something). If you know where to start
the search, use string.find(something, pos) with a specific pos.

5.5 Longest Substring Without Repeating Characters

5.5.1 Problem statement

1
Given a string s, your task is to determine the length of the longest substring within
s that does not contain any repeating characters.

Example 1

Input: s = "abcabcbb"
Output: 3
Explanation: The answer is "abc", with a length of 3.

Example 2

Input: s = "bbbbb"
Output: 1
Explanation: The answer is "b", with the length of 1.

4 https://en.cppreference.com/w/cpp/container/unordered_set
5 https://en.cppreference.com/w/cpp/container/unordered_map
1 https://leetcode.com/problems/longest-substring-without-repeating-characters/

96 Chapter 5. String
Example 3

Input: s = "pwwkew"
Output: 3
Explanation: The answer is "wke", with a length of 3.
Notice that the answer must be a substring, "pwke" is a subsequence and␣
˓→not a substring.

Constraints

• 0 <= s.length <= 5 * 10^4.

• s consists of English letters, digits, symbols and spaces.

5.5.2 Solution: Store the position of the visited characters

Whenever you meet a visited character s[i] == s[j] for some 0 <= i < j < s.length,
the substring "s[i]...s[j - 1]" might be valid, i.e., it consists of only nonrepeating
characters.
But in case you meet another visited character s[x] == s[y] where x < i < j < y,
the substring "s[x]...s[y - 1]" is not valid because it consists of repeated character
s[i] == s[j].
That shows the substring "s[i]...s[j - 1]" is not always a valid one. You might
need to find the right starting position start >= i for the valid substring "s[start].
..s[j - 1]".

Example 4

For the string s = "babba":

• When you visit the second letter 'b', the substring "ba" is a valid one.
• When you visit the third letter 'b', the substring of interest should be started
by the second letter 'b'. It gives you the substring "b".
• When you visit the second letter 'a', the substring "abb" is not a valid one since
'b' is repeated. To ensure no repetition, the starting position for this substring
should be the latter 'b', which leads to the valid substring "b".
• The final longest valid substring is "ba" with length 2.

5.5. Longest Substring Without Repeating Characters 97

Example 4 shows the starting position start for the substring of interest "s[i]...s[j
- 1]" should be:

this_start = max(previous_start, i).

Code

#include <iostream>
#include <unordered_map>
using namespace std;
int lengthOfLongestSubstring(const string& s) {
// keep track latest index of a character in s
unordered_map<char, int> position;

// posible maximum length of the longest substring

int maxLen = 0;

// starting index of current substring

int start = -1;
for (int i = 0; i < s.length(); i++) {
auto it = position.find(s.at(i));
// found this s[i] has appeared in the map
if (it != position.end()) {
// start a new substring from this index i
start = max(start, it->second);

// update latest found position of character s[i] to be i

it->second = i;
} else {

// keep track index of s[i] in the map

position.insert({s.at(i), i});
}
// update the maximum length
maxLen = max(maxLen, i - start);
}
return maxLen;
}
int main() {
cout << lengthOfLongestSubstring("abcabcbb") << endl;
(continues on next page)

98 Chapter 5. String
 (continued from previous page)
cout << lengthOfLongestSubstring("bbbbb") << endl;
cout << lengthOfLongestSubstring("pwwkew") << endl;
}

Output:
3
1
3

Complexity

• Runtime: O(N), where N = s.length.

• Extra space: O(N).

5.5.3 Conclusion

This solution utilizes a sliding window approach to track the starting index of the
current substring and an unordered map to store the position of the characters en-
countered so far. By updating the starting index when a repeating character is en-
countered, it ensures that the current substring contains only unique characters.
This approach optimizes the computation of the length of the longest substring by
handling the sliding window and updating the length accordingly, resulting in an
overall efficient solution.

5.5.4 Exercise

• Optimal Partition of String2 .

2 https://leetcode.com/problems/optimal-partition-of-string/

5.5. Longest Substring Without Repeating Characters 99

5.6 Compare Version Numbers

5.6.1 Problem statement

1
Given two version numbers, version1 and version2, your task is to compare them.
Version numbers consist of one or more revisions joined by a dot '.'. Each revision
is composed of digits and may contain leading zeros. Each revision has at least one
character. Revisions are indexed from left to right, with the leftmost revision being
revision 0, the next revision being revision 1, and so on.
For instance, 2.5.33 and 0.1 are valid version numbers.
To compare version numbers, you should compare their revisions in left-to-right order.
Revisions are compared using their integer value, ignoring any leading zeros. This
means that revisions 1 and 001 are considered equal. If a version number does not
specify a revision at a particular index, treat that revision as 0. For example, version
1.0 is less than version 1.1 because their revision 0s are the same, but their revision
1s are 0 and 1 respectively, and 0 is less than 1.
The function should return the following:
• If version1 is less than version2, return -1.
• If version1 is greater than version2, return 1.
• If version1 and version2 are equal, return 0.

Example 1

Input: version1 = "1.01", version2 = "1.001"

Output: 0
Explanation: Ignoring leading zeroes, both "01" and "001" represent the␣
˓→same integer "1".

1 https://leetcode.com/problems/compare-version-numbers/

100 Chapter 5. String

Example 2

Input: version1 = "1.0", version2 = "1.0.0"

Output: 0
Explanation: version1 does not specify revision 2, which means it is␣
˓→treated as "0".

Example 3

Input: version1 = "0.1", version2 = "1.1"

Output: -1
Explanation: version1's revision 0 is "0", while version2's revision 0␣
˓→is "1". 0 < 1, so version1 < version2.

Constraints

• 1 <= version1.length, version2.length <= 500.

• version1 and version2 only contain digits and '.'.
• version1 and version2 are valid version numbers.
• All the given revisions in version1 and version2 can be stored in a 32-bit inte-
ger.

5.6.2 Solution

Each version can be considered as an array of revisions.

version = revisions[0].revisions[1].revisions[2]....

The problem is to compare each revisions[i] between two versions.

For example, revisions[0] of version1 is less than of version2 in Example 3. So the
result is -1.
All revisions[i] of version1 and version2 are equal in Example 1. So the result is
0.
The number of revisions between the versions might not be equal (like in Example
2).

5.6. Compare Version Numbers 101

If all revisions of the shorter version are equal to the corresponding revisions of the
longer one, the version having extra revisions and there exists a non-zero revision
among them is the bigger one. Otherwise, the two versions are equal.

Code

#include <iostream>
#include <vector>
#include <string>
#include <numeric>
using namespace std;
//! @return the vector of revisions of the version
//! @example if version = "1.02.11", return {1,2,11}
vector<int> toVector(const string& version) {
vector<int> revisions;
string revision;
for (auto& c : version) {
if (c != '.') {
// continue to build current revision
revision += c;
} else {
// current revision completes
// uses stoi() to ignore leading zeros
revisions.push_back(stoi(revision));

// start a new revision

revision = "";
}
}
revisions.push_back(stoi(revision));
return revisions;
}

int compareVersion(const string& version1, const string& version2) {

vector<int> r1 = toVector(version1);
vector<int> r2 = toVector(version2);

int i = 0;
// perform the comparison on the revisions
while (i < r1.size() && i < r2.size()) {
if (r1[i] < r2[i]) {
(continues on next page)

102 Chapter 5. String

 (continued from previous page)
return -1;
} else if (r1[i] > r2[i]) {
return 1;
}
i++;
}
if (i == r1.size()) {
// if version1 is not longer than version2
// and version2 still has some valid revisions remain
if (accumulate(r2.begin() + i, r2.end(), 0) > 0) {
return -1;
}
} else if (accumulate(r1.begin() + i, r1.end(), 0) > 0) {
// if version2 is not longer than version1
// and version1 still has some valid revisions remain
return 1;
}
return 0;
}
int main() {
cout << compareVersion("1.01", "1.001") << endl;
cout << compareVersion("1.0", "1.0.0") << endl;
cout << compareVersion("0.1", "1.1") << endl;
}

Output:
0
0
-1

Complexity

• Runtime: O(N) where N = max(version1.length, version2.length).

• Extra space: O(N).

5.6. Compare Version Numbers 103

5.6.3 Conclusion

This solution first converts the version strings into vectors of integers representing the
individual components of the version numbers. This conversion is done by iterating
through each character of the version string, accumulating digits until encountering
a dot, at which point the accumulated integer is added to the revisions vector.
Once both version strings are converted into vectors, the function iterates through the
vectors, comparing corresponding elements to determine the relationship between
the versions. Additionally, it accounts for any remaining digits in the longer version
string after the common components by summing them up and comparing the totals.
This approach simplifies the comparison process by breaking down the version strings
into easily comparable components.

C++ Notes

• std::stoi2 is used to convert a string to an int. It ignores the leading zeros for
you.
• std::accumulate3 is used to compute the sum of a container.

2 https://en.cppreference.com/w/cpp/string/basic_string/stol
3 https://en.cppreference.com/w/cpp/algorithm/accumulate

104 Chapter 5. String

 CHAPTER

SIX

STACK

This chapter explores the stack data structure, a useful tool for managing data in a
Last-In-First-Out (LIFO) way. We’ll investigate the basics of stacks and examine how
they work using C++’s std::stack and std::vector from the Standard Template
Library (STL).
Stacks in programming are like a stack of books where you add and remove books
from the top. They provide a structured way to manage data, making them ideal for
handling temporary information, tracking function calls, and solving various algorith-
mic challenges.
What this chapter covers:
1. Introduction to Stacks: Begin by understanding the core principles of stacks,
their fundamental operations, and their real-world applications.
2. Leveraging std::stack: Dive into the STL’s powerful std::stack container,
mastering its usage and versatility for stack-based operations.
3. Exploring std::vector: Discover the capabilities of std::vector in context with
stacks, exploiting its dynamic array nature to create flexible stack structures.
4. Stack Operations: Explore operations such as push and pop, understanding
their impact on the stack’s state and memory usage.
5. Balancing Parentheses: Tackle the classic problem of parentheses balancing
using stacks, a prime example of their utility in parsing and validation.
As you progress through this chapter, you’ll learn about the importance of stacks and
how std::stack and std::vector can help you solve problems more efficiently. By
the end of the chapter, you’ll thoroughly understand the stack data structure’s Last-In-
First-Out (LIFO) principle and how you can leverage std::stack and std::vector to
manage data effectively. Let’s embark on this enlightening journey through stacks and
uncover their potential for simplifying complex operations and algorithmic problems!

105
6.1 Baseball Game

6.1.1 Problem statement

1
You are responsible for keeping score in a unique baseball game with special rules.
The game involves multiple rounds where the scores of previous rounds can influence
the scores of future rounds.
At the beginning of the game, your record is empty. You are given a list of operations
called ops, where each ops[i] is one of the following:
1. An integer x - This represents recording a new score of x.
2. "+" - This represents recording a new score that is the sum of the previous two
scores. It is guaranteed that there will always be two previous scores.
3. "D" - This represents recording a new score that is double the previous score. It
is guaranteed that there will always be a previous score.
4. "C" - This represents invalidating the previous score, removing it from the
record. It is guaranteed that there will always be a previous score.
Your task is to calculate and return the sum of all the scores in the record after per-
forming all the operations.

Example 1

Input: ops = ["5","2","C","D","+"]

Output: 30
Explanation:
"5" - Add 5 to the record; the record is now [5].
"2" - Add 2 to the record; the record is now [5, 2].
"C" - Invalidate and remove the previous score; the record is now [5].
"D" - Add 2 * 5 = 10 to the record; the record is now [5, 10].
"+" - Add 5 + 10 = 15 to the record, record is now [5, 10, 15].
The total sum is 5 + 10 + 15 = 30.

1 https://leetcode.com/problems/baseball-game/

106 Chapter 6. Stack

Example 2

Input: ops = ["5","-2","4","C","D","9","+","+"]

Output: 27
Explanation:
"5" - Add 5 to the record; the record is now [5].
"-2" - Add -2 to the record; the record is now [5, -2].
"4" - Add 4 to the record; the record is now [5, -2, 4].
"C" - Invalidate and remove the previous score; the record is now [5, -
˓→2].

"D" - Add 2 * -2 = -4 to the record; the record is now [5, -2, -4].
"9" - Add 9 to the record; the record is now [5, -2, -4, 9].
"+" - Add -4 + 9 = 5 to the record, record is now [5, -2, -4, 9, 5].
"+" - Add 9 + 5 = 14 to the record, record is now [5, -2, -4, 9, 5, 14].
The total sum is 5 + -2 + -4 + 9 + 5 + 14 = 27.

Example 3

Input: ops = ["1"]

Output: 1

Constraints

• 1 <= ops.length <= 1000.

• ops[i] is "C", "D", "+", or a string representing an integer in the range [-3 *
10^4, 3 * 10^4].
• For operation "+", there will always be at least two previous scores on the
record.
• For operations "C" and "D", there will always be at least one previous score on
the record.

6.1. Baseball Game 107

6.1.2 Solution

Code

#include <vector>
#include <iostream>
#include <string>
#include <numeric>
using namespace std;
int calPoints(const vector<string>& ops) {
vector<int> stk;
for (auto& s : ops) {
if (s == "C") {
stk.pop_back();
} else if (s == "D") {
stk.push_back(stk.back()*2);
} else if (s == "+") {
stk.push_back(stk[stk.size() - 1] + stk[stk.size() - 2]);
} else { // s is an integer
stk.push_back(stoi(s));
}
}
// compute the sum
return accumulate(stk.begin(), stk.end(), 0);
}
int main() {
vector<string> ops{"5","2","C","D","+"};
cout << calPoints(ops) << endl;
ops = {"5","-2","4","C","D","9","+","+"};
cout << calPoints(ops) << endl;
}

Output:
30
27

This solution simulates the baseball game by processing each round’s operation and
maintaining a stack of valid points. It accurately calculates the final sum of valid
points based on the given operations.

108 Chapter 6. Stack

Complexity

• Runtime: O(N), where N = ops.length.

• Extra space: O(N).

6.1.3 Implementation tips

1. The data structure stk you might need to solve this problem is a stack. But here
are the reasons you had better use std::vector:
• std::vector has also methods push_back(value) and pop_back() like the
ones in stack.
• On the other hand, a stack does not give easy access to the second last
element for the operator "+" in this problem.
2. accumulate(stk.begin(), stk.end(), 0)2 computes the sum of the vector stk.

6.1.4 Exercise

• Crawler Log Folder3 .

6.2 Valid Parentheses

6.2.1 Problem statement

1
You are given a string s containing only the characters '(', ')', '{', '}', '[', and
']'. Your task is to check if the input string is valid.
A string is considered valid if the following conditions are satisfied:
1. Opening brackets must be closed by the same type of brackets.
2. Opening brackets must be closed in the correct order, meaning that the inner-
most opening bracket should be closed before its surrounding brackets.
2 https://en.cppreference.com/w/cpp/algorithm/accumulate
3 https://leetcode.com/problems/crawler-log-folder/
1 https://leetcode.com/problems/valid-parentheses/

6.2. Valid Parentheses 109

Example 1

Input: s = "()"
Output: true

Example 2

Input: s = "()[]{}"
Output: true

Example 3

Input: s = "(]"
Output: false

Constraints

• 1 <= s.length <= 10^4.

• s consists of parentheses only '()[]{}'.

6.2.2 Solution: Using a stack

For each character c of s:

1. If it is an open parenthesis ('(', '{', or '['), push it into the stack.
2. If it is a closed parenthesis (')', '}', or ']') but its previous character is not
the corresponding open one, return false. End.
3. Otherwise (i.e. match open-closed), erase the pair.
4. Continue the process until all characters of s are visited.
5. Return true if the stack is empty, i.e. all valid pairs are erased.

110 Chapter 6. Stack

Code

#include <iostream>
#include <stack>
using namespace std;
bool isValid(const string& s) {
stack<char> stk;
for (auto& c : s) {
if (c == '(' || c == '[' || c == '{') {
stk.push(c);
} else if (stk.empty()) {
// start with a non-open parenthesis is invalid
return false;
} else if (c == ')' && stk.top() != '('
|| c == ']' && stk.top() != '['
|| c == '}' && stk.top() != '{') {
// the last open parenthesis does not match this closed one
return false;
} else {
// open-close match
stk.pop();
}
}
return stk.empty();
}
int main() {
cout << isValid("()") << endl;
cout << isValid("(){}[]") << endl;
cout << isValid("(]") << endl;
cout << isValid("([)]") << endl;
}

Output:
1
1
0
0

6.2. Valid Parentheses 111

Complexity:

• Runtime: O(N), where N = s.length.

• Extra space: O(N).

6.2.3 Conclusion

This solution efficiently checks the validity of a string of parentheses, brackets, and
curly braces by using a stack to ensure that each opening bracket is correctly matched
with its corresponding closing bracket.

6.2.4 Exercise

• Check If Word Is Valid After Substitutions2 .

6.3 Backspace String Compare

6.3.1 Problem statement

1
You are provided with two strings, s and t. Your task is to determine if these two
strings are equal when typed into an empty text editor, where the character '#' rep-
resents a backspace action.
Note that applying a backspace action to an empty text does not change the text; it
remains empty. Your function should return true if the two strings become equal after
considering the backspace actions, otherwise return false.

Example 1

Input: s = "ab#c", t = "ad#c"

Output: true
Explanation: Both s and t become "ac".

2 https://leetcode.com/problems/check-if-word-is-valid-after-substitutions/
1 https://leetcode.com/problems/backspace-string-compare/

112 Chapter 6. Stack

Example 2

Input: s = "ab##", t = "c#d#"

Output: true
Explanation: Both s and t become "".

Example 3

Input: s = "a#c", t = "b"

Output: false
Explanation: s becomes "c" while t becomes "b".

Constraints

• 1 <= s.length, t.length <= 200.

• s and t only contain lowercase letters and '#' characters.

Follow up

• Can you solve it in O(n) time and O(1) space?

6.3.2 Solution: Build and clean the string using the stack’s behaviors

Code

#include <iostream>
#include <vector>
using namespace std;
string cleanString(const string &s) {
vector<char> v;
for (int i = 0; i < s.length(); i++) {
if (s[i] != '#') {
// s[i] is a normal letter
v.push_back(s[i]);
} else {
if (!v.empty()) {
(continues on next page)

6.3. Backspace String Compare 113

 (continued from previous page)
// perform the backspace
v.pop_back();
}
}
}
// create a string from a vector of char
return string(v.begin(), v.end());
}
bool backspaceCompare(const string& s, const string& t) {
return cleanString(s) == cleanString(t);
}
int main() {
cout << backspaceCompare("ab#c", "ad#c") << endl;
cout << backspaceCompare("ab##", "c#d#") << endl;
cout << backspaceCompare("a#c", "b") << endl;
}

Output:
1
1
0

This solution effectively handles backspace characters ('#') in input strings s and t by
constructing cleaned versions of the strings and then comparing the cleaned strings
for equality.

Complexity

• Runtime: O(n), where n = max(s.length, t.length).

• Extra space: O(n).

6.3.3 Implementation notes

Why vector instead of stack?

You can use the methods push and pop of the data structure stack2 to build and clean
the strings.
2 https://en.cppreference.com/w/cpp/container/stack

114 Chapter 6. Stack

But vector has also such methods: push_back and pop_back.
On the other hand, using vector it is easier to construct a string by constructor than
using stack after cleaning.

Can you solve it in O(n) time and O(1) space?

Yes, you can.

The simplest way is just to perform the erasure directly on strings s and t. But the
run time complexity of string::erase3 is not constant.

6.3.4 Exercise

• Removing Stars From a String4 .

6.4 Remove All Adjacent Duplicates in String II

6.4.1 Problem statement

1
You are given a string s and an integer k. A k duplicate removal operation involves
selecting k adjacent and identical letters from s and removing them, causing the
remaining portions on the left and right of the removed substring to join together.
You need to perform the k duplicate removal operation on s repeatedly until it is no
longer possible. After completing all such operations, return the resulting string. It is
guaranteed that the answer will be unique.
3 https://en.cppreference.com/w/cpp/string/basic_string/erase
4 https://leetcode.com/problems/removing-stars-from-a-string/
1 https://leetcode.com/problems/remove-all-adjacent-duplicates-in-string-ii/

6.4. Remove All Adjacent Duplicates in String II 115

Example 1

Input: s = "abcd", k = 2
Output: "abcd"
Explanation: There is nothing to delete.

Example 2

Input: s = "deeedbbcccbdaa", k = 3
Output: "aa"
Explanation:
First delete "eee" and "ccc", get "ddbbbdaa"
Then delete "bbb", get "dddaa"
Finally delete "ddd", get "aa"

Example 3

Input: s = "pbbcggttciiippooaais", k = 2
Output: "ps"

Constraints

• 1 <= s.length <= 10^5.

• 2 <= k <= 10^4.
• s only contains lower case English letters.

6.4.2 Solution: Strings of adjacent equal letters

Construct a stack of strings that has adjacent equal letters and perform the removal
during building those strings.

116 Chapter 6. Stack

Example 2

For s = "deeedbbcccbdaa" and k = 3:

• The first built string is "d".
• Then "eee" with the exact length k, remove this string.
• The next character is 'd', which equals the last character of the last string "d",
merge them together. The first string becomes "dd".
• The next string is "bb".
• Then "ccc" is removed.
• The next character 'b' is merged with the last string ("bb") to become "bbb"
and be removed.
• The next character 'd' is merged with the last string ("dd") to become "ddd"
and be removed.
• The remaining string is "aa".

Code

#include <iostream>
#include <vector>
using namespace std;
string removeDuplicates(string& s, int k) {
// stk is used as a stack
// all letters in each string a of stk are equal
// every a's length is less than k
vector<string> stk;
int i = 0;
while (i < s.length()) {
// a represents the current string with duplicate letters
string a;

if (!stk.empty() && s[i] == stk.back().back()) {

// let a be the latest string in stk
// because its letters are equal to s[i]
a = move(stk.back());
stk.pop_back();
}
int j = i;
(continues on next page)

6.4. Remove All Adjacent Duplicates in String II 117

 (continued from previous page)
// iterate all adjacent duplicates of s[i]
while (j < s.length() && s[j] == s[i]) {
a += s[j];
// remove the k-duplicate
if (a.length() == k) {
a = "";
}
j++;
}
// after the loop, the number of duplicates in a is less than k
if (!a.empty()) {
stk.push_back(a);
}
i = j;
}
// create the final result from stk
s = "";
for (auto& str : stk) {
s += str;
}
return s;
}
int main() {
cout << removeDuplicates("abcd", 2) << endl;
cout << removeDuplicates("deeedbbcccbdaa", 3) << endl;
cout << removeDuplicates("pbbcggttciiippooaais", 2) << endl;
}

Output:
abcd
aa
ps

This solution efficiently removes substrings of consecutive duplicate characters of

length k from a given string by using a stack to keep track of the adjacent charac-
ters.

118 Chapter 6. Stack

Complexity

• Runtime: O(N), where N = s.length.

• Extra space: O(N).

6.4.3 Implementation tips

• The data structure stk you might need to solve this problem is a stack. But here
are the reasons you had better use std::vector:
• std::vector also has methods push_back(value) and pop_back() like the ones
in a stack.
• On the other hand, it is faster for a vector to perform the string concatenation
at the end.
• In the expression stk.back().back(): stk.back() is the latest string a of stk.
Then stk.back().back() = a.back() is the last character of a.

6.4.4 Exercise

• Remove All Adjacent Duplicates In String2 .

2 https://leetcode.com/problems/remove-all-adjacent-duplicates-in-string/

6.4. Remove All Adjacent Duplicates in String II 119

120 Chapter 6. Stack
 CHAPTER

SEVEN

PRIORITY QUEUE (HEAP)

This chapter explores priority queues (or heaps), the fascinating data structures de-
signed to manage elements with distinct levels of importance. In this chapter, we’ll
focus on harnessing the capabilities of C++’s std::priority_queue from the Stan-
dard Template Library (STL).
Think of a priority queue as a line at a theme park, where individuals with priority
passes are served before others. Similarly, a priority queue ensures that elements
with higher priority are processed ahead of those with lower priority, enabling us to
address a wide range of problems that involve ordering and selection.
What this chapter covers:
1. Understanding Priority Queues: Begin by grasping the essence of priority
queues, their underlying mechanisms, and the significance of their unique or-
dering.
2. Leveraging std::priority_queue: Dive into the versatile std::priority_queue
container provided by the STL, mastering its usage for managing priorities ef-
fectively.
3. Operations and Methods: Explore the operations available in
std::priority_queue, including insertion, and extraction while maintain-
ing optimal order.
4. Custom Comparators: Customize the behavior of your priority queue by uti-
lizing custom comparators, tailoring it to handle diverse data types and priority
criteria.
5. Problem-Solving with Priority Queues: Learn strategies for tackling problems
where prioritization is key, from scheduling tasks to efficient data retrieval.

121
7.1 Last Stone Weight

7.1.1 Problem statement

1
You are given an array of integers called stones, where each stones[i] represents
the weight of the i-th stone.
A game is played with these stones as follows: In each turn, we choose the two
heaviest stones and smash them together. Let us say the weights of the two heaviest
stones are x and y, where x <= y. The outcome of this smash operation is:
1. If x is equal to y, both stones are destroyed.
2. If x is not equal to y, the stone with weight x is destroyed, and the stone with
weight y now has a new weight of y - x.
The game continues until there is at most one stone left. Your task is to determine the
smallest possible weight of the remaining stone after the game ends. If there are no
stones left, return 0.

Example 1

Input: stones = [2,7,4,1,8,1]

Output: 1
Explanation:
We combine 7 and 8 to get 1, so the array converts to [2,4,1,1,1] then,
we combine 2 and 4 to get 2, so the array converts to [2,1,1,1] then,
we combine 2 and 1 to get 1, so the array converts to [1,1,1] then,
we combine 1 and 1 to get 0, so the array converts to [1] then that's␣
˓→the value of the last stone.

Example 2

Input: stones = [1]

Output: 1

1 https://leetcode.com/problems/last-stone-weight/

122 Chapter 7. Priority Queue (Heap)

Constraints

• 1 <= stones.length <= 30.

• 1 <= stones[i] <= 1000.

7.1.2 Solution: Keeping the heaviest stones on top

The only things you want at any time are the two heaviest stones. One way of keeping
this condition is by using std::priority_queue.

Code

#include <vector>
#include <iostream>
#include <queue>
using namespace std;
int lastStoneWeight(vector<int>& stones) {
priority_queue<int> q(stones.begin(), stones.end());
while (q.size() >= 2) {
int y = q.top();
q.pop();
int x = q.top();
q.pop();
// compare two heaviest stones
if (y != x) {
q.push(y - x);
}
}
return q.empty() ? 0 : q.top();
}
int main() {
vector<int> stones{2,7,4,1,8,1};
cout << lastStoneWeight(stones) << endl;
stones = {1};
cout << lastStoneWeight(stones) << endl;
}

7.1. Last Stone Weight 123

Output:
1
1

Complexity

• Runtime: O(n*logn), where n = stones.length.

• Extra space: O(n).

7.1.3 Conclusion

This solution efficiently simulates the process of smashing stones and finding the last
remaining stone by using a max-heap (priority queue) to always select the heaviest
stones to smash together.

7.2 Kth Largest Element in a Stream

7.2.1 Problem statement

1
Create a class that can find the k-th largest element in a stream of integers. This
is the k-th largest element when the elements are arranged in sorted order, not the
k-th distinct element.
The KthLargest class needs to support the following operations:
1. KthLargest(int k, int[] nums): This initializes the object with an integer k
and a stream of integers nums.
2. int add(int val): This method adds the integer val to the stream and returns
the element representing the k-th largest element in the stream.
1 https://leetcode.com/problems/kth-largest-element-in-a-stream/

124 Chapter 7. Priority Queue (Heap)

Example 1

Input
["KthLargest", "add", "add", "add", "add", "add"]
[[3, [4, 5, 8, 2]], [3], [5], [10], [9], [4]]
Output
[null, 4, 5, 5, 8, 8]

Explanation
KthLargest kthLargest = new KthLargest(3, [4, 5, 8, 2]);
kthLargest.add(3); // return 4
kthLargest.add(5); // return 5
kthLargest.add(10); // return 5
kthLargest.add(9); // return 8
kthLargest.add(4); // return 8

Constraints

• 1 <= k <= 10^4.

• 0 <= nums.length <= 10^4.
• -10^4 <= nums[i] <= 10^4.
• -10^4 <= val <= 10^4.
• At most 10^4 calls will be made to add.
• It is guaranteed that there will be at least k elements in the array when you
search for the k-th element.

7.2.2 Solution 1: Sort and Append

Sort the stream when initialization. And keep it sorted whenever you append a new
value.

7.2. Kth Largest Element in a Stream 125

Example 1

For nums = [4, 5, 8, 2] and k = 3.

• Sort nums = [8, 5, 4, 2].
• Adding 3 to nums. It becomes [8, 5, 4, 3, 2]. The third largest element is 4.
• Adding 5 to nums. It becomes [8, 5, 5, 4, 3, 2]. The third largest element is
5.
• Adding 10 to nums. It becomes [10, 8, 5, 5, 4, 3, 2]. The third largest
element is 5.
• So on and so on.

Code

#include <vector>
#include <algorithm>
#include <iostream>
using namespace std;
class KthLargest {
vector<int> _nums;
int _k;
public:
KthLargest(int k, vector<int>& nums) : _nums(nums), _k(k) {
// sort the nums when constructed
sort(_nums.begin(), _nums.end(), std::greater());
}

int add(int val) {

auto it = _nums.begin();
// find the position to insert val
while (it != _nums.end() && val < *it) {
it++;
}
_nums.insert(it, val);
// return the k-th largest element
return *(_nums.begin() + _k - 1);
}
};
int main() {
(continues on next page)

126 Chapter 7. Priority Queue (Heap)

 (continued from previous page)
vector<int> nums{4,5,8,2};
KthLargest a(3, nums);
cout << a.add(3) << endl;
cout << a.add(5) << endl;
cout << a.add(10) << endl;
cout << a.add(9) << endl;
cout << a.add(4) << endl;
}

Output:
4
5
5
8
8

This solution maintains a sorted vector _nums in non-ascending order upon initializa-
tion, which stores the elements. When adding a new element val, it inserts it into
_nums while maintaining the sorted order.
Since _nums is sorted in non-ascending order, the k-th largest element is always at
index _k - 1. Thus, upon adding a new element, it returns the value at index _k - 1
as the k-th largest element in the collection.
This approach optimizes the add operation by leveraging the sorted nature of the data
structure, resulting in efficient retrieval of the k-th largest element.

Complexity

• Runtime: for the constructor O(N*logN), where N = nums.length. For the add
method, O(N).
• Extra space: O(1).

7.2. Kth Largest Element in a Stream 127

7.2.3 Solution 2: Priority queue

There is a data structure that has the property you want in this problem.
It is std::priority_queue, which keeps its top element is always the largest one
according to the comparison you define for the queue.
By default, the “less than” comparison is used for std::priority_queue (heap) and the
top one is always the biggest element.
If you want the top one is always the smallest element, you can use the comparison
“greater than” for your heap.

Code

#include <vector>
#include <queue>
#include <iostream>
using namespace std;
class KthLargest {
priority_queue<int, vector<int>, greater<int>> _q;
int _k;
public:
KthLargest(int k, vector<int>& nums)
// create the heap when constructed
: _q(nums.begin(), nums.end()), _k(k) {

int add(int val) {

_q.push(val);

// remove elements until _q remains k elements

while (_q.size() > _k) {
_q.pop();
}
return _q.top();
}
};
int main() {
vector<int> nums{4,5,8,2};
KthLargest a(3, nums);
(continues on next page)

128 Chapter 7. Priority Queue (Heap)

 (continued from previous page)
cout << a.add(3) << endl;
cout << a.add(5) << endl;
cout << a.add(10) << endl;
cout << a.add(9) << endl;
cout << a.add(4) << endl;
}

Output:
4
5
5
8
8

Complexity

• Runtime: for the constructor, O(N*logN), where N = nums.length. For the add
method, O(logN).
• Extra space: O(1).

7.2.4 Conclusion

The key insight of Solution 2 is utilizing a min-heap (priority queue with the greater
comparator) to find the kth largest element in a collection.
Upon initialization, the constructor populates the priority queue with the elements
from the input vector nums. When adding a new element val, it inserts it into the pri-
ority queue and then removes elements until the size of the priority queue is reduced
to _k, ensuring that only the k largest elements are retained in the queue.
Finally, it returns the top element of the priority queue, which represents the kth
largest element. This approach leverages the properties of a min-heap to track the
kth largest element in the collection, resulting in an overall efficient solution.

7.2. Kth Largest Element in a Stream 129

7.2.5 Exercise

• Kth Largest Element in an Array2 .

7.3 Kth Smallest Element in a Sorted Matrix

7.3.1 Problem statement

1
You are given an n x n matrix where each row and column is sorted in ascending
order. Your task is to find the k-th smallest element in this matrix.
Please note that we are looking for the k-th smallest element based on its position in
the sorted order, and not counting distinct elements.
Additionally, it is required to find a solution with a memory complexity better than
O(n^2).

Example 1

Input: matrix = [[1,5,9],[10,11,13],[12,13,15]], k = 8

Output: 13
Explanation: The elements in the matrix are [1,5,9,10,11,12,13,13,15],␣
˓→and the 8th smallest number is 13

Example 2

Input: matrix = [[-5]], k = 1

Output: -5

2 https://leetcode.com/problems/kth-largest-element-in-an-array/
1 https://leetcode.com/problems/kth-smallest-element-in-a-sorted-matrix/

130 Chapter 7. Priority Queue (Heap)

Constraints

• n == matrix.length == matrix[i].length.
• 1 <= n <= 300.
• -10^9 <= matrix[i][j] <= 10^9.
• All the rows and columns of matrix are guaranteed to be sorted in non-
decreasing order.
• 1 <= k <= n^2.

Follow up

• Could you solve the problem with a constant memory (i.e., O(1) memory com-
plexity)?
• Could you solve the problem in O(n) time complexity? The solution may be too
advanced for an interview but you may find reading this paper fun.

7.3.2 Solution 1: Transform the 2-D matrix into a 1-D vector then sort

You can implement exactly what Example 1 has explained.

Code

#include <iostream>
#include <vector>
#include <algorithm>
using namespace std;
int kthSmallest(const vector<vector<int>>& matrix, int k) {
vector<int> m;
// transform the 2D matrix into a 1D array m
for (auto& row : matrix) {
m.insert(m.end(), row.begin(), row.end());
}
// sort the array m
sort(m.begin(), m.end());
return m.at(k - 1);
}
int main() {
(continues on next page)

7.3. Kth Smallest Element in a Sorted Matrix 131

 (continued from previous page)
vector<vector<int>> matrix{{1,5,9},{10,11,13},{12,13,15}};
cout << kthSmallest(matrix, 8) << endl;
matrix = {{-5}};
cout << kthSmallest(matrix, 1) << endl;
}

Output:
13
-5

The core idea behind this solution is to transform the 2D matrix into a 1D sorted array,
making it easier to find the k-th smallest element efficiently. The time complexity of
this solution is dominated by the sorting step, which is O(N*logN), where N is the total
number of elements in the matrix.

Complexity

• Runtime: O(N*logN), where N = n^2 is the total number of elements in the

matrix.
• Extra space: O(N).

7.3.3 Solution 2: Build the max heap and keep it ungrown

Instead of sorting after building the vector in Solution 1, you can do the other way
around. It means building up the vector from scratch and keeping it sorted.
Since you need only the k-th smallest element, std::priority_queue2 can be used for
this purpose.

Code

#include <iostream>
#include <vector>
#include <queue>
using namespace std;
int kthSmallest(const vector<vector<int>>& matrix, int k) {
(continues on next page)
2 https://en.cppreference.com/w/cpp/container/priority_queue

132 Chapter 7. Priority Queue (Heap)

 (continued from previous page)
priority_queue<int> q;
for (int row = 0; row < matrix.size(); row++) {
for (int col = 0; col < matrix[row].size(); col++) {
q.push(matrix[row][col]);
// maintain q's size does not exceed k
if (q.size() > k) {
q.pop();
}
}
}
return q.top();
}
int main() {
vector<vector<int>> matrix{{1,5,9},{10,11,13},{12,13,15}};
cout << kthSmallest(matrix, 8) << endl;
matrix = {{-5}};
cout << kthSmallest(matrix, 1) << endl;
}

Output:
13
-5

Complexity

• Runtime: O(N*logk), where N = n^2 is the total number of elements of the

matrix.
• Extra space: O(k).

7.3.4 Conclusion

Solution 2 maintains a priority queue of size k, allowing it to efficiently keep track of

the k-th smallest element encountered while iterating through the matrix.
This approach is handy for large matrices, as it doesn’t require sorting the entire
matrix.

7.3. Kth Smallest Element in a Sorted Matrix 133

7.3.5 Exercise

• Find K Pairs with Smallest Sums3 .

7.4 Construct Target Array With Multiple Sums

7.4.1 Problem statement

1
You are provided with an array of integers called target with n elements. You start
with another array, arr, consisting of n elements, all initialized to 1. You have the
ability to perform the following operation:
1. Calculate the sum of all elements in your current array arr, let’s call it x.
2. Choose an index i where 0 <= i < n, and update the value at index i in arr to
be x.
You can repeat this operation as many times as needed. Your task is to determine
whether it’s possible to transform the initial array arr into the given target array
using this operation. If it’s possible, return true; otherwise, return false.

Example 1

Input: target = [9,3,5]

Output: true
Explanation: Start with arr = [1, 1, 1]
[1, 1, 1], sum = 3 choose index 1
[1, 3, 1], sum = 5 choose index 2
[1, 3, 5], sum = 9 choose index 0
[9, 3, 5] Done

3 https://leetcode.com/problems/find-k-pairs-with-smallest-sums/
1 https://leetcode.com/problems/construct-target-array-with-multiple-sums/

134 Chapter 7. Priority Queue (Heap)

Example 2

Input: target = [1,1,1,2]

Output: false
Explanation: Impossible to create target array from [1,1,1,1].

Example 3

Input: target = [8,5]

Output: true

Constraints

• n == target.length.
• 1 <= n <= 5 * 10^4.
• 1 <= target[i] <= 10^9.

7.4.2 Solution 1: Going backward

If you start from arr = [1,1,...,1] and follow the required procedure, the new
element x you get for the next state is always the max element of arr.
To solve this problem, you can start from the max element of the given target to
compute its previous state until you get the arr = [1,1,...,1].

Example 1

For target = [9,3,5]:

• The max element is 9, subtract it from the remaining sum: 9 - (3 + 5) = 1,
you get target = [1,3,5].
• The max element is 5, subtract it from the remaining sum: 5 - (1 + 3) = 1,
you get target = [1,3,1].
• The max element is 3, subtract it from the remaining sum: 3 - (1 + 1) = 1,
you get target = [1,1,1].
• Return true.

7.4. Construct Target Array With Multiple Sums 135

Notes

• If target = [m,1] or target = [1,m] for any m >= 1, you can always turn it to
arr = [1,1].
• If the changed value after the subtraction is still the max element of the previous
state, you need to redo the subtraction at the same position. In this case, the
modulo might be used instead of subtraction.

Code

#include <iostream>
#include <numeric>
#include <algorithm>
#include <vector>
using namespace std;
bool isPossible(const vector<int>& target) {
// compute sum of all target's elements
unsigned long sum = accumulate(target.begin(),
target.end(),
(unsigned long) 0);
// find the max element in the target
// pmax is the pointer to the max element,
// *pmax is the value that pointer points to
auto pmax = max_element(target.begin(), target.end());
while (*pmax > 1) {
// compute the remaining sum
sum -= *pmax;
if (sum == 1) {
// This is the case target = [m,1],
// which you can always turn it to [1,1].
return true;
}
if (*pmax <= sum) {
// the next subtraction leads to non-positive values
return false;
}
if (sum == 0) {
// cannot change target
return false;
}
(continues on next page)

136 Chapter 7. Priority Queue (Heap)

 (continued from previous page)
// perform the subtraction as much as possible
// and update new value for the pointer pmax
*pmax %= sum;
if (*pmax == 0) {
return false;
}
// compute the sum of the subtracted target
sum += *pmax;
// find the max element in the subtracted target
pmax = max_element(target.begin(), target.end());
}
// if the final target = [1, .., 1],
// its sum equals to its length
return sum == target.size();
}
int main() {
vector<int> target{9,3,5};
cout << isPossible(target) << endl;
target = {1,1,1,2};
cout << isPossible(target) << endl;
target = {8,5};
cout << isPossible(target) << endl;
}

Output:
1
0
1

This solution iteratively reduces the maximum element in the target array while
keeping track of the total sum. It checks various conditions to determine whether
it’s possible to reach an array consisting of only 1s. If all conditions are satisfied, it
returns true; otherwise, it returns false.

7.4. Construct Target Array With Multiple Sums 137

Complexity

• Runtime: O(log N), where N = max(target).

• Extra space: O(1).

7.4.3 Solution 2: Using priority_queue

In the solution above, the position of the max element in each state is not so important
as long as you update exactly it, not the other ones.
That might lead to the usage of the std::priority_queue.

Code

#include <iostream>
#include <numeric>
#include <queue>
#include <vector>
using namespace std;
bool isPossible(const vector<int>& target) {
// create a heap from the target
priority_queue<int> q(target.begin(), target.end());
// compute the sum of all elements
unsigned long sum = accumulate(target.begin(),
target.end(),
(unsigned long) 0);
while (q.top() > 1) {
// compute the remaining sum
sum -= q.top();
if (sum == 1) {
return true;
}
if (q.top() <= sum) {
return false;
}
if (sum == 0) {
return false;
}
// perform the subtraction as much as possible
int pre = q.top() % sum;
(continues on next page)

138 Chapter 7. Priority Queue (Heap)

 (continued from previous page)
if (pre == 0) {
return false;
}
// remove the old max element
q.pop();
// add subtracted element to the heap
q.push(pre);
// compute the sum of the subtracted target
sum += pre;
}
return sum == target.size();
}
int main() {
vector<int> target{9,3,5};
cout << isPossible(target) << endl;
target = {1,1,1,2};
cout << isPossible(target) << endl;
target = {8,5};
cout << isPossible(target) << endl;
}

Output:
1
0
1

Complexity

• Runtime: O(logN), where N = max(target).

• Extra space: O(n), where n = target.length.

7.4. Construct Target Array With Multiple Sums 139

7.4.4 Conclusion

Solution 2 uses a max heap (priority_queue) to efficiently find and process the max-
imum element in the target array while keeping track of the total sum. It checks
various conditions to determine whether it’s possible to reach an array consisting of
only 1s.

7.4.5 Exercise

• Minimum Amount of Time to Fill Cups2 .

2 https://leetcode.com/problems/minimum-amount-of-time-to-fill-cups/

140 Chapter 7. Priority Queue (Heap)

 CHAPTER

EIGHT

BIT MANIPULATION

In this chapter, we’re diving deep into Bit Manipulation, a fascinating computer sci-
ence and programming area that manipulates individual bits within data.
Bit Manipulation is crucial in various programming tasks, from optimizing algorithms
to working with hardware-level data. Whether you’re a seasoned programmer looking
to expand your skill set or a newcomer eager to delve into the intricacies of bits and
bytes, this chapter has something valuable for you.
Here’s what you can expect to explore in this chapter:
1. Understanding the Basics: We’ll start by demystifying bits and binary numbers,
ensuring you’re comfortable with the fundamentals. You’ll learn to convert be-
tween decimal and binary, perform basic bit operations, and understand two’s
complement representation.
2. Bitwise Operators: We’ll delve into the world of bitwise operators in program-
ming languages like C++. You’ll get hands-on experience with AND, OR, XOR,
and other essential operators, seeing how they can be applied to practical cod-
ing problems.
3. Bit Hacks: Discover the art of Bit Hacks – clever and often elegant tricks pro-
grammers use to solve specific problems efficiently. You’ll learn to perform tasks
like counting bits, finding the rightmost set bit, and swapping values without
temporary variables.
4. Bit Manipulation Techniques: We’ll explore techniques and patterns for com-
mon bit manipulation tasks, such as setting, clearing, or toggling specific bits,
checking if a number is a power of two, or extracting subsets of bits from a
larger number.
By the end of this chapter, you’ll have a solid foundation in Bit Manipulation and
the ability to harness the power of bits to optimize your code and tackle complex
problems. So, let’s embark on this exciting journey into the realm of Bit Manipulation

141
and discover how the smallest data units can have a massive impact on your coding
skills and efficiency!

8.1 Hamming Distance

8.1.1 Problem statement

1
The Hamming distance2 between two integers is the number of positions at which
the corresponding bits are different.
Given two integers x and y, return the Hamming distance between them.

Example 1

Input: x = 1, y = 4
Output: 2
Explanation:
1 (0 0 0 1)
4 (0 1 0 0)
^ ^
The above arrows point to positions where the corresponding bits are␣
˓→different.

Example 2

Input: x = 3, y = 1
Output: 1

1 https://leetcode.com/problems/hamming-distance/
2 https://en.wikipedia.org/wiki/Hamming_distance

142 Chapter 8. Bit Manipulation

Constraints

• 0 <= x, y <= 2^31.

8.1.2 Solution: Using bitwise operator XOR

You could use bitwise XOR (^) to get the bit positions where x and y are different.
Then use bitwise AND operator (&) at each position to count them.

Example 1

x = 1 (0 0 0 1)
y = 4 (0 1 0 0)
z = x^y (0 1 0 1)

Code

#include <iostream>
int hammingDistance(int x, int y) {
// compute the bit difference
int z = x ^ y;
int count = 0;
while (z) {
count += z & 1; // e.g. '0101' & '0001'
// shift z to the right one position
z = z >> 1; // e.g. z = '0101' >> '0010'
}
return count;
}
int main() {
std::cout << hammingDistance(1,4) << std::endl; // 2
std::cout << hammingDistance(1,3) << std::endl; // 1
}

Output:
2
1

8.1. Hamming Distance 143

Complexity

• Runtime: O(1) as the number of bits is at most 32 as constrained.

• Extra space: O(1).

8.1.3 Conclusion

Utilizing bitwise operations, such as XOR (^) and bitwise AND (&), allows for ef-
ficient computation of the Hamming distance between two integers. This approach
provides a straightforward and efficient method for calculating the Hamming distance
without the need for complex logic or additional data structures.

8.1.4 Exercise

• Number of 1 Bits3 .

8.2 Power of Four

8.2.1 Problem statement

1
Given an integer n, return true if it is a power of four. Otherwise, return false.
An integer n is a power of four if there exists an integer x such that n == 4^x.

Example 1

Input: n = 16
Output: true

3 https://leetcode.com/problems/number-of-1-bits/
1 https://leetcode.com/problems/power-of-four/

144 Chapter 8. Bit Manipulation

Example 2

Input: n = 5
Output: false

Example 3

Input: n = 1
Output: true

Constraints

• -2^31 <= n <= 2^31 - 1.

Follow up

• Could you solve it without loops/recursion?

8.2.2 Solution 1: Division by four

Code

#include <iostream>
using namespace std;
bool isPowerOfFour(int n) {
// perform the divison by 4 repeatedly
while (n % 4 == 0 && n > 0) {
n /= 4;
}
// if n % 4 != 0, then n > 1
return n == 1;
}
int main()
{
cout << isPowerOfFour(16) << endl;
cout << isPowerOfFour(5) << endl;
(continues on next page)

8.2. Power of Four 145

 (continued from previous page)
cout << isPowerOfFour(1) << endl;
}

Output:
1
0
1

This solution repeatedly divides the given number n by 4 until n becomes either 1 or
a number that is not divisible by 4. If n becomes 1 after this process, it means that n
was originally a power of 4.

Complexity

• Runtime: O(logn).
• Extra space: O(1).

8.2.3 Solution 2: Binary representation

You can write down the binary representation of the powers of four to find the pattern.

1 : 1
4 : 100
16 : 10000
64 : 1000000
256 : 100000000
...

You might notice the patterns are n is a positive integer having only one bit 1 in
its binary representation and it is located at the odd positions (starting from the
right).
How can you formulate those conditions?
If n has only one bit 1 in its binary representation 10...0, then n - 1 has the complete
opposite binary representation 01...1.
You can use the bit operator AND to formulate this condition

n & (n - 1) == 0

146 Chapter 8. Bit Manipulation

Let A be the number whose binary representation has only bits 1 at all odd positions,
then n & A is never 0.
In this problem, A < 2^31. You can chooseA = 0x55555555, the hexadecimal of 0101
0101 0101 0101 0101 0101 0101 0101.

Code

#include <iostream>
using namespace std;
bool isPowerOfFour(int n) {
// the condition of the pattern "n is a positive integer
// having only one bit 1 in its binary representation and
// it is located at the odd positions"
return n > 0 && (n & (n - 1)) == 0 && (n & 0x55555555) != 0;
}
int main() {
cout << isPowerOfFour(16) << endl;
cout << isPowerOfFour(5) << endl;
cout << isPowerOfFour(1) << endl;
}

Output:
1
0
1

Complexity

• Runtime: O(1).
• Extra space: O(1).

8.2. Power of Four 147

8.2.4 Conclusion

Recognizing the unique properties of powers of four, such as their binary representa-
tion, can lead to efficient solutions. Solution 2 leverages bitwise operations to check
if a number meets the criteria of being a power of four.
By examining the binary representation and ensuring that the only set bit is located
at an odd position, Solution 2 effectively determines whether the number is a power
of four in constant time complexity, without the need for division operations.
But in term of readable code, Solution 2 is not easy to understand like Solution 1,
where complexity of O(logn) is not too bad.

8.2.5 Exercise

• Power of Two.2 .

8.3 Find the Duplicate Number

8.3.1 Problem statement

1
You have an array of integers called nums that contains n + 1 integers. Each integer
in the array falls within the range [1, n] inclusive.
Within this array, there is only one number that appears more than once. Your task is
to find and return this repeated number.
Importantly, you must solve this problem without making any modifications to the
original array nums, and you are only allowed to use a constant amount of extra
space.
2 https://leetcode.com/problems/power-of-two/
1 https://leetcode.com/problems/find-the-duplicate-number/

148 Chapter 8. Bit Manipulation

Example 1

Input: nums = [1,3,4,2,2]

Output: 2

Example 2

Input: nums = [3,1,3,4,2]

Output: 3

Constraints

• 1 <= n <= 10^5.

• nums.length == n + 1.
• 1 <= nums[i] <= n.
• All the integers in nums appear only once except for precisely one integer which
appears two or more times.

Follow up

• How can we prove that at least one duplicate number must exist in nums?
• Can you solve the problem in linear runtime complexity?

8.3.2 Solution 1: Sorting

Code

#include <vector>
#include <iostream>
#include <algorithm>
using namespace std;
int findDuplicate(vector<int>& nums) {
sort(nums.begin(), nums.end());
for (int i = 0; i < nums.size() - 1; i++) {
if (nums[i] == nums[i + 1]) {
(continues on next page)

8.3. Find the Duplicate Number 149

 (continued from previous page)
return nums[i];
}
}
return 0;
}
int main() {
vector<int> nums{1,3,4,2,2};
cout << findDuplicate(nums) << endl;
nums = {3,1,3,4,2};
cout << findDuplicate(nums) << endl;
}

Output:
2
3

The code relies on sorting to bring duplicate elements together, making it easy to
identify them during the linear pass.

Complexity

• Runtime: O(n*logn) (sorting).

• Extra space: O(1).

8.3.3 Follow up

How can we prove that at least one duplicate number must exist in nums?

Due to Pigeonhole principle2 :

Here there are n + 1 pigeons in n holes. The pigeonhole principle says that at least
one hole has more than one pigeon.
2 https://en.wikipedia.org/wiki/Pigeonhole_principle

150 Chapter 8. Bit Manipulation

Can you solve the problem in linear runtime complexity?

Here are a few solutions.

8.3.4 Solution 2: Marking the visited numbers

Code

#include <vector>
#include <iostream>
using namespace std;
int findDuplicate(const vector<int>& nums) {
// initialize n + 1 elements false
vector<bool> visited(nums.size());
for (auto& a : nums) {
if (visited.at(a)) {
return a;
}
visited[a] = true;
}
return 0;
}
int main() {
vector<int> nums{1,3,4,2,2};
cout << findDuplicate(nums) << endl;
nums = {3,1,3,4,2};
cout << findDuplicate(nums) << endl;
}

Output:
2
3

This solution utilizes a boolean array visited to track visited elements.

It iterates through the elements of the input vector nums, marking each element as
visited by setting the corresponding index in the visited array to true. Upon encoun-
tering a visited element, it immediately returns it as the duplicate element.
This approach efficiently identifies the duplicate element in the vector by tracking
visited elements without requiring additional space proportional to the size of the
input vector.

8.3. Find the Duplicate Number 151

Complexity

• Runtime: O(n).
• Extra space: much less than O(n) since std::vector<bool> is optimized for
space-efficient.

8.3.5 Solution 3: Marking with std::bitset

Since n <= 10^5, you can use this size for a std::bitset3 to do the marking.

Code

#include <vector>
#include <iostream>
#include <bitset>
using namespace std;
int findDuplicate(const vector<int>& nums) {
// initialize visited = '000..0' with 100001 bits 0
bitset<100001> visited;
for (auto& a : nums) {
if (visited[a]) {
return a;
}
// set a-th bit to 1
visited[a] = 1;
}
return 0;
}
int main() {
vector<int> nums{1,3,4,2,2};
cout << findDuplicate(nums) << endl;
nums = {3,1,3,4,2};
cout << findDuplicate(nums) << endl;
}

Output:
2
3

3 https://en.cppreference.com/w/cpp/utility/bitset

152 Chapter 8. Bit Manipulation

This code uses a bitset to keep track of visited elements and quickly detects any du-
plicate element encountered during the iteration.

Complexity

• Runtime: O(n).
• Extra space: O(1).

8.3.6 Key Takeaway

Different strategies can be employed to detect duplicates in an array, such as sorting

followed by linear search, using a hash set to track visited elements, or utilizing a
bitset to mark visited indices.
Each approach has its trade-offs in terms of time complexity, space complexity, and
implementation simplicity. To me, Solution 2 balances all the trade-offs. What do you
think?

8.3.7 Exercise

• Missing Number4 .

8.4 Maximum Product of Word Lengths

8.4.1 Problem statement

1
Given a string array words, return the maximum value of length(word[i]) *
length(word[j]) where the two words do not share common letters. If no such two
words exist, return 0.
4 https://leetcode.com/problems/missing-number/
1 https://leetcode.com/problems/maximum-product-of-word-lengths/

8.4. Maximum Product of Word Lengths 153

Example 1

Input: words = ["abcw","baz","foo","bar","xtfn","abcdef"]

Output: 16
Explanation: The two words can be "abcw", "xtfn".

Example 2

Input: words = ["a","ab","abc","d","cd","bcd","abcd"]

Output: 4
Explanation: The two words can be "ab", "cd".

Example 3

Input: words = ["a","aa","aaa","aaaa"]

Output: 0
Explanation: No such pair of words.

Constraints

• 2 <= words.length <= 1000.

• 1 <= words[i].length <= 1000.
• words[i] consists only of lowercase English letters.

8.4.2 Solution 1: Bruteforce

For each words[i], for all words[j] with j > i, check if they do not share common
letters and compute the product of their lengths.

154 Chapter 8. Bit Manipulation

Code

#include <vector>
#include <iostream>
using namespace std;
int maxProduct(const vector<string>& words) {
int maxP = 0;
for (int i = 0; i < words.size(); i++) {
// visited marks all letters that appear in words[i]
// words[i] consists of only 26 lowercase English letters
vector<bool> visited(26, false);
for (auto& c : words[i]) {
// map 'a'->0, 'b'->1, .., 'z'->25
visited[c - 'a'] = true;
}
// compare with all other words[j]
for (int j = i + 1; j < words.size(); j++) {
bool found = false;
for (auto& c : words[j]) {
if (visited[c - 'a']) {
// this words[j] has common letter with words[i]
found = true;
break;
}
}
// if words[j] disjoints words[i]
if (!found) {
// compute and update max product of their lengths
maxP = max(maxP, (int) (words[i].length() * words[j].
˓→length()));

}
}
}
return maxP;
}
int main() {
vector<string> words{"abcw","baz","foo","bar","xtfn","abcdef"};
cout << maxProduct(words) << endl;
words = {"a","ab","abc","d","cd","bcd","abcd"};
cout << maxProduct(words) << endl;
words = {"a","aa","aaa","aaaa"};
(continues on next page)

8.4. Maximum Product of Word Lengths 155

 (continued from previous page)
cout << maxProduct(words) << endl;
}

Output:
16
4
0

This solution checks for common characters between pairs of words to determine
their product of lengths.
It iterates through each pair of words in the input vector words, maintaining a boolean
array visited to mark the presence of characters in each word. By comparing the
characters of each pair of words, it identifies whether there are any common charac-
ters. If no common characters are found, it computes the product of the lengths of
the two words and updates the maximum product accordingly.
This approach optimizes the computation of the maximum product by efficiently
checking for common characters between pairs of words without requiring additional
space proportional to the length of the words.

Complexity

• Runtime: O(n^2 * m).

• Extra space: O(1).

8.4.3 Solution 2: Checking common letters using bit masking

You can map a words[i] to the bit representation of an integer n by their characters
like the following:
• If the word words[i] contains the letter 'a', the bit at position 0 of n is 1.
• If the word words[i] contains the letter 'b', the bit at position 1 of n is 1.
• ...
• If the word words[i] contains the letter 'z', the bit at position 25 of n is 1.
Then to check if two words have common letters, you just perform the bitwise opera-
tor AND on them.

156 Chapter 8. Bit Manipulation

Example 1:

• The word "abcw" is mapped to 00010000000000000000000111.

• The word "baz" is mapped to 10000000000000000000000011.
• "abcw" & "baz" = 00000000000000000000000011. This value is not zero, which
means they have common letters.
This technique is called bit masking.

Code

#include <vector>
#include <iostream>
using namespace std;
int maxProduct(const vector<string>& words) {
int maxP = 0;
// initialize all elements of mask to 0
vector<int> mask(words.size());
for (int i = 0; i < words.size(); i++) {
// mark all characters of word[i]
for (auto& c : words[i]) {
// map 'a'->0, 'b'->1, .., 'z'->25
// set the bit at that mapped position of mask[i] to 1
mask[i] |= 1 << (c - 'a');
}
for (int j = 0; j < i; j++) {
if ((mask[j] & mask[i]) == 0) {
// there is no common bit between mask[j] and mask[i]
maxP = max(maxP, (int) (words[i].length() * words[j].
˓→length()));

}
}
}
return maxP;
}
int main() {
vector<string> words{"abcw","baz","foo","bar","xtfn","abcdef"};
cout << maxProduct(words) << endl;
words = {"a","ab","abc","d","cd","bcd","abcd"};
cout << maxProduct(words) << endl;
(continues on next page)

8.4. Maximum Product of Word Lengths 157

 (continued from previous page)
words = {"a","aa","aaa","aaaa"};
cout << maxProduct(words) << endl;
}

Output:
16
4
0

This solution represents each word in the input vector words as a bitmask, where each
bit represents the presence or absence of a character in the word.
By iterating through the words and constructing their corresponding bitmasks, it en-
codes the character information. Then, by comparing the bitmasks of pairs of words,
it identifies whether there are any common characters between them. If no common
characters are found, it computes the product of the lengths of the two words and
updates the maximum product accordingly.
This approach optimizes the computation of the maximum product by using bitwise
operations to efficiently check for common characters between pairs of words without
requiring additional space proportional to the length of the words.

Complexity

• Runtime: O(n^2), where n is the number of words in the input vector.

• Extra space: O(n).

8.4.4 Tips

Utilizing bit manipulation techniques, such as bitmasking, can significantly optimize

the comparison process for determining common characters between words.
Solution 2 reduces the time complexity compared to brute force methods, particularly
when dealing with large datasets, as it avoids nested loops and unnecessary character
comparisons.

158 Chapter 8. Bit Manipulation

 CHAPTER

NINE

SORTING

The arrangement of the elements in an array can hold the key to improved efficiency
and a deeper understanding of your code, which is explored in this chapter as it delves
into the usage of sorting algorithms.
Sorting is similar to putting puzzle pieces to reveal the overall structure. Rearrang-
ing elements makes it possible to retrieve data more quickly, conduct searches more
quickly, and even discover patterns and linkages that might otherwise go unnoticed.
What this chapter covers:
1. Introduction to Sorting: Establish a strong foundation by understanding the
significance of sorting, its impact on algorithmic performance, and the role of
ordering in data analysis.
2. Stability and Uniqueness: Learn about the concepts of stability and uniqueness
in sorting and how they can impact the integrity and usefulness of sorted data.
3. Insights through Sorting: Discover scenarios where sorted data provides valu-
able insights, such as identifying trends, anomalies, or patterns that inform
decision-making.

9.1 Majority Element

9.1.1 Problem statement

1
You’re given an array nums with a total of n elements. Your task is to find and return
the majority element.
1 https://leetcode.com/problems/majority-element/

159
The majority element is the element that appears more frequently in the array than
any other element, specifically, it appears more than n / 2 times.
You can assume that the majority element always exists in the given array.

Example 1

Input: nums = [3,2,3]

Output: 3

Example 2

Input: nums = [2,2,1,1,1,2,2]

Output: 2

Constraints

• n == nums.length.
• 1 <= n <= 5 * 10^4.
• -2^31 <= nums[i] <= 2^31 - 1.

Follow-up:

Could you solve the problem in linear time and in O(1) space?

9.1.2 Solution 1: Counting the frequency

Code

#include <vector>
#include <iostream>
#include <unordered_map>
using namespace std;
int majorityElement(const vector<int>& nums) {
unordered_map<int,int> freq;
(continues on next page)

160 Chapter 9. Sorting

 (continued from previous page)
const int HALF = nums.size() / 2;
for (auto& a : nums) {
// count a's occurrences
freq[a]++;
if (freq[a] > HALF) {
return a;
}
}
return nums[0];
}
int main() {
vector<int> nums{3,2,3};
cout << majorityElement(nums) << endl;
nums = {2,2,1,1,1,2,2};
cout << majorityElement(nums) << endl;
}

Output:
3
2

The code effectively counts the occurrences of each integer in the array and checks if
any integer appears more than n/2 times. If so, it returns that integer as the majority
element; otherwise, it defaults to the first element of the array.

Complexity

• Runtime: O(n), where n = nums.length.

• Extra space: O(n).

9.1.3 Solution 2: Sorting and picking the middle element

Code

#include <vector>
#include <iostream>
#include <algorithm>
using namespace std;
(continues on next page)

9.1. Majority Element 161

 (continued from previous page)
int majorityElement(vector<int>& nums) {
sort(nums.begin(), nums.end());
return nums[nums.size()/2];
}
int main() {
vector<int> nums{3,2,3};
cout << majorityElement(nums) << endl;
nums = {2,2,1,1,1,2,2};
cout << majorityElement(nums) << endl;
}

Output:
3
2

This code leverages the property of a majority element, which guarantees that it oc-
cupies the middle position in the sorted list of elements. Sorting the array allows us
to easily access this middle element.

Complexity

• Runtime: O(n*logn), where n = nums.length.

• Extra space: O(1).

9.1.4 Solution 3: Partial sort

Since you are interested in only the middle element after sorting, the partial sorting
algorithm std::nth_element can be used in this case to reduce the cost of the full
sorting.

Code

#include <vector>
#include <iostream>
#include <algorithm>
using namespace std;
int majorityElement(vector<int>& nums) {
(continues on next page)

162 Chapter 9. Sorting

 (continued from previous page)
const int mid = nums.size() / 2;
// rearrange nums such that all elements less than or equal␣
˓→nums[mid]

// are placed before nums[mid]

nth_element(nums.begin(), nums.begin() + mid, nums.end());
return nums[mid];
}
int main() {
vector<int> nums{3,2,3};
cout << majorityElement(nums) << endl;
nums = {2,2,1,1,1,2,2};
cout << majorityElement(nums) << endl;
}

Output:
3
2

The code uses the std::nth_element function to rearrange the elements in the nums
vector such that the element at index mid will be in its correct sorted position, and all
elements before it will be less than or equal to it, while all elements after it will be
greater than or equal to it.

Complexity

• Runtime: O(n), where n = nums.length.

• Extra space: O(1).

9.1.5 Modern C++ tips

In the code of Solution 3, the partial sorting algorithm std::nth_element will make
sure for all indices i and j that satisfy 0 <= i <= mid <= j < nums.length,

nums[i] <= nums[mid] <= nums[j].

In other words, nums[mid] divides the array nums into two groups: all elements that
are less than or equal to nums[mid] and the ones that are greater than or equal to
nums[mid].
Those two groups are unsorted. That is why the algorithm is called partial sorting.

9.1. Majority Element 163

9.1.6 Exercise

• Most Frequent Even Element2 .

9.2 Merge Sorted Array

9.2.1 Problem statement

1
You’re given two integer arrays, nums1 and nums2, both sorted in non-decreasing or-
der. Additionally, you have two integers, m and n, representing the number of elements
in nums1 and nums2, respectively.
Your task is to merge the elements from nums2 into nums1 in a way that the resulting
array is sorted in non-decreasing order.
However, the sorted array should not be returned as a separate result. Instead, the
merged elements should be stored inside the nums1 array. Here’s the setup for that
purpose:
• nums1 has a total length of m + n, where the first m elements represent the
elements that should be merged, and the last n elements are initialized to 0 and
should be ignored.
• The nums2 array has a length of n, representing the elements to be merged from
nums2 into the final nums1 array.

Example 1

Input: nums1 = [1,2,3,0,0,0], m = 3, nums2 = [2,5,6], n = 3

Output: [1,2,2,3,5,6]
Explanation: The arrays we are merging are [1,2,3] and [2,5,6].
The result of the merge is [1,2,2,3,5,6] with the underlined elements␣
˓→coming from nums1.

2 https://leetcode.com/problems/most-frequent-even-element/
1 https://leetcode.com/problems/merge-sorted-array/

164 Chapter 9. Sorting

Example 2

Input: nums1 = [1], m = 1, nums2 = [], n = 0

Output: [1]
Explanation: The arrays we are merging are [1] and [].
The result of the merge is [1].

Example 3

Input: nums1 = [0], m = 0, nums2 = [1], n = 1

Output: [1]
Explanation: The arrays we are merging are [] and [1].
The result of the merge is [1].
Note that because m = 0, there are no elements in nums1. The 0 is only␣
˓→there to ensure the merge result can fit in nums1.

Constraints

• nums1.length == m + n.
• nums2.length == n.
• 0 <= m, n <= 200.
• 1 <= m + n <= 200.
• -10^9 <= nums1[i], nums2[j] <= 10^9.

Follow up

• Can you come up with an algorithm that runs in O(m + n) time?

9.2.2 Solution 1: Store the result in a new container

Code

9.2. Merge Sorted Array 165

#include <iostream>
#include <vector>
using namespace std;
void merge(vector<int>& nums1, int m, vector<int>& nums2, int n)
{
vector<int> result;
int i = 0;
int j = 0;
while (i < m || j < n) {
if (j == n) {
// nums2 is done, only nums1 still runs
result.push_back(nums1[i++]);
} else if (i == m) {
// nums1 is done, only nums2 still runs
result.push_back(nums2[j++]);
} else if (nums1[i] < nums2[j]) {
result.push_back(nums1[i++]);
} else {
result.push_back(nums2[j++]);
}
}
nums1.swap(result);
}
void printResult(const vector<int>& nums1) {
cout << "[";
for (auto& n : nums1) {
cout << n << ",";
}
cout << "]\n";
}
int main() {
vector<int> nums1 = {1,2,3,0,0,0};
vector<int> nums2 = {2,5,6};
merge(nums1, 3, nums2, 3);
printResult(nums1);
nums1 = {1};
nums2 = {};
merge(nums1, 1, nums2, 0);
printResult(nums1);
nums1 = {0};
nums2 = {1};
(continues on next page)

166 Chapter 9. Sorting

 (continued from previous page)
merge(nums1, 0, nums2, 1);
printResult(nums1);
}

Output:
[1,2,2,3,5,6,]
[1,]
[1,]

This solution merges two sorted arrays nums1 and nums2 into nums1 while maintaining
sorted order. It iterates through both arrays, comparing elements and adding them to
a temporary result vector. After the merging is complete, it replaces the contents of
nums1 with the merged result.

Complexity

• Runtime: O(m+n), where m = nums1.length and n = nums2.length.

• Extra space: O(m+n).

9.2.3 Solution 2: Reassigning nums1 backward

Code

#include <iostream>
#include <vector>
using namespace std;
void merge(vector<int>& nums1, int m, vector<int>& nums2, int n)
{
int k = m + n - 1;
int i = m - 1;
int j = n - 1;
while (k >= 0) {
if (j < 0) {
// nums2 is done
nums1[k--] = nums1[i--];
} else if (i < 0) {
// nums1 is done
nums1[k--] = nums2[j--];
(continues on next page)

9.2. Merge Sorted Array 167

 (continued from previous page)
} else if (nums1[i] > nums2[j]) {
nums1[k--] = nums1[i--];
} else {
nums1[k--] = nums2[j--];
}
}
}
void printResult(const vector<int>& nums1) {
cout << "[";
for (auto& n : nums1) {
cout << n << ",";
}
cout << "]\n";
}
int main() {
vector<int> nums1 = {1,2,3,0,0,0};
vector<int> nums2 = {2,5,6};
merge(nums1, 3, nums2, 3);
printResult(nums1);
nums1 = {1};
nums2 = {};
merge(nums1, 1, nums2, 0);
printResult(nums1);
nums1 = {0};
nums2 = {1};
merge(nums1, 0, nums2, 1);
printResult(nums1);
}

Output:
[1,2,2,3,5,6,]
[1,]
[1,]

168 Chapter 9. Sorting

Complexity

• Runtime: O(m+n), where m = nums1.length and n = nums2.length.

• Extra space: O(1).

9.2.4 Conclusion

Solution 2 efficiently merges two sorted arrays, nums1 and nums2, into nums1 while
preserving the sorted order. It uses three pointers (k, i, and j) to perform the merge
in reverse order, which helps avoid the need for additional space.

9.2.5 Exercise

• Squares of a Sorted Array2 .

9.3 Remove Covered Intervals

9.3.1 Problem statement

1
You’re given an array called intervals, where each element intervals[i] is a pair
[li, ri] representing a half-open interval [li, ri).
Your task is to remove all intervals from the list that are completely covered by another
interval. An interval [a, b) is considered covered by the interval [c, d) if and only
if c is less than or equal to a, and b is less than or equal to d.
After removing the covered intervals, you need to return the number of remaining
intervals.
2 https://leetcode.com/problems/squares-of-a-sorted-array/
1 https://leetcode.com/problems/remove-covered-intervals/

9.3. Remove Covered Intervals 169

Example 1

Input: intervals = [[1,4],[3,6],[2,8]]

Output: 2
Explanation: Interval [3,6] is covered by [2,8], therefore it is removed.

Example 2

Input: intervals = [[1,4],[2,3]]

Output: 1

Constraints

• 1 <= intervals.length <= 1000.

• intervals[i].length == 2.
• 0 <= li <= ri <= 10^5.
• All the given intervals are unique.

9.3.2 Solution 1: Bruteforce

For each interval i, find if any other interval j such that j covers i or i covers j then
remove the smaller one from intervals.

Example 1

For intervals = [[1,4],[3,6],[2,8]].

• With interval i = [1,4], there is no other interval j such that covers i or j
covers i.
• With interval i = [3,6], there is interval j = [2,8] convering i. Remove [3,6]
from intervals.
Final intervals = [[1,4],[2,8]].

170 Chapter 9. Sorting

Code

#include <vector>
#include <iostream>
using namespace std;
//! @return true if the interval i is covered by j
inline bool isCovered(const vector<int>& i, const vector<int>& j) {
return j[0] <= i[0] && i[1] <= j[1];
}
int removeCoveredIntervals(vector<vector<int>>& intervals) {
int i = 0;
while (i < intervals.size() - 1) {
int j = i + 1;
bool erase_i = false;
while (j < intervals.size()) {
if (isCovered(intervals[i], intervals[j])) {
// remove intervals[i] from intervals
intervals.erase(intervals.begin() + i);
erase_i = true;
break;
} else if (isCovered(intervals[j], intervals[i])) {
// remove intervals[j] from intervals
intervals.erase(intervals.begin() + j);
} else {
j++;
}
}
if (!erase_i) {
i++;
}
}
return intervals.size();
}
int main() {
vector<vector<int>> intervals{{1,4},{3,6},{2,8}};
cout << removeCoveredIntervals(intervals) << endl;
intervals = {{1,4},{2,3}};
cout << removeCoveredIntervals(intervals) << endl;
}

Output:
(continues on next page)

9.3. Remove Covered Intervals 171

 (continued from previous page)
2
1

This solution effectively removes covered intervals and retains only those that do not
have others covering them. The time complexity of this solution is O(N^3), where N is
the number of intervals, as it involves nested loops and potential removal of intervals
from the list.

Complexity

• Runtime: O(N^3), where N = intervals.length.

• Extra space: O(1).

9.3.3 Solution 2: Using dictionary order

You might know how to look up words in a dictionary.

The word apple appears before candy in the dictionary because the starting letter a
of apple appears before c of candy in the English alphabet.
And apple appears after animal since the next letter p appears after n.
The C++ Standard Library uses that dictionary order to compare two std::vectors.

Example 1

Rewriting intervals = [[1,4],[3,6],[2,8]] in dictionary order you get intervals

= [[1,4],[2,8],[3,6]]. In this order, the left bounds of the intervals are sorted
first.
If intervals is sorted like that, you can avoid bruteforce in Solution 1 by a simpler
algorithm.
Check if each interval i covers or is covered by some of the previous ones.
Remember that the left bound of interval i is always bigger than or equal to all left
bounds of the previous ones. So,
1. i is covered by some previous interval if the right bound of i is less than some
of the right bounds before.
2. Otherwise i can only cover its exact previous one that has the same left bound.

172 Chapter 9. Sorting

Code

#include <vector>
#include <iostream>
#include <algorithm>
using namespace std;
int removeCoveredIntervals(vector<vector<int>>& intervals) {
// sort the intervals using dictionary order
sort(intervals.begin(), intervals.end());
// count the intervals to be removed
int count = 0;
// keep track max right bound of all previous intervals
int maxRight = -1;
// log the left bound of the previous interval
int preLeft = -1;
for (auto& i : intervals) {
if (i[1] <= maxRight) {
// i's right bound is less than some previous one's
count++;
} else if (i[0] == preLeft) {
// i's left bound is the same as exact previous one's
count++;
} else {
// update previous interval's left bound
preLeft = i[0];
}
// update max right bound
maxRight = max(maxRight, i[1]);
}
return intervals.size() - count;
}
int main() {
vector<vector<int>> intervals{{1,4},{3,6},{2,8}};
cout << removeCoveredIntervals(intervals) << endl;
intervals = {{1,4},{2,3}};
cout << removeCoveredIntervals(intervals) << endl;
}

Output:
2
1

9.3. Remove Covered Intervals 173

Complexity

• Runtime: O(N*logN), where N = intervals.length.

• Extra space: O(1).

9.3.4 Key takeaway

• Two std::vectors can be compared using dictionary order.

• Solution 2 first sorts the intervals and then iterates through them while keeping
track of whether each interval is covered by others or not. The final result is the
count of intervals that are not covered.

9.4 My Calendar I

9.4.1 Problem statement

1
You’re creating a program to use as your calendar. You can add new events to the
calendar, but only if adding the event will not lead to a double booking.
A double booking occurs when two events have some time overlap, meaning there’s
a shared time period between them.
An event is represented as a pair of integers: start and end, which represent the
booking on a half-open interval [start, end). This interval includes all real numbers
x such that start <= x < end.
You need to implement the MyCalendar class, which has the following functions:
1. MyCalendar(): Initializes the calendar object.
2. boolean book(int start, int end): This function checks if the event with
the given start and end can be added to the calendar without causing a double
booking. If it’s possible to add the event without a double booking, the function
returns true. Otherwise, it returns false, and the event is not added to the
calendar.
1 https://leetcode.com/problems/my-calendar-i/

174 Chapter 9. Sorting

Example 1

Input
["MyCalendar", "book", "book", "book"]
[[], [10, 20], [15, 25], [20, 30]]
Output
[null, true, false, true]

Explanation
MyCalendar myCalendar = new MyCalendar();
myCalendar.book(10, 20); // return True
myCalendar.book(15, 25); // return False. It can not be booked because␣
˓→time 15 is already booked by another event.

myCalendar.book(20, 30); // return True, The event can be booked, as the␣

˓→first event takes every time less than 20, but not including 20.

Constraints

• 0 <= start < end <= 10^9.

• At most 1000 calls will be made to book.

9.4.2 Solution 1: Vector

You can store the booked events in a vector and check the intersection condition
whenever you add a new event.

Code

#include <iostream>
#include <vector>
using namespace std;
class MyCalendar {
private:
vector<pair<int,int>> _events;
public:
MyCalendar() {}
bool book(int start, int end) {
for (auto& e : _events) {
(continues on next page)

9.4. My Calendar I 175

 (continued from previous page)
// check for overlap
if (!(e.second <= start || end <= e.first)) {
return false;
}
}
_events.push_back({start, end});
return true;
}
};
int main() {
MyCalendar c;
std::cout << c.book(10, 20) << std::endl;
std::cout << c.book(15, 25) << std::endl;
std::cout << c.book(20, 30) << std::endl;
}

Output:
1
0
1

This code essentially maintains a list of events and checks for overlaps when booking
a new event. If no overlaps are found, it adds the new event to the list and allows the
booking.

Complexity

For the book method:

• Runtime: O(n), where n= _events.length.
• Extra space: O(1).

176 Chapter 9. Sorting

9.4.3 Solution 2: Set

Since the events have no intersection, they can be sorted. You can also consider two
events to be the same if they intersect.
With that in mind, you can use std::set2 to store the sorted unique events.

Code

#include <iostream>
#include <set>
using namespace std;
using Event = pair<int,int>;
struct EventCmp {
bool operator()(const Event& lhs, const Event& rhs) const {
return lhs.second <= rhs.first;
}
};
class MyCalendar {
private:
// declare a set with custom comparison operator
set<Event, EventCmp> _events;
public:
MyCalendar() {}
bool book(int start, int end) {
auto result = _events.insert({start, end});
// result.second stores a bool indicating
// if the insertion was actually performed
return result.second;
}
};
int main() {
MyCalendar c;
std::cout << c.book(10, 20) << std::endl;
std::cout << c.book(15, 25) << std::endl;
std::cout << c.book(20, 30) << std::endl;
}

2 https://en.cppreference.com/w/cpp/container/set

9.4. My Calendar I 177

Output:
1
0
1

Complexity

For the book method:

• Runtime: O(logn), where n = _events.length.
• Extra space: O(1).

9.4.4 Key Takeaway

Solution 2 efficiently handles event bookings by maintaining a sorted set of events

based on their end times, allowing for quick overlap checks when booking new events.

9.4.5 Exercise

• Determine if Two Events Have Conflict3 .

9.5 Remove Duplicates from Sorted Array II

9.5.1 Problem statement

1
Given an integer array nums already sorted in non-decreasing order, you must remove
duplicates so that each unique element appears at most twice. The relative order of
the elements should remain unchanged.
Since changing the array’s length in some programming languages is impossible, you
must place the result in the first part of the nums array. In other words, if there are
k elements after removing the duplicates, the first k elements of nums should contain
the final result. Anything beyond the first k elements is not important.
3 https://leetcode.com/problems/determine-if-two-events-have-conflict/
1 https://leetcode.com/problems/remove-duplicates-from-sorted-array-ii/

178 Chapter 9. Sorting

You should return the value of k after placing the final result in the first k slots of the
nums array.
The key requirement is to accomplish this task without using extra space for another
array. It must be done by modifying the input array nums in-place, using only O(1)
extra memory.

Example 1

Input: nums = [1,1,1,2,2,3]

Output: 5, nums = [1,1,2,2,3,_]
Explanation: Your function should return k = 5, with the first five␣
˓→elements of nums being 1, 1, 2, 2, and 3, respectively.

What you leave does not matter beyond the returned k (hence, they are␣
˓→underscores).

Example 2

Input: nums = [0,0,1,1,1,1,2,3,3]

Output: 7, nums = [0,0,1,1,2,3,3,_,_]
Explanation: Your function should return k = 7, with the first seven␣
˓→elements of nums being 0, 0, 1, 1, 2, 3, and 3, respectively.

What you leave does not matter beyond the returned k (hence, they are␣
˓→underscores).

Constraints

• 1 <= nums.length <= 3 * 10^4.

• -10^4 <= nums[i] <= 10^4.
• nums is sorted in non-decreasing order.

9.5. Remove Duplicates from Sorted Array II 179

9.5.2 Solution 1: Erasing the duplicates

In order for each unique element to appear at most twice, you have to erase the
further appearances if they exist.
Since the array nums is sorted, you can determine that existence by checking if nums[i]
== nums[i-2] for 2 <= i < nums.length.

Code

#include <vector>
#include <iostream>
using namespace std;
int removeDuplicates(vector<int>& nums) {
int i = 2;
while (i < nums.size()) {
// find the element appearing more than twice
if (nums[i] == nums[i-2]) {
int j = i;
// find all duplicates
while (j < nums.size() && nums[j] == nums[i]) {
j++;
}
// keep nums[i-2] and nums[i-1] remove all later duplicates
nums.erase(nums.begin() + i, nums.begin() + j);
} else {
i++;
}
}
return nums.size();
}
void printResult(const int k, const vector<int>& nums) {
cout << k << ", [";
for (int i = 0; i < k ; i++) {
cout << nums[i] << ",";
}
cout << "]\n";
}
int main() {
vector<int> nums{1,1,1,2,2,3};
printResult(removeDuplicates(nums), nums);
(continues on next page)

180 Chapter 9. Sorting

 (continued from previous page)
nums = {0,0,1,1,1,1,2,3,3};
printResult(removeDuplicates(nums), nums);
}

Output:
5, [1,1,2,2,3,]
7, [0,0,1,1,2,3,3,]

This solution efficiently removes duplicates from the sorted array by checking for
duplicates and erasing the excess occurrences while preserving two instances of each
unique element. It then returns the length of the modified array.

Complexity

• Runtime:
– Worst case: O(N^2/3), where N = nums.size(). The complexity of the
erase()2 method is linear in N. The worst case is when erase() is called
maximum N/3 times.

Example of the worst case:

nums = [1,1,1,2,2,2,3,3,3,4,4,4,5,5,5,6,6,6].

– On average: O(N) since the number of erase() calls is O(1).

• Extra space: O(1).

9.5.3 Solution 2: Reassigning the satisfying elements

You might need to avoid the erase() method in the solution above to reduce the
complexity. Moreover, after removing the duplicates, the problem only cares about
the first k elements of the array nums.
If you look at the final result after removing duplication, the expected nums satisfies

nums[i] > nums[i-2] for 2 <= i < nums.length.

You can use this invariant to reassign the array nums only the satisfied elements.
2 https://en.cppreference.com/w/cpp/container/vector/erase

9.5. Remove Duplicates from Sorted Array II 181

Code

#include <vector>
#include <iostream>
using namespace std;
int removeDuplicates(vector<int>& nums) {
if (nums.size() <= 2) {
return nums.size();
}
int k = 2;
int i = 2;
while (i < nums.size()) {
if (nums[i] > nums[k - 2]) {
// make sure nums[k] != nums[k-2]
nums[k++] = nums[i];
}
i++;
}
return k;
}
void printResult(const int k, const vector<int>& nums) {
cout << k << ", [";
for (int i = 0; i < k ; i++) {
cout << nums[i] << ",";
}
cout << "]\n";
}
int main() {
vector<int> nums{1,1,1,2,2,3};
printResult(removeDuplicates(nums), nums);
nums = {0,0,1,1,1,1,2,3,3};
printResult(removeDuplicates(nums), nums);
}

Output:
Output:
5, [1,1,2,2,3,]
7, [0,0,1,1,2,3,3,]

182 Chapter 9. Sorting

Complexity

• Runtime: O(N), where N = nums.size().

• Extra space: O(1).

9.5.4 Conclusion

Solution 2 effectively modifies the input array in-place, removing duplicates that oc-
cur more than twice while maintaining the desired order of unique elements. It does
so in a single pass through the array, resulting in a time complexity of O(N), where N
is the number of elements in the array.

9.5.5 Exercise

• Remove Duplicates from Sorted Array3 .

3 https://leetcode.com/problems/remove-duplicates-from-sorted-array/

9.5. Remove Duplicates from Sorted Array II 183

184 Chapter 9. Sorting
 CHAPTER

TEN

GREEDY ALGORITHM

This chapter will explore a fascinating and highly practical problem-solving approach
known as greedy algorithms. Greedy algorithms are powerful tools for making de-
cisions at each step of an optimization problem, often leading to efficient and near-
optimal solutions.
In this chapter, we’ll dive deep into the world of greedy algorithms, learning how to
apply them to a wide range of real-world scenarios. Here’s what you can look forward
to:
1. Understanding Greedy Algorithms: We’ll begin by establishing a solid foun-
dation in greedy algorithms. You’ll understand this approach’s key principles,
advantages, and limitations.
2. The Greedy Choice Property: Discover the core characteristic of greedy algo-
rithms—the greedy choice property. Learn how it guides us in making locally
optimal decisions at each step.
3. Greedy for Searching: Greedy algorithms can also be applied to search prob-
lems. We’ll delve into graph traversal algorithms and heuristic search methods.
4. Exercises and Problems: Reinforce your understanding of greedy algorithms
with exercises and LeetCode problems covering a wide range of greedy-based
challenges. Practice is essential for mastering this problem-solving technique.
By the end of this chapter, you’ll have a comprehensive understanding of greedy algo-
rithms and the ability to apply them to a wide range of problems, from optimization
to search. Greedy algorithms are valuable tools in your problem-solving toolkit, and
this chapter will equip you with the skills needed to confidently tackle complex opti-
mization challenges. So, let’s dive in and explore the world of greedy algorithms!

185
10.1 Can Place Flowers

10.1.1 Problem statement

1
You are presented with a long flowerbed containing plots, some of which are planted
with flowers (denoted by 1) and some are empty (denoted by 0). Flowers cannot be
planted in adjacent plots. You are given an integer array flowerbed representing the
layout of the flowerbed, and an integer n representing the number of new flowers you
want to plant.
Your task is to determine if it is possible to plant n new flowers in the flowerbed with-
out violating the rule of no-adjacent-flowers. If it is possible, return true; otherwise,
return false.

Example 1

Input: flowerbed = [1,0,0,0,1], n = 1

Output: true

Example 2

Input: flowerbed = [1,0,0,0,1], n = 2

Output: false

Constraints

• 1 <= flowerbed.length <= 2 * 10^4.

• flowerbed[i] is 0 or 1.
• There are no two adjacent flowers in flowerbed.
• 0 <= n <= flowerbed.length.
1 https://leetcode.com/problems/can-place-flowers/

186 Chapter 10. Greedy Algorithm

10.1.2 Solution: Check the no-adjacent-flowers rule

A new flower can be planted at position i only if

flowerbed[i - 1] == 0 && flowerbed[i] == 0 && flowerbed[i + 1] == 0.

If the condition is satisfied, the flower can be planted at position i. flowerbed[i] is

now assigned to 1. Then you can skip checking the rule for the position i + 1 and go
directly to position i + 2.

Code

#include <iostream>
#include <vector>
using namespace std;
bool canPlaceFlowers(vector<int>& flowerbed, int n) {
if (n == 0) {
return true;
}
flowerbed.insert(flowerbed.begin(), 0);
flowerbed.push_back(0);
int i = 1;
while (i < flowerbed.size() - 1) {
if (flowerbed[i - 1] == 0 && flowerbed[i] == 0 && flowerbed[i +␣
˓→1] == 0) {

// plant i if it satisfies the no-adjacent condition

flowerbed[i] = 1;
n--;
i+=2;
} else {
i++;
}
}
return n <= 0; // have planted all n
}
int main() {
vector<int> flowerbed{1,0,0,0,1};
cout << canPlaceFlowers(flowerbed, 1) << endl;
flowerbed = {1,0,0,0,1};
cout << canPlaceFlowers(flowerbed, 2) << endl;
}

10.1. Can Place Flowers 187

Output:
1
0

This solution efficiently iterates through the flowerbed, planting flowers wherever
possible while adhering to the constraints. It returns true if it’s possible to plant all
the required flowers and false otherwise.

Complexity

• Runtime: O(N), where N = flowerbed.length.

• Extra space: O(1).

10.1.3 Implementation tips

• In this implementation, you could insert element 0 to the front and the back
of vector flowerbed to avoid writing extra code for checking the no-adjacent-
flowers rule at i = 0 and i = flowerbed.size() - 1.
• There are a few ways to insert an element to a vector. Here you can see an
example of using the methods insert and push_back of a std::vector.

10.1.4 Ecercise

• Teemo Attacking2 .

10.2 Minimum Deletions to Make Character Frequencies

Unique

10.2.1 Problem statement

1
A string s is considered “good” if there are no two different characters in the string
that have the same frequency, meaning each character appears a unique number of
times.
2 https://leetcode.com/problems/teemo-attacking/
1 https://leetcode.com/problems/minimum-deletions-to-make-character-frequencies-unique/

188 Chapter 10. Greedy Algorithm

You’re given a string s, and your task is to determine the minimum number of char-
acters you need to delete from s to make it a “good” string.
The frequency of a character in a string is the count of times that character appears
in the string. For instance, in the string "aab", the frequency of 'a' is 2, and the
frequency of 'b' is 1.

Example 1

Input: s = "aab"
Output: 0
Explanation: s is already good.

Example 2

Input: s = "aaabbbcc"
Output: 2
Explanation: You can delete two 'b's resulting in the good string "aaabcc
˓→".

Another way is to delete one 'b' and one 'c' resulting in the good␣
˓→string "aaabbc".

Example 3

Input: s = "ceabaacb"
Output: 2
Explanation: You can delete both 'c's resulting in the good string
˓→"eabaab".

Note that we only care about characters that are still in the string at␣
˓→the end (i.e. frequency of 0 is ignored).

10.2. Minimum Deletions to Make Character Frequencies Unique 189

Constraints

• 1 <= s.length <= 10^5.

• s contains only lowercase English letters.

10.2.2 Solution: Delete the frequencies in sorted order

Your goal is to make all the frequencies be different.

One way of doing that is by sorting the frequencies and performing the deletion.

Example 4

For s = "ceaacbb", the frequencies of the characters are: freq['a'] = 2, freq['b']

= 2, freq['c'] = 2 and freq['e'] = 1. They are already in sorted order.
• Let the current frequency be the first frequency freq['a'] = 2.
• The next frequency is freq['b'] = 2, equal to the current frequency. Delete
one appearance to make the current frequency be 1.
• The next frequency is freq['c'] = 2, bigger than the current frequency. Delete
two appearances to make the current frequency be 0.
• Because the current frequency is 0, delete all appearances of the remaining
frequencies, which is freq['e'] = 1.
• In total, there are 4 deletions.

Code

#include <algorithm>
#include <iostream>
#include <vector>
using namespace std;
int minDeletions(string& s) {
// map 'a'->0, 'b'->1, ..,'z'->25
vector<int> freq(26, 0);
for (char& c: s) {
// count the frequency of character c
freq[c - 'a']++;
}
(continues on next page)

190 Chapter 10. Greedy Algorithm

 (continued from previous page)
// sort freq in descending order
sort(freq.begin(), freq.end(), greater<int>());
int deletion = 0;
int currentFreq = freq.at(0); // start with the max frequency
for (int i = 1; i < freq.size() && freq.at(i) > 0; i++) {
if (currentFreq == 0) {
// delete all remaining characters
deletion += freq.at(i);
} else if (freq[i] >= currentFreq) {
// delete just enough to make the freq[i] < currentFreq
deletion += freq.at(i) - currentFreq + 1;
currentFreq--;
} else {
// do not delete on freq[i] < currentFreq
currentFreq = freq.at(i);
}
}
return deletion;
}
int main() {
cout << minDeletions("aab") << endl;
cout << minDeletions("aaabbbcc") << endl;
cout << minDeletions("ceabaacb") << endl;
}

Output:
0
2
2

Complexity

• Runtime: O(N), where N = s.length;

• Extra space: O(1).

10.2. Minimum Deletions to Make Character Frequencies Unique 191

10.2.3 Conclusion

The problem of determining the minimum number of deletions required to make

character frequencies unique can be efficiently solved by counting the frequencies of
characters and iteratively adjusting the frequencies to ensure uniqueness.
This solution achieves this by first counting the frequencies of characters and then
sorting them in descending order. By iteratively processing the sorted frequencies,
the solution ensures that each character frequency is unique while minimizing the
number of deletions required.

10.2.4 Exercise

• Minimum Deletions to Make Array Beautiful2 .

10.3 Wiggle Subsequence

10.3.1 Problem statement

1
A wiggle sequence is a sequence where the differences between successive numbers
strictly alternate between positive and negative. The first difference (if one exists)
may be either positive or negative. A sequence with one element and a sequence with
two non-equal elements are trivially wiggle sequences.
• For example, [1, 7, 4, 9, 2, 5] is a wiggle sequence because the differences
(6, -3, 5, -7, 3) alternate between positive and negative.
• In contrast, [1, 4, 7, 2, 5] and [1, 7, 4, 5, 5] are not wiggle sequences.
The first is not because its first two differences are positive, and the second is
not because its last difference is zero.
A subsequence is obtained by deleting some elements (possibly zero) from the origi-
nal sequence, leaving the remaining elements in their original order.
Given an integer array nums, return the length of the longest wiggle subsequence of
nums.
2 https://leetcode.com/problems/minimum-deletions-to-make-array-beautiful/
1 https://leetcode.com/problems/wiggle-subsequence/

192 Chapter 10. Greedy Algorithm

Example 1

Input: nums = [1,7,4,9,2,5]

Output: 6
Explanation: The entire sequence is a wiggle sequence with differences␣
˓→(6, -3, 5, -7, 3).

Example 2

Input: nums = [1,17,5,10,13,15,10,5,16,8]

Output: 7
Explanation: There are several subsequences that achieve this length.
One is [1, 17, 10, 13, 10, 16, 8] with differences (16, -7, 3, -3, 6, -
˓→8).

Example 3

Input: nums = [1,2,3,4,5,6,7,8,9]

Output: 2

Constraints

• 1 <= nums.length <= 1000.

• 0 <= nums[i] <= 1000.

Follow up

• Could you solve this in O(n) time?

10.3. Wiggle Subsequence 193

10.3.2 Solution: Counting the local extrema of nums

First, if you pick all local extrema (minima and maxima) of nums to form a subse-
quence e, then it is wiggle. Let us call it an extrema subsequence.

Example 2

For nums = [1,17,5,10,13,15,10,5,16,8], the local extrema are [1,17,5,15,5,16,

8]. It is wiggle and called extrema subsequence.
Note that if nums.length = n then nums[0] and nums[n - 1] are always the first and
the last extremum.
Second, given any two successive local extrema a and b, you cannot have any wiggle
subsequence between them. Because the elements between them are either mono-
tonic increasing or decreasing.
That proves the extrema subsequence is the longest wiggle one.

Code

#include <iostream>
#include <vector>
using namespace std;
int wiggleMaxLength(const vector<int>& nums) {
// nums[0] is always the first extremum
// start to find the second extremum
int i = 1;
while (i < nums.size() && nums[i] == nums[i - 1]) {
i++;
}
if (i == nums.size()) {
// all nums[i] are equal
return 1;
}
int sign = nums[i] > nums[i - 1] ? 1 : -1;
int count = 2;
i++;
while (i < nums.size()) {
if ((nums[i] - nums[i - 1]) * sign < 0) {
// nums[i] is an extremum
(continues on next page)

194 Chapter 10. Greedy Algorithm

 (continued from previous page)
count++;
sign = -sign;
}
i++;
}
return count;
}
int main() {
vector<int> nums{1,7,4,9,2,5};
cout << wiggleMaxLength(nums) << endl;
nums = {1,17,5,10,13,15,10,5,16,8};
cout << wiggleMaxLength(nums) << endl;
nums = {1,2,3,4,5,6,7,8,9};
cout << wiggleMaxLength(nums) << endl;
}

Output:
6
7
2

Complexity

• Runtime: O(n), where n = nums.length.

• Extra space: O(1).

10.3.3 Conclusion

The problem of finding the length of the longest wiggle subsequence can be efficiently
solved using a greedy approach. The solution iterates through the input array, identi-
fying alternating extremums (peaks and valleys) to form the wiggle subsequence.
By keeping track of the current trend direction (increasing or decreasing), the solution
efficiently identifies extremums and increments the count accordingly. This greedy
approach ensures that each extremum contributes to increasing the length of the
wiggle subsequence, maximizing its overall length.

10.3. Wiggle Subsequence 195

10.4 Partitioning Into Minimum Number Of Deci-Binary
Numbers

10.4.1 Problem statement

1
A decimal number is called deci-binary if each of its digits is either 0 or 1 without
any leading zeros. For example, 101 and 1100 are deci-binary, while 112 and 3001 are
not.
Given a string n that represents a positive decimal integer, return the minimum num-
ber of positive deci-binary numbers needed so that they sum up to n.

Example 1

Input: n = "32"
Output: 3
Explanation: 10 + 11 + 11 = 32

Example 2

Input: n = "82734"
Output: 8

Example 3

Input: n = "27346209830709182346"
Output: 9

1 https://leetcode.com/problems/partitioning-into-minimum-number-of-deci-binary-numbers/

196 Chapter 10. Greedy Algorithm

Constraints

• 1 <= n.length <= 10^5.

• n consists of only digits.
• n does not contain any leading zeros and represents a positive integer.

10.4.2 Solution: Identify the maximum digit of n

Any digit d can be obtained by summing the digit 1 d times.

The problem turns into identifying the maximum digit of n.

Example 2

For n = "82734" the answer is 8 because:

82734
= 11111
+ 11111
+ 10111
+ 10101
+ 10100
+ 10100
+ 10100
+ 10000

Code

#include <iostream>
using namespace std;
int minPartitions(const string& n) {
char maxDigit = '0';
for (auto& d : n) {
maxDigit = max(maxDigit, d);
}
return maxDigit - '0';
}
int main() {
(continues on next page)

10.4. Partitioning Into Minimum Number Of Deci-Binary Numbers 197

 (continued from previous page)
cout << minPartitions("32") << endl;
cout << minPartitions("82734") << endl;
cout << minPartitions("27346209830709182346") << endl;
}

Output:
3
8
9

Complexity

• Runtime: O(N), where N = n.length.

• Extra space: O(1).

10.4.3 Conclusion

This problem can be efficiently solved by identifying the maximum digit in the string.
Since each deci-binary number can only contain digits from 0 to 9, the maximum digit
determines the minimum number of deci-binary numbers needed.
By finding the maximum digit in the string and converting it to an integer, the solution
effectively determines the minimum number of deci-binary numbers required.

10.5 Maximum Units on a Truck

10.5.1 Problem statement

1
You are assigned to put some amount of boxes onto one truck. You are given a 2D
array boxTypes, where boxTypes[i] = [numberOfBoxes_i, numberOfUnitsPerBox_i]:
• numberOfBoxes_i is the number of boxes of type i.
• numberOfUnitsPerBox_i is the number of units in each box of the type i.
1 https://leetcode.com/problems/maximum-units-on-a-truck/

198 Chapter 10. Greedy Algorithm

You are also given an integer truckSize, which is the maximum number of boxes that
can be put on the truck. You can choose any boxes to put on the truck as long as the
number of boxes does not exceed truckSize.
Return the maximum total number of units that can be put on the truck.

Example 1

Input: boxTypes = [[1,3],[2,2],[3,1]], truckSize = 4

Output: 8
Explanation: There are:
- 1 box of the first type that contains 3 units.
- 2 boxes of the second type that contain 2 units each.
- 3 boxes of the third type that contain 1 unit each.
You can take all the boxes of the first and second types, and one box of␣
˓→the third type.

The total number of units will be = (1 * 3) + (2 * 2) + (1 * 1) = 8.

Example 2

Input: boxTypes = [[5,10],[2,5],[4,7],[3,9]], truckSize = 10

Output: 91
Explanation: (5 * 10) + (3 * 9) + (2 * 7) = 91.

Constraints

• 1 <= boxTypes.length <= 1000.

• 1 <= numberOfBoxes_i, numberOfUnitsPerBox_i <= 1000.
• 1 <= truckSize <= 10^6.

10.5.2 Solution: Greedy algorithm

Put the boxes having more units first.

10.5. Maximum Units on a Truck 199

Code

#include <iostream>
#include <vector>
#include <algorithm>
using namespace std;
int maximumUnits(vector<vector<int>>& boxTypes, int truckSize) {
// sort for the boxes based on their number of units
sort(boxTypes.begin(), boxTypes.end(),
[](const vector<int>& a, const vector<int>& b) {
return a[1] > b[1];
});
int maxUnits = 0;
int i = 0;
while (truckSize > 0 && i < boxTypes.size()) {
if (boxTypes[i][0] <= truckSize) {
// put all boxTypes[i] if there is still room
maxUnits += boxTypes[i][0] * boxTypes[i][1];
truckSize -= boxTypes[i][0];
} else {
// can put only truckSize < boxTypes[i][0] of boxTypes[i]
maxUnits += truckSize * boxTypes[i][1];
break;
}
i++;
}
return maxUnits;
}
int main() {
vector<vector<int>> boxTypes{{1,3},{2,2},{3,1}};
cout << maximumUnits(boxTypes, 4) << endl;
boxTypes = {{5,10},{2,5},{4,7},{3,9}};
cout << maximumUnits(boxTypes, 10) << endl;
}

Output:
8
91

This solution optimally loads boxes onto a truck to maximize the total number of
units that can be transported, considering both the number of boxes available and
their units per box.

200 Chapter 10. Greedy Algorithm

Complexity

• Runtime: O(N*logN), where N = boxTypes.length.

• Extra space: O(1).

10.5.3 Modern C++ STL notes

Note that two vectors2 can be compared. That is why you can sort them.
But in this case you want to sort them based on the number of units. That is why you
need to define the comparison function like the code above. Otherwise, the std::sort3
algorithm will use the dictionary order to sort them by default.

10.5.4 Exercise

• Maximum Bags With Full Capacity of Rocks4 .

2 https://en.cppreference.com/w/cpp/container/vector
3 https://en.cppreference.com/w/cpp/algorithm/sort
4 https://leetcode.com/problems/maximum-bags-with-full-capacity-of-rocks/

10.5. Maximum Units on a Truck 201

202 Chapter 10. Greedy Algorithm
 CHAPTER

ELEVEN

DYNAMIC PROGRAMMING

This chapter explains dynamic programming, a method for solving complex prob-
lems with strategic optimization. Elegant and efficient solutions can be found by
breaking down problems into smaller subproblems and using memorization and re-
cursion. It’s like solving a puzzle by solving smaller pieces and putting them together
to form the larger picture.
What this chapter covers:
1. Introduction to Dynamic Programming: Establish a solid foundation by un-
derstanding the core principles of dynamic programming, its advantages, and
the problems it best suits.
2. Overlapping Subproblems and Optimal Substructure: Delve into the key
concepts that underlie dynamic programming, namely identifying overlapping
subproblems and exploiting optimal substructure.
3. Fibonacci Series and Beyond: Begin with classic examples like solving the
Fibonacci series and gradually progress to more intricate problems that involve
complex optimization.
4. Efficiency and Trade-offs: Understand the trade-offs involved in dynamic pro-
gramming, including the balance between time and space complexity.
5. Problem-Solving Strategies: Develop systematic strategies for approaching dy-
namic programming problems, from identifying subproblems to deriving recur-
rence relations.

203
11.1 Fibonacci Number

11.1.1 Problem statement

1
The Fibonacci numbers make up a sequence denoted as F(n), known as the Fibonacci
sequence. Each number in this sequence is the sum of the two preceding numbers,
with the sequence starting from 0 and 1. In other words:

F(0) = 0, F(1) = 1
F(n) = F(n - 1) + F(n - 2), for n > 1.

Your task is to calculate the value of F(n) given an integer n.

Example 1

Input: n = 2
Output: 1
Explanation: F(2) = F(1) + F(0) = 1 + 0 = 1.

Example 2

Input: n = 3
Output: 2
Explanation: F(3) = F(2) + F(1) = 1 + 1 = 2.

Example 3

Input: n = 4
Output: 3
Explanation: F(4) = F(3) + F(2) = 2 + 1 = 3.

1 https://leetcode.com/problems/fibonacci-number/

204 Chapter 11. Dynamic Programming

Constraints

• 0 <= n <= 30.

11.1.2 Solution 1: Recursive

Code

#include <iostream>
int fib(int n) {
if (n <= 1) {
return n;
}
return fib(n - 1) + fib(n - 2);
}
int main() {
std::cout << fib(2) << std::endl;
std::cout << fib(3) << std::endl;
std::cout << fib(4) << std::endl;
}

Output:
1
2
3

This solution computes the nth Fibonacci number using a recursive approach.

Complexity

The time complexity of this solution is exponential, specifically O(2^n). This is be-
cause it repeatedly makes two recursive calls for each n, resulting in an exponential
number of function calls and calculations. As n grows larger, the execution time in-
creases significantly.
The space complexity of the given recursive Fibonacci solution is O(n). This space
complexity arises from the function call stack when making recursive calls.
When you call the fib function with a value of n, it generates a call stack with a depth
of n, as each call to fib leads to two more recursive calls (one for n - 1 and one for n

11.1. Fibonacci Number 205

- 2) until n reaches the base cases (0 or 1). The space required to store the function
call stack is proportional to the depth of the recursion, which is n.
Therefore, the space complexity is linearly related to the input value n, making it
O(n). This can be a concern for large values of n because it could lead to a stack
overflow if n is too large.
• Runtime: O(2^n).
• Extra space: O(n).

11.1.3 Solution 2: Dynamic programming

#include <iostream>
#include <vector>
int fib(int n) {
if (n <= 1) {
return n;
}
// store all computed Fibonacci numbers
std::vector<int> f(n + 1);
f[0] = 0;
f[1] = 1;
for (int i = 2; i <= n; i++) {
f[i] = f[i - 1] + f[i - 2];
}
return f[n];
}
int main() {
std::cout << fib(2) << std::endl;
std::cout << fib(3) << std::endl;
std::cout << fib(4) << std::endl;
}

Output:
1
2
3

This solution uses dynamic programming to avoid redundant calculations by storing

and reusing previously computed Fibonacci numbers.

206 Chapter 11. Dynamic Programming

Complexity

• Runtime: O(n).
• Extra space: O(n).

11.1.4 Solution 3: Reduce space for dynamic programming

Code

#include <iostream>
int fib(int n) {
if (n <= 1) {
return n;
}
// store only two previous Fibonacci numbers
int f0 = 0;
int f1 = 1;
for (int i = 2; i <= n; i++) {
int f2 = f1 + f0;
// update for next round
f0 = f1;
f1 = f2;
}
return f1;
}
int main() {
std::cout << fib(2) << std::endl;
std::cout << fib(3) << std::endl;
std::cout << fib(4) << std::endl;
}

Output:
1
2
3

This solution calculates the nth Fibonacci number iteratively using two variables to
keep track of the last two Fibonacci numbers.

11.1. Fibonacci Number 207

Complexity

• Runtime: O(n).
• Extra space: O(1).

11.1.5 Conclusion

The Fibonacci sequence can be efficiently computed using various techniques, includ-
ing recursion with memoization, bottom-up dynamic programming, or even optimiz-
ing space usage by storing only the necessary previous Fibonacci numbers.
Solutions 2 and 3 demonstrate dynamic programming approaches, where Fibonacci
numbers are computed iteratively while storing intermediate results to avoid redun-
dant computations.
Solution 3 further optimizes space usage by only storing the necessary previous Fi-
bonacci numbers, resulting in a space complexity of O(1). Understanding these dif-
ferent approaches and their trade-offs is essential for selecting the most appropriate
solution based on the problem constraints and requirements.

11.1.6 Exercise

• N-th Tribonacci Number2 .

11.2 Unique Paths

11.2.1 Problem statement

1
A robot starts at the top-left corner of a grid with dimensions m x n. It can move
either down or right at each step. The robot’s goal is to reach the bottom-right corner
of the grid.
The problem is to determine the number of unique paths the robot can take to reach
the bottom-right corner.
2 https://leetcode.com/problems/n-th-tribonacci-number/
1 https://leetcode.com/problems/unique-paths/

208 Chapter 11. Dynamic Programming

Example 1

Input: m = 3, n = 7
Output: 28

Example 2

Input: m = 3, n = 2
Output: 3
Explanation:
From the top-left corner, there are a total of 3 ways to reach the␣
˓→bottom-right corner:

1. Right -> Down -> Down

2. Down -> Down -> Right
3. Down -> Right -> Down

11.2. Unique Paths 209

Example 3

Input: m = 7, n = 3
Output: 28

Example 4

Input: m = 3, n = 3
Output: 6

Constraints

• 1 <= m, n <= 100.

• It’s guaranteed that the answer will be less than or equal to 2*10^9.

11.2.2 Solution 1: Recursive

At each point, the robot has two ways of moving: right or down. Let P(m,n) is the
wanted result. Then you have a recursive relationship:

P(m,n) = P(m-1, n) + P(m, n-1)

If the grid has only one row or only one column, then there is only one possible path.

P(1, n) = P(m, 1) = 1.

We have a recursive implementation.

Code

#include <iostream>
#include <vector>
using namespace std;
int uniquePaths(int m, int n) {
if (m == 1 || n == 1) {
return 1;
}
(continues on next page)

210 Chapter 11. Dynamic Programming

 (continued from previous page)
return uniquePaths(m - 1, n) + uniquePaths(m, n - 1);
}
int main() {
std::cout << uniquePaths(3,7) << std::endl;
std::cout << uniquePaths(7,3) << std::endl;
std::cout << uniquePaths(3,2) << std::endl;
std::cout << uniquePaths(3,3) << std::endl;
}

Output:
28
28
3
6

This is a recursive solution that breaks down the problem into two subproblems:
• uniquePaths(m-1, n)
• uniquePaths(m, n-1)
Each recursive call reduces either the m or n value by 1.
The base case is when m == 1 or n == 1, where there is only 1 unique path.

Complexity

To calculate the complexity, let’s look at the recursion tree:

• uniquePaths(m,n) calls:
– uniquePaths(m-1,n)
– uniquePaths(m, n-1)
• Each of those calls two more calls and so on.
The height of the tree will be max(m,n). At each level, there are 2 branches.
So the total number of nodes in the tree will be 2^max(m,n).
Since each node represents a function call, the runtime complexity is O(2^max(m,n)).
The space complexity is O(max(m,n)) due to the call stack.
In summary, the complexities are:

11.2. Unique Paths 211

 • Runtime: O(2^max(m,n), where m x n is the size of the grid.
• Extra space: O(max(m,n).

11.2.3 Solution 2: Dynamic programming

The recursive implementation repeats a lot of computations.

For example, uniquePaths(2,2) was recomputed in both uniquePaths(2,3) and
uniquePaths(3,2) when you compute uniquePaths(3,3).
One way of storing what has been computed is by using dynamic programming.

Code

#include <iostream>
#include <vector>
using namespace std;
int uniquePaths(int m, int n) {
// store what have been calculated in dp
vector<vector<int> > dp(m, vector<int>(n,1));
for (int i = 1; i < m; i++) {
for (int j = 1; j < n; j++) {
dp[i][j] = dp[i - 1][j] + dp[i][j - 1];
}
}
return dp[m - 1][n - 1];
}
int main() {
std::cout << uniquePaths(3,7) << std::endl;
std::cout << uniquePaths(7,3) << std::endl;
std::cout << uniquePaths(3,2) << std::endl;
std::cout << uniquePaths(3,3) << std::endl;
}

Output:
28
28
3
6

212 Chapter 11. Dynamic Programming

This solution uses dynamic programming to efficiently compute the number of unique
paths in an m x n grid. It creates a 2D vector dp of size m x n and initializes all its
values to 1 because there’s exactly one way to reach any cell in the first row or first
column.
Then, it iterates through the grid, starting from the second row and second column
(i.e., indices (1, 1)), and for each cell, it calculates the number of unique paths to
reach that cell. This is done by summing the number of paths from the cell above it
and the cell to the left of it, as these are the only two possible directions to reach the
current cell.
Finally, the value at dp[m-1][n-1] contains the total number of unique paths to reach
the bottom-right corner of the grid, which is returned as the result.

Complexity

• Runtime: O(m*n), where m x n is the size of the grid.

• Extra space: O(m*n).

11.2.4 Solution 3: Reduced dynamic programming

You can rephrase the relationship inside the loop like this:
“new value” = “old value” + “previous value”;
Then you do not have to store all values of all rows.

Code

#include <iostream>
#include <vector>
using namespace std;
int uniquePaths(int m, int n) {
// store the number of unique paths for each column in each row
vector<int> dp(n, 1);
for (int i = 1; i < m; i++) {
for (int j = 1; j < n; j++) {
dp[j] += dp[j - 1];
}
}
return dp[n - 1];
(continues on next page)

11.2. Unique Paths 213

 (continued from previous page)
}
int main() {
std::cout << uniquePaths(3,7) << std::endl;
std::cout << uniquePaths(7,3) << std::endl;
std::cout << uniquePaths(3,2) << std::endl;
std::cout << uniquePaths(3,3) << std::endl;
}

Output:
28
28
3
6

Complexity

• Runtime O(m*n), where m x n is the size of the grid.

• Memory O(n).

11.2.5 Conclusion

Solution 3 uses only a 1D vector dp of size n to store the number of unique paths for
each column.
First, it initializes all elements of dp to 1, as there’s exactly one way to reach any cell
in the first row or first column.
Then, it iterates through the grid, starting from the second row and second column
(i.e., indices (1, 1)). For each cell, it updates the value in dp by adding the value
from the cell directly above it and the value from the cell to the left of it. This step
efficiently accumulates the number of unique paths to reach the current cell.
Finally, the value at dp[n-1] contains the total number of unique paths to reach the
bottom-right corner of the grid, which is returned as the result.

214 Chapter 11. Dynamic Programming

A bit of wonder

I am wondering if there is some mathematics behind this problem. Please share your
finding if you find a formula for the solution to this problem.

11.2.6 Exercise

• Minimum Path Sum2 .

11.3 Largest Divisible Subset

11.3.1 Problem statement

1
You have a collection of positive integers called nums, where each integer is distinct.
Your task is to find the largest subset answer from this collection, such that for every
pair of elements (answer[i], answer[j]) within this subset:
• Either answer[i] is a multiple of answer[j] (i.e., answer[i] % answer[j] ==
0), or
• answer[j] is a multiple of answer[i] (i.e., answer[j] % answer[i] == 0).
If there are multiple possible solutions, you can return any of them.

Example 1

Input: nums = [1,2,3]

Output: [1,2]
Explanation: [1,3] is also accepted.

2 https://leetcode.com/problems/minimum-path-sum/
1 https://leetcode.com/problems/largest-divisible-subset/

11.3. Largest Divisible Subset 215

Example 2

Input: nums = [1,2,4,8]

Output: [1,2,4,8]

Constraints

• 1 <= nums.length <= 1000.

• 1 <= nums[i] <= 2 * 10^9.
• All the integers in nums are unique.

11.3.2 Solution 1: Bruteforce with Dynamic programming

Note that the condition a % b == 0 is called a is divisible by b. In mathematics, it can

also be called b divides a and be written as b | a.
The symmetry of the divisibility criteria means it does not count the ordering of the
answer. You could sort the vector nums before trying to find the longest subset answer
= [answer[0], answer[1], ..., answer[m]] where answer[i] | answer[j] with all
0 <= i <= j <= m.
Now assuming the nums were sorted. For each i, you need to find the largest subset
maxSubset[i] starting from nums[i]. And the final answer is the largest one among
them.

Example 3

Input: nums = [2, 4, 3, 9, 8].

Sorted nums = [2, 3, 4, 8, 9].
maxSubset[0] = [2, 4, 8].
maxSubset[1] = [3, 9].
maxSubset[2] = [4, 8].
maxSubset[3] = [8].
maxSubset[4] = [9].
Output: [2, 4, 8].

Note that for a sorted nums, if nums[i] | nums[j] for some i < j, then maxSubset[j]
is a subset of maxSubset[i].

216 Chapter 11. Dynamic Programming

For example, maxSubset[2] is a subset of maxSubset[0] in Example 3 because nums[0]
= 2 | 4 = nums[2].
That might lead to some unnecessary recomputing. To avoid it, you could use dynamic
programming to store the maxSubset[j] you have already computed.

Code

#include <iostream>
#include <vector>
#include <unordered_map>
#include <algorithm>
using namespace std;

//! @return the max divisible subset starting from nums[i]

//! and store it to _map[i]
//! @param nums a sorted array of unique positive integers
vector<int> largestDivisibleSubsetOf(vector<int>& nums,
int i, unordered_map<int, vector<int> >& _map) {
if (_map.find(i) != _map.end()) {
// already computed!
return _map[i];
}
vector<int> maxSubset{nums[i]}; // start with nums[i]
if (i == nums.size() - 1) {
// largest value in nums
_map.insert({i, maxSubset});
return maxSubset;
}
for (int j = i + 1; j < nums.size(); j++) {
if (nums[j] % nums[i] == 0) {
// compute the max divisble subset starting from nums[j]
auto subset = largestDivisibleSubsetOf(nums, j, _map);

// add nums[i] to subset as it might become maxSubset

subset.push_back(nums[i]);
if (maxSubset.size() < subset.size()) {
// update maxSubset
maxSubset = subset;
}
}
(continues on next page)

11.3. Largest Divisible Subset 217

 (continued from previous page)
}
// store what have been calculated in _map
_map.insert({i, maxSubset});
return maxSubset;
}
vector<int> largestDivisibleSubset(vector<int>& nums) {
if (nums.size() <= 1) {
return nums;
}
unordered_map<int, vector<int> > _map;
sort(nums.begin(), nums.end());
vector<int> answer;
for (int i = 0; i < nums.size(); i++) {
auto maxSubset = largestDivisibleSubsetOf(nums, i, _map);
if (answer.size() < maxSubset.size()) {
// update the maximal subset
answer = maxSubset;
}
}
return answer;
}
void printSolution(const vector<int>& result) {
cout << "[";
for (auto& v : result) {
cout << v << ",";
}
cout << "]" << endl;
}
int main() {
vector<int> nums{2,1,3};
auto answer = largestDivisibleSubset(nums);
printSolution(answer);
nums = {1,2,4,8};
answer = largestDivisibleSubset(nums);
printSolution(answer);
}

Output:
[2,1,]
[8,4,2,1,]

218 Chapter 11. Dynamic Programming

This solution uses dynamic programming with memoization to find the largest divisi-
ble subset of a given set of numbers.
The largestDivisibleSubsetOf function recursively computes the largest divisible
subset starting from a particular index i in the sorted array nums. It memoizes the
computed subsets in an unordered map _map to avoid redundant computations. By
iteratively calling largestDivisibleSubsetOf for each index i in the sorted array and
updating the answer with the largest subset found so far, the largestDivisibleSubset
function computes the largest divisible subset of the input array nums.
This approach optimizes the computation by avoiding repeated calculations and lever-
aging dynamic programming techniques to efficiently explore the solution space.

Complexity

• Runtime: O(n^2), where n is the number of elements in the nums vector.

• Extra space: O(n^2).

11.3.3 Solution 2: Store only the representative of the maxSubset

In the brute-force solution above, you used a big map to log all maxSubset[i] though
you need only the largest one at the end.
One way to save memory (and eventually improve performance) is just storing the
representative of the chain relationship between the values nums[i] of the maxSubset
through their indices mapping.
That means if maxSubset[i] = [nums[i0] | nums[i1] | ... | nums[iN1]] |
nums[iN]], you could log pre[iN] = iN1, . . . , prev[i1] = i0.
Then all you need to find is only the last index iN of the largest maxSubset.

Example 3

Input: nums = [2, 4, 3, 9, 8].

sorted nums = [2, 3, 4, 8, 9].
pre[0] = -1 (there is no nums[i] | nums[0]).
pre[1] = -1 (there is no nums[i] | nums[1]).
pre[2] = 0 (nums[0] is the only divisor of nums[2]).
pre[3] = 2 (for the largest subset though nums[0] and nums[2] are both␣
˓→divisors of nums[3]).

(continues on next page)

11.3. Largest Divisible Subset 219

 (continued from previous page)
pre[4] = 1 (nums[1] is the only divisor of nums[4]).
iN = 3 ([2 | 4 | 8] is the largest maxSubset).
Output: [8, 4, 2].

Code

#include <iostream>
#include <vector>
#include <algorithm>
using namespace std;
vector<int> largestDivisibleSubset(vector<int>& nums) {
if (nums.size() <= 1) {
return nums;
}
sort(nums.begin(), nums.end());
// the size of the resulting subset
int maxSize = 0;

// nums[maxindex] is the largest value of the resulting subset

int maxindex = 0;

// subsetSize[i] stores the size of the largest subset

// having the biggest number nums[i]
vector<int> subsetSize(nums.size(), 1);

// pre[i] stores the previous index of i in their largest subset

vector<int> pre(nums.size(), -1);
for (int i = 0; i < nums.size(); i++) {
// find the previous nums[j] that make subsetSize[i] largest
for (int j = i - 1; j >= 0; j--) {

if (nums[i] % nums[j] == 0 &&

subsetSize[j] + 1 > subsetSize[i])
{
subsetSize[i] = subsetSize[j] + 1;
pre[i] = j;
}
}
// update the largest subset
(continues on next page)

220 Chapter 11. Dynamic Programming

 (continued from previous page)
if (maxSize < subsetSize[i]) {
maxSize = subsetSize[i];
maxindex = i;
}
}
vector<int> result;
while (maxindex != -1) {
result.push_back(nums[maxindex]);
maxindex = pre[maxindex];
}
return result;
}
void printSolution(const vector<int>& result) {
cout << "[";
for (auto& v : result) {
cout << v << ",";
}
cout << "]" << endl;
}
int main() {
vector<int> nums{2,1,3};
auto result = largestDivisibleSubset(nums);
printSolution(result);
nums = {1,2,4,8};
result = largestDivisibleSubset(nums);
printSolution(result);
}

Output:
[2,1,]
[8,4,2,1,]

This solution finds the largest divisible subset of a given set of numbers by dynamically
updating the size of the subsets and maintaining the previous index of each element
in their largest subset.
It iterates through the sorted array of numbers, updating the size of the largest subset
that ends with each element by considering the previous elements that are factors of
the current element. By keeping track of the maximum subset size and the index of
the largest element in the subset, it constructs the largest divisible subset.
This approach optimizes the computation by avoiding redundant calculations and

11.3. Largest Divisible Subset 221

leveraging dynamic programming techniques to efficiently explore the solution space.

Complexity

• Runtime: O(n^2), where n is the number of elements in the nums vector. The
nested loop searches for previous elements with divisibility relationships, which
may lead to quadratic time complexity in the worst case. However, it maintains
information about subset sizes and elements, reducing redundant calculations
and improving performance.
• Extra space: O(n).

11.3.4 Key takeaway

In this interesting problem, we use index mapping to simplify everything. That im-
proves the performance in both runtime and memory.

11.4 Triangle

11.4.1 Problem statement

1
You’re provided with a triangle array. Your goal is to find the smallest possible sum
of a path from the top of the triangle to the bottom.
At each step, you have the option to move to an adjacent number in the row below.
Specifically, if you’re at index i in the current row, you can move to either index i or
index i + 1 in the next row.

Example 1

Input: triangle = [[2],[3,4],[6,5,7],[4,1,8,3]]

Output: 11
Explanation: The triangle looks like:
2
3 4
(continues on next page)
1 https://leetcode.com/problems/triangle/

222 Chapter 11. Dynamic Programming

 (continued from previous page)
6 5 7
4 1 8 3
The minimum path sum from top to bottom is 2 + 3 + 5 + 1 = 11␣
˓→(underlined above).

Example 2

Input: triangle = [[-10]]

Output: -10

Constraints

• 1 <= triangle.length <= 200.

• triangle[0].length == 1.
• triangle[i].length == triangle[i - 1].length + 1.
• -10^4 <= triangle[i][j] <= 10^4.

Follow up

• Could you do this using only O(n) extra space, where n is the total number of
rows in the triangle?

11.4.2 Solution 1: Store all minimum paths

You can store all minimum paths at every positions (i,j) so you can compute the
next ones with this relationship.

minPath[i][j] = triangle[i][j] + min(minPath[i - 1][j - 1], minPath[i -␣

˓→1][j]);

11.4. Triangle 223

Code

#include <iostream>
#include <vector>
#include <algorithm>
using namespace std;
int minimumTotal(const vector<vector<int>>& triangle) {
const int n = triangle.size(); // triangle's height
vector<vector<int>> minPath(n);
minPath[0] = triangle[0];
for (int i = 1; i < n; i++) {
const int N = triangle[i].size();
minPath[i].resize(N);
// left most number
minPath[i][0] = triangle[i][0] + minPath[i-1][0];
for (int j = 1; j < N - 1; j++) {
minPath[i][j] = triangle[i][j] + min(minPath[i-1][j-1],␣
˓→minPath[i-1][j]);

}
// right most number
minPath[i][N-1] = triangle[i][N-1] + minPath[i-1][N-2];

}
// pick the min path among the ones (begin -> end)
// go to the bottom (n-1)
return *min_element(minPath[n-1].begin(), minPath[n-1].end());
}
int main() {
vector<vector<int>> triangle{{2},{3,4},{6,5,7},{4,1,8,3}};
cout << minimumTotal(triangle) << endl;
triangle = {{-10}};
cout << minimumTotal(triangle) << endl;
}

Output:
11
-10

This solution finds the minimum path sum from the top to the bottom of a triangle,
represented as a vector of vectors. It uses dynamic programming to calculate the
minimum path sum.

224 Chapter 11. Dynamic Programming

The algorithm initializes a minPath vector of vectors to store the minimum path sum
for each element in the triangle. It starts by setting the first row of minPath to be
the same as the first row of the triangle.
Then, it iterates through the rows of the triangle starting from the second row. For
each element in the current row, it calculates the minimum path sum by considering
the two possible paths from the previous row that lead to that element. It takes the
minimum of the two paths and adds the value of the current element. This way, it
accumulates the minimum path sum efficiently.
The algorithm continues this process until it reaches the last row of the triangle. Fi-
nally, it returns the minimum element from the last row of minPath, which represents
the minimum path sum from top to bottom.

Complexity

• Runtime: O(n^2), where n is the number of rows in the triangle.

• Extra space: O(n^2).

11.4.3 Solution 2: Store only the minimum paths of each row

You do not need to store all paths for all rows. The computation of the next row only
depends on its previous one.

Code

#include <iostream>
#include <vector>
#include <algorithm>
using namespace std;
int minimumTotal(const vector<vector<int>>& triangle) {
const int n = triangle.size();
// store only min path for each row
vector<int> minPath(n);
minPath[0] = triangle[0][0];
for (int i = 1; i < n; i++) {
// right most number
minPath[i] = triangle[i][i] + minPath[i - 1];
for (int j = i - 1; j > 0; j--) {
minPath[j] = triangle[i][j] + min(minPath[j - 1],␣
(continues on next page)

11.4. Triangle 225

 (continued from previous page)
˓→ minPath[j]);
}
// left most number
minPath[0] = triangle[i][0] + minPath[0];
}
return *min_element(minPath.begin(), minPath.end());
}
int main() {
vector<vector<int>> triangle{{2},{3,4},{6,5,7},{4,1,8,3}};
cout << minimumTotal(triangle) << endl;
triangle = {{-10}};
cout << minimumTotal(triangle) << endl;
}

Output:
11
-10

Complexity

• Runtime: O(n^2), where n is the number of rows in the triangle.

• Extra space: O(n).

11.5 Unique Paths II

11.5.1 Problem statement

1
You’re given an m x n grid represented as an integer array called grid. In this
grid, there is a robot initially located at the top-left corner (i.e., grid[0][0]). The
robot’s goal is to move to the bottom-right corner (i.e., grid[m-1][n-1]). The robot
is allowed to move only downwards or to the right at any given point.
The grid contains obstacles and empty spaces, which are marked as 1 or 0 respectively.
The robot cannot pass through squares marked as obstacles.
1 https://leetcode.com/problems/unique-paths-ii/

226 Chapter 11. Dynamic Programming

Your task is to determine the number of unique paths the robot can take to reach the
bottom-right corner while avoiding obstacles.
It’s important to note that the test cases are designed in such a way that the answer
will always be less than or equal to 2 * 10^9.

Example 1

Input: obstacleGrid = [[0,0,0],[0,1,0],[0,0,0]]

Output: 2
Explanation: There is one obstacle in the middle of the 3x3 grid above.
There are two ways to reach the bottom-right corner:
1. Right -> Right -> Down -> Down
2. Down -> Down -> Right -> Right

Example 2

Input: obstacleGrid = [[0,1],[0,0]]

Output: 1

11.5. Unique Paths II 227

Constraints

• m == obstacleGrid.length.
• n == obstacleGrid[i].length.
• 1 <= m, n <= 100.
• obstacleGrid[i][j] is 0 or 1.

11.5.2 Solution: Dynamic programming in place

Let us find the relationship between the positions.

If there is no obstacle at the position (row = i, col = j), the number of paths
np[i][j] that the robot can take to reach this position is:

np[i][j] = np[i - 1][j] + np[i][j - 1]

• As long as there is no obstacle in the first row, np[0][j] = 1. Otherwise,

np[0][k] = 0 for all k >= j0, where (0, j0) is the position of the first obstacle
in the first row.
• Similarly, as long as there is no obstacle in the first column, np[i][0] = 1.
Otherwise, np[k][0] = 0 for all k >= i0, where (i0, 0) is the position of the
first obstacle in the first column.

Code

#include <vector>
#include <iostream>
using namespace std;
int uniquePathsWithObstacles(const vector<vector<int>>& obstacleGrid) {
const int row = obstacleGrid.size();
const int col = obstacleGrid[0].size();
vector<vector<int>> np(row, vector<int>(col, 0));
for (int i = 0; i < row && obstacleGrid[i][0] == 0; i++) {
// can move as long as there is no obstacle
np[i][0] = 1;
}
for (int j = 0; j < col && obstacleGrid[0][j] == 0; j++) {
// can move as long as there is no obstacle
(continues on next page)

228 Chapter 11. Dynamic Programming

 (continued from previous page)
np[0][j] = 1;
}
for (int i = 1; i < row; i++) {
for (int j = 1; j < col; j++) {
if (obstacleGrid[i][j] == 0) {
// can move since there is obstacle
np[i][j] = np[i - 1][j] + np[i][j - 1];
}
}
}
return np[row - 1][col - 1];
}
int main() {
vector<vector<int>> obstacleGrid = {{0,0,0},{0,1,0},{0,0,0}};
cout << uniquePathsWithObstacles(obstacleGrid) << endl;
obstacleGrid = {{0,1},{0,0}};
cout << uniquePathsWithObstacles(obstacleGrid) << endl;
}

Output:
2
1

Complexity

• Runtime: O(m*n), where m and n are the dimensions of the grid.

• Extra space: O(m*n).

11.5.3 Conclusion

This solution computes the number of unique paths in an m x n grid with obsta-
cles using dynamic programming. It initializes a 2D vector np of the same size as
obstacleGrid to store the number of unique paths for each cell.
First, it initializes the top row and left column of np. If there are no obstacles in the
top row or left column of obstacleGrid, it sets the corresponding cells in np to 1
because there’s only one way to reach any cell in the top row or left column.

11.5. Unique Paths II 229

Then, it iterates through the grid starting from the second row and second column
(i.e., indices (1, 1)). For each cell, if there’s no obstacle (obstacleGrid[i][j] == 0),
it updates the value in np by summing up the values from the cell directly above it
and the cell to the left of it. This step efficiently accumulates the number of unique
paths while avoiding obstacles.
Finally, the value at np[row-1][col-1] contains the total number of unique paths to
reach the bottom-right corner of the grid, which is returned as the result.

11.5.4 Exercise

• Minimum Path Cost in a Grid2 .

2 https://leetcode.com/problems/minimum-path-cost-in-a-grid/

230 Chapter 11. Dynamic Programming

 CHAPTER

TWELVE

COUNTING

In this chapter, we will explore the benefits of counting elements and how it can en-
hance the efficiency of different algorithms and operations. By tallying occurrences,
you can gain valuable insights that simplify computations and give you a better un-
derstanding of your data.
Counting elements is like organizing a messy room. When you categorize items, it
becomes easier to access and make decisions. In algorithms, counting allows you
to optimize processes by identifying the most frequent elements or solving complex
problems more efficiently.
What this chapter covers:
1. Introduction to Counting: Lay the foundation by understanding the signifi-
cance of counting elements, its role in performance enhancement, and the var-
ious scenarios where counting is crucial.
2. Frequency Counting: Explore the technique of tallying element occurrences,
enabling you to identify the most frequent items within a dataset quickly.
3. Counting Sort: Delve into the world of counting sort, a specialized sorting algo-
rithm that capitalizes on the power of element counting to achieve exceptional
performance.
4. Problem-Solving with Counts: Develop approaches to solve problems that
benefit from element counting, from optimizing search operations to identifying
anomalies.

231
12.1 Single Number

12.1.1 Problem statement

1
You’re provided with a non-empty array of integers called nums. In this array, every
element occurs twice except for one element that appears only once. Your task is to
identify and find that unique element.
To solve this problem, your solution needs to have a linear runtime complexity and
utilize only a constant amount of extra space.

Example 1

Input: nums = [2,2,1]

Output: 1

Example 2

Input: nums = [4,1,2,1,2]

Output: 4

Example 3

Input: nums = [1]

Output: 1

Constraints

• 1 <= nums.length <= 3 * 10^4.

• -3 * 10^4 <= nums[i] <= 3 * 10^4.
• Each element in the array appears twice except for one element which appears
only once.
1 https://leetcode.com/problems/single-number/description/

232 Chapter 12. Counting

12.1.2 Solution 1: Counting the appearances

Count how many times each element appears in the array. Then return the one ap-
pearing only once.

Code

#include <vector>
#include <iostream>
#include <unordered_map>
using namespace std;
int singleNumber(const vector<int>& nums) {
unordered_map<int, int> count;
for (auto& n : nums) {
count[n]++;
}
int single;
for (auto& pair : count) {
if (pair.second == 1) {
single = pair.first;
break;
}
}
return single;
}
int main() {
vector<int> nums{2,2,1};
cout << singleNumber(nums) << endl;
nums = {4,1,2,1,2};
cout << singleNumber(nums) << endl;
nums = {1};
cout << singleNumber(nums) << endl;
}

Output:
1
4
1

This solution effectively finds the single number by counting the occurrences of each
element in the array and selecting the one with a count of 1.

12.1. Single Number 233

Complexity

• Runtime: O(N).
• Extra space: O(N).

12.1.3 Solution 2: Bitwise exclusive OR

You can also use the bitwise XOR operator to cancel out the duplicated elements in
the array. The remain element is the single one.

a XOR a = 0.
a XOR 0 = a.

Code

#include <vector>
#include <iostream>
using namespace std;
int singleNumber(const vector<int>& nums) {
int single = 0;
for (auto& n : nums) {
single ^= n;
}
return single;
}
int main() {
vector<int> nums{2,2,1};
cout << singleNumber(nums) << endl;
nums = {4,1,2,1,2};
cout << singleNumber(nums) << endl;
nums = {1};
cout << singleNumber(nums) << endl;
}

Output:
1
4
1

234 Chapter 12. Counting

Complexity

• Runtime: O(N).
• Extra space: O(1).

12.1.4 Conclusion

Leveraging bitwise XOR (^) operations offers an efficient solution to find the single
number in an array. Solution 2 utilizes the property of XOR where XORing a number
with itself results in 0.
By XORing all the numbers in the array, Solution 2 effectively cancels out pairs of
identical numbers, leaving only the single number behind. This approach achieves a
linear time complexity without the need for additional data structures, providing a
concise and efficient solution.

12.1.5 Exercise

• Missing Number2 .

12.2 First Unique Character in a String

12.2.1 Problem statement

1
You have a string called s. Your objective is to locate the index of the first character
in the string that does not repeat anywhere else in the string. If such a character
doesn’t exist, return -1.
2 https://leetcode.com/problems/missing-number/
1 https://leetcode.com/problems/first-unique-character-in-a-string/

12.2. First Unique Character in a String 235

Example 1

Input: s = "leetcode"
Output: 0

Example 2

Input: s = "loveleetcode"
Output: 2

Example 3

Input: s = "aabb"
Output: -1

Constraints

• 1 <= s.length <= 10^5.

• s consists of only lowercase English letters.

12.2.2 Solution 1: Using a map to store the appearances

Code

#include <iostream>
#include <unordered_map>
using namespace std;
int firstUniqChar(const string& s) {
unordered_map<char, int> count;
for (auto& c : s) {
count[c]++;
}
for (int i = 0; i < s.length(); i++) {
if (count[s[i]] == 1) {
return i;
(continues on next page)

236 Chapter 12. Counting

 (continued from previous page)
}
}
return -1;
}
int main() {
cout << firstUniqChar("leetcode") << endl;
cout << firstUniqChar("loveleetcode") << endl;
cout << firstUniqChar("aabb") << endl;
}

Output:
0
2
-1

This solution finds the index of the first non-repeating character in a string by using
an unordered map to count the occurrences of each character.
By iterating through the string and populating the unordered map with the count of
each character, it constructs the character count. Then, it iterates through the string
again and returns the index of the first character with a count of 1, indicating that it
is non-repeating.
This approach optimizes the computation by efficiently tracking the count of each
character and identifying the first non-repeating character without requiring addi-
tional space proportional to the length of the string.

Complexity

• Runtime: O(n), where n is the length of the string s.

• Extra space: O(1) as the problem considers only 26 lowercase English letters.

12.2.3 Solution 2: Using an array to store the appearances

From the constraints “s consists of only lowercase English letters”, you can use an
array of 26 elements to store the counts.

12.2. First Unique Character in a String 237

Code

#include <iostream>
#include <vector>
using namespace std;
int firstUniqChar(const string& s) {
// map 'a'->0, 'b'->1, .., 'z'->25
// initializes an array of 26 elements, all set to zero
std::array<int, 26> count{};
for (auto& c : s) {
count[c - 'a']++;
}
for (int i = 0; i < s.length(); i++) {
if (count[s[i] - 'a'] == 1) {
return i;
}
}
return -1;
}
int main() {
cout << firstUniqChar("leetcode") << endl;
cout << firstUniqChar("loveleetcode") << endl;
cout << firstUniqChar("aabb") << endl;
}

Output:
0
2
-1

Complexity

• Runtime: O(n), where n = s.length.

• Extra space: O(1) as the array is fixed regardless of how big n is.

238 Chapter 12. Counting

12.2.4 Conclusion

Utilizing hash maps or arrays to count the frequency of characters in a string provides
an efficient way to identify the first unique character. Both solutions use this approach
to iterate through the string and count the occurrences of each character.
By storing the counts in a data structure indexed by the character value, the solutions
achieve a linear time complexity proportional to the length of the string. Solution 2
further optimizes memory usage by employing an array with a fixed size correspond-
ing to the lowercase English alphabet, avoiding the overhead associated with hash
maps.

12.2.5 Exercise

• First Letter to Appear Twice2 .

12.3 Max Number of K-Sum Pairs

12.3.1 Problem statement

1
You’re provided with an array of integers called nums and an integer k. Each opera-
tion involves selecting two numbers from the array whose sum is equal to k, and then
removing them from the array. Your goal is to determine the maximum count of such
operations you can perform on the array.

Example 1

Input: nums = [1,2,3,4], k = 5

Output: 2
Explanation: Starting with nums = [1,2,3,4]:
- Remove numbers 1 and 4, then nums = [2,3]
- Remove numbers 2 and 3, then nums = []
There are no more pairs that sum up to 5, hence a total of 2 operations.

2 https://leetcode.com/problems/first-letter-to-appear-twice/
1 https://leetcode.com/problems/max-number-of-k-sum-pairs/

12.3. Max Number of K-Sum Pairs 239

Example 2

Input: nums = [3,1,3,4,3], k = 6

Output: 1
Explanation: Starting with nums = [3,1,3,4,3]:
- Remove the first two 3's, then nums = [1,4,3]
There are no more pairs that sum up to 6, hence a total of 1 operation.

Constraints

• 1 <= nums.length <= 10^5.

• 1 <= nums[i] <= 10^9.
• 1 <= k <= 10^9.

12.3.2 Solution: Count the appearances

You can use a map to count the appearances of the elements of nums.

Example 2

For nums = [3,1,3,4,3] and k = 6:

• Initialize count = 0.
• For i = 0: m[3] = 1; k - 3 = 3 but m[3] is only 1, not enough to have two
numbers.
• For i = 1: m[1] = 1; k - 1 = 5 and m[5] = 0.
• For i = 2: m[3] = 2; k - 3 = 3 and m[3] = 2 just enough to have two numbers
to perform the sum. count = 1. Erase those two values 3’s from the map: m[3]
= 0.
• For i = 3: m[4] = 1; k - 4 = 2 and m[2] = 0.
• For i = 4: m[3] = 1; k - 3 = 3 but m[3] is only 1, not enough to have two
numbers.
• Final count = 1.

240 Chapter 12. Counting

Code

#include <vector>
#include <iostream>
#include <unordered_map>
using namespace std;
int maxOperations(const vector<int>& nums, int k) {
unordered_map<int,int> m;
int count = 0;
for (auto& a : nums) {
m[a]++; // count a's occurences
if (m[k - a] > 0) {
// k-a appears in nums
if (a != k - a || m[a] >= 2) {
// if a == k - a, a is required to appear at least twice
count++;
m[a]--;
m[k - a]--;
}
}
}
return count;
}
int main() {
vector<int> nums{1,2,3,4};
cout << maxOperations(nums, 5) << endl;
nums = {3,1,3,4,3};
cout << maxOperations(nums, 6) << endl;
}

Output:
2
1

12.3. Max Number of K-Sum Pairs 241

Complexity

• Runtime: O(n), where n is the number of elements in the nums array.

• Extra space: O(n).

12.3.3 Conclusion

This solution utilizes an unordered map to store the frequency of each element en-
countered while iterating through nums.
By examining each element a in nums, it checks if k - a exists in the map and if its
frequency is greater than 0. If so, it increments the count of pairs and decrements the
frequency of both a and k - a, ensuring that each pair is counted only once.
This approach optimizes the computation by efficiently tracking the frequencies of
elements and identifying valid pairs whose sum equals the target value without re-
quiring additional space proportional to the size of the array.

12.3.4 Exercise

• Two Sum2 .

2 https://leetcode.com/problems/two-sum/

242 Chapter 12. Counting

 CHAPTER

THIRTEEN

PREFIX SUMS

This chapter will introduce you to a technique called prefix sums. This technique
can make calculations much faster and more efficient. The chapter will explain how
cumulative aggregation works and can help optimize your operations.
Prefix sums are like building blocks that can create many different algorithms. They
make it easier to handle cumulative values and allow you to solve complex problems
much more efficiently than before.
What this chapter covers:
1. Introduction to Prefix Sums: Establish the groundwork by understanding the
essence of prefix sums, their role in performance enhancement, and the scenar-
ios where they shine.
2. Prefix Sum Array Construction: Dive into the mechanics of constructing a
prefix sum array, unlocking the potential to access cumulative values efficiently.
3. Range Sum Queries: Explore how prefix sums revolutionize calculating sums
within a given range, enabling quick and consistent results.
4. Subarray Sum Queries: Delve into the technique’s application in efficiently
determining the sum of elements within any subarray of an array.
5. Prefix Sum Variants: Discover the versatility of prefix sums in solving problems
related to averages, running maximum/minimum values, and more.
6. Problem-Solving with Prefix Sums: Develop strategies for solving diverse
problems by incorporating prefix sums, from optimizing sequence operations
to speeding up specific algorithms.

243
13.1 Running Sum of 1d Array

13.1.1 Problem statement

1
Given an array called nums, calculate the running sum of its elements and return the
resulting array. The running sum at index i is the sum of elements from index 0 to i
in the nums array.

Example 1

Input: nums = [1,2,3,4]

Output: [1,3,6,10]
Explanation: Running sum is obtained as follows: [1, 1+2, 1+2+3,␣
˓→1+2+3+4].

Example 2

Input: nums = [1,1,1,1,1]

Output: [1,2,3,4,5]
Explanation: Running sum is obtained as follows: [1, 1+1, 1+1+1, 1+1+1+1,
˓→ 1+1+1+1+1].

Example 3

Input: nums = [3,1,2,10,1]

Output: [3,4,6,16,17]

Constraints

• 1 <= nums.length <= 1000.

• -10^6 <= nums[i] <= 10^6.
1 https://leetcode.com/problems/running-sum-of-1d-array/

244 Chapter 13. Prefix Sums

13.1.2 Solution 1: Unchange nums

Code

#include <vector>
#include <iostream>
using namespace std;
vector<int> runningSum(const vector<int>& nums) {
vector<int> rs;
int s = 0;
for (auto& n : nums) {
s += n;
rs.push_back(s);
}
return rs;
}
void printResult(const vector<int>& sums) {
cout << "[";
for (auto& s: sums) {
cout << s << ",";
}
cout << "]\n";
}
int main() {
vector<int> nums{1,2,3,4};
auto rs = runningSum(nums);
printResult(rs);
nums = {1,1,1,1,1};
rs = runningSum(nums);
printResult(rs);
nums = {3,1,2,10,1};
rs = runningSum(nums);
printResult(rs);
}

Output:
[1,3,6,10,]
[1,2,3,4,5,]
[3,4,6,16,17,]

This solution iterates through the input array nums, calculates the running sum at
each step, and appends the running sums to a result vector. This approach efficiently

13.1. Running Sum of 1d Array 245

computes the running sums in a single pass through the array.

Complexity

• Runtime: O(n), where n = nums.length.

• Extra space: O(1).

13.1.3 Solution 2: Change nums

If nums is allowed to be changed, you could use it to store the result directly.

Code

#include <vector>
#include <iostream>
using namespace std;
vector<int> runningSum(vector<int>& nums) {
for (int i = 1; i < nums.size(); i++) {
nums[i] += nums[i - 1];
}
return nums;
}
void printResult(const vector<int>& sums) {
cout << "[";
for (auto& s: sums) {
cout << s << ",";
}
cout << "]\n";
}
int main() {
vector<int> nums{1,2,3,4};
auto rs = runningSum(nums);
printResult(rs);
nums = {1,1,1,1,1};
rs = runningSum(nums);
printResult(rs);
nums = {3,1,2,10,1};
rs = runningSum(nums);
(continues on next page)

246 Chapter 13. Prefix Sums

 (continued from previous page)
printResult(rs);
}

Output:
[1,3,6,10,]
[1,2,3,4,5,]
[3,4,6,16,17,]

Complexity

• Runtime: O(n), where n = nums.length.

• Extra space: O(1).

13.1.4 Conclusion

Solution 2 directly modifies the input array nums to store the running sums by itera-
tively updating each element with the cumulative sum of the previous elements. This
approach efficiently calculates the running sums in a single pass through the array.

13.2 Maximum Subarray

13.2.1 Problem statement

1
You’re provided with an array of integers called nums. Your task is to identify a
subarray (a consecutive sequence of numbers) that has the highest sum. Once you
find this subarray, return the sum of its elements.
1 https://leetcode.com/problems/maximum-subarray/

13.2. Maximum Subarray 247

Example 1

Input: nums = [-2,1,-3,4,-1,2,1,-5,4]

Output: 6
Explanation: [4,-1,2,1] has the largest sum = 6.

Example 2

Input: nums = [1]

Output: 1

Example 3

Input: nums = [5,4,-1,7,8]

Output: 23

Constraints

• 1 <= nums.length <= 10^5.

• -10^4 <= nums[i] <= 10^4.

13.2.2 Solution

The subarrays you want to find should not have negative prefix sums. A negative
prefix sum would make the sum of the subarray smaller.

Example 1

For nums = [-2,1,-3,4,-1,2,1,-5,4], [-2] or [-2,1] or [-2,1,-3] should not be a

prefix of the subarrays you want to find. Since it makes the sum of the result smaller.

248 Chapter 13. Prefix Sums

Code

#include <vector>
#include <iostream>
using namespace std;
int maxSubArray(const vector<int>& nums) {
int maxSum = -10000; // just chose some negative number to start
int currSum = 0; // sum of current subarray
for (auto& num : nums) {
if (currSum < 0) {
// start a new subarray from this num
currSum = num;
} else {
currSum = currSum + num;
}
// update max sum so far
maxSum = max(maxSum, currSum);
}
return maxSum;
}
int main() {
vector<int> nums = {-2,1,-3,4,-1,2,1,-5,4};
cout << maxSubArray(nums) << endl;
nums = {1};
cout << maxSubArray(nums) << endl;
nums = {5,4,-1,7,8};
cout << maxSubArray(nums) << endl;
}

Output:
6
1
23

13.2. Maximum Subarray 249

Complexity

• Runtime O(n), where n = nums.length.

• Memory O(1).

13.2.3 Conclusion

This solution is the Kadane’s algorithm2 to find the maximum sum of a contiguous
subarray in the given array nums.
It iterates through the elements of the array, updating currSum to either the current
element or the sum of the current element and the previous currSum, whichever is
greater. By considering whether adding the current element improves the overall
sum, it effectively handles both positive and negative numbers in the array. Finally, it
updates maxSum with the maximum value encountered during the iteration, ensuring
it holds the maximum sum of any contiguous subarray within the given array.
This approach optimizes the computation by tracking the maximum sum and dynam-
ically updating it as it iterates through the array.

13.2.4 Exercise

• Maximum Product Subarray3 .

13.3 Product of Array Except Self

13.3.1 Problem statement

1
Given an integer array nums, return an array answer such that answer[i] is equal to
the product of all the elements of nums except nums[i].
The product of any prefix or suffix of nums is guaranteed to fit in a 32-bit integer.
You must write an algorithm that runs in O(n) time and without using the division
operation.
2 https://en.wikipedia.org/wiki/Maximum_subarray_problem#Kadane’s_algorithm
3 https://leetcode.com/problems/maximum-product-subarray/
1 https://leetcode.com/problems/product-of-array-except-self/

250 Chapter 13. Prefix Sums

Example 1

Input: nums = [1,2,3,4]

Output: [24,12,8,6]

Example 2

Input: nums = [-1,1,0,-3,3]

Output: [0,0,9,0,0]

Constraints

• 2 <= nums.length <= 10^5.

• -30 <= nums[i] <= 30.
• The product of any prefix or suffix of nums is guaranteed to fit in a 32-bit integer.

Follow up

• Can you solve the problem in O(1) extra space complexity? (The output array
does not count as extra space for space complexity analysis.)

13.3.2 Solution 1: Compute the prefix and suffix products

To avoid division operation, you can compute the prefix product and the suffix one of
nums[i].

Code

#include <vector>
#include <iostream>
using namespace std;
vector<int> productExceptSelf(const vector<int>& nums) {
const int n = nums.size();
vector<int> prefix(n);
prefix[0] = 1;
(continues on next page)

13.3. Product of Array Except Self 251

 (continued from previous page)
// compute all prefix products nums[0]*nums[1]*..*nums[i-1]
for (int i = 1; i < n; i++) {
prefix[i] = prefix[i - 1] * nums[i - 1];
}
vector<int> suffix(n);
suffix[n - 1] = 1;
// compute all suffix products nums[i+1]*nums[i+2]*..*nums[n-1]
for (int i = n - 2; i >= 0; i--) {
suffix[i] = suffix[i + 1] * nums[i + 1];
}
vector<int> answer(n);
for (int i = 0; i < n; i++) {
answer[i] = prefix[i] * suffix[i];
}
return answer;
}
void print(const vector<int>& nums) {
for (auto& v : nums) {
cout << v << " ";
}
cout << endl;
}
int main() {
vector<int> nums = {1, 2, 3, 4};
auto answer = productExceptSelf(nums);
print(answer);
nums = {-1, 1, 0, -3, 3};
answer = productExceptSelf(nums);
print(answer);
}

Output:
24 12 8 6
0 0 9 0 0

This solution computes the product of all elements in an array except for the current
element.
It accomplishes this by first computing two arrays: prefix and suffix. The prefix
array stores the product of all elements to the left of the current element, while the
suffix array stores the product of all elements to the right of the current element. By

252 Chapter 13. Prefix Sums

multiplying the corresponding elements from prefix and suffix arrays, it effectively
computes the product of all elements except for the current element at each index.
This approach optimizes the computation by breaking down the problem into smaller
subproblems and leveraging the precomputed prefix and suffix arrays to efficiently
compute the final result.

Complexity

• Runtime: O(n), where n = nums.length.

• Extra space: O(n).

13.3.3 Solution 2: Use directly vector answer to store the prefix prod-
uct

In the solution above you can use directly vector answer for prefix and merge the
last two loops into one.

Code

#include <vector>
#include <iostream>
using namespace std;
vector<int> productExceptSelf(const vector<int>& nums) {
const int n = nums.size();
vector<int> answer(n);
answer[0] = 1;
// compute all prefix products nums[0]*nums[1]*..*nums[i-1]
for (int i = 1; i < n; i++) {
answer[i] = answer[i - 1] * nums[i - 1];
}
int suffix = 1;
for (int i = n - 2; i >= 0; i--) {
// compute suffix product and the final product the same time
suffix *= nums[i + 1];
answer[i] *= suffix;
}
return answer;
}
(continues on next page)

13.3. Product of Array Except Self 253

 (continued from previous page)
void print(const vector<int>& nums) {
for (auto& v : nums) {
cout << v << " ";
}
cout << endl;
}
int main() {
vector<int> nums = {1, 2, 3, 4};
auto answer = productExceptSelf(nums);
print(answer);
nums = {-1, 1, 0, -3, 3};
answer = productExceptSelf(nums);
print(answer);
}

Output:
24 12 8 6
0 0 9 0 0

This code efficiently calculates the products of all elements in the nums vector except
for the element at each index using two passes through the array. The first pass cal-
culates products to the left of each element, and the second pass calculates products
to the right of each element.

Complexity

• Runtime: O(n), where n = nums.length.

• Extra space: O(1).

13.3.4 Conclusion

The problem of computing the product of all elements in an array except the element
at the current index can be efficiently solved using different approaches. Solution
1 utilizes two separate passes through the array to compute prefix and suffix prod-
ucts independently. By first computing prefix products from left to right and then
suffix products from right to left, this solution efficiently calculates the product of all
elements except the one at the current index.

254 Chapter 13. Prefix Sums

Solution 2 offers a more concise approach by combining the computation of prefix
and suffix products into a single pass through the array. By iteratively updating a
variable to compute suffix products while simultaneously updating the elements of
the answer array, this solution achieves the desired result more efficiently with only
one pass through the array.

13.3.5 Exercise

• Construct Product Matrix2 .

13.4 Subarray Sum Equals K

13.4.1 Problem Statement

1
You have an array of integers called nums and an integer k. Your task is to determine
the count of contiguous subarrays within this array, where the sum of elements in
each subarray is equal to the value of k.

Example 1

Input: nums = [1,1,1], k = 2

Output: 2

Example 2

Input: nums = [1,2,3], k = 3

Output: 2

2 https://leetcode.com/problems/construct-product-matrix/
1 https://leetcode.com/problems/subarray-sum-equals-k/

13.4. Subarray Sum Equals K 255

Constraints

• 1 <= nums.length <= 2 * 10^4.

• -1000 <= nums[i] <= 1000.
• -10^7 <= k <= 10^7.

13.4.2 Solution 1: Bruteforce

For each element, for all subarrays starting from it, choose the satisfied ones.

Example 3

For nums = [1, -1, 0] and k = 0, you get 3 subarrays for the result:
• There are three subarrays starting from 1, which are [1], [1, -1], and [1, -1,
0]. Only the last two are satisfied.
• There are two subarrays starting from -1, which are [-1] and [-1, 0]. None is
satisfied.
• Only [0] is the subarray starting from 0. It is satisfied.

Code

#include <iostream>
#include <vector>
using namespace std;
int subarraySum(const vector<int>& nums, int k) {
int count = 0;
for (int i = 0; i < nums.size(); i++) {
int sum = 0;
for (int j = i; j < nums.size(); j++) {
sum += nums[j];
if (sum == k) {
count++;
}
}
}
return count;
}
(continues on next page)

256 Chapter 13. Prefix Sums

 (continued from previous page)
int main() {
vector<int> nums{1,1,1};
cout << subarraySum(nums, 2) << endl;
nums = {1,2,3};
cout << subarraySum(nums, 3) << endl;
nums = {1,-1,0};
cout << subarraySum(nums, 0) << endl;
}

Output:
2
2
3

This solution employs a brute-force approach by considering all possible subarrays

and checking whether their sum equals the target k. The time complexity of this
solution is relatively high due to the nested loops, resulting in an inefficient algorithm
for larger input sizes.

Complexity

• Runtime: O(n^2), where n = nums.length.

• Extra space: O(1).

13.4.3 Solution 2: Prefix sum

In the solution above, many sums can be deducted from the previous ones.

Example 4

For nums = [1, 2, 3, 4]. Assume the sum of the subarrays [1], [1, 2], [1, 2, 3],
[1, 2, 3, 4] were computed in the first loop. Then the sum of any other subarray
can be deducted from those values.
• sum([2, 3]) = sum([1, 2, 3]) - sum([1]).
• sum([2, 3, 4]) = sum([1, 2, 3, 4]) - sum([1]).
• sum([3, 4]) = sum(1, 2, 3, 4) - sum(1, 2).

13.4. Subarray Sum Equals K 257

In general, assume you have computed the sum sum[i] for the subarray [nums[0],
nums[1], ..., nums[i]] for all 0 <= i < nums.length. Then the sum of the subarray
[nums[j+1], nums[j+2], ..., nums[i]] for any 0 <= j <= i can be computed as
sum[i] - sum[j].

Code

#include <iostream>
#include <vector>
using namespace std;
int subarraySum(const vector<int>& nums, int k) {
const int n = nums.size();
vector<int> sum(n);
sum[0] = nums[0];
// compute all prefix sums nums[0] + .. + nums[i]
for (int i = 1; i < n; i++) {
sum[i] = sum[i-1] + nums[i];
}
int count = 0;
for (int i = 0; i < n; i++) {
if (sum[i] == k) {
// nums[0] + .. + nums[i] = k
count++;
}
for (int j = 0; j < i; j++) {
if (sum[i] - sum[j] == k) {
// nums[j+1] + nums[j+2] + .. + nums[i] = k
count++;
}
}
}
return count;
}
int main() {
vector<int> nums{1,1,1};
cout << subarraySum(nums, 2) << endl;
nums = {1,2,3};
cout << subarraySum(nums, 3) << endl;
nums = {1,-1,0};
cout << subarraySum(nums, 0) << endl;
}

258 Chapter 13. Prefix Sums

Output:
2
2
3

This solution uses the concept of prefix sum to efficiently calculate the sum of subar-
rays. It then iterates through the array to find subarrays with a sum equal to k, and
the nested loop helps in calculating the sum of various subarray ranges. The time
complexity of this solution is improved compared to the brute-force approach.

Complexity

• Runtime: O(n^2), where n = nums.length.

• Extra space: O(n).

13.4.4 Solution 3: Faster lookup

You can rewrite the condition sum[i] - sum[j] == k in the inner loop of the Solution
2 to sum[i] - k == sum[j].
Then that loop can rephrase to “checking if sum[i] - k was already a value of some
computed sum[j]”.
Now you can use an unordered_map to store the sums as indices for the fast lookup.

Code

#include <iostream>
#include <vector>
#include <unordered_map>
using namespace std;
int subarraySum(const vector<int>& nums, int k) {
int count = 0;
// count the frequency of all subarrays' sums
unordered_map<int, int> sums;
int sumi = 0;
for (int i = 0; i < nums.size(); i++) {
sumi += nums[i];
if (sumi == k) {
(continues on next page)

13.4. Subarray Sum Equals K 259

 (continued from previous page)
count++;
}
auto it = sums.find(sumi - k);
if (it != sums.end()) {
// it->second is the count of j so far
// having sum[j] = sum[i] - k
count += it->second;
}
// store the count of prefix sum sumi
sums[sumi]++;
}
return count;
}
int main() {
vector<int> nums{1,1,1};
cout << subarraySum(nums, 2) << endl;
nums = {1,2,3};
cout << subarraySum(nums, 3) << endl;
nums = {1,-1,0};
cout << subarraySum(nums, 0) << endl;
}

Output:
2
2
3

Complexity

• Runtime: O(n), where n = nums.length.

• Extra space: O(n).

260 Chapter 13. Prefix Sums

13.4.5 Conclusion

Solution 3 uses an unordered map to efficiently track the frequency of cumulative

sums. It utilizes the concept of complement sums to identify subarrays with the de-
sired sum and adds their counts to the overall count. This approach significantly
improves the time complexity compared to the brute-force solution.

13.4.6 Exercise

• Find Pivot Index2 .

2 https://leetcode.com/problems/find-pivot-index/

13.4. Subarray Sum Equals K 261

262 Chapter 13. Prefix Sums
 CHAPTER

FOURTEEN

TWO POINTERS

This chapter will explore the Two Pointers technique, a strategic approach that can
help solve complex problems quickly and effectively. We’ll show you how to use simul-
taneous traversal to streamline operations, optimize algorithms, and extract solutions
from complicated scenarios.
The Two Pointers technique is like exploring a cryptic map from both ends to find
the treasure. It can enhance your problem-solving skills and help you tackle intricate
challenges with a broader perspective.
What this chapter covers:
1. Introduction to Two Pointers: Lay the foundation by understanding the
essence of the Two Pointers technique, its adaptability, and its role in unrav-
eling complex problems.
2. Two Pointers Approach: Dive into the mechanics of the technique, exploring
scenarios where two pointers traverse a sequence to locate solutions or patterns.
3. Collision and Separation: Discover the duality of the technique, where point-
ers can converge to solve particular problems or diverge to address different
aspects of a challenge.
4. Optimal Window Management: Explore how the Two Pointers technique opti-
mizes sliding window problems, facilitating efficient substring or subarray anal-
ysis.
5. Intersection and Union: Uncover the technique’s versatility in solving prob-
lems that involve intersecting or uniting elements within different sequences.
6. Problem-Solving with Two Pointers: Develop strategies to address diverse
problems through the Two Pointers technique, from array manipulation to string
analysis.

263
14.1 Middle of the Linked List

14.1.1 Problem statement

1
Given the head of a singly linked list, return the middle node of the linked list.
If there are two middle nodes, return the second middle node.

Example 1

Input: head = [1,2,3,4,5]

Output: [3,4,5]
Explanation: The middle node of the list is node 3.

Example 2

Input: head = [1,2,3,4,5,6]

Output: [4,5,6]
Explanation: Since the list has two middle nodes with values 3 and 4, we␣
˓→return the second one.

1 https://leetcode.com/problems/middle-of-the-linked-list/

264 Chapter 14. Two Pointers

Constraints

• The number of nodes in the list is in the range [1, 100].

• 1 <= Node.val <= 100.

14.1.2 Solution 1: Counting the number of nodes

Code

#include <iostream>
struct ListNode {
int val;
ListNode *next;
ListNode() : val(0), next(nullptr) {}
ListNode(int x) : val(x), next(nullptr) {}
ListNode(int x, ListNode *next) : val(x), next(next) {}
};
ListNode* middleNode(ListNode* head) {
ListNode *node = head;
int count = 0;
while (node) {
count++;
node = node->next;
}
int i = 1;
node = head;
while (i <= count/2) {
node = node->next;
i++;
}
return node;
}
void print(const ListNode *head) {
ListNode *node = head;
std::cout << "[";
while (node) {
std::cout << node->val << ",";
node = node->next;
}
std::cout << "]\n";
(continues on next page)

14.1. Middle of the Linked List 265

 (continued from previous page)
}
int main() {
ListNode five(5);
ListNode four(4, &five);
ListNode three(3, &four);
ListNode two(2, &three);
ListNode one(1, &two);
auto result = middleNode(&one);
print(result);

ListNode six(6);
five.next = &six;
result = middleNode(&one);
print(result);
}

Output:
[3,4,5,]
[4,5,6,]

This solution first counts the total number of nodes in the linked list, and then it
iterates to the middle node using the count variable.

Complexity

• Runtime: O(N), where N is the number of nodes in the linked list.

• Extra space: O(1).

14.1.3 Solution 2: Fast and Slow pointers

Use two pointers to go through the linked list.

One goes one step at a time. The other goes two steps at a time. When the faster
reaches the end, the slower reaches the middle.

266 Chapter 14. Two Pointers

Code

#include <iostream>
struct ListNode {
int val;
ListNode *next;
ListNode() : val(0), next(nullptr) {}
ListNode(int x) : val(x), next(nullptr) {}
ListNode(int x, ListNode *next) : val(x), next(next) {}
};
ListNode* middleNode(ListNode* head) {
ListNode *slow = head;
ListNode *fast = head;
while (fast && fast->next) {
slow = slow->next;
fast = fast->next->next;
}
return slow;
}
void print(const ListNode *head) {
ListNode *node = head;
std::cout << "[";
while (node) {
std::cout << node->val << ",";
node = node->next;
}
std::cout << "]\n";
}
int main() {
ListNode five(5);
ListNode four(4, &five);
ListNode three(3, &four);
ListNode two(2, &three);
ListNode one(1, &two);
auto result = middleNode(&one);
print(result);

ListNode six(6);
five.next = &six;
result = middleNode(&one);
print(result);
}

14.1. Middle of the Linked List 267

Output:
[3,4,5,]
[4,5,6,]

This solution uses two pointers, a slow pointer and a fast pointer, to find the middle
node of a linked list. Both pointers start from the head of the list, and in each iteration,
the slow pointer moves one step forward while the fast pointer moves two steps
forward. This ensures that the slow pointer reaches the middle node of the list when
the fast pointer reaches the end.
By advancing the pointers at different speeds, the algorithm identifies the middle
node of the linked list. If the list has an odd number of nodes, the slow pointer will
be positioned at the middle node. If the list has an even number of nodes, the slow
pointer will be positioned at the node closer to the middle of the list.
Finally, the algorithm returns the slow pointer, which points to the middle node of the
linked list.
This approach optimizes the computation by traversing the linked list only once and
using two pointers to efficiently locate the middle node.

Complexity

• Runtime: O(N), where N is the number of nodes in the linked list.

• Extra space: O(1).

14.1.4 OBS!

• The approach using slow and fast pointers looks very nice and faster. But it
is not suitable to generalize this problem to any relative position (one-third, a
quarter, etc.). Moreover, long expressions like fast->next->...->next are not
recommended.
• Though the counting nodes approach does not seem optimized, it is more read-
able, scalable and maintainable.

268 Chapter 14. Two Pointers

14.1.5 Exercise

• Delete the Middle Node of a Linked List2 .

14.2 Linked List Cycle

14.2.1 Problem statement

1
Given head, the head of a linked list, determine if the linked list has a cycle in it.
Return true if there is a cycle in the linked list. Otherwise, return false.

Example 1

Input: head = [3,2,0,-4], where -4 links next to 2.

Output: true

Example 2

Input: head = [1,2], where 2 links next to 1.

Output: true

2 https://leetcode.com/problems/delete-the-middle-node-of-a-linked-list/
1 https://leetcode.com/problems/linked-list-cycle/

14.2. Linked List Cycle 269

Example 3

Input: head = [1], and 1 links to NULL.

Output: false
Explanation: There is no cycle in this linked list.

Constraints

• The number of the nodes in the list is in the range [0, 10^4].
• -10^5 <= Node.val <= 10^5.

Follow up

• Can you solve it using O(1) (i.e. constant) memory?

14.2.2 Solution 1: Storing the visited nodes

Code

#include <unordered_map>
#include <iostream>
struct ListNode {
int val;
ListNode *next;
ListNode(int x) : val(x), next(NULL) {}
};
bool hasCycle(ListNode *head) {
std::unordered_map<ListNode*, bool> m;
while (head) {
if (m[head]) {
// found this node marked in the map
return true;
}
m[head] = true; // mark this node visited
(continues on next page)

270 Chapter 14. Two Pointers

 (continued from previous page)
head = head->next;
}
return false;
}
int main() {
{
ListNode three(3);
ListNode two(2);
three.next = &two;
ListNode zero(0);
two.next = &zero;
ListNode four(4);
zero.next = &four;
four.next = &two;
std::cout << hasCycle(&three) << std::endl;
}
{
ListNode one(1);
ListNode two(2);
one.next = &two;
two.next = &one;
std::cout << hasCycle(&one) << std::endl;
}
{
ListNode one(1);
std::cout << hasCycle(&one) << std::endl;
}
}

Output:
1
1
0

This solution uses a hash map to track visited nodes while traversing the linked list.
By iterating through the linked list and marking pointers to visited nodes in the hash
map, it detects cycles in the linked list. If a node is found marked true in the map, it
indicates the presence of a cycle, and the function returns true. Otherwise, if the end
of the linked list is reached without finding any node marked, it confirms the absence
of a cycle, and the function returns false.

14.2. Linked List Cycle 271

This approach optimizes the computation by leveraging the hash map to efficiently
detect cycles in the linked list without requiring additional space proportional to the
length of the list.

Complexity

• Runtime: O(N), where N is the length of the linked list.

• Extra space: O(N).

14.2.3 Solution 2: Fast and Slow runners

Imagine there are two runners both start to run along the linked list from the head.
One runs twice faster than the other.
If the linked list has a cycle in it, they will meet at some point. Otherwise, they never
meet each other.

Example 1

The slower runs [3,2,0,-4,2,0,...] while the faster runs [3,0,2,-4,0,2,...].

They meet each other at node -4 after three steps.

Example 2

The slower runs [1,2,1,2,...] while the faster runs [1,1,1,...]. They meet each
other at node 1 after two steps.

Code

#include <iostream>
struct ListNode {
int val;
ListNode *next;
ListNode(int x) : val(x), next(NULL) {}
};
bool hasCycle(ListNode *head) {
if (head == nullptr) {
return false;
(continues on next page)

272 Chapter 14. Two Pointers

 (continued from previous page)
}
ListNode* fast = head;
ListNode* slow = head;
while (fast && fast->next) {
fast = fast->next->next;
slow = slow->next;
if (fast == slow) {
return true;
}
}
return false;
}
int main() {
{
ListNode three(3);
ListNode two(2);
three.next = &two;
ListNode zero(0);
two.next = &zero;
ListNode four(4);
zero.next = &four;
four.next = &two;
std::cout << hasCycle(&three) << std::endl;
}
{
ListNode one(1);
ListNode two(2);
one.next = &two;
two.next = &one;
std::cout << hasCycle(&one) << std::endl;
}
{
ListNode one(1);
std::cout << hasCycle(&one) << std::endl;
}
}

Output:
1
1
(continues on next page)

14.2. Linked List Cycle 273

 (continued from previous page)
0

Complexity

• Runtime: O(N), where N is the number of nodes in the linked list.

• Extra space: O(1).

14.2.4 Conclusion

Solution 2 uses two pointers, a fast pointer and a slow pointer, to detect cycles in a
linked list.
Both pointers start from the head of the list, and the fast pointer moves two steps for-
ward while the slow pointer moves one step forward in each iteration. By comparing
the positions of the fast and slow pointers, the algorithm detects cycles in the linked
list.
If the fast pointer catches up with the slow pointer at any point during traversal, it
indicates the presence of a cycle, and the function returns true. Otherwise, if the
fast pointer reaches the end of the list without intersecting with the slow pointer, it
confirms the absence of a cycle, and the function returns false.
This approach optimizes the computation by simultaneously advancing two pointers
at different speeds to efficiently detect cycles in the linked list.

14.2.5 Exercise

• Linked List Cycle II2 .

2 https://leetcode.com/problems/linked-list-cycle-ii/

274 Chapter 14. Two Pointers

14.3 Sort Array By Parity II

14.3.1 Problem statement

1
Given an array of integers nums, half of the integers in nums are odd, and the other
half are even.
Sort the array so that whenever nums[i] is odd, i is odd, and whenever nums[i] is
even, i is even.
Return any answer array that satisfies this condition.

Example 1

Input: nums = [4,2,5,7]

Output: [4,5,2,7]
Explanation: [4,7,2,5], [2,5,4,7], [2,7,4,5] would also have been␣
˓→accepted.

Example 2

Input: nums = [2,3]

Output: [2,3]

Constraints:

• 2 <= nums.length <= 2 * 10^4.

• nums.length is even.
• Half of the integers in nums are even.
• 0 <= nums[i] <= 1000.
1 https://leetcode.com/problems/sort-array-by-parity-ii/

14.3. Sort Array By Parity II 275

14.3.2 Solution 1: Bubble Sort

For each 0 <= i < nums.length, if nums[i] has the same parity with i, you do nothing.
Otherwise you need to find another nums[j] that has the same parity with i to swap
with nums[i].

Example 1

For nums = [4,2,5,7]:

• nums[0] = 4 is even like i = 0.
• nums[1] = 2 is even, unlike i = 1 is odd. Found nums[2] = 5 is odd. Swap
nums[1] <-> nums[2]. nums[2] becomes 2 while nums[1] becomes 5 is odd like
i = 1.
• nums[2] = 2 is even, like i = 2.
• nums[3] = 7 is odd like i = 3.

Code

#include<vector>
#include<iostream>
using namespace std;
vector<int> sortArrayByParityII(vector<int>& nums) {
for (int i = 0; i < nums.size(); i++) {
if (i % 2 != nums[i] % 2) {
// find suitable nums[j] to swap
for (int j = i + 1; j < nums.size(); j++) {
if (nums[j] % 2 == i % 2) {
swap(nums[i], nums[j]);
break;
}
}
}
}
return nums;
}
void print(vector<int>& nums) {
for (auto num : nums) {
cout << num << " ";
(continues on next page)

276 Chapter 14. Two Pointers

 (continued from previous page)
}
cout << endl;
}
int main() {
vector<int> nums = {4,2,5,7};
auto result = sortArrayByParityII(nums);
print(result);
nums = {1,0,7,3,8,9,2,5,4,1,2,4};
result = sortArrayByParityII(nums);
print(result);
nums = {3,4};
result = sortArrayByParityII(nums);
print(result);
nums = {648,831,560,986,192,424,997,829,897,843};
result = sortArrayByParityII(nums);
print(result);
}

Output:
4 5 2 7
0 1 8 3 2 9 4 5 2 1 4 7
4 3
648 831 560 997 192 829 986 897 424 843

This solution iteratively scans through the array and swap elements to ensure that the
parity (even or odd) of each element matches its index modulo 2.
The algorithm iterates over each index of the array. For each index i, if the parity
of the element at index i does not match i % 2, it implies that the element is in the
wrong position. In such cases, the algorithm searches for the next element with the
correct parity (i.e., even or odd) starting from index i + 1. Once found, it swaps the
elements at indices i and j, where j is the index of the next element with the correct
parity.
By performing these swaps, the algorithm ensures that each element is at the correct
position based on its parity.
This approach optimizes the sorting process by performing a single pass through the
array and minimizing the number of swaps required to achieve the desired parity
arrangement.

14.3. Sort Array By Parity II 277

Complexity

• Runtime: O(N^2), where N = nums.length.

• Extra space: O(1).

14.3.3 Solution 2: Two pointers - Make use of the problem’s con-

straints

In the Bubble Sort approach, you do not make use of the constraint that half of the
integers in nums are even. Because of that, these are unnecessary things:
1. The loops scan through full nums.
2. The loops are nested. That increases the complexity.
3. The swap(nums[i], nums[j]) happens even when nums[j] was already in place,
i.e. nums[j] had the same parity with j (Why to move it?).
Here is a two-pointer approach which takes the important constraint into account.

Code

#include<vector>
#include<iostream>
#include <algorithm>
using namespace std;
vector<int> sortArrayByParityII(vector<int>& nums) {
int N = nums.size();
int evenPos = 0;
int oddPos = N - 1;
while (evenPos < N) {
// find the nums[evenPos] that is odd for swapping
while (evenPos < N && nums[evenPos] % 2 == 0) {
evenPos += 2;
}
// If not found, it means all even nums are in place. Done!
if (evenPos >= N) {
break;
}
// Otherwise, the problem's constraint makes sure
// there must be some nums[oddPos] that is even for swapping
(continues on next page)

278 Chapter 14. Two Pointers

 (continued from previous page)
while (oddPos >= 0 && nums[oddPos] % 2 == 1) {
oddPos -= 2;
}
swap(nums[evenPos], nums[oddPos]);
}
return nums;
}
void print(vector<int>& nums) {
for (auto num : nums) {
cout << num << " ";
}
cout << endl;
}
int main() {
vector<int> nums = {4,2,5,7};
auto result = sortArrayByParityII(nums);
print(result);
nums = {1,0,7,3,8,9,2,5,4,1,2,4};
result = sortArrayByParityII(nums);
print(result);
nums = {3,4};
result = sortArrayByParityII(nums);
print(result);
nums = {648,831,560,986,192,424,997,829,897,843};
result = sortArrayByParityII(nums);
print(result);
}

Output:
4 5 2 7
0 1 8 3 2 9 4 5 2 1 4 7
4 3
648 831 560 997 192 829 986 897 424 843

14.3. Sort Array By Parity II 279

Complexity

• Runtime: O(N), where N = nums.length.

• Extra space: O(1).

14.3.4 Conclusion

Solution 2 uses two pointers, one starting from the beginning of the array (evenPos)
and the other starting from the end (oddPos), to efficiently identify misplaced ele-
ments.
By incrementing evenPos by 2 until an odd element is found and decrementing oddPos
by 2 until an even element is found, the algorithm can swap these elements to ensure
that even-indexed elements contain even values and odd-indexed elements contain
odd values. This process iterates until all even and odd elements are correctly posi-
tioned.

14.3.5 Exercise

• Rearrange Array Elements by Sign2 .

14.4 Container With Most Water

14.4.1 Problem statement

1
You are given an integer array height of length n. There are n vertical lines drawn
such that the two endpoints of the i-th line are (i, 0) and (i, height[i]).
Find two lines that together with the x-axis form a container, such that the container
contains the most water.
Return the maximum amount of water a container can store.
Notice that you may not slant the container.
2 https://leetcode.com/problems/rearrange-array-elements-by-sign/
1 https://leetcode.com/problems/container-with-most-water/

280 Chapter 14. Two Pointers

Example 1

Input: height = [1,8,6,2,5,4,8,3,7]

Output: 49
Explanation: The above vertical lines are represented by array [1,8,6,2,
˓→5,4,8,3,7]. In this case, the max area of water (blue/grey section)␣

˓→the container can contain is 49.

Example 2

Input: height = [1,1]

Output: 1

Constraints

• n == height.length.
• 2 <= n <= 10^5.
• 0 <= height[i] <= 10^4.

14.4.2 Solution 1: Bruteforce

For each line i, find the line j > i such that it gives the maximum amount of water
the container (i, j) can store.

14.4. Container With Most Water 281

Code

#include <iostream>
#include <vector>
using namespace std;
int maxArea(const vector<int>& height) {
int maxA = 0;
for (int i = 0; i < height.size() - 1; i++) {
for (int j = i + 1; j < height.size(); j++) {
maxA = max(maxA, min(height[i], height[j]) * (j - i));
}
}
return maxA;
}
int main() {
vector<int> height{1,8,6,2,5,4,8,3,7};
cout << maxArea(height) << endl;
height = {1,1};
cout << maxArea(height) << endl;
}

Output:
49
1

This solution computes the maximum area of water that can be trapped between
two vertical lines by iterating through all possible pairs of lines. By considering all
combinations of lines and calculating the area using the formula (min(height[i],
height[j]) * (j - i)), where height[i] and height[j] represent the heights of the
two lines and (j - i) represents the width between them, it effectively evaluates the
area formed by each pair and updates maxA with the maximum area encountered.
This approach optimizes the computation by exhaustively considering all possible
pairs of lines and efficiently computing the area without requiring additional space.

282 Chapter 14. Two Pointers

Complexity

• Runtime: O(n^2), where n = height.length. This is because it checks all possi-

ble pairs of vertical lines, resulting in a quadratic time complexity.
• Extra space: O(1).

14.4.3 Solution 2: Two pointers

Any container has left line i and right line j satisfying 0 <= i < j < height.length.
The biggest container you want to find satisfies that condition too.
You can start from the broadest container with the left line i = 0 and the right line j
= height.length - 1. Then by moving i forward and j backward, you can narrow
down the container to find which one will give the maximum amount of water it can
store.
Depending on which line is higher, you can decide which one to move next. Since
you want a bigger container, you should move the shorter line.

Example 1

For height = [1,8,6,2,5,4,8,3,7]:

• Starting with i = 0 and j = 8.

area = min(height[i], height[j]) * (j - i) = min(1, 7) * (8 - 0) =␣

˓→8.

maxArea = 8.

• height[i] = 1 < 7 = height[j], move i to 1.

area = min(8, 7) * (8 - 1) = 49.

maxArea = 49.

• height[i] = 8 > 7 = height[j], move j to 7.

area = min(8, 3) * (7 - 1) = 18.

maxArea = 49.

• So on and so on. Final maxArea = 49.

14.4. Container With Most Water 283

Code

#include <iostream>
#include <vector>
using namespace std;
int maxArea(const vector<int>& height) {
int maxA = 0;
int i = 0;
int j = height.size() - 1;
while (i < j) {
if (height[i] < height[j]) {
maxA = max(maxA, height[i] * (j - i) );
i++;
} else {
maxA = max(maxA, height[j] * (j - i) );
j--;
}
}
return maxA;
}
int main() {
vector<int> height{1,8,6,2,5,4,8,3,7};
cout << maxArea(height) << endl;
height = {1,1};
cout << maxArea(height) << endl;
}

Output:
49
1

284 Chapter 14. Two Pointers

Complexity

• Runtime: O(n), where n = height.length.

• Extra space: O(1).

14.4.4 Conclusion

Solution 2 is the two-pointer approach. By initializing two pointers i and j at the

beginning and end of the array respectively, and iteratively moving them towards
each other until they converge, it evaluates all possible pairs of lines. At each step, it
calculates the area. By moving the pointer corresponding to the shorter line inward
at each step, it ensures that the maximum possible area is considered.
This approach optimizes the computation by avoiding redundant calculations and
efficiently exploring the solution space using two pointers.

14.5 Remove Nth Node From End of List

14.5.1 Problem statement

1
Given the head of a linked list, remove the n-th node from the end of the list and
return its head.

Example 1

1 https://leetcode.com/problems/remove-nth-node-from-end-of-list/

14.5. Remove Nth Node From End of List 285

Input: head = [1,2,3,4,5], n = 2
Output: [1,2,3,5]

Example 2

Input: head = [1], n = 1

Output: []

Example 3

Input: head = [1,2], n = 1

Output: [1]

Constraints

• The number of nodes in the list is sz.

• 1 <= sz <= 30.
• 0 <= Node.val <= 100.
• 1 <= n <= sz.

Follow up

• Could you do this in one pass?

14.5.2 Solution 1: Store the nodes

Code

#include <iostream>
#include <vector>
struct ListNode {
int val;
ListNode *next;
(continues on next page)

286 Chapter 14. Two Pointers

 (continued from previous page)
ListNode() : val(0), next(nullptr) {}
ListNode(int x) : val(x), next(nullptr) {}
ListNode(int x, ListNode *next) : val(x), next(next) {}
};
using namespace std;
ListNode* removeNthFromEnd(ListNode* head, int n) {
vector<ListNode*> nodes;
ListNode* node = head;
while (node)
{
nodes.push_back(node);
node = node->next;
}
node = nodes[nodes.size() - n];

if (node == head) {
// remove head if n == nodes.size()
head = node->next;
} else {
ListNode* pre = nodes[nodes.size() - n - 1];
pre->next = node->next;
}
return head;
}
void printList(const ListNode *head) {
ListNode* node = head;
cout << "[";
while (node) {
cout << node->val << ",";
node = node->next;
}
cout << "]\n";
}
int main() {
ListNode five(5);
ListNode four(4, &five);
ListNode three(3, &four);
ListNode two(2, &three);
ListNode one(1, &two);
auto head = removeNthFromEnd(&one, 2);
(continues on next page)

14.5. Remove Nth Node From End of List 287

 (continued from previous page)
printList(head);
head = removeNthFromEnd(&five, 1);
printList(head);
head = removeNthFromEnd(&four, 1);
printList(head);
}

Output:
[1,2,3,5,]
[]
[4,]

This solution uses a vector to store pointers to all nodes in the linked list, enabling
easy access to the node to be removed and its predecessor.
By iterating through the linked list and storing pointers to each node in the vector,
it constructs a representation of the linked list in an array-like structure. Then, it
retrieves the node to be removed using its index from the end of the vector. Finally,
it handles the removal of the node by updating the next pointer of its predecessor or
updating the head pointer if the node to be removed is the head of the linked list.
This approach optimizes the computation by sacrificing space efficiency for simplicity
of implementation and ease of manipulation of linked list elements.

Complexity

• Runtime: O(N), where N is the number of nodes in the list.

• Extra space: O(N).

14.5.3 Solution 2: Two pointers

The distance between the removed node and the end (nullptr) of the list is always
n.
You can apply the two-pointer technique as follows.
Let the slower runner start after the faster one n nodes. Then when the faster reaches
the end of the list, the slower reaches the node to be removed.

288 Chapter 14. Two Pointers

Code

#include <iostream>
#include <vector>
struct ListNode {
int val;
ListNode *next;
ListNode() : val(0), next(nullptr) {}
ListNode(int x) : val(x), next(nullptr) {}
ListNode(int x, ListNode *next) : val(x), next(next) {}
};
using namespace std;
ListNode* removeNthFromEnd(ListNode* head, int n) {
ListNode* fast = head;
// let fast goes ahead n nodes
for (int i = 0; i < n; i++) {
fast = fast->next;
}
if (fast == nullptr) {
// remove head if n equals the list's length
return head->next;
}
ListNode* slow = head;
while (fast->next) {
slow = slow->next;
fast = fast->next;
}
// remove slow
slow->next = slow->next->next;
return head;
}
void printList(const ListNode *head) {
ListNode* node = head;
cout << "[";
while (node) {
cout << node->val << ",";
node = node->next;
}
cout << "]\n";
}
int main() {
(continues on next page)

14.5. Remove Nth Node From End of List 289

 (continued from previous page)
ListNode five(5);
ListNode four(4, &five);
ListNode three(3, &four);
ListNode two(2, &three);
ListNode one(1, &two);
auto head = removeNthFromEnd(&one, 2);
printList(head);
head = removeNthFromEnd(&five, 1);
printList(head);
head = removeNthFromEnd(&four, 1);
printList(head);
}

Output:
[1,2,3,5,]
[]
[4,]

Complexity

• Runtime: O(N), where N is the number of nodes in the list.

• Extra space: O(1).

14.5.4 Conclusion

Solution 2 uses two pointers, a fast pointer and a slow pointer, to remove the nth
node from the end of a linked list.
Initially, both pointers start from the head of the list. The fast pointer moves n steps
ahead, effectively positioning itself n nodes ahead of the slow pointer. Then, while the
fast pointer is not at the end of the list, both pointers move forward simultaneously.
This ensures that the slow pointer stays n nodes behind the fast pointer, effectively
reaching the node preceding the nth node from the end when the fast pointer reaches
the end of the list. Finally, the nth node from the end is removed by updating the
next pointer of the node preceding it.
This approach optimizes the computation by traversing the linked list only once and
using two pointers to efficiently locate the node to be removed.

290 Chapter 14. Two Pointers

14.5.5 Exercise

• Swapping Nodes in a Linked List2 .

14.6 Shortest Unsorted Continuous Subarray

14.6.1 Problem statement

1
Given an integer array nums, you need to find one continuous subarray that if you
only sort this subarray in ascending order, then the whole array will be sorted in
ascending order.
Return the shortest such subarray and output its length.

Example 1

Input: nums = [2,6,4,8,10,9,15]

Output: 5
Explanation: You need to sort [6, 4, 8, 10, 9] in ascending order to␣
˓→make the whole array sorted in ascending order.

Example 2

Input: nums = [1,2,3,4]

Output: 0

Example 3

Input: nums = [1]

Output: 0

2 https://leetcode.com/problems/swapping-nodes-in-a-linked-list/
1 https://leetcode.com/problems/shortest-unsorted-continuous-subarray/

14.6. Shortest Unsorted Continuous Subarray 291

Constraints:

• 1 <= nums.length <= 10^4.

• -10^5 <= nums[i] <= 10^5.

Follow up

• Can you solve it inO(n) time complexity?

14.6.2 Solution 1: Sort and compare the difference

Example 1

Comparing nums = [2,6,4,8,10,9,15] with its sorted one sortedNums = [2,4,6,8,

9,10,15]:
• The first position that makes the difference is left = 1, where 6 != 4.
• The last (right) position that makes the difference is right = 5, where 9 != 10.
• The length of that shortest subarray is right - left + 1 = 5.

Code

#include <vector>
#include <iostream>
#include <algorithm>
using namespace std;
int findUnsortedSubarray(const vector<int>& nums) {
vector<int> sortedNums = nums;
sort(sortedNums.begin(), sortedNums.end());
int left = 0;
while (left < nums.size() && nums[left] == sortedNums[left]) {
left++;
}
int right = nums.size() - 1;
while (right >= 0 && nums[right] == sortedNums[right]) {
right--;
}
return left >= right ? 0 : right - left + 1;
(continues on next page)

292 Chapter 14. Two Pointers

 (continued from previous page)
}
int main() {
vector<int> nums{2,6,4,8,10,9,15};
cout << findUnsortedSubarray(nums) << endl;
nums = {1,2,3,4};
cout << findUnsortedSubarray(nums) << endl;
nums = {1};
cout << findUnsortedSubarray(nums) << endl;
}

Output:
5
0
0

This solution compares the original array with a sorted version of itself to identify the
unsorted boundaries efficiently.

Complexity

• Runtime: O(N*logN) due to the sorting step, where N is the number of elements
in the nums vector.
• Extra space: O(N).

14.6.3 Solution 2: Comparing only maximum and minimum elements

Assume the subarray A = [nums[0], ..., nums[i - 1]] is sorted. What would be the
wanted right position for the subarray B = [nums[0], ..., nums[i - 1], nums[i]]?
If nums[i] is smaller than max(A), the longer subarray B is not in ascending order. You
might need to sort it, which means right = i.
Similarly, assume the subarray C = [nums[j + 1], ..., nums[n - 1]] is sorted.
What would be the wanted left position for the subarray D = [nums[j], nums[j +
1], ..., nums[n - 1]]?
If nums[j] is bigger than min(C), the longer subarray D is not in ascending order. You
might need to sort it, which means left = j

14.6. Shortest Unsorted Continuous Subarray 293

Code

#include <vector>
#include <iostream>
#include <algorithm>
using namespace std;
int findUnsortedSubarray(const vector<int>& nums) {
const int n = nums.size();
int right = 0;
int max = nums[0];
for (int i = 0; i < nums.size(); i++) {
if (nums[i] < max) {
right = i;
} else {
max = nums[i];
}
}
int left = n - 1;
int min = nums[n - 1];
for (int j = n - 1; j >= 0; j--) {
if (nums[j] > min) {
left = j;
} else {
min = nums[j];
}
}
return left >= right ? 0 : right - left + 1;
}
int main() {
vector<int> nums{2,6,4,8,10,9,15};
cout << findUnsortedSubarray(nums) << endl;
nums = {1,2,3,4};
cout << findUnsortedSubarray(nums) << endl;
nums = {1};
cout << findUnsortedSubarray(nums) << endl;
}

Output:
5
0
0

294 Chapter 14. Two Pointers

This solution determines the boundaries of the unsorted subarray by iterating through
the array from both ends. It starts by initializing the right boundary to the beginning
of the array and tracks the maximum element encountered so far. It iterates from
the beginning of the array towards the end, updating the right boundary whenever
an element smaller than the current maximum is encountered. This identifies the
rightmost position where the array is unsorted.
Similarly, it initializes the left boundary to the end of the array and tracking the
minimum element encountered so far. It iterates from the end of the array towards
the beginning, updating the left boundary whenever an element greater than the
current minimum is encountered. This identifies the leftmost position where the array
is unsorted.
Finally, it returns the length of the unsorted subarray, calculated as right - left +
1, unless the left boundary is greater than or equal to the right boundary, in which
case the array is already sorted, and it returns 0.
This approach optimizes the computation by traversing the array only twice, once
from the end and once from the beginning, to efficiently determine the boundaries of
the unsorted subarray.

Complexity

• Runtime: O(N), where N is the number of elements in the nums vector.

• Extra space: O(1).

14.6.4 Key Takeaway

Solution 2 helped you identify the shortest subarray (by the left and right indices)
needed to be sorted in order to sort the whole array.
That means in some cases you can sort an array with complexity O(N + m*logm) <
O(N*logN) where N is the length of the whole array and m is the length of the shortest
subarray.

14.6. Shortest Unsorted Continuous Subarray 295

296 Chapter 14. Two Pointers
 CHAPTER

FIFTEEN

MATHEMATICS

This chapter will explore how mathematics and programming create efficient solu-
tions. We’ll cover mathematical concepts and show how they can be integrated into
coding to enhance problem-solving skills.
Mathematics and programming complement each other and can lead to innovative
outcomes. By applying mathematical principles, you can refine algorithms, identify
patterns, streamline processes, and better understand your code’s underlying logic.
What this chapter covers:
1. Introduction to Mathematics in Coding: Set the stage by understanding the
symbiotic relationship between mathematics and programming and how math-
ematical concepts enrich your coding toolkit.
2. Number Theory and Modular Arithmetic: Delve into number theory, under-
standing modular arithmetic and its applications.
3. Combinatorics and Probability: Uncover the power of combinatorial mathe-
matics and probability theory in solving problems related to permutations, com-
binations, and statistical analysis.
4. Problem-Solving with Mathematics: Develop strategies for leveraging mathe-
matical concepts to solve problems efficiently and elegantly, from optimization
tasks to simulation challenges.

297
15.1 Excel Sheet Column Number

15.1.1 Problem statement

1
Given a string columnTitle that represents the column title as appears in an Excel
sheet, return its corresponding column number.
For example:

A -> 1
B -> 2
C -> 3
...
Z -> 26
AA -> 27
AB -> 28
...

Example 1

Input: columnTitle = "A"

Output: 1

Example 2

Input: columnTitle = "AB"

Output: 28

Example 3

Input: columnTitle = "ZY"

Output: 701

1 https://leetcode.com/problems/excel-sheet-column-number/

298 Chapter 15. Mathematics

Constraints

• 1 <= columnTitle.length <= 7.

• columnTitle consists only of uppercase English letters.
• columnTitle is in the range ["A", "FXSHRXW"].

15.1.2 Solution: Finding The Pattern

Let us write down some other columnTitle strings and its value.

"A" = 1
"Z" = 26
"AA" = 27
"AZ" = 52
"ZZ" = 702
"AAA" = 703

Then try to find the pattern

"A" = 1 = 1
"Z" = 26 = 26
"AA" = 27 = 26 + 1
"AZ" = 52 = 26 + 26
"ZZ" = 702 = 26*26 + 26
"AAA" = 703 = 26*26 + 26 + 1

If you map 'A' = 1, ..., 'Z' = 26, the values can be rewritten as

"A" = 1 = 'A'
"Z" = 26 = 'Z'
"AA" = 27 = 26*'A' + 'A'
"AZ" = 52 = 26*'A' + 'Z'
"ZZ" = 702 = 26*'Z' + 'Z'
"AAA" = 703 = 26*26*'A' + 26*'A' + 'A'

In general the formula for a string columnTitle = abcd is

abcd = 26^3*a + 26^2*b + 26*c + d,

where a, b, c, d are some uppercase English letters A, ..., Z.

Longer columnTitles will have bigger leading exponents of 26.

15.1. Excel Sheet Column Number 299

Code

#include <iostream>
using namespace std;
int titleToNumber(const string& columnTitle) {
int column = 0;
for (auto& c : columnTitle) {
// The ASCII value of 'A' is 65.
column = 26*column + (c - 64);
}
return column;
}
int main() {
cout << titleToNumber("A") << endl;
cout << titleToNumber("AB") << endl;
cout << titleToNumber("ZY") << endl;
}

Output:
1
28
701

The solution calculates the decimal representation of the Excel column title by pro-
cessing each character and updating the result.

Complexity

• Runtime: O(N), where N = columnTitle.length.

• Extra space: O(1).

Implementation notes

1. There are many ways to compute the series

26^3*a + 26^2*b + 26*c + d.

If you write it as

300 Chapter 15. Mathematics

 26*(26*(26*(0 + a) + b) + c) + d,

you get the loop in the code above.

2. To map 'A' = 1, ..., 'Z' = 26, you can use their ASCII values ('A' = 65,
..., 'Z' = 90) minus 64.
3. The parentheses around (c - 64) is needed. Otherwise the value of
columnTitle = "FXSHRXW" makes 26*column + c exceed the limit of int be-
fore it subtracts 64.

15.1.3 Exercise

• Excel Sheet Column Title2 .

15.2 Power of Three

15.2.1 Problem statement

1
Given an integer n, return true if it is a power of three. Otherwise, return false.
An integer n is a power of three, if there exists an integer x such that n == 3^x.

Example 1

Input: n = 27
Output: true
Explanation: 27 = 3^3.

2 https://leetcode.com/problems/excel-sheet-column-title/
1 https://leetcode.com/problems/power-of-three/

15.2. Power of Three 301

Example 2

Input: n = 0
Output: false
Explanation: There is no x where 3^x = 0.

Example 3

Input: n = -1
Output: false
Explanation: There is no x where 3^x = (-1).

Constraints

• -2^31 <= n <= 2^31 - 1.

Follow up

• Could you solve it without loops/recursion?

15.2.2 Solution 1: Repeat the division

Code

#include <iostream>
using namespace std;
bool isPowerOfThree(int n) {
while (n % 3 == 0 && n > 0) {
n /= 3;
}
return n == 1;
}
int main() {
cout << isPowerOfThree(27) << endl;
cout << isPowerOfThree(0) << endl;
cout << isPowerOfThree(-1) << endl;
}

302 Chapter 15. Mathematics

Output:
1
0
0

This solution repeatedly divides the input by 3 until it either becomes 1 (indicating
that it was a power of three) or cannot be divided further by 3.

Complexity

• Runtime: O(logn).
• Extra space: O(1).

15.2.3 Solution 2: Mathematics and the constraints of the problem

A power of three must divide another bigger one, i.e. 3𝑥 |3𝑦 where 0 ≤ 𝑥 ≤ 𝑦.
Because the constraint of the problem is 𝑛 ≤ 231 − 1, you can choose the biggest
power of three in this range to test the others.
It is 319 = 1162261467. The next power will exceed 231 = 2147483648.

Code

#include <iostream>
using namespace std;
bool isPowerOfThree(int n) {
return n > 0 && 1162261467 % n == 0;
}
int main() {
cout << isPowerOfThree(27) << endl;
cout << isPowerOfThree(0) << endl;
cout << isPowerOfThree(-1) << endl;
}

Output:
1
0
0

15.2. Power of Three 303

This solution effectively checks whether n is a power of three by verifying if it is a
divisor of the largest power of three that fits within 32 bits. If the condition is met, it
returns true, indicating that n is a power of three; otherwise, it returns false.

Complexity

• Runtime: O(1).
• Extra space: O(1).

15.2.4 Readable code

Though Solution 2 offers a direct approach without the need for iteration, it is not
easy to understand like Solution 1, where complexity of O(logn) is not too bad.

15.2.5 Exercise

• Check if Number is a Sum of Powers of Three2 .

15.3 Best Time to Buy and Sell Stock

15.3.1 Problem statement

1
You are given an array prices where prices[i] is the price of a given stock on the
i-th day.
You want to maximize your profit by choosing a single day to buy one stock and
choosing a different day in the future to sell that stock.
Return the maximum profit you can achieve from this transaction. If you cannot
achieve any profit, return 0.
2 https://leetcode.com/problems/check-if-number-is-a-sum-of-powers-of-three/
1 https://leetcode.com/problems/best-time-to-buy-and-sell-stock/

304 Chapter 15. Mathematics

Example 1

Input: prices = [7,1,5,3,6,4]

Output: 5
Explanation: Buy on day 2 (price = 1) and sell on day 5 (price = 6),␣
˓→profit = 6-1 = 5.

Note that buying on day 2 and selling on day 1 is not allowed because␣
˓→you must buy before you sell.

Example 2

Input: prices = [7,6,4,3,1]

Output: 0
Explanation: In this case, no transactions are done and the max profit =␣
˓→0.

Constraints

• 1 <= prices.length <= 10^5.

• 0 <= prices[i] <= 10^4.

15.3.2 Solution 1: Bruteforce

For each day i, find the day j > i that gives maximum profit.

Code

#include <vector>
#include <iostream>
using namespace std;
int maxProfit(const vector<int>& prices) {
int maxProfit = 0;
for (int i = 0; i < prices.size(); i++) {
for (int j = i + 1; j < prices.size(); j++) {
if (prices[j] > prices[i]) {
maxProfit = max(maxProfit, prices[j] - prices[i]);
(continues on next page)

15.3. Best Time to Buy and Sell Stock 305

 (continued from previous page)
}
}
}
return maxProfit;
}
int main() {
vector<int> prices{7,1,5,3,6,4};
cout << maxProfit(prices) << endl;
prices = {7,6,4,3,1};
cout << maxProfit(prices) << endl;
}

Output:
5
0

This solution uses a brute force approach to find the maximum profit. It compares the
profit obtained by buying on each day with selling on all subsequent days and keeps
track of the maximum profit found.

Complexity

• Runtime: O(N^2), where N = prices.length.

• Extra space: O(1).

15.3.3 Solution 2: Smallest and largest prices

Given a past day i, the future day j > i that gives the maximum profit is the day that
has the largest price which is bigger than prices[i].
Conversely, given a future day j, the past day i < j that gives the maximum profit is
the day with the smallest price.

306 Chapter 15. Mathematics

Code

#include <vector>
#include <iostream>
#include <algorithm>
using namespace std;
int maxProfit(const vector<int>& prices) {
int maxProfit = 0;
int i = 0;
while (i < prices.size()) {
// while prices are going down,
// find the bottommost one to start
while (i < prices.size() - 1 && prices[i] >= prices[i + 1]) {
i++;
}
// find the largest price in the future
auto imax = max_element(prices.begin() + i, prices.end());
// find the smallest price in the past
auto imin = min_element(prices.begin() + i, imax);
maxProfit = max(maxProfit, *imax - *imin);
// next iteration starts after the found largest price
i = distance(prices.begin(), imax) + 1;
}
return maxProfit;
}
int main() {
vector<int> prices{7,1,5,3,6,4};
cout << maxProfit(prices) << endl;
prices = {7,6,4,3,1};
cout << maxProfit(prices) << endl;
prices = {2,4,1,7};
cout << maxProfit(prices) << endl;
prices = {2,4,1};
cout << maxProfit(prices) << endl;
}

Output:
5
0
6
2

15.3. Best Time to Buy and Sell Stock 307

This solution optimally finds the maximum profit by iterating through the array only
once, avoiding the need for nested loops.

Complexity

• Runtime: O(N), where N = prices.length.

• Extra space: O(1).

15.3.4 Solution 3: Only the smallest price

Given a future day j, the past day i that gives the maximum profit is the day with
minimum price.

Code

#include <vector>
#include <iostream>
using namespace std;
int maxProfit(const vector<int>& prices) {
int maxProfit = 0;
// keep track the minimum price so fat
int minPrice = prices[0];
for (int i = 1; i < prices.size(); i++) {
// update the minimum price
minPrice = min(minPrice, prices[i]);
maxProfit = max(maxProfit, prices[i] - minPrice);
}
return maxProfit;
}
int main() {
vector<int> prices{7,1,5,3,6,4};
cout << maxProfit(prices) << endl;
prices = {7,6,4,3,1};
cout << maxProfit(prices) << endl;
prices = {2,4,1,7};
cout << maxProfit(prices) << endl;
prices = {2,4,1};
cout << maxProfit(prices) << endl;
}

308 Chapter 15. Mathematics

Output:
5
0
6
2

This solution efficiently computes the maximum profit by iterating through the array
only once, maintaining the minimum buying price and updating the maximum profit
accordingly.

Complexity

• Runtime: O(N), where N = prices.length.

• Extra space: O(1).

15.3.5 Conclusion

The problem of finding the maximum profit that can be achieved by buying and selling
a stock can be efficiently solved using different approaches. Solutions 1, 2, and 3 each
offer a different approach to solving the problem, including brute-force iteration,
finding local minima and maxima, and maintaining a running minimum price.
Solution 3 stands out as the most efficient approach, achieving a linear time complex-
ity by iterating through the prices only once and updating the minimum price seen so
far. This approach avoids unnecessary comparisons and achieves the desired result in
a single pass through the array.

15.3.6 Exercise

• Best Time to Buy and Sell Stock II2 .

2 https://leetcode.com/problems/best-time-to-buy-and-sell-stock-ii/

15.3. Best Time to Buy and Sell Stock 309

15.4 Subsets

15.4.1 Problem Statement

1
Given an integer array nums of unique elements, return all possible subsets (the
power set).
The solution set must not contain duplicate subsets. Return the solution in any order.

Example 1

Input: nums = [1,2,3]

Output: [[],[1],[2],[1,2],[3],[1,3],[2,3],[1,2,3]]

Example 2

Input: nums = [1]

Output: [[],[1]]

Constraints

• 1 <= nums.length <= 10.

• -10 <= nums[i] <= 10.
• All the numbers of nums are unique.

15.4.2 Solution

You might need to find the relationship between the result of the array nums with the
result of itself without the last element.
1 https://leetcode.com/problems/subsets/

310 Chapter 15. Mathematics

Example 3

Input: nums = [1,2]

Output: [[],[1],[2],[1,2]]

You can see the powerset of Example 3 was obtained from the one in Example 2 with
additional subsets [2], [1,2]. These new subsets were constructed from subsets [],
[1] of Example 2 appended with the new element 2.
Similarly, the powerset of Example 1 was obtained from the one in Example 3 with the
additional subsets [3], [1,3], [2,3], [1,2,3]. These new subsets were constructed
from the ones of Example 3 appended with the new element 3.

Code

#include <vector>
#include <iostream>
using namespace std;
vector<vector<int>> subsets(const vector<int>& nums) {
vector<vector<int>> powerset = {{}};
int i = 0;
while (i < nums.size()) {
vector<vector<int>> newSubsets;
for (auto subset : powerset) {
subset.push_back(nums[i]);
newSubsets.push_back(subset);
}
powerset.insert(powerset.end(), newSubsets.begin(), newSubsets.
˓→end());

i++;
}
return powerset;
}
void print(const vector<vector<int>>& powerset) {
for (auto& set : powerset ) {
cout << "[";
for (auto& element : set) {
cout << element << ",";
}
cout << "]";
}
(continues on next page)

15.4. Subsets 311

 (continued from previous page)
cout << endl;
}
int main() {
vector<int> nums{1,2,3};
auto powerset = subsets(nums);
print(powerset);
nums = {1};
powerset = subsets(nums);
print(powerset);
}

Output:
[][1,][2,][1,2,][3,][1,3,][2,3,][1,2,3,]
[][1,]

Complexity

• Runtime: O(2^N), where N is the number of elements in nums, as it generates all

possible subsets.
• Extra space: O(2^N) due to the space required to store the subsets.

15.4.3 Conclusion

This solution generates subsets by iteratively adding each element of nums to the
existing subsets and accumulating the results.
Note that in for (auto subset : powerset) you should not use reference auto&
because we do not want to change the subsets that have been created.

15.4.4 Exercise

• Subsets II2 .

2 https://leetcode.com/problems/subsets-ii/

312 Chapter 15. Mathematics

15.5 Minimum Moves to Equal Array Elements II

15.5.1 Problem statement

1
Given an integer array nums of size n, return the minimum number of moves required
to make all array elements equal.
In one move, you can increment or decrement an element of the array by 1.

Example 1

Input: nums = [1,2,3]

Output: 2
Explanation:
Only two moves are needed (remember each move increments or decrements␣
˓→one element):

[1,2,3] => [2,2,3] => [2,2,2]

Example 2

Input: nums = [1,10,2,9]

Output: 16

Constraints

• n == nums.length.
• 1 <= nums.length <= 10^5.
• -10^9 <= nums[i] <= 10^9.
1 https://leetcode.com/problems/minimum-moves-to-equal-array-elements-ii/

15.5. Minimum Moves to Equal Array Elements II 313

15.5.2 Solution 1: Median - The math behind the problem

You are asked to move all elements of an array to the same value M. The problem can
be reduced to identifying what M is.
First, moving elements of an unsorted array and moving a sorted one are the same.
So you can assume nums is sorted in some order. Let us say it is sorted in ascending
order.
Second, M must be in between the minimum element and the maximum one. Appar-
ently!
We will prove that M will be the median2 of nums, which is nums[n/2] of the sorted
nums.
In other words, we will prove that if you choose M a value different from nums[n/2],
then the number of moves will be increased.
In fact, if you choose M = nums[n/2] + x, where x > 0, then:
• Each element nums[i] that is less than M needs more x moves, while each
nums[j] that is greater than M can reduce x moves.
• But the number of nums[i] is bigger than the number of nums[j].
• So the total number of moves is bigger.
The same arguments apply for x < 0.

Example 3

For nums = [0,1,2,2,10]. Its median is 2. The minimum number of moves is 2 + 1

+ 0 + 0 + 8 = 11.
If you choose M = 3 (the average value, the mean), the total number of moves is 3 +
2 + 1 + 1 + 7 = 14.

Code

#include <iostream>
#include <vector>
#include <algorithm>
using namespace std;
(continues on next page)
2 https://en.wikipedia.org/wiki/Median

314 Chapter 15. Mathematics

 (continued from previous page)
int minMoves2(vector<int>& nums) {
sort(nums.begin(), nums.end());
const int median = nums[nums.size() / 2];
int moves = 0;
for (int& a: nums) {
moves += abs(a - median);
}
return moves;
}
int main() {
vector<int> nums{1,2,3};
cout << minMoves2(nums) << endl;
nums = {1,10,2,9};
cout << minMoves2(nums) << endl;
}

Output:
2
16

This solution leverages the concept of the median to minimize the total absolute
differences between each element and the median, resulting in the minimum number
of moves to equalize the array.

Complexity

• Runtime: O(n*logn) due to the sorting step, where n is the number of elements
in the nums array.
• Extra space: O(1).

15.5.3 Solution 2: Using std::nth_element to compute the median

What you only need in Solution 1 is the median value. Computing the total number
of moves in the for loop does not require the array nums to be fully sorted.
In this case, you can use std::nth_element3 to reduce the runtime complexity.
3 https://en.cppreference.com/w/cpp/algorithm/nth_element

15.5. Minimum Moves to Equal Array Elements II 315

Code

#include <iostream>
#include <vector>
#include <algorithm>
using namespace std;
int minMoves2(vector<int>& nums) {
const int mid = nums.size() / 2;
// make sure all elements that are less than or equals to nums[mid]
// are on the left
std::nth_element(nums.begin(), nums.begin() + mid, nums.end());
const int median = nums[mid];
int moves = 0;
for (int& a: nums) {
moves += abs(a - median);
}
return moves;
}
int main() {
vector<int> nums{1,2,3};
cout << minMoves2(nums) << endl;
nums = {1,10,2,9};
cout << minMoves2(nums) << endl;
}

Output:
2
16

This solution efficiently finds the median of the nums array in linear time using
std::nth_element and then calculates the minimum number of moves to make all
elements equal to this median.

Complexity

• Runtime: O(n), where n = nums.length.

• Extra space: O(1).

316 Chapter 15. Mathematics

15.5.4 Modern C++ tips

In the code of Solution 2, the partial sorting algorithm std::nth_element will make
sure for all indices i and j that satisfy 0 <= i <= mid <= j < nums.length, then

nums[i] <= nums[mid] <= nums[j].

With this property, if mid = nums.length / 2, then the value of nums[mid] is un-
changed no matter how nums is sorted or not.

15.5.5 Exercise

• Minimum Moves to Equal Array Elements4 .

15.6 Array Nesting

15.6.1 Problem statement

1
You are given an integer array nums of length n where nums is a permutation of the
numbers in the range [0, n - 1].
You should build a set s[k] = {nums[k], nums[nums[k]], nums[nums[nums[k]]],
... } subjected to the following rule:
• The first element in s[k] starts with the element nums[k].
• The next element in s[k] should be nums[nums[k]], and then
nums[nums[nums[k]]], and so on.
• We stop adding elements before a duplicate element occurs in s[k].
Return the length of the longest set s[k].
4 https://leetcode.com/problems/minimum-moves-to-equal-array-elements/
1 https://leetcode.com/problems/array-nesting/

15.6. Array Nesting 317

Example 1

Input: nums = [5,4,0,3,1,6,2]

Output: 4
Explanation:
nums[0] = 5, nums[1] = 4, nums[2] = 0, nums[3] = 3, nums[4] = 1, nums[5]␣
˓→= 6, nums[6] = 2.

One of the longest sets s[k]:

s[0] = {nums[0], nums[5], nums[6], nums[2]} = {5, 6, 2, 0}

Example 2

Input: nums = [0,1,2]

Output: 1

Constraints:

• 1 <= nums.length <= 10^5.

• 0 <= nums[i] < nums.length.
• All the values of nums are unique.

15.6.2 Solution: Understanding the math behind

A permutation2 is a one-to-one mapping from a set of integers to itself.

The permutation on the set nums in this problem is defined by the mapping i ->
nums[i]. For instance in Example 1, the permutation is defined as following:

0 -> 5,
1 -> 4,
2 -> 0,
3 -> 3,
4 -> 1,
5 -> 6,
6 -> 2.

2 https://en.wikipedia.org/wiki/Permutation

318 Chapter 15. Mathematics

You can always rearrange the definition of a permutation into groups of cyclic chains
(factors).

0 -> 5, 5 -> 6, 6 -> 2, 2 -> 0,

1 -> 4, 4 -> 1,
3 -> 3

The set s[k] in this problem is such a chain. In mathematics, it is called a cycle;
because the chain (0, 5, 6, 2) is considered the same as (5, 6, 2, 0), (6, 2, 0,
5) or (2, 0, 5, 6) in Example 1.
Assume you have used some elements of the array nums to construct some cycles. To
construct another one, you should start with the unused elements.
The problem leads to finding the longest cycle of a given permutation.

Code

#include <vector>
#include <iostream>
#include <algorithm>
using namespace std;
int arrayNesting(const vector<int>& nums) {
int maxLen{0};
vector<bool> visited(nums.size());
for (auto& i : nums) {
if (visited[i]) {
continue;
}
int len{0};
// visit the cycle starting from i
while (!visited[i]) {
visited[i] = true;
i = nums[i];
len++;
}
maxLen = max(len, maxLen);
}
return maxLen;
}

int main() {
(continues on next page)

15.6. Array Nesting 319

 (continued from previous page)
vector<int> nums = {5,4,0,3,1,6,2};
cout << arrayNesting(nums) << endl;
nums = {0,1,2};
cout << arrayNesting(nums) << endl;
nums = {0,2,1};
cout << arrayNesting(nums) << endl;
nums = {2,0,1};
cout << arrayNesting(nums) << endl;
}

Output:
4
1
2
3

Complexity

• Runtime: O(n) where n is the size of the nums array.

• Extra space: much less than O(n) since vector<bool> is optimized for space
efficiency.

15.6.3 Conclusion

The problem of finding the length of the longest cycle in an array can be efficiently
solved using a cycle detection approach. This solution efficiently detects cycles in the
array by using a boolean array to mark visited elements.
By iterating through each element in the array and visiting the cycle starting from
each unvisited element, the solution identifies the length of each cycle and updates
the maximum length accordingly. This approach ensures that each cycle is visited
only once and maximizes the length of the longest cycle in the array.

320 Chapter 15. Mathematics

15.7 Count Sorted Vowel Strings

15.7.1 Problem statement

1
Given an integer n, return the number of strings of length n that consist only of
vowels (a, e, i, o, u) and are lexicographically sorted.
A string s is lexicographically sorted if for all valid i, s[i] is the same as or comes
before s[i+1] in the alphabet.

Example 1

Input: n = 1
Output: 5
Explanation: The 5 sorted strings that consist of vowels only are ["a","e
˓→","i","o","u"].

Example 2

Input: n = 2
Output: 15
Explanation: The 15 sorted strings that consist of vowels only are
["aa","ae","ai","ao","au","ee","ei","eo","eu","ii","io","iu","oo","ou",
˓→"uu"].

Note that "ea" is not a valid string since 'e' comes after 'a' in the␣
˓→alphabet.

Example 3

Input: n = 33
Output: 66045

1 https://leetcode.com/problems/count-sorted-vowel-strings/

15.7. Count Sorted Vowel Strings 321

Constraints

• 1 <= n <= 50.

15.7.2 Solution 1: Finding the pattern

Let us find the relationship of the strings between the vowels.

Example 3

For n = 3:
• There is (always) only one string starting from u, which is uuu.
• There are 3 strings starting from o: ooo, oou and ouu.
• There are 6 strings starting from i: iii, iio, iiu, ioo, iou, iuu.
• There are 10 strings starting from e: eee, eei, eeo, eeu, eii, eio, eiu, eoo, eou,
euu.
• There are 15 strings starting from a: aaa, aae, aai, aao, aau, aee, aei, aeo, aeu,
aii, aio, aiu, aoo, aou, auu.
• In total: there are 35 strings that satisfy the problem.

Findings

In Example 3, if you ignore the leading vowel of those strings, then the shorted strings
of the line above all appear in the ones of the line below and the remaining strings of
the line below come from n = 2.
More precisely:
• All the shorted strings oo, ou and uu starting from o appear on the ones starting
from i. The remaining ii, io, iu starting from i come from the strings of length
n = 2 (see Example 2).
• Similarly, all shorted strings ii, io, iu, oo, ou, uu starting from i appear on the
ones starting from e. The remaining ee, ei, eo, eu come from n = 2.
• And so on.
That leads to the following recursive relationship.
Let S(x, n) be the number of strings of length n starting from a vowel x. Then

322 Chapter 15. Mathematics

 • S('o', n) = S('o', n - 1) + S('u', n) for all n > 1.
• S('i', n) = S('i', n - 1) + S('o', n) for all n > 1.
• S('e', n) = S('e', n - 1) + S('i', n) for all n > 1.
• S('a', n) = S('a', n - 1) + S('e', n) for all n > 1.
• S(x, 1) = 1 for all vowels x.
• S('u', n) = 1 for all n >= 1.
For this problem, you want to compute

S(n) = S('a', n) + S('e', n) + S('i', n) + S('o', n) + S('u', n).

Code

#include <iostream>
using namespace std;
int countVowelStrings(int n) {
int a, e, i, o, u;
a = e = i = o = u = 1;
while (n > 1) {
o += u;
i += o;
e += i;
a += e;
n--;
}
return a + e + i + o + u;
}
int main() {
cout << countVowelStrings(1) << endl;
cout << countVowelStrings(2) << endl;
cout << countVowelStrings(33) << endl;
}

Output:
5
15
66045

15.7. Count Sorted Vowel Strings 323

This solution efficiently computes the count of vowel strings of length n using dynamic
programming, updating the counts based on the previous lengths to avoid redundant
calculations.

Complexity

• Runtime: O(n).
• Extra space: O(1).

15.7.3 Solution 2: The math behind the problem

The strings of length n you want to count are formed by a number of 'a', then some
number of 'e', then some number of 'i', then some number of 'o' and finally some
number of 'u'.
So it looks like this

s = "aa..aee..eii..ioo..ouu..u".

And you want to count how many possibilities of such strings of length n.
One way to count it is using combinatorics in mathematics.
If you separate the groups of vowels by '|' like this

s = "aa..a|ee..e|ii..i|oo..o|uu..u",

the problem becomes counting how many ways of putting those 4 separators '|' to
form a string of length n + 4.
In combinatorics, the solution is 𝑛+4
(︀ )︀ (︀𝑛)︀ 2
4 , where 𝑘 is the binomial coefficient :
(︂ )︂
𝑛 𝑛!
= .
𝑘 𝑘!(𝑛 − 𝑘)!

The final number of strings is

(︂ )︂
𝑛+4 (𝑛 + 4)! (𝑛 + 1)(𝑛 + 2)(𝑛 + 3)(𝑛 + 4)
= = .
4 4!𝑛! 24
2 https://en.wikipedia.org/wiki/Binomial_coefficient

324 Chapter 15. Mathematics

Code

#include <iostream>
using namespace std;
int countVowelStrings(int n) {
return (n + 1) * (n + 2) * (n + 3) * (n + 4) / 24;
}
int main() {
cout << countVowelStrings(1) << endl;
cout << countVowelStrings(2) << endl;
cout << countVowelStrings(33) << endl;
}

Output:
5
15
66045

Complexity

• Runtime: O(1).
• Extra space: O(1).

15.7.4 Conclusion

The problem of counting the number of strings of length n that consist of the vowels
‘a’, ‘e’, ‘i’, ‘o’, and ‘u’ in sorted order can be efficiently solved using combinatorial
techniques. Solution 1 uses dynamic programming to iteratively calculate the count
of strings for each length up to n, updating the counts based on the previous counts.
This approach efficiently computes the count of sorted vowel strings for the given
length n without requiring excessive memory usage or computational overhead.
Solution 2 offers a more direct approach by utilizing a combinatorial formula to cal-
culate the count of sorted vowel strings directly based on the given length n. By
leveraging the combinatorial formula, this solution avoids the need for iterative cal-
culations and achieves the desired result more efficiently.

15.7. Count Sorted Vowel Strings 325

15.8 Concatenation of Consecutive Binary Numbers

15.8.1 Problem statement

1
Given an integer n, return the decimal value of the binary string formed by concate-
nating the binary representations of 1 to n in order, modulo 10^9 + 7.

Example 1

Input: n = 1
Output: 1
Explanation: "1" in binary corresponds to the decimal value 1.

Example 2

Input: n = 3
Output: 27
Explanation: In binary, 1, 2, and 3 corresponds to "1", "10", and "11".
After concatenating them, we have "11011", which corresponds to the␣
˓→decimal value 27.

Example 3

Input: n = 12
Output: 505379714
Explanation: The concatenation results in
˓→"1101110010111011110001001101010111100".

The decimal value of that is 118505380540.

After modulo 10^9 + 7, the result is 505379714.

1 https://leetcode.com/problems/concatenation-of-consecutive-binary-numbers/

326 Chapter 15. Mathematics

Constraints

• 1 <= n <= 10^5.

15.8.2 Solution: Recursive

There must be some relationship between the result of n and the result of n - 1.
First, let us list some first values of n.
• For n = 1: the final binary string is "1", its decimal value is 1.
• For n = 2: the final binary string is "110", its decimal value is 6.
• For n = 3: the final binary string is "11011", its decimal value is 27.
Look at n = 3, you can see the relationship between the decimal value of "11011" and
the one of "110" (of n = 2) is:

27 = 6 * 2^2 + 3
Dec("11011") = Dec("110") * 2^num_bits("11") + Dec("11")
Result(3) = Result(2) * 2^num_bits(3) + 3.

The same equation for n = 2:

6 = 1 * 2^2 + 2
Dec("110") = Dec("1") * 2^num_bits("10") + Dec("10")
Result(2) = Result(1) * 2^num_bits(2) + 2.

In general, the recursive relationship between n and n - 1 is:

Result(n) = Result(n - 1) * 2^num_bits(n) + n.

Code

#include <cmath>
#include <iostream>
int concatenatedBinary(int n) {
unsigned long long result = 1;
for (int i = 2; i <= n; i++) {
const int num_bits = std::log2(i) + 1;
result = ((result << num_bits) + i) % 1000000007;
(continues on next page)

15.8. Concatenation of Consecutive Binary Numbers 327

 (continued from previous page)
}
return result;
}
int main() {
std::cout << concatenatedBinary(1) << std::endl;
std::cout << concatenatedBinary(3) << std::endl;
std::cout << concatenatedBinary(12) << std::endl;
}

Output:
1
27
505379714

Complexity

• Runtime: O(n*logn).
• Extra space: O(1).

15.8.3 Conclusion

This solution efficiently calculates the concatenated binary representation of integers

from 1 to n, using bitwise operations and modular arithmetic. Note that a << t is
equivalent to a * 2^t.

15.9 Perfect Squares

15.9.1 Problem statement

1
Given an integer n, return the least number of perfect square numbers that sum to n.
A perfect square is an integer that is the square of an integer; in other words, it is the
product of some integer with itself. For example, 1, 4, 9, and 16 are perfect squares
while 3 and 11 are not.
1 https://leetcode.com/problems/perfect-squares/

328 Chapter 15. Mathematics

Example 1

Input: n = 9
Output: 1
Explanation: 9 is already a perfect square.

Example 2

Input: n = 13
Output: 2
Explanation: 13 = 4 + 9.

Example 3

Input: n = 7
Output: 4
Explanation: 7 = 4 + 1 + 1 + 1.

Example 4

Input: n = 12
Output: 3
Explanation: 12 = 4 + 4 + 4.

Constraints

• 1 <= n <= 10^4.

15.9.2 Solution 1: Dynamic Programming

Let us call the function to be computed numSquares(n), which calculates the least
number of perfect squares that sum to n.
Here are the findings.
1. If n is already a perfect square then numSquares(n) = 1.

15.9. Perfect Squares 329

 2. Otherwise, it could be written as n = 1 + (n-1), or n = 4 + (n-4), or n = 9
+ (n-9), etc. which means n is a sum of a perfect square (1, 4 or 9, etc.) and
another number m < n. That leads to the problems numSquares(m) of smaller
values m.
3. If you have gotten the results of the smaller problems numSquares(n-1),
numSquares(n-4), numSquares(n-9), etc. then numSquares(n) = 1 + the
minimum of those results.

Example 4

n = 12 is not a perfect square. It can be written as n = 1 + 11 = 4 + 8 = 9 + 3.

• For m = 11, it is not a perfect square and can be written as m = 1 + 10 = 4 + 7
= 9 + 2.
• For m = 8, it is not a perfect square and can be written as m = 1 + 7 = 4 + 4
(matched). You get numSquares(8) = 2.
• For m = 3, it is not a perfect square and can be written as m = 1 + 2.
You can continue to compute numSquares(m) for other values m in this recursive
process. But you can see the case of m = 8 was already the best solution. And
numSquares(12) = 1 + numSquares(8) = 1 + 2 = 3, which is the case of n = 12 = 4
+ 4 + 4.
To improve runtime, you can apply dynamic programming to cache the numSquares(n)
that you have computed.

Code

#include <iostream>
#include <cmath>
#include <unordered_map>
using namespace std;
//! @return the least number of perfect squares that sum to n
//! @param[out] ns a map stores all intermediate results
int nsq(int n, unordered_map<int, int>& ns) {
auto it = ns.find(n);
if (it != ns.end()) {
return it->second;
}
const int sq = sqrt(n);
(continues on next page)

330 Chapter 15. Mathematics

 (continued from previous page)
if (sq * sq == n) {
// n is already a perfect square
ns[n] = 1;
return 1;
}
// if n is written as 1 + 1 + .. + 1,
// maximum of result is n
int result = n;
// finding the minimum nsq(n - i*i) across all i <= sqrt(n)
for (int i = 1; i <= sq; i++) {
//
result = min(result, nsq(n - i*i, ns));
}
// write n as imin^2 + (n - imin^2)
ns[n] = result + 1;
return ns[n];
}
int numSquares(int n) {
unordered_map<int, int> ns;
return nsq(n, ns);
}
int main() {
cout << numSquares(12) << endl;
cout << numSquares(13) << endl;
}

Output:
3
2

The key idea of this algorithm is to build the solution incrementally, starting from
the smallest perfect squares, and use memoization to store and retrieve intermediate
results. By doing this, it efficiently finds the minimum number of perfect squares
required to sum up to n.

15.9. Perfect Squares 331

Complexity

• Runtime: O(n*sqrt(n)) = O(n^(3/2)) due to the nested loops and recursive

calls.
• Extra space: O(n).

15.9.3 Solution 2: Number Theory

The dynamic programming solution above is good enough. But for those who are
interested in Algorithmic Number Theory, there is a very interesting theorem that can
solve the problem directly without recursion.
It is called Lagrange’s Four-Square Theorem2 , which states
every natural number can be represented as the sum of four integer squares.
It was proven by Lagrange in 1770.

Example 4

n = 12 = 4 + 4 + 4 + 0 or 12 = 1 + 1 + 1 + 9.
Applying to our problem, numSquares(n) can only be 1, 2, 3, or 4. Not more.
It turns into the problem of
identifying when numSquares(n) returns 1, 2, 3, or 4.
Here are the cases.
1. If n is a perfect square, numSquares(n) = 1.
2. There is another theorem, Legendre’s Three-Square Theorem3 , which states that
numSquares(n) cannot be 1, 2, or 3 if n can be expressed as
𝑛 = 4𝑎 (8 · 𝑏 + 7),
where 𝑎, 𝑏 are nonnegative integers.
In other words, numSquares(n) = 4 if n is of this form.
2 https://en.wikipedia.org/wiki/Lagrange%27s_four-square_theorem
3 https://en.wikipedia.org/wiki/Legendre%27s_three-square_theorem

332 Chapter 15. Mathematics

Example 3

𝑛 = 7 = 40 (8 · 0 + 7). It can only be written as 7 = 4 + 1 + 1 + 1.

Code

#include <iostream>
#include <cmath>
using namespace std;
bool isSquare(int n) {
int sq = sqrt(n);
return sq * sq == n;
}
int numSquares(int n) {
if (isSquare(n)) {
return 1;
}
// Legendre's three-square theorem
int m = n;
while (m % 4 == 0) {
m /= 4;
}
if (m % 8 == 7) {
return 4;
}
const int sq = sqrt(n);
for (int i = 1; i <= sq; i++) {
if (isSquare(n - i*i)) {
return 2;
}
}
return 3;
}
int main() {
cout << numSquares(12) << endl;
cout << numSquares(13) << endl;
}

Output:
3
2

15.9. Perfect Squares 333

This solution finds the minimum number of perfect squares required to sum up to the
given integer n by first applying mathematical properties and Legendre’s three-square
theorem to simplify the problem and then using a loop to find possible combinations
of two perfect squares.

Complexity

• Runtime: O(sqrt(n)) = O(n^(1/2)), and it does not require dynamic program-

ming or memoization.
• Extra space: O(1).

15.9.4 Solution 3: Further performance improvement

Lagrange’s Four-Square Theorem and Legendre’s Three-Square Theorem are so pow-

erful to solve this problem. But you can still do a little more algebra to improve
further the runtime of the implementation above.
Instead of looping over sqrt(n) in the final for loop, we will prove that this loop over
sqrt(m) is enough. That will improve runtime a lot since m is much less than n.
Let m be the reduced value of n after the Legendre’s while loop. It satisfies

𝑛 = 4𝑎 · 𝑚.

We will prove that numSquares(n) = numSquares(m).

In fact, if m is written as 𝑚 = 𝑥2 + 𝑦 2 + 𝑧 2 , where 𝑥, 𝑦, 𝑧 are nonnegative integers.
Then

𝑛 = 4𝑎 · 𝑚 = (2𝑎 )2 · 𝑚 = (2𝑎 · 𝑥)2 + (2𝑎 · 𝑦)2 + (2𝑎 · 𝑧)2 .

In other words, numSquares(n) = numSquares(m).

Now you can change directly the value n during the Legendre’s while loop without
affecting the final result.

334 Chapter 15. Mathematics

Code

#include <iostream>
#include <cmath>
using namespace std;
bool isSquare(int n) {
int sq = sqrt(n);
return sq * sq == n;
}
int numSquares(int n) {
if (isSquare(n)) {
return 1;
}
// Legendre's three-square theorem
while (n % 4 == 0) {
n /= 4;
}
if (n % 8 == 7) {
return 4;
}
const int sq = sqrt(n);
for (int i = 1; i <= sq; i++) {
if (isSquare(n - i*i)) {
return 2;
}
}
return 3;
}
int main() {
cout << numSquares(12) << endl;
cout << numSquares(13) << endl;
}

Output:
3
2

15.9. Perfect Squares 335

Complexity

• Runtime: O(sqrt(n)) = O(n^(1/2)).

• Extra space: O(1).

15.9.5 Conclusion

• The title of this coding challenge (Perfect squares) gives you a hint it is more
about mathematics than coding technique.
• It is amazing from Lagrange’s Four-Square Theorem there are only four possi-
bilities for the answer to the problem. Not many people knowing it.
• You can get an optimal solution to a coding problem when you know something
about the mathematics behind it.
Hope you learn something interesting from this code challenge.
Have fun with coding and mathematics!

15.9.6 Exercise

• Ways to Express an Integer as Sum of Powers4 .

4 https://leetcode.com/problems/ways-to-express-an-integer-as-sum-of-powers/

336 Chapter 15. Mathematics

 CHAPTER

SIXTEEN

CONCLUSION

Congratulations! You have made it to the end of this book! I hope you have enjoyed
and learned from the coding challenges and solutions presented in this book.
Through these challenges, you have not only improved your coding skills but also
your problem-solving abilities, logical thinking, and creativity. You have been
exposed to different programming techniques and algorithms, which have broad-
ened your understanding of the programming world. These skills and knowledge will
undoubtedly benefit you in your future coding endeavors.
Remember, coding challenges are not only a way to improve your coding skills but
also a fun and engaging way to stay up-to-date with the latest technology trends.
They can also help you to prepare for technical interviews, which are a crucial part of
landing a programming job.
In conclusion, I encourage you to continue exploring the world of coding challenges,
as there is always something new to learn and discover. Keep practicing, keep learn-
ing, and keep challenging yourself. With hard work and dedication, you can be-
come an expert in coding and a valuable asset to any team.

337
338 Chapter 16. Conclusion
 APPENDIX

CODING CHALLENGE BEST PRACTICES

Here are some best practices to keep in mind when working on coding challenges:

A.1 Read the problem carefully

Before jumping into writing code, take the time to read and understand the prob-
lem statement. Make sure you understand the input and output requirements, any
constraints or special cases, and the desired algorithmic approach.

A.2 Plan and pseudocode

Once you understand the problem, take some time to plan and sketch out a high-level
algorithmic approach. Write pseudocode to help break down the problem into smaller
steps and ensure that your solution covers all cases.

A.3 Test your code

After writing your code, test it thoroughly to make sure it produces the correct output
for a range of test cases. Consider edge cases, large inputs, and unusual scenarios to
make sure your solution is robust.

339
A.4 Optimize for time and space complexity

When possible, optimize your code for time and space complexity. Consider the Big
O notation of your solution and try to reduce it if possible. This can help your code
to run faster and more efficiently.

A.5 Write clean, readable code

Make sure your code is easy to read and understand. Use meaningful variable names,
indent properly, and comment your code where necessary. This will make it easier for
other programmers to read and understand your code, and will help prevent errors
and bugs.

A.6 Submit your code and learn from feedback

Once you have a working solution, submit it for review and feedback. Pay attention
to any feedback you receive and use it to improve your coding skills and approach for
future challenges.

A.7 Keep practicing

The more coding challenges you complete, the better you will become. Keep prac-
ticing and challenging yourself to learn new techniques and approaches to problem-
solving.
In conclusion, coding challenges are a great way to improve your coding skills and
prepare for technical interviews. By following these best practices, you can ensure
that you approach coding challenges in a structured and efficient manner, producing
clean and readable code that is optimized for time and space complexity.

340
 THANK YOU!

Dear Readers,
Thank you for reading my book! Your support means the world to me. If you enjoyed
the book and would like to share your thoughts, I would greatly appreciate it if you
could leave a review and join our vibrant reader community. Your feedback and
engagement help shape the future of our community, and I’m excited to connect with
you further.
Again, Thank you for your support and for participating in this journey.
I hope it has been a valuable experience and that you are excited to continue your
coding journey. Best of luck with your coding challenges, and remember to have fun
along the way!
Warm regards,
Nhut Nguyen, Ph.D.

341
342
 ABOUT THE AUTHOR

Nhut Nguyen is a seasoned software engineer and career coach with nearly a decade
of experience in the tech industry.
He was born and grew up in Ho Chi Minh City, Vietnam. In 2012, he moved to
Denmark for a Ph.D. in mathematics at the Technical University of Denmark. After
his study, Nhut Nguyen switched to the industry in 2016 and has worked at various
tech companies in Copenhagen, Denmark.
Nhut’s passion for helping aspiring software engineers succeed led him to write sev-
eral articles and books where he shares his expertise and practical strategies to help
readers navigate their dream job and work efficiently.
With a strong background in mathematics and computer science and a deep under-
standing of the industry, Nhut is dedicated to empowering individuals to unlock their
full potential, land their dream jobs and live happy lives.
Learn more at nhutnguyen.com.

343
 INDEX

A P
algorithm complexity, 2 Partial sort, 162
permutation, 318
B power set, 310
binomial coefficient, 324
bit masking, 156 R
bitwise AND, 143 readable code, 4
bitwise XOR, 143, 234
S
C sliding window, 74, 99
Coding challenges, 1 Sorting, 149
std::accumulate, 104
D std::bitset, 152
dictionary order, 172 std::nth_element, 163
dummy node, 45 std::priority_queue, 123, 128, 132,
138
F std::set, 177
Fast and Slow, 266, 272 std::sort, 201
Fibonacci Number, 204 std::stoi, 104
std::swap, 26
K
Kadane's algorithm, 250

L
LeetCode, 1

M
memoization, 219
Morse Code, 87

344