JavaScript Algorithms

JavaScript Algorithms
The Web Developer’s Guide to Data Structures and Algorithms
Written by Oleksii Trekhleb and Sophia Shoemaker

Edited by Nate Murray
© 2019 Fullstack.io
All rights reserved. No portion of the book manuscript may be reproduced, stored in a retrieval
system, or transmitted in any form or by any means beyond the number of purchased copies,
except for a single backup or archival copy. The code may be used freely in your projects,
commercial or otherwise.
The authors and publisher have taken care in preparation of this book, but make no expressed
or implied warranty of any kind and assume no responsibility for errors or omissions. No
liability is assumed for incidental or consequential damagers in connection with or arising out
of the use of the information or programs container herein.
Published by Fullstack.io.
FULLSTACK .io
Contents
Book Revision . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
EARLY RELEASE VERSION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Join Our Discord . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Bug Reports . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Be notified of updates via Twitter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
We’d love to hear from you! . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
PRE-RELEASE VERSION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Join Our Discord . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
How To Read This Book . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Algorithms and Their Complexities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

What is an Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
Algorithm Complexity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
Big O Notation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
Quiz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
Linked List . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
Linked list and its common operations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
Implementation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
Complexities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
Problems Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
Quiz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
Queue . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
Queue and its common operations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
When to use a Queue . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
Usage Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
Implementation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
Complexities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
Quiz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
CONTENTS
Stack . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
Stack and its common operations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
Usage Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
Implementation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
Complexities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
Quiz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
Hash Table . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
Hash Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
Collision Resolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
Implementation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
Operations Time Complexity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
Quiz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
Binary Search Tree (BST) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

Tree . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
Binary Tree . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
Binary Search Tree . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
Application . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
Basic Operations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
Usage Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
Implementation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
Quiz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
Binary Heap . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
Application . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
Usage Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
Implementation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
Complexities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
Quiz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
Priority Queue . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105

Application . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
CONTENTS

Usage Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
Implementation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
Complexities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
Quiz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
Graphs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
Application . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
Graph Representation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
Usage Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
Implementation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
Quiz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
Bit Manipulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133

Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
Code . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
Quiz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
Factorial . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
Intro . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
Recursion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
Code . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
Quiz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152
Fibonacci Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153

Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
Code . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
Primality Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158

Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160
Code . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160
CONTENTS

Quiz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
Is a power of two . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163

The Task . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
Naive solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
Bitwise solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
Quiz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 166
Search. Linear Search. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167

The Task . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
The Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
Application . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
Usage Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
Implementation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
Complexity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
Quiz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
Search. Binary Search. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175

The Task . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
The Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
Algorithm Complexities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176
Application . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177
Usage Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177
Implementation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
Quiz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
Sets. Cartesian Product. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181

Sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
Cartesian Product . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
Usage Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
Implementation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
Complexities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184
Quiz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
CONTENTS
Sets. Power Set. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186

Usage Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188
Implementation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189
Complexities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194
Quiz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
Sets. Permutations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196

Permutations With Repetitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
Permutations Without Repetitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
Application . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
Usage Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
Implementation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
Quiz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206
Sets. Combinations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207

Combinations Without Repetitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
Combinations With Repetitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
Application . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210
Usage Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210
Implementation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212
Quiz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
Sorting: Quicksort . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217

The Task . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
The Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
Usage Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218
Implementation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220
Complexities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
Quiz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222
Trees. Depth-First Search. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223

The Task . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223
The Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225
Usage Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226
Implementation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227
Complexities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229
CONTENTS

Quiz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 230
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 230
Trees. Breadth-First Search. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231

The Task . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
The Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 233
Usage Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 233
Implementation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235
Complexities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237
Quiz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
Graphs. Depth-First Search. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239

The Task . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239
The Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 240
Application . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 241
Usage Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 241
Implementation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 244
Complexities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 246
Quiz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247
Graphs. Breadth-First Search. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 248

The Task . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 248
The Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 248
Application . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 250
Usage Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 250
Implementation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253
Complexities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255
Quiz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 256
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 256
Dijkstra’s Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 257

The Task . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 257
The Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 258
Application . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 264
Usage Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 265
Implementation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 268
Complexities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 272
CONTENTS
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273
Appendix A: Quiz Answers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 274
Appendix B: Big O Times Comparison . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 278
Appendix C: Data Structures Operations Complexities . . . . . . . . . . . . . . . . . . . . . . . 280

Common Data Structures Operations Complexities . . . . . . . . . . . . . . . . . . . . . . . . 280
Graph Operations Complexities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 280
Heap Operations Complexities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 280
Appendix D: Array Sorting Algorithms Complexities . . . . . . . . . . . . . . . . . . . . . . . . 281
Changelog . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 282
Revision 2 (11-25-2019) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 282
Revision 1 (10-29-2019) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 282
CONTENTS 1
Book Revision
Revision 1 - EARLY RELEASE - 2019-11-25
EARLY RELEASE VERSION

This version of the book is an early release. Most, but not all, of the chapters have been edited.
Contents will change before First Edition.
Join Our Discord

Come chat with other readers of the book in the official newline Discord channel:
Join here: https://newline.co/discord/algorithms¹
Bug Reports
If you’d like to report any bugs, typos, or suggestions just email us at: us@fullstack.io.
Be notified of updates via Twitter

If you’d like to be notified of updates to the book on Twitter, follow us at @fullstackio².
We’d love to hear from you!

Did you like the book? Did you find it helpful? We’d love to add your face to our list of testimonials
on the website! Email us at: us@fullstack.io³.
¹https://newline.co/discord/algorithms
²https://twitter.com/fullstackio
³mailto:us@fullstack.io
CONTENTS 1
PRE-RELEASE VERSION
This is a pre-release version of the book.
Most, but not all, of the chapters have been edited.
Join Our Discord

Come chat with other readers of the book in the official newline Discord channel:
Join here: https://newline.co/discord/algorithms⁴
Introduction
As a JavaScript developer you might think you don’t need to learn about data structures or
algorithms.
Did you know JavaScript itself (runtime + browser) uses things like a stack, a heap and a queue? The
more you understand the JavaScript you use on a daily basis – really understand it, the better you
can wield it’s power and make less mistakes. If you thought algorithms and data structures were
just a computer science course that wasn’t necessary on a day to day basis as JavaScript developer,
think again!
How To Read This Book

Each chapter covers an algorithm you might encounter in your work or at an interview. You can
work through the algorithms one at a time or feel free to jump around.
⁴https://newline.co/discord/algorithms
Algorithms and Their Complexities
What is an Algorithm
In this book, we’re going to write algorithms - the sets of steps that will solve specific problems for
us.
We constantly use algorithms in our everyday life. For example what if we’re somewhere on a street
and want to get home to our apartment that is on the 20th floor? In order to achieve that we could
do the following:
1. Take a WALK to your home.

2. Use STAIRS to go up to the 20th floor.
This is an algorithm. We’ve defined and used the set of steps that solve our task of getting home.
Algorithm Complexity
Let’s take the example of “getting home” issue from the previous section. Is there another way of
solving it? Yes, we implement different steps (read “algorithm”) to achieve that goal:
1. Call a CAB that will drive you home.

2. Use the ELEVATOR to get upstairs to the 20th floor.
Now, let’s think about which of these two algorithms we should choose:
First, we need to clarify our restrictions (how much money do we have, how urgently we need to
get home, how healthy we are) and evaluate algorithms from the perspective of these restrictions.
As you can see each of these two algorithms requires a different amount of time and resources. The
first one will require more time and energy but less money. The second one will require less time
and energy but more money. And if we don’t have money and are pretty healthy we need to choose
the first algorithm and enjoy the walk. Otherwise, if we have enough money and need to get home
urgently the second option is better.
What we’ve just done here is we’ve evaluated the complexity of each algorithm.
Algorithms and Their Complexities 3
Time and Space Complexities

In the previous example, we’ve evaluated time and resources (i.e. money and/or health) that each
algorithm requires. We will refer to these evaluations as time complexity and resource complexity.
These two characteristics are crucial for us to decide which algorithm suits our needs.
In this book, we won’t deal with “getting home” problems from the perspective of daily routine
planning where we may apply our intuition only. Rather, we will deal with computational issues
that are being solved on computer devices (i.e. finding the shortest path home in GPS navigator)
where we must express our intuition about algorithm complexity in a much more strict, objective
and mathematical way. When we speak about algorithms we normally use such metrics as time
complexity (how much time the algorithm requires to solve a problem) and space complexity (how
much memory the algorithm requires to solve a problem). These two metrics need to be evaluating
to make a decision of which algorithm is better.
The question is how we may express the values of time and space complexities in a more formal
mathematical way. And here is where big O notation comes to the rescue…
Big O Notation
Big O Notation Definition

Big O notation is used to classify algorithms according to how their running time or space
requirements grow as the input size grows.
The key here is that Big O notation is a relative metric that shows how fast execution time of the
algorithm or its consumed memory will grow depending on input size growth.
Absolute values of time and memory vary on different hardware so we need to use a relative metric.
The same program may be executed in 1 second on one computer and in 10 seconds on another
one. Thus when you compare time complexities of two algorithms it is required for them to be
running on the same predefined hardware configuration which is not convenient and sometimes is
not reproducible.
Big O notation defines relationship input size � spent time for time complexity and input size
� consumed memory for space complexity. It characterizes programs (functions) according to their
growth rates: different programs with the same growth rate may be represented using the same O
notation.
The letter O is used because the growth rate of a program (function) is also referred to as the order
of the function.
Big O Notation Explanation

Let’s illustrate the Big O concept on the “getting home” algorithm that was mentioned above,
specifically the “stair climbing” portion of the algorithm.
Let’s imagine we have two men: one is a young and fast Olympic champion and the other one is
your typical middle-aged male. We’ve just agreed that we want the complexity of “stairs climbing”
algorithm to be the same for both men so it should not depend on a man’s speed but rather depends
on how high the two men climb. In terms of a computer, our algorithm analysis will not depend on
the hardware, but on the software problem itself. How many steps would these two men need to
take to get to the 20ʰ floor? Well, let’s assume that we have 20 steps between each floor. That means
the two men need to take 20 * 20 = 400 steps to get to the 20ʰ floor. Let’s generalize it for n's floor:
Input: (20 * n) steps

Output: (20 * n) steps (moves)
Do you see the pattern here? The time complexity of climbing the stairs has a linear dependency on
the number of floors. In Big O syntax, this is written as follows:
O(20 * n)
One important rule to understand here:
We must always get rid of non-dominant parts when describing complexity using Big O
notation: O(n + 1) becomes O(n), O(n + n^2) becomes O(n^2), O(n + log(n)) becomes
O(n) and so on.
Why is that? Because Big O must describe the order of the function and not an exact number of
milliseconds or megabytes it consumes. It describes the trend and not absolute values. Therefore
non-dominant constants do not matter. Non-dominant constants are the ones we may discard for
huge values of n. For example in a situation when n = 10000 which of these two parts will influence
the final number more: 20 or n = 10000? It is obvious that the influence of n to the final number is
500 times bigger than the constant 20. Therefore, we may discard it and the time complexity of the
“stairs climbing” algorithm is as follows:
O(n)
This was an example of linear dependency, but there are other types of dependency exists as well:
O(log(n)), O(n * log(n)), O(n^2), O(2^n), O(n!) etc.
Take a look at the graphs of functions commonly used in the analysis of algorithms, showing the
number of operations N versus input size n for each function.
All these graphs show us how fast our function’s memory and time consumption will grow
depending on the input. Two different array sorting algorithms may accomplish the same task of
sorting but one will do it in O(n * log(n)) and another one in O(n^2) time. If we have a choice, we
would prefer the first one over the second one.
With Big O notation we may compare algorithms (functions, programs) based on their Big O value
independently of the hardware they are running on.
Take a look at the list of some of the most common Big O notations and their performance
comparisons against different sizes of the input data. This will give you the feeling of how different
algorithms complexities (time and memory consumptions) may be.
Big O Notation Computations for Computations for Computations for

10 elements 100 elements 1000 elements
O(1) 1 1 1
O(log(n)) 3 6 9
O(n) 10 100 1000
O(n log(n)) 30 600 9000
O(n²) 100 10000 1000000
O(2ⁿ) 1024 1.26e+29 1.07e+301
O(n!) 3628800 9.3e+157 4.02e+2567
Big O Notation Examples

Let’s move on and see some other examples of big O notation.
O(1) example
Let’s write a function that raises a number to a specified power:
01-algorithms-and-their-complexities/fastPower.js
1 /**
2 * Raise number to the power.
3 *
4 * Example:
5 * number = 3
6 * power = 2
7 * output = 3^2 = 9
8 *
9 * @param {number} number
10 * @param {number} power
11 * @return {number}
12 */
13 export function fastPower(number, power) {
14 return number ** power;
15 }
What would be its time and space complexities?

Let’s think about how the input of the function will affect the number of operations it will
accomplish. For every input, the function does exactly one operation. That would mean that time
complexity of this function is:
Time complexity: O(1)
Now let’s roughly analyze how much memory it will require to evaluate this function. In this case,
the number of variables we operate with doesn’t depend on the input. So the space complexity is:
Space complexity: O(1)
O(n) example
And now let’s implement the same power function but in an iterative way:
01-algorithms-and-their-complexities/iterativePower.js
1 /**
2 * Raise number to the power.
3 *
4 * Example:
5 * number = 3
6 * power = 2
7 * output = 3^2 = 9
8 *
10 * @param {number} power
12 */
13 export function iterativePower(number, power) {
14 let result = 1;
15
16 for (let i = 0; i < power; i += 1) {
17 result *= number;
18 }
19
20 return result;
21 }
The function still does the same. But let’s see how many operations it will perform depending on
the input.
All operations outside and inside the for() loop require constant time and space. There are just
assignment, multiplication and return operations. But for() loop itself acts as a multiplier for time
complexity of the code inside of it because it will make the body of the loop to be executed power
times. So the time complexity of this function may be expressed as a sum of time complexities of
its code parts like O(1) + power * O(1) + O(1). Where the first and last O(1) in the equation are
for assignment and return operations. After we drop non-dominant parts like O(1) and move power
into O() brackets we will get our time complexity as:
Time complexity: O(power)
Notice that we’ve written O(power) and not O(number). This is important since we’re showing that
the execution time of our function directly depends on the power input. A big number input won’t
slow the function down but big power input will.
From the perspective of space complexity, we may conclude that our function won’t create additional
variables or stack layers with bigger input. Amount of memory it consumes does not change.
Space complexity: O(1)
O(n) recursive example

In mathematics, the factorial of a non-negative integer n, denoted by n!, is the product of all positive
integers less than or equal to n. For example:
5! = 1 * 2 * 3 * 4 * 5 = 120
Let’s write recursive factorial function:

01-algorithms-and-their-complexities/factorial.js
1 /**
2 * Calculate factorial.
3 *
4 * Example:
5 * number = 5
6 * output = 120
7 *
10 */
11 export function factorial(number) {
12 if (number === 0) {
13 return 1;
14 }
15
16 return factorial(number - 1) * number;
17 }
If we were not dealing with recursion we would say that all operations inside the function have O(1)
time complexity and thus overall time complexity is also O(1) (the sum of all O(1)s with discarded
non-dominant parts). But we do have recursion call here at the very last line of the function. Let’s
analyze it and see the tree of recursion calls for 4!:
factorial(4)
� factorial(3) * 4
You see that from time perspective our recursion acts like a for() loop from the previous example
multiplying time complexity of function body by number. Thus time complexity is calculated as
follows:
Time complexity: O(number)
You may also notice that recursion makes the stack of function calls grow proportionally to the
number. This happens because to calculate factorial(4) computer needs to calculate factorial(3)
first and so on. Thus all intermediate calculations and variables must be stored in memory before the
next one in the stack will be calculated. This will lead us to the conclusion that space complexity is
in linear proportion to the number. The space complexity of one function call must be multiplied by
the number of function calls. Since one call of factorial() function would cost us constant memory
(because the number of variables and their sizes is constant) then we may calculate overall recursive
space complexity as follows:
Space complexity: O(number)
O(n²) example
Let’s write a function that generates all possible pairs out of provided letters:
01-algorithms-and-their-complexities/pairs.js
1 /**
2 * Get all possible pairs out of provided letters.
3 *
4 * Example:
5 * letter = ['a', 'b']
6 * output = ['aa', 'ab', 'ba', 'bb']
7 *
8 * @param {string[]} letters
9 * @return {string[]}
10 */
11 export function pairs(letters) {
12 const result = [];
13
14 for (let i = 0; i < letters.length; i += 1) {
15 for (let j = 0; j < letters.length; j += 1) {
16 result.push(`${letters[i]}${letters[j]}`);
17 }
18 }
19
20 return result;
21 }
Now we have two nested for() loops that acts like multipliers for the code inside the loops. Inside
the loops we have a code with constant execution time O(1) - it is just a pushing to the array.
So let’s calculate our time complexity as a sum of our function parts time complexities: O(1) +
letters.length * letters.length * O(1) + O(1). After dropping non-dominant parts we’ll get
our final time complexity:
Time complexity: O(letters.length^2)
This would mean that our function will slow down proportionally to the square of letters number
we will provide for it as an input.
You may also notice that the more letters we will provide for the function as an input the more pairs
it will generate. All those pairs are stored in pairs array. Thus this array will consume the memory
that is proportional to the square of letters number.
Space complexity: O(letters.length^2)
Space complexity and auxiliary space complexity example

Additionally, to space complexity, there is also auxiliary space complexity. These two terms are often
misused.
The difference between them is that auxiliary space complexity does not include the amount of
memory we need to store input data. Auxiliary space complexity only shows how much additional
memory the algorithm needs to solve the problem. Space complexity on contrary takes into account
both: the amount of memory we need to store the input data and also the amount of additional
memory the algorithm requires.
Space complexity = Input + Auxiliary space complexity
Let’s see it from the following example. Here is a function that multiplies all array elements by
specific value:
01-algorithms-and-their-complexities/multiplyArrayInPlace.js
1 /**
2 * Multiply all values of the array by certain value in-place.
3 *
4 * Example:
5 * array = [1, 2, 3]
6 * multiplier = 2
7 * output = [2, 4, 6]
8 *
9 * @param {number[]} array
10 * @param {number} multiplier
11 * @return {number[]}
12 */
13 export function multiplyArrayInPlace(array, multiplier) {
14 for (let i = 0; i < array.length; i += 1) {
15 array[i] *= multiplier;
16 }
17
18 return array;
19 }
The space complexity of this function is in linear proportion to input array size. So the bigger the
input array size is the more memory this function will consume.
Space complexity: O(array.length)
Not let’s analyze additional space. We may see that function does all of its operations in-place
without allocating any additional memory.
Auxiliary space complexity: O(1)
Sometimes it may be unacceptable to modify function arguments. And instead of modifying the
input array parameter in-place we might want to clone it first. Let’s see how we change the array
multiplication function to prevent input parameters modification.
01-algorithms-and-their-complexities/multiplyArray.js
1 /**
2 * Multiply all values of the array by certain value with allocation
3 * of additional memory to prevent input array modification.
4 *
5 * Example:
6 * array = [1, 2, 3]
7 * multiplier = 2
8 * output = [2, 4, 6]
9 *
10 * @param {number[]} array
11 * @param {number} multiplier
12 * @return {number[]}
13 */
14 export function multiplyArray(array, multiplier) {
15 const multipliedArray = [...array];
16
17 for (let i = 0; i < multipliedArray.length; i += 1) {
18 multipliedArray[i] *= multiplier;
19 }
20
21 return multipliedArray;
22 }
In this version of the function, we’re allocating additional memory for cloning the input array. Thus
our space complexity becomes equal to O(2 * array.length). After discarding non-dominant parts
we get the following:
Space complexity: O(array.length)
Because our algorithm now allocates additional memory the auxiliary space complexity has also
changed:
Auxiliary space complexity: O(array.length)
Understanding the difference is important so as not to get confused by space complexity evaluations.
Auxiliary space complexity might also become handy for example if we want to compare standard
sorting algorithms on the basis of space. It is better to use auxiliary space than space complexity
because it will make the difference between algorithms clearer. Merge Sort uses O(n) auxiliary space,
Insertion Sort and Heap Sort use O(1) auxiliary space. But at the same time, the space complexity
of all these sorting algorithms is O(n).
Other examples
You will find other big O examples like O(log(n)), O(n * log(n)) in the next chapters of this book.
Quiz
Q1: Does time complexity answer the question of how many milliseconds an algorithm would take
to finish?
Q2: Which time complexity means faster execution time: O(n) or O(log(n))?
Q3: Is it true that algorithms with better time complexity also have better space complexity?
Linked List
• Difficulty: easy
Linked list and its common operations

A linked list is a collection of entities which are not stored in sequential order. Instead, each entity
has a pointer to the next entity. Each entity, also referred to as a node, is composed of data and a
reference (in other words, a link) to the next node in the sequence.
This structure allows for efficient insertion or removal of nodes from any position in the sequence
during iteration. More complex implementations of linked lists add additional links, allowing
efficient insertion or removal from arbitrary node references. A drawback of linked lists is that
access time is linear. Faster access, such as random access, is not possible.
The main operations on linked lists are:
• prepend - add a node to the beginning of the list

• append - add a node to the end of the list
• delete - remove a node from the list
• deleteTail - remove the last node from the list
• deleteHead - remove the first node from the list
• find - find a node in the list
Applications
A linked list is used as a good foundation for many other data structures like queues, stacks and
hash tables. The linked list implementation we will create in JavaScript will be used for many of the
other data structures we will learn about in this book.
Linked List 15
Implementation
Our LinkedList implementation will consist of two parts:
• A LinkedListNode class
• A LinkedList class
We will explore the details of each of these classes in the following sections.
LinkedListNode
Our LinkedListNode class is a class that will represent one item in our collection of items. The
LinkedListNode has two important properties:
1. this.value
2. this.next
this.value contains the actual value we want to store in our node (a number, another object, a
string, etc). this.next is a pointer to the next node our list of nodes.
What is a pointer?
Objects in JavaScript are dictionaries with key-value pairs. Keys are always strings and values can be
a boolean or strings, numbers and objects. When we say our next property in our LinkedListNode
class is a “pointer”, what we really mean is that the value of this.next is a reference to another
JavaScript object – another LinkedListNode object.
constructor
Our constructor method for our LinkedListNode class sets the initial values for this.value and
this.next – both set to null.
02-linked-list/LinkedListNode.js
2 constructor(value, next = null) {
3 this.value = value;
4 this.next = next;
5 }
toString
Our toString method is a convenience method when we want to see what is contained in each node
of our linked list.
Linked List 16
02-linked-list/LinkedListNode.js
7 toString(callback) {
8 return callback ? callback(this.value) : `${this.value}`;
9 }
LinkedList
The LinkedList class contains some important methods to be able to manipulate our collection of
LinkedListNode objects. Let’s take a look at what each method does:
constructor
Our LinkedList class contains two properties: this.head and this.tail. Each of these properties
points to a LinkedListNode object. They are the beginning and the end of our collection of
LinkedListNodes.
02-linked-list/LinkedList.js
4 constructor() {
5 /** @var LinkedListNode */
6 this.head = null;
7
8 /** @var LinkedListNode */
9 this.tail = null;
10 }
prepend
The prepend method adds a LinkedListNode object to the beginning of our LinkedList. First, we
create a new LinkedListNode object.
16 prepend(value) {
17 // Make new node to be a head.
18 const newNode = new LinkedListNode(value, this.head);
Then we set our this.head to point to our newly created LinkedListNode object. If this is the first
item in our LinkedList, we must also set this.tail to point to the new object, as it is also the last
item in our LinkedList.
Linked List 17
19 this.head = newNode;
20
21 // If there is no tail yet let's make new node a tail.
22 if (!this.tail) {
23 this.tail = newNode;
24 }
Finally, we return the value that was added to our LinkedList

25 return this;
26 }
append
The append method adds a LinkedListNode object to the end of our LinkedList. First, we create a
new LinkedListNode object.
33 append(value) {
34 const newNode = new LinkedListNode(value);
If there are no other items in our LinkedList, then we need to set this.head to our new node.
35 // If there is no head yet let's make new node a head.

36 if (!this.head) {
37 this.head = newNode;
39
40 return this;
41 }
Then, we get the value of this.tail and set that objects next property to be our newly created
LinkedListNode object.
Linked List 18
44 const currentTail = this.tail;

45 currentTail.next = newNode;
46 // Attach new node to the end of linked list.
Finally, we set this.tail to be the new LinkedListNode object.

48
49 return this;
delete
Our delete method finds an item in the LinkedList and removes it. First, we check to see if our
LinkedList is empty by checking if this.head is null. If it is, we return null.
56 delete(value) {
58 return null;
59 }
If our LinkedList contains items in it, then we create a deletedNode variable which will serve as a
placeholder for the node that we will eventually return as the result from our delete method.
61 let deletedNode = null;
Before we traverse our LinkedList, we first check to see if the value of this.head matches the value
we are trying to delete. If so, we set our deletedNode variable to this.head and then move this.head
to be the next item in our LinkedList.
Linked List 19
62 // If the head must be deleted then make next node that is different
63 // from the head to be a new head.
64 while (this.head && this.head.value === value) {
65 deletedNode = this.head;
66 this.head = this.head.next;
67 }
In order to traverse our LinkedList, we need to create a placeholder variable, currentNode to keep
track of where we are in our LinkedList. We set the value of currentNode equal to this.head to
ensure we start at the beginning of our list.
70 let currentNode = this.head;
We then check to make sure that our currentNode is not null. If it isn’t, we kick off our while loop.
We will break out of our while loop when we reach a node where its this.next property is null,
meaning it doesn’t point to another LinkedListNode object.
72 if (currentNode !== null) {

73 while (currentNode.next) {
74 if (currentNode.next.value === value) {
Next, we check to see if the value of currentNode.next object is equal to the value we want to
delete. If it is, then we set our deletedNode variable equal to currentNode.next. We also set the
currentNode.next pointer equal to the object that the currentNode.next object points to.
If the values don’t match, we just move on to the next item in our LinkedList.
72 if (currentNode !== null) {

74 if (currentNode.next.value === value) {
Finally, we must check if the value of the tail of our LinkedList matches the value we are trying to
remove. If it does match, we’ll set this.tail equal to currentNode.
See the following diagrams for a visual representation of what happens when the delete method is
called:
Linked List 20
Linked list before delete method is called
delete method is called
Linked list after delete method is called
deleteTail
For our deleteTail method, we first create a variable deletedTail and set it to this.tail so we
can return the object that is deleted when we return from this method.
95 deleteTail() {
96 const deletedTail = this.tail;
We have two scenarios for which we need to write code:

Linked List 21
1. We only have one item in our LinkedList

2. We have many items in our LinkedList
For scenario 1, this.head.value and this.tail.value are the same, so we set both this.head and
this.tail to null and return deleteTail.
98 if (this.head === this.tail) {

99 // There is only one node in linked list.
100 this.head = null;
101 this.tail = null;
102
103 return deletedTail;
104 }
For scenario 2, we need to traverse the entire list to get to the next to last LinkedListNode in
the LinkedList. We will set the next to last LinkedListNode to this.tail by checking if the
currentNode.next.next value is null. If currentNode.next.next is null, we know the currentNode.next
is the object we want to set to this.tail.
106 // If there are many nodes in linked list...

107 // Rewind to the last node and delete "next" link for the node before the last o\
108 ne.
111 if (!currentNode.next.next) {
112 currentNode.next = null;
113 } else {
114 currentNode = currentNode.next;
115 }
116 }
117
118 this.tail = currentNode;
119
120 return deletedTail;
121 }
And lastly, we return our deletedTail object from the method:

Linked List 22
121
deleteHead
Our deleteHead method removes the first item in the list. Similar to our deleteTail method, we
have a few scenarios for which we need to write code to delete the head our LinkedList:
1. There are no items in our list

2. There is only 1 item in our list
3. There are many items in our list
For scenario 1, we check if this.head is null, if it is null we return null from the method
125 deleteHead() {
127 return null;
128 }
For scenarios 2 and 3, we must first create a variable: deletedHead and set it to this.head so we can
return the object that is deleted when we return from this method.
130 const deletedHead = this.head;
Also for scenarios 2 and 3, we combine the logic into an if/else statement. If this.head.next is not
null, we know there is more than 1 item in the list and we set this.head.next to equal this.head
which removes the first item in the list. If this.head.next is null, then we know there is only 1
item in the list and we set both this.head and this.tail to null as there are no more items in our
LinkedList.
Linked List 23
131 if (this.head.next) {
132 this.head = this.head.next;
133 } else {
134 this.head = null;
135 this.tail = null;
136 }
Finally, we return our deletedHead object from the method:

139 return deletedHead;
140 }
find
The find method traverses the items in the LinkedList and locates the first LinkedListNode object
that matches the value we want. The method takes an object that has two key/value pairs, a value
and a callback:
148 find({ value = undefined, callback = undefined }) {
As in our previous methods, we first check if this.head is null. If it is null, we return from the
method because the LinkedList is empty.
150 return null;
151 }
Next, we create a variable currentNode to keep track of where we are in the LinkedList and set it to
this.head to start at the beginning of the LinkedList.
152
Then we create a while loop to loop over all the nodes in the LinkedList and either call the callback
function sent as a parameter or check the value of the current node to see if it matches the value we
are looking for. If there is a match, we return the node.
Linked List 24
155 while (currentNode) {
156 // If callback is specified then try to find node by callback.
157 if (callback && callback(currentNode.value)) {
158 return currentNode;
159 }
160
161 // If value is specified then try to compare by value..
162 if (value !== undefined && currentNode.value === value) {
163 return currentNode;
164 }
165
167 }
Finally, if we go through the entire LinkedList and do not find the value we are looking for, we
return null from our method.
169 return null;
At this point, we’ve implemented the essential LinkedList methods. We’ll now implement two helper
methods that will make it more convenient to test and debug our LinkedList.
toArray
The toArray method takes all the nodes in the LinkedList and puts them in an array. To put them
in an array, we create a new array called nodes. Then we loop over all the nodes in the LinkedList
and push each one on to the array.
175 toArray() {
176 const nodes = [];
177
179 while (currentNode) {
180 nodes.push(currentNode);
182 }
183
184 return nodes;
185 }
Linked List 25
toString
The toString method takes the LinkedList and prints out a string representation of the LinkedList.
The method takes the LinkedList and first converts it to an array (using the toArray method
described above). Then the map array method is called to process each LinkedListNode in the list. The
toString method for each LinkedListNode is called and then the toString array method is called
to print out the entire array of LinkedListNodes.
192 return this.toArray().map(node => node.toString(callback)).toString();
193 }
Complexities
Access Search Insertion Deletion

O(n) O(n) O(1) O(1)
Problems Examples
Here are some related problems that you might encounter during the interview:
• Cycle in a Linked List⁵

• Intersection of Two Linked LIsts⁶
Quiz
Q1: What does this.next refer to in a linked list in JavaScript?
Q2: What is the difference between a deleteTail and deleteHead method in a linked list?
References
Linked List on Wikipedia https://en.wikipedia.org/wiki/Linked_list⁷
Linked List on YouTube https://www.youtube.com/watch?v=njTh_OwMljA⁸
⁵https://leetcode.com/problems/linked-list-cycle-ii/
⁶https://leetcode.com/problems/intersection-of-two-linked-lists/
⁷https://en.wikipedia.org/wiki/Linked_list
⁸https://www.youtube.com/watch?v=njTh_OwMljA
Linked List 26
Linked List example and test cases https://github.com/trekhleb/javascript-algorithms/tree/master/s-

rc/data-structures/linked-list⁹
Doubly Linked List implementation https://github.com/trekhleb/javascript-algorithms/tree/mas-
ter/src/data-structures/doubly-linked-list¹⁰
⁹https://github.com/trekhleb/javascript-algorithms/tree/master/src/data-structures/linked-list
¹⁰https://github.com/trekhleb/javascript-algorithms/tree/master/src/data-structures/doubly-linked-list
Queue
Queue and its common operations

A queue is a collection of entities in which the entities are kept in First-In-First-Out (FIFO) order.
In a FIFO data structure, the first element added to the queue is the first one removed. When an
element is added to the queue, all the elements that were added before it must be removed before
the new element is removed. A commonly seen implementation of this in the real world are lines in
a supermarket where the first shoppers in line are served first, with everyone else behind the first
shopper served after.
The main operations on queues are: - enqueue - add an entity to the rear position (joining the line
at the supermarket), - dequeue - remove an entity from the front position (serving the first shopper
in line), - peek - reading the entity from the front of the queue without dequeuing it (asking who
the next shopper in line will be).
Operations such as searching or updating elements are not common for queues. There are only two
things we’re interested in when working with queues: the beginning and the end of it.
Queue 28
When to use a Queue

Queues are found in different messaging applications where messages are often generated faster
than they are processed. For example, when you want to parse a lot of web-pages or send a lot of
emails, you might want to put those web-pages or emails into a queue for further processing and
process them one by one without worrying that some of the data being lost or messages dropped.
Queues are also used as a component in many other algorithms. Queues are used in graph or tree
breadth-first search, for example, when we need to store a list of the nodes that we need to process
later. Each time we process a node, we add its adjacent nodes or children to the end of the queue.
Using a queue allows us to process nodes in the order they are visited.
Usage Example
Before we continue with the queue class implementation, let’s imagine that we already have one
implemented and use its basic methods mentioned above. This will help us get a better understanding
of what we’re going to implement.
Let’s implement a simple messaging system where we store all the messages that need to be
processed in a queue. In the code below, we instantiate a new Queue object and use the methods
associated with that object to enqueue and dequeue messages in our queue as well as peek to see
which message will be next. We also use the toString and isEmpty methods to see what currently
is in our queue and check to see if it is empty or not.
03-queue/example.js
1 // First, we need to import our Queue class.

2 import { Queue } from './Queue';
3
4 // Then we need to instantiate a queue.
5 const messageQueue = new Queue();
6
7 // Queue now is empty and there is no messages to process.
8 messageQueue.isEmpty(); // => true
9 messageQueue.toString(); // => ''
10
11 // Let's add new message to the queue.
12 messageQueue.enqueue('message_1');
13 messageQueue.isEmpty(); // => false
14 messageQueue.toString(); // => 'message_1'
15 messageQueue.peek(); // => 'message_1'
16
17 // Let's add one more message.
Queue 29
18 messageQueue.enqueue('message_2');
19 messageQueue.isEmpty(); // => false
20 messageQueue.toString(); // => 'message_1,message_2'
21 messageQueue.peek(); // => 'message_2'
22
23 // Now let's process this messages.
24 messageQueue.dequeue(); // => 'message_1'
25 messageQueue.dequeue(); // => 'message_2'
26
27 messageQueue.isEmpty(); // => true
Implementation
A queue can be implemented with a Linked List. The structure and operations on a Linked List are
similar to a Queue. Our Linked List class has an append method and a delete head method that are
both useful for implementing a Queue. We will use these methods for our enqueue and de-queue
methods in our Queue.
First, let’s import our LinkedList class:
03-queue/Queue.js
1 import LinkedList from '../02-linked-list/LinkedList';
constructor
In the constructor for the Queue class we create a new LinkedList object that we will use to store our
data:
03-queue/Queue.js
4 constructor() {
5 // We're going to implement Queue based on LinkedList.
6 this.linkedList = new LinkedList();
7 }
enqueue
Now let’s implement enqueue() operation. This method will append a new element to the end of the
queue. We will call our LinkedList object’s append method to add our item to the end of the queue:
Queue 30
03-queue/Queue.js
30 /**
31 * @param {*} value
32 */
33 enqueue(value) {
34 // Enqueueing means to stand in the line. Therefore let's just add
35 // new value at the beginning of the linked list. It will need to wait
36 // until all previous nodes will be processed.
37 this.linkedList.append(value);
38 }
dequeue
When we want to dequeue() element, we take the first element from our LinkedList (the head of
the list) and return it. We call the LinkedList object’s removeHead method, and if the node that is
returned from the method is null we return null. Otherwise, we return the value of the node.
03-queue/Queue.js
40 /**
41 * @return {*}
42 */
43 dequeue() {
44 // Let's try to delete the first node from linked list (the head).
45 // If there is no head in linked list (it is empty) just return null.
46 const removedHead = this.linkedList.deleteHead();
47 return removedHead ? removedHead.value : null;
48 }
peek
To implement the peek() operation, we won’t remove any elements from our list, we only get the
value from the head of our LinkedList object and return it.
Queue 31
03-queue/Queue.js
17 /**
18 * @return {*}
19 */
20 peek() {
21 if (!this.linkedList.head) {
22 // If linked list is empty then there is nothing to peek.
23 return null;
24 }
25
26 // Just read the value from the end of linked list without deleting it.
27 return this.linkedList.head.value;
28 }
At this point, we’ve implemented the major queue methods. Let’s implement several helper methods
that will make our queue usage more convenient.
isEmpty
Let’s implement a method that checks if the queue is empty. To do that, we just need to check the
head of the linked list and make sure that no element exists:
03-queue/Queue.js
9 /**
10 * @return {boolean}
11 */
12 isEmpty() {
13 // The queue is empty in case if its linked list don't have tail.
14 return !this.linkedList.head;
15 }
toString
Let’s also implement toString() method to make testing and debugging of our Queue class more
convenient. This method will re-use toString() method of our LinkedList class. It will accept
optional callback() function that may be used to convert each linked list element to a string. It
may be useful if we’re going to store objects in the queue.
Queue 32
03-queue/Queue.js
50 /**
51 * @param [callback]
52 * @return {string}
53 */
55 // Return string representation of the queue's linked list.
56 return this.linkedList.toString(callback);
57 }
Our basic implementation of the Queue class is ready at this point. You may find full class code
example in code/src/03-queue/Queue.js file.
Complexities
Access Search Insertion (enqueue) Deletion (dequeue)

O(n) O(n) O(1) O(1)
When the queue is implemented with a LinkedList as in the example above, then accessing and
searching of the items are done through the Linked List search method which traverses all the
elements in the list. Therefore, these operations are done in O(n) time.
Insertion (enqueuing) and deletion (dequeuing) operations are fast. For a Linked List it will take a
constant amount of time O(1) to prepend a new value to the beginning of the list or to delete the
list’s tail.
Problems Examples
• Implement Queues Using Stacks¹¹

• Design A Circular Queue¹²
Quiz
Q1: Is a Queue first-in-first-out or first-in-last-out?
¹¹https://leetcode.com/problems/implement-queue-using-stacks/
¹²https://leetcode.com/problems/design-circular-queue/
Queue 33
Q2: What basic operations does Queue have: peek() / enqueue() / dequeue() or peek() / push()
/ pop()?
Q3: What is the time complexity of enqueue() operation in a Queue?
References
Queue on Wikipedia <a href=”https://en.wikipedia.org/wiki/Queue(abstract_data_type)”>https://en.wikipedia.org/w
data_type)</a>
Queue on YouTube https://www.youtube.com/watch?v=wjI1WNcIntg¹³
Queue example and test cases https://github.com/trekhleb/javascript-algorithms/tree/master/src/-
data-structures/queue¹⁴
¹³https://www.youtube.com/watch?v=wjI1WNcIntg
¹⁴https://github.com/trekhleb/javascript-algorithms/tree/master/src/data-structures/queue
Stack
Stack and its common operations

A stack is a linear data structure that keeps its data in a stack or pile manner. A data structure is
said to be linear if its elements form a sequence. Let’s think about a deck of playing cards that is face
down. Each card is a unit of data that the deck (or “Stack”) is holding. They are face down because
you don’t normally access the cards in the middle of the deck but you work with the top of the stack
instead. The basic operations you may perform with such deck of playing cards are the following:
• You may push a new card on top of the deck,

• you may pop a card from the top of the deck,
• you may peek a card from the top of the deck but leave it on the deck.
These are the three most essential operations you normally use with a stack.
A stack uses LIFO (last-in-first-out) ordering which means that the latest item you pushed to the
stack is processed first. All the old items that are pushed to the stack will need to wait until all the
new ones are processed.
Operations such as searching or updating the elements are not typical for stacks. We’re only
interested in working with the top of the stack.
Stack 35
Applications
The Stack data structure is used to solve many algorithmic tasks such as:
• graph topological sorting,

• finding strongly connected components in graphs,
• The Tower of Hanoi Problem,
• remembering partially completed tasks,
• undoing (backtracking from) an action
• and many more.
Stacks are often useful in recursive algorithms when you need to push temporary data onto a stack
as you recurse, but then remove them as you backtrack. A stack offers an intuitive way to do this.
Low-level programming languages also use the stack data structure to store function parameters and
local variables in memory when calling one function from another for example. You might probably
know this from the notorious “Stack Overflow” error. The error is connected to stack usage and may
be caused by excessively deep or infinite recursion, in which a function calls itself so many times
that the space needed to store the variables and information associated with each call is more than
can fit on the stack.
Usage Example
Before we continue with the stack class implementation, let’s imagine that we already have one
implemented and try to use its basic methods that were mentioned above. This will help us to get a
better understanding of what we’re going to implement. Let’s say we want to use a stack to reverse
an array:
04-stack/example.js
1 // Import the Stack class.

2 import { Stack } from './Stack';
3
4 // Create stack instance.
5 const stack = new Stack();
6
7 // Create an array we want to reverse.
8 const arrayToReverse = ['a', 'b', 'c'];
9
10 stack.isEmpty(); // => true
11 stack.toString(); // => ''
12 stack.toArray(); // => []
Stack 36
13
14 // Let's add items to stack.
15 stack.push(arrayToReverse[0]);
16 stack.isEmpty(); // => false
17 stack.toString(); // => 'a'
18 stack.toArray(); // => ['a']
19
20 // Add more items to stack.
23 stack.toString(); // => 'c,b,a'
24 stack.toArray(); // ['c', 'b', 'a']
25
26 // Check what next.
27 stack.peek(); // => 'c'
28
29 // Pop from stack.
30 stack.pop(); // => 'c'
31 stack.pop(); // => 'b'
32 stack.pop(); // => 'a'
33
34 stack.isEmpty(); // => true
Implementation
A stack can be implemented with a linked list. We can use our append and deleteTail operations
from our LinkedList class to implement the push and pop operations of a stack. The idea is that you
append the item to the linked list when you want to push it to the stack and you delete the tail of
linked list when you want to pop an item from the stack.
Since we’re going to build a stack based on a linked list let’s start by importing it:
04-stack/Stack.js
constructor
Let’s create the Stack class:

Stack 37
04-stack/Stack.js
3 export class Stack {
Next we create our LinkedList object that will hold all stack data:
04-stack/Stack.js
4 constructor() {
5 // We're going to implement Queue based on LinkedList since this
6 // structures a quite similar. Compare push/pop operations of the Stack
7 // with prepend/deleteHead operations of LinkedList.
8 this.linkedList = new LinkedList();
9 }
push
If we want to push something to the stack, we need to append the data to the end of the linked list.
In this case, the top of the stack would be the tail of the linked list:
04-stack/Stack.js
32 /**
34 */
35 push(value) {
36 // Pushing means to lay the value on top of the stack. Therefore let's just add
37 // new value at the head of the linked list.
38 this.linkedList.prepend(value);
39 }
pop
If we want to pop the element from the stack we need to delete the linked list tail and return it:
Stack 38
04-stack/Stack.js
41 /**
42 * @return {*}
43 */
44 pop() {
45 // Let's try to delete the first node from linked list (the head).
46 // If there is no head in linked list (it is empty) just return null.
47 const removedHead = this.linkedList.deleteHead();
48 return removedHead ? removedHead.value : null;
49 }
peek
Another common stack operation is checking what is on top of the stack without actually extracting
the data. In this case, we just need to read linked list tail:
04-stack/Stack.js
19 /**
20 * @return {*}
21 */
22 peek() {
23 if (this.isEmpty()) {
24 // If linked list is empty then there is nothing to peek from.
25 return null;
26 }
27
28 // Just read the value from the head of linked list without deleting it.
29 return this.linkedList.head.value;
30 }
At this point, we’re done with mandatory stack operations. Now let’s move on and implement some
helper methods that will make the usage of Stack class to be more convenient.
isEmpty
Let’s implement isEmpty() method that will check if the stack is empty or not. To do this we just
need to check if linked list has a head:
Stack 39
04-stack/Stack.js
11 /**
13 */
14 isEmpty() {
15 // The queue is empty in case if its linked list don't have head.
16 return !this.linkedList.head;
17 }
toArray
Another useful method is toArray() which converts our stack to an array. To do this, we need to:
1. Convert the linked list to an array,

2. Extract the values of each linked list nodes since we don’t want our Stack class consumers to
deal with linked list nodes directly. By doing this, we encapsulate the implementation of the
class,
3. Reverse the linked list array so that the tail of the list would become the first element in the
array and the head would become the last element of the array. This is not mandatory but is
done for convenience reasons only.
The code looks like this:

04-stack/Stack.js
51 /**
52 * @return {*[]}
53 */
54 toArray() {
55 return this.linkedList
56 .toArray()
57 .map(linkedListNode => linkedListNode.value);
58 }
toString
And the last method we will implement is toString() that will be based on toArray() and will
convert the stack to its string representation:
Stack 40
04-stack/Stack.js
60 /**
61 * @param {function} [callback]
63 */
65 return this.linkedList.toString(callback);
66 }
67 }
The last curly bracket is the closing bracket of Stack class.

Our basic implementation of the Stack class is ready at this point. You may find full class code
example in code/src/04-stack/Stack.js file.
Complexities
Access Search Insertion (push) Deletion (pop)

O(n) O(n) O(1) O(1)
In case if Stack is implemented via LinkedList as in example above then accessing and searching of
the item are done through the linked list search method which goes through all the elements in the
list. Therefore these operations are done in O(n) time.
Insertion (pushing) and deletion (popping) operations are fast since for linked list it will take a
constant amount of time O(1) to append the new value to the end of the list or to delete the list’s
tail.
Problems Examples
• Implement Stacks Using Queues¹⁵

• Climbing Stairs¹⁶
Quiz
Q1: Is a Stack first-in-first-out or first-in-last-out?
¹⁵https://leetcode.com/problems/implement-stack-using-queues/
¹⁶https://leetcode.com/problems/climbing-stairs/
Stack 41
Q2: What basic operations does Stack have: peek() / enqueue() / dequeue() or peek() / push()
/ pop()?
Q3: What is the time complexity of push() operation in a Stack?
References
Stack on Wikipedia <a href=”https://en.wikipedia.org/wiki/Stack(abstract_data_type)”>https://en.wikipedia.org/wik
data_type)</a>
Stack on YouTube https://www.youtube.com/watch?v=wjI1WNcIntg¹⁷
Stack example and test cases https://github.com/trekhleb/javascript-algorithms/tree/master/src/-
data-structures/stack¹⁸
¹⁷https://www.youtube.com/watch?v=wjI1WNcIntg
¹⁸https://github.com/trekhleb/javascript-algorithms/tree/master/src/data-structures/stack
Hash Table
• Difficulty: medium
A hash table (or hash map) is a data structure that implements an associative array abstract data
type, a structure that can map keys to values. For example, your name (the key) may be used as a
unique identifier for your phone number (the value) in a phone book (the hash map). The common
operations for working with your phone in a phone book would look like this:
src/05-hash-table/HashTable.js
1 // Import HashTable class.

2 import { HashTable } from './HashTable';
3
4 // Create hash table instance.
5 const phoneBook = new HashTable();
6
7 // Add several contacts to the phone book.
8 phoneBook.set('John Smith', '+112342345678');
9 phoneBook.set('Bill Jones', '+111111111111');
10
11 // Now we may access each contact's phone in O(1) time.
12 phoneBook.get('John Smith'); // => '+112342345678'
13 phoneBook.get('Bill Jones'); // => ''+111111111111'
14
15 // Delete phone number.
16 phoneBook.delete('John Smith');
17 phoneBook.get('John Smith'); // => undefined
Ideally, hash tables allow us to do all of these operations in O(1) time. Compare that with a Linked
List which requires us to go through all the nodes in the list to fetch John Smith’s phone number,
with an O(n) lookup time.
The highly efficient lookup is the main advantage of hash tables.
In the worst case scenario though it is possible for hash tables to have O(n) lookup time. We’ll get
to this scenario later in the chapter.
So, how do we achieve a fast O(1) lookup in a hash table? We allocate an array of buckets or slots
that will hold all the hash table data. Then we need to map the key of the value to a specific number
Hash Table 43
(index or address) of the bucket where the value will be stored. By doing that we will be able to
create, read and delete records in the hash table by simply accessing it using the bucket address in
O(1) time.
There are certain limitations in building a hash table. The number of buckets will have a fixed
size. This is because any program has certain memory limitations and we can’t assume we have an
unlimited amount of memory for our hash table.
The key that we’re going to map to a bucket address may be of arbitrary size (i.e. John Smith or just
Jimmy). How do we map a key of arbitrary size to a number (index or address) when the number of
buckets is limited? The hint is already hidden in the “Hash Table” data structure name! And it is by
using a hash function.
Hash Function
A hash function is a function that can be used to map data of arbitrary size to data of a fixed size.
The values returned by a hash function are called hash values, hash codes, digests, or simply hashes.
In our case, our hash function must accept the key (i.e. John Smith) as an input and output the index
of a bucket in a given range (i.e. from 0 to 5 if we allocate an array of size 5 for the buckets). In
the real world, the number of buckets is much bigger. It should be proportional to the amount of
information we’re going to store in our hash table.
The hash function consists of two parts:
1. Convert the string to a hash code – ideally to a unique numeric representation of the key.
2. Map the hash code to a specific bucket cell using the modulo operator (since we have a limited
bucket size).
Hash Table 44
Hash function example

There are a number of ways to hash a string into an index or address. One of the simplest ways to
calculate the hash of a string is to sum up the ASCII codes of all the characters in a string.
What is an ASCII code?

ASCII stands for American Standard Code for Information Interchange and it is a code
that represents characters in English as numbers. Computers only understand 1s and
0s and this code translates the letters we type into a number that the computer can
understand.
Let’s create a JavaScript function called hash that takes the key and number of buckets as parameters.
05-hash-table/hash.js
13 export function hash(key, bucketsNumber) {
We’ll loop over each character in our string and use the JavaScript string method charCodeAt to get
an integer value for each character. The charCodeAt method returns the ASCII code of a character.
We add up all the codes of each character to calculate our hash code.
14 let hashCode = 0;
15 // Let's go through all key characters and add their code to hash
16 for (let characterIndex = 0; characterIndex < key.length; characterIndex += 1) {
17 hashCode += key.charCodeAt(characterIndex);
18 }
Then, in order to make sure our index is a valid index for our hash table, we use the modulo operator
to calculate the correct bucket index and return the value.
20 // Reduce hash number so it would fit buckets table size.
21 return hashCode % bucketsNumber;
22 }
For example, say we called our function using ab as our string and 100 as our bucket size. What
result would get returned from calling our hash function?
hash('ab',100)
The ASCII code for a is 97 and the ASCII code for b is 98. 97 + 98 = 195. 195 % 100 is 95 and that is
what is returned from our hash function. 95 is the array index where the value will be stored in our
hash table.
Hash Table 45
You may also want to use more sophisticated approaches like polynomial string hash to make the
hash code more unique. We won’t cover these more sophisticated approaches, but you can check
the footnotes at the end of this chapter if you’d like to learn more about different hash function
techniques.
Collision Resolution
Ideally, the hash function will assign each key to a unique bucket, but most hash table designs employ
an imperfect hash function, which might cause hash collisions where the hash function generates
the same index for more than one key.
For example:
ASCII for 'a' is 97, for 'd' is 100.
Hash for `a`: (97) % 100 = 97

Hash for `ad`: (97 + 100) % 100 = 97
These collisions must be accommodated in some way otherwise the write operation will override
the values for keys a and ad You might get the value for a when you wanted the ad value and
vice-versa.
Separate chaining
Separate chaining is a technique where we convert each bucket to a linked list. Then each bucket
will be able to store many values instead of just one. If a collision happens we’ll attach new value
to the tail of bucket’s linked list.
The read operation will calculate the bucket number first and then do a lookup in the bucket’s linked
list by key. Here are the required steps to fetch Billy Black’s phone number from the example above
using a linked list for our bucket:
1. Calculate hash code for the Billy Black key. Let’s assume that hash code for the Billy Black
key is 11.
Hash Table 46
2. Calculate the bucket index by doing a modulo operation. With 5 buckets available we’ll get
bucket index equal to 11 % 5 = 1.
3. Fetch the linked list from the cell with index 1 and go through each of its values comparing
each value key with the key we need (in our case is Billy Black). Once we’ve found the linked
list node with Billy Black key we just return it.
Hash function collisions are the main reason why time complexities for basic hash table
operations like read, write, update and delete may be downgraded from O(1) to O(n). In
the worst case scenario, with a small number of buckets and with imperfect hash function
all values might be stored in just one bucket. This will make the hash table act like a linked
list.
Implementation
Let’s take the hash table basic operations that have been mentioned above along with separate
chaining collision resolution approach and implement the hash table using the LinkedList class from
the Linked List chapter.
First, we’ll need to import our Linked List class to be used in our hash table. Each bucket in our hash
table will have a Linked List as the value where we will place the values for each key/value pair.
Next, we’ll set our bucket size:

3 // Buckets number directly affects the number of collisions.

4 // The bigger the bucket number the less collisions we'll get.
5 // For demonstration purposes we set the hash table size to a
6 // smaller number to show how collisions are handled.
7 const defaultBucketsNumber = 32;
constructor
Let’s start by creating HashTable class:

Hash Table 47
4-stack/Stack.js
9 export class HashTable {
In our constructor method, we have quite a few functions chained together. Let’s break down what’s
going on in each piece so we have a better understanding of what is happening.
10 /**
11 * @param {number} bucketsNumber - Number of buckets that will hold hash table dat\
12 a.
13 */
14 constructor(bucketsNumber = defaultBucketsNumber) {
15 // Create hash table of certain size and fill each bucket with empty linked list.
16 this.buckets = Array(bucketsNumber).fill(null).map(() => new LinkedList());
17 }
First, we create a new array and initialize it with the bucketsNumber value:
this.buckets = new Array(bucketsNumber)
Then, we fill the array with null values so we can loop over it:
this.buckets = new Array(bucketsNumber).fill(null)
Finally we set all of the newly created values in our array to a LinkedList object
this.buckets = new Array(bucketsNumber).fill(null).map(() => new LinkedList());
hash
The hash() method will convert the string key to the index of a bucket we need to use to store the
value for the specified key. As it was described above we will use the simple approach of calculating
the hash based on the sum of ASCII code values of each character in a key.
Hash Table 48
18 /**
19 * Converts key string to hash number.
20 *
21 * @param {string} key
23 */
24 hash(key) {
25 let hashCode = 0;
26
27 // Let's go through all key characters and add their code to hash
28 for (let characterIndex = 0; characterIndex < key.length; characterIndex += 1) {
29 hashCode += key.charCodeAt(characterIndex);
30 }
31
32 // Reduce hash number so it would fit buckets table size.
33 return hashCode % this.buckets.length;
34 }
First we go through all the characters using a for loop and calculate their sum.
for (let characterIndex = 0; characterIndex < key.length; characterIndex += 1) {

hashCode += key.charCodeAt(characterIndex);
}
Next, we apply the modulo operator % that will limit the range of hashes we generate to the number
of available buckets:
// Reduce hash number so it would fit buckets table size.

return hashCode % this.buckets.length;
If our hashCode value equals 97 and this.buckets.length equals 50 the function will return 97 %
50 = 47. We will use the bucket with index 47 to store the value for the key ‘ab’.
set
We add elements to a hash table based on it’s key and value.

Hash Table 49
36 /**
37 * ADD or UPDATE the value by its key.
38 *
41 */
42 set(key, value) {
43 // Calculate bucket index.
44 const keyHash = this.hash(key);
45 // Fetch linked list for specific bucket index.
46 const bucketLinkedList = this.buckets[keyHash];
47 // Perform 'find by key' operation in linked list.
48 const node = bucketLinkedList.find({ callback: nodeValue => nodeValue.key === ke\
49 y });
50
51 // Check the value with specified key is already exists in linked list.
52 if (!node) {
53 // Insert new linked list node.
54 bucketLinkedList.append({ key, value });
55 } else {
56 // Update value of existing linked list node.
57 node.value.value = value;
58 }
59 }
We start by calculating the hash for the provided key. This hash value will point to the specific
bucket index.
const keyHash = this.hash(key);
Next, we get the linked list that is stored in the bucket we need.
const bucketLinkedList = this.buckets[keyHash];
At this point, we need to check whether we already have the value with the specified key in our
hash table or not. First, we’ll try to find it in our linked list by using the key/value pair. We’ll use the
LinkedList find method with a custom callback. This custom callback will compare the key/value
pair stored in the LinkedList node with the key that we’re trying to use to store the new value.
Hash Table 50
const node = bucketLinkedList.find({ callback: nodeValue => nodeValue.key === key });
Now we need to decide whether we need to create a new node in our LinkedList or update an
existing one. If the value we’re trying to set already exists in the LinkedList then we need to update
it. Otherwise, we need to append the new value to the end of the LinkedList.
Each LinkedList node has a value property which can be any type of data (strings, object, numbers,
etc). For our hash table, the value property for the nodes in our LinkedList will be an object that
contains the key/value pair. We must save both the key and the value in order to be able to distinguish
one node from another. If we have two keys that resolve to the same bucket when the hash function
is applied to them, we need to know which value to retrieve based on the key.
if (!node) {
bucketLinkedList.append({ key, value });
} else {
node.value.value = value;
}
get
Getting a value from the hash table by using its key is a similar operation to the set method. The
only difference is that after checking if the value exists in the LinkedList we either return its value
or undefined.
60 /**
61 * GET the value by its key.s
62 *
64 * @return {*}
65 */
66 get(key) {
73 y });
74
75 // Check the value with specified key is already exists in linked list.
76 return node ? node.value.value : undefined;
77 }
Hash Table 51
delete
When we delete a value from our hash table, we use some of the same steps as the set method. We
first calculate the hash for the provided key, then fetch the bucket’s LinkedList. Finally, we check if
the value we want to delete exists in the LinkedList. If it exists, we use the LinkedList delete method
and return true. Otherwise, we return false indicating there was nothing to delete.
78 /**
79 * DELETE the value by its key.
80 *
83 */
84 delete(key) {
91 y });
92
93 // Delete node from linked list if it exists there.
94 if (node) {
95 bucketLinkedList.delete(node.value);
96
97 return true;
98 }
99
100 return false;
101 }
Our basic implementation of the HashTable class is ready at this point. You may find full class code
example in code/src/05-hash-table/HashTable.js file.
Operations Time Complexity
Case Search Insertion Deletion Comments

Average O(1) O(1) O(1) With good hash function and
sufficient buckets number
Worst Case O(n) O(n) O(n) With bad hash function and too
small buckets number
Hash Table 52
Problems Examples
• Design a HashSet¹⁹
• Design a HashMap²⁰
Quiz
Q1: What is the main advantage of a hash table compared to a linked list?
Q2: What is the time complexity of get() operation in a hash table?
Q3: What are collisions in hash tables?
Q4: What way(s) of handling the collision do you know?
References
Hash table on Wikipedia https://en.wikipedia.org/wiki/Hash_table²¹
Hash table on YouTube https://www.youtube.com/watch?v=shs0KM3wKv8²²
Hash table example and test cases https://github.com/trekhleb/javascript-algorithms/tree/master/s-
rc/data-structures/hash-table²³
¹⁹https://leetcode.com/problems/design-hashset/
²⁰https://leetcode.com/problems/design-hashmap/
²¹https://en.wikipedia.org/wiki/Hash_table
²²https://www.youtube.com/watch?v=shs0KM3wKv8
²³https://github.com/trekhleb/javascript-algorithms/tree/master/src/data-structures/hash-table
Binary Search Tree (BST)
Before we dive into the topic of binary search trees, we first need to understand what tree and binary
tree data structures are.
Tree
A tree data structure is defined as a collection of nodes (starting at the root node), where each
node is a data structure consisting of a value and a list of references to nodes (the “children”), with
the constraints that there are no duplicate references, and no references point to the root. These
references are often referred to as edges.
What is a node?
A node is a basic unit used in computer science. Nodes are individual parts of a larger data
structure, such as linked lists and tree data structures. Nodes contain data and also may
link to other nodes. Links between nodes are often implemented by pointers. -Wikipedia²⁴
Every node in a tree data structure (excluding the root node) is connected by an edge from exactly
one other node. All nodes form a parent -> child relationship, and there are no cycles in the tree.
Here is an example of a tree:
²⁴https://en.wikipedia.org/wiki/Node_(computer_science)
Binary Search Tree (BST) 54
In this example, the tree consists of a set of nodes {“a”, “b”, “c”, …, “i”} and set of edges {“ab”, “ac”,
“bd”, …, “fi”} that form a parent-child relationship between the nodes. The node “a” is the root of
the tree and it has no parent but has two child - nodes “b” and “c”. The node “f” in this example has
one parent “c” and three child nodes “g”, “h” and “i”. The nodes that don’t have any children are
called leaf nodes. In this example the nodes “d”, “e”, “g”, “h”, “i” are the leaves.
The size of the tree is the number of nodes it contains. In the example above the size of the tree is 9.
The height of the node is the longest path (the number of edges) from it to the farthest leaf. The
height of the root node is the height of the tree. In the example above the height of the node “c” is
2 since there are two edges from it to the deepest leaves (i.e., “cf” and “fg”). And the height of the
tree is 3.
Here are several examples that are not trees because some nodes have more than two parents and
some nodes form a cycle.
Binary Tree
A binary tree is a tree in which each node has at most two children, which are referred to as the
left child and the right child.
Here is an example of a binary tree:
The root node “5” has left child “15” and right child “21”. The node “21” has only one right child
(the node “3”) which does not violate the binary tree definition.
Types of Binary Trees

There are several types of binary trees:
• A full binary tree is a binary tree with all nodes having either 0 or 2 children.
• A complete binary tree is a binary tree in which all levels (except possibly the deepest one)
are entirely filled with nodes. The deepest level may be partially filled, but all nodes must be
filled starting from the left without forming any gaps between them.
• A perfect binary tree is a binary tree in which all internal nodes have two children and all
leaves have the same depth.
• A balanced binary tree is a binary tree in which for all the nodes the left and right subtrees
differ in height by no more than 1.
• A pathological (or degenerate) binary tree is a binary tree in which each node has at most
one child. This type of tree will perform similarly to linked lists.
Binary Search Tree

A binary search tree (or BST) is a binary tree in which each node (except leaf nodes) satisfies the
following rules:
• all values of the right subtree of a node must be greater than the node value itself,
• all values of the left subtree of a node must be less than the node value itself.
The leaf nodes children are null therefore we’re not applying these rules to them.
Here is an example of a binary search tree:
The root value for this tree is 9 and the root’s right child value is 13 which is greater than the root
value. As we travel farther down the right subtree, we see that all the values in the right subtree are
greater than the root node value. The root’s left child value is 5 which is less than the root value. As
we travel down the left subtree, we see that all the values in the left subtree are less than the root
node value. These rules may be applied to all of the nodes in the tree recursively.
Application
Binary search trees may be used in cases when we need to maintain dynamically changing datasets
in sorted order, for some sortable data.
Basic Operations
Searching (Lookup)
The fact that a binary search tree is a sorted tree makes it possible to apply a binary search approach
to the tree and perform a fast lookup.
Let’s imagine that we want to find the node with the value 7 in binary search tree mentioned above.
The lookup algorithm looks like this:
1. Start from the root node (treat the root node as current node).
2. If current node value is greater than 7, then go left (make the left child the current node).
Otherwise, go right (make the right child the current node).
3. Repeat steps 1 and 2 until the node is found or until there is no children of the current node
are left.
After every lookup iteration, we’re ignoring about half of the tree (or subtree) and focusing on the
other half. By focusing only on one half of the tree and shrinking the search area by two on every
iteration the time complexity of lookup operation is O(log(n)).
Self-Balancing
The performance of lookup operation in a binary search tree tightly depends on the shape of the
tree. For example, a balanced binary search tree takes O(log(n)) time to look for a specific value in
a tree as explained above. But for pathological binary search trees, the same operation takes O(n)
time because the tree is similar to a linked list.
Keeping binary search trees balanced while adding or removing nodes maintains the quick
performance of the lookup operation. Several data structures are built on top of binary search trees
but with the additional feature of balancing the tree after each modification: the AVL Tree²⁵, Red-
Black Tree²⁶ and others.
In this chapter, we will cover the basic implementation of a binary search tree but without the re-
balancing feature.
Inserting
To insert a new node into a binary search tree, we need to recursively compare the current node
value with a new node value. If the new node value is greater than current node value we need to go
right (make the right child the current node). Otherwise, we need to go left (make the left child the
current node). Once the next child node is empty, it means we’ve reached the leaf node and we’re
ready to place a new node into its proper position.
²⁵https://github.com/trekhleb/javascript-algorithms/tree/master/src/data-structures/tree/avl-tree
²⁶https://github.com/trekhleb/javascript-algorithms/tree/master/src/data-structures/tree/red-black-tree
Removing.
Removing the node from a binary search tree is a little bit trickier. First, we need to find the node
we want to remove (see the lookup operation above). Several cases are possible:
• Removing a leaf node - most straightforward case, we remove the node.

• Removing a node with one child - we make the child node become a child node for the current
node’s parent
• Removing the node with two children - we need to find the next largest value (minimum
value in the right branch) and replace current node with that next largest value node.
Here is an illustration for the last case when the node has two children.
Traversing
Traversing is visiting all the nodes of a tree in a specific order.
In this chapter, we will implement in-order traversal to convert the binary search tree to a sorted
array.
Usage Example
Before we move on to the exact JavaScript implementation of the binary search tree, let’s look at a
use case for our BinarySearchTree class and describe how we’re going to use it.
Say we have a collection of shopping items. And we want to make all of the following operations to
perform fast (in O(log(n)) time): - getting items by specific price, - inserting new items, - removing
existing items, - getting the cheapest item from the collection, - getting the most expensive item
from the collection.
06-binary-search-tree/example.js
1 // Import dependencies
2 import BinarySearchTree from './BinarySearchTree';
3
4 // Create binary search tree instance.
5 const bstCollection = new BinarySearchTree();
6
7 // Add shopping items to our binary search tree collection.
8 // We will use items price as a keys.
9 bstCollection.insert(1220, { title: 'Phone' });
10 bstCollection.insert(3455, { title: 'TV' });
11 bstCollection.insert(8200, { title: 'Notebook' });
12 bstCollection.insert(120, { title: 'Remote Control' });
13
14 // Let's find the cheapest and most expensive items so far.
15 const minItem = bstCollection.findMin().data; // -> { title: 'Remote Control' }
16 const maxItem = bstCollection.findMax().data; // -> { title: 'Notebook' }
17
18 // Let's fins the item with the price 8200.

19 const item = bstCollection.find(3455).data; // -> { title: 'TV' }
20
21 // Remove the item from collection by price.
22 bstCollection.remove(120);
23 // Check what is the cheapest item at the moment.
24 const newMinItem = bstCollection.findMin().data; // -> { title: 'Phone' }
25
26 // Now we may use the data we've just fetched anyhow.
27 // eslint-disable-next-line no-console
28 console.log(minItem, maxItem, item, newMinItem);
Implementation
We’ll split our implementation into two parts. First, we will first implement BinarySearchTreeNode
class which will represent nodes of the binary search tree and their functionality. This class will
contain the core functionality. Then, we’ll implement BinarySearchTree class that will be just a
wrapper over the BinarySearchTreeNode class and will make working with a binary search tree a
little bit easier.
BinarySearchTreeNode
constructor
Let’s create the class.

06-binary-search-tree/BinarySearchTreeNode.js
1 export default class BinarySearchTreeNode {
Our constructor will accept the value parameter that we will use as a key to define where in the tree
the current node should be placed. All node comparisons will be made using this value parameter.
The data parameter serves as a data bag that contains any type of data. For example, it may contain
a linked list with all shopping items of a specific price, or it may contain a user object with a name,
telephone number, etc.
In our constructor, we’ll also create pointers to the left and right child as well as a pointer to the
parent node. For the root node in a tree the parent will always be null.
2 /**
3 * @param {*} [value] - node value.
4 * @param {*} [data] - node data (could be anything).
5 */
6 constructor(value = null, data = null) {
7 this.left = null; // Pointer to the left child.
8 this.right = null; // Pointer to the right child.
9 this.parent = null; // Pointer to the parent node.
11 this.data = data;
12 }
find()
This method finds the node in a whole tree by its value (key). It does a binary search tree lookup
operation that was in the explanation above. As a reminder, to find a node by its key in the binary
search tree we need to recursively compare its value with the current node and then go left (to the
left child) if current node value is bigger than the one we want to find. Otherwise, we need to go
right (to the right node). The same process is repeated for the next node (left or right child).
168 /**
170 * @return {BinarySearchTreeNode}
171 */
172 find(value) {
173 // Check the root.
174 if (this.value === value) {
175 return this;
176 }
177
178 if (value < this.value && this.left) {
179 // Check left nodes.
180 return this.left.find(value);
181 }
182
183 if (value > this.value && this.right) {
184 // Check right nodes.
185 return this.right.find(value);
186 }
187
188 return null;
189 }
insert()
To insert a new node to the binary search tree we perform a similar function as the lookup - we
compare the new node value with the current node value and if the new value is bigger than the
current node value, we move right (to the right node). Otherwise, we move left (to the left node).
We repeat this process until the next node cannot be found (it is null). When we have reached a
null value, we have found the proper place for the new node to be inserted and we assign this null
pointer to the new node. The new node instance is created using BinarySearchTreeNode constructor
with value and data parameters.
This method uses other helper methods like setLeft() and setRight() that will be implemented
below.
128 /**
132 */
133 insert(value, data) {
134 if (this.value === null) {
137
138 return this;
139 }
140
141 if (value < this.value) {
142 // Insert to the left.
143 if (this.left) {
144 return this.left.insert(value, data);
145 }
146
147 const newNode = new BinarySearchTreeNode(value, data);
148 this.setLeft(newNode);
149
150 return newNode;
151 }
152
153 if (value > this.value) {
154 // Insert to the right.
155 if (this.right) {
156 return this.right.insert(value, data);
157 }
158
159 const newNode = new BinarySearchTreeNode(value, data);
160 this.setRight(newNode);
161
162 return newNode;
163 }
164
165 return this;
166 }
remove()
Before removing the node, we need to find it in the tree using find() function described above.
After we find the correct node, several cases are possible:
• Removing the leaf node - remove a pointer to this node from the parent node.
• Removing the node with one child - make one and only child to be a direct ancestor for the
current node’s parent.
• Removing the node with two children - find the next biggest value (minimum value in the
right branch) and replace current node with that next biggest value node.
This method uses other helper methods like removeChild(), findMin(), setValue(), setRight(),
replaceChild(), copyNode() that will be implemented below.
221 /**
224 */
225 remove(value) {
226 const nodeToRemove = this.find(value);
227
228 if (!nodeToRemove) {
229 throw new Error('Item not found in the tree');
230 }
231
232 const { parent } = nodeToRemove;
233
234 if (!nodeToRemove.left && !nodeToRemove.right) {
235 // Node is a leaf and thus has no children.
236 if (parent) {
237 // Node has a parent. Just remove the pointer to this node from the parent.
238 parent.removeChild(nodeToRemove);
239 } else {
240 // Node has no parent. Just erase current node value.
241 nodeToRemove.setValue(undefined);
242 }
243 } else if (nodeToRemove.left && nodeToRemove.right) {
244 // Node has two children.
245 // Find the next biggest value (minimum value in the right branch)
246 // and replace current value node with that next biggest value.
247 const nextBiggerNode = nodeToRemove.right.findMin();
248 if (nextBiggerNode !== nodeToRemove.right) {
249 this.remove(nextBiggerNode.value);
250 nodeToRemove.setValue(nextBiggerNode.value);
251 } else {
252 // In case if next right value is the next bigger one and it doesn't have le\
253 ft child
254 // then just replace node that is going to be deleted with the right node.
255 nodeToRemove.setValue(nodeToRemove.right.value);
256 nodeToRemove.setRight(nodeToRemove.right.right);
257 }
258 } else {
259 // Node has only one child.
260 // Make this child to be a direct child of current node's parent.
261 /** @var BinarySearchTreeNode */
262 const childNode = nodeToRemove.left || nodeToRemove.right;
263
264 if (parent) {
265 parent.replaceChild(nodeToRemove, childNode);
266 } else {
267 BinarySearchTreeNode.copyNode(childNode, nodeToRemove);
268 }
269 }
270
271 // Clear the parent of removed node.
272 nodeToRemove.parent = null;
273
274 return true;

275 }
traverseInOrder()
This method does an in-order traversal of the BST recursively. It first traverses the left sub-tree of
the root node. Then it concatenates the root node value to the traversed array. Then it continues
traversing the right sub-tree of the root node. This method will return an array of BST node values
in sorted order.
276 /**
277 * @return {*[]}
278 */
279 traverseInOrder() {
280 let traverse = [];
281
282 // Add left node.
283 if (this.left) {
284 traverse = traverse.concat(this.left.traverseInOrder());
285 }
286
287 // Add root.
288 traverse.push(this.value);
289
290 // Add right node.
292 traverse = traverse.concat(this.right.traverseInOrder());
293 }
294
295 return traverse;
296 }
setLeft()
This method will set the left child for the current node. It will also take care of parent pointer for
the left node by pointing it to the current node.
34 /**
35 * @param {BinarySearchTreeNode} node
37 */
38 setLeft(node) {
39 // Reset parent for left node since it is going to be detached.
40 if (this.left) {
41 this.left.parent = null;
42 }
43
44 // Attach new node to the left.
45 this.left = node;
46
47 // Make current node to be a parent for new left one.
48 if (this.left) {
49 this.left.parent = this;
50 }
51
52 return this;
53 }
setRight()
This method will set the right child for the current node. It will also take care of parent pointer for
the right node by pointing it to the current node.
55 /**
56 * @param {BinarySearchTreeNode} node
58 */
59 setRight(node) {
60 // Reset parent for right node since it is going to be detached.
62 this.right.parent = null;
63 }
64
65 // Attach new node to the right.
66 this.right = node;
67
68 // Make current node to be a parent for new right one.

69 if (node) {
70 this.right.parent = this;
71 }
72
73 return this;
74 }
setData()
setData is a helper function that we will use for setting the node data. We return this object to
allow us to chain methods. This method makes working with the tree more convenient. We can do
something like this: bst.setValue(5).setData('User name').
24 /**
25 * @param {*} data
27 */
28 setData(data) {
30
31 return this;
32 }
setValue()
This is a helper function that we will use for setting the node value. We return this object to allow
us to chain methods.
14 /**
17 */
18 setValue(value) {
20
21 return this;
22 }
contains()
This is a helper method that returns true or false showing if the node with the specified value
exists in the tree.
191 /**
194 */
195 contains(value) {
196 return !!this.find(value);
197 }
removeChild()
This method removes child node from the current node. It accepts the node that is going to be
removed by using the nodeToRemove parameter. Then it will compare the node that is going to be
removed with the left and right child of the current node to decide which one should be removed. If
the nodeToRemove is found among the children then the child is set to null and the function returns
true as a sign the removal process was successful. Otherwise this method returns false which means
that nodeToRemove is not among the children of the current node.
76 /**
77 * @param {BinarySearchTreeNode} nodeToRemove
79 */
80 removeChild(nodeToRemove) {
81 if (this.left && this.left === nodeToRemove) {
82 this.left = null;
83 return true;
84 }
85
86 if (this.right && this.right === nodeToRemove) {
87 this.right = null;
88 return true;
89 }
90
91 return false;
92 }
replaceChild()
This is a helper method that replaces a child node nodeToReplace with replacementNode. First, it
tries to find the nodeToReplace among children of the current node. If the node can’t be found the
method will return false. Otherwise, it will change child pointer to point to the new node.
94 /**
95 * @param {BinarySearchTreeNode} nodeToReplace
96 * @param {BinarySearchTreeNode} replacementNode
98 */
99 replaceChild(nodeToReplace, replacementNode) {
100 if (!nodeToReplace || !replacementNode) {
101 return false;
102 }
103
104 if (this.left && this.left === nodeToReplace) {
105 this.left = replacementNode;
106 return true;
107 }
108
109 if (this.right && this.right === nodeToReplace) {
110 this.right = replacementNode;
111 return true;
112 }
113
114 return false;
115 }
copyNode()
This is a static helper method that clones one node attributes to another.
117 /**
118 * @param {BinarySearchTreeNode} sourceNode
119 * @param {BinarySearchTreeNode} targetNode
120 */
121 static copyNode(sourceNode, targetNode) {
122 targetNode.setValue(sourceNode.value);
123 targetNode.setData(sourceNode.data);
124 targetNode.setLeft(sourceNode.left);
125 targetNode.setRight(sourceNode.right);
126 }
findMin()
This method recursively finds the node with the minimum value. The node may be found recursively
by following the left child of every node starting from the root. This will lead us to the very left
bottom node in a tree.
199 /**
201 */
202 findMin() {
203 if (!this.left) {
204 return this;
205 }
206
207 return this.left.findMin();
208 }
findMax()
This method recursively finds the node with the maximum value. The node may be found recursively
by following the right child of every node starting from the root. This will lead us to the very right
bottom node in a tree.
210 /**
212 */
213 findMax() {
214 if (!this.right) {
215 return this;
216 }
217
218 return this.right.findMax();
219 }
toString()
This is a helper method that returns a string representation of the BST. It does in-order traversal
first to convert the tree into a sorted array of values. Then it converts the sorted array of values to
a string.
298 /**
300 */
301 toString() {
302 return this.traverseInOrder().toString();
303 }
BinarySearchTree
BinarySearchTree class is just a thin wrapper on top of the BinarySearchTreeNode class that hides
all the helper methods and makes the work with BST more semantically correct.
06-binary-search-tree/example.js
1 import BinarySearchTreeNode from './BinarySearchTreeNode';
2
3 export default class BinarySearchTree {
4 constructor() {
5 this.root = new BinarySearchTreeNode(null);
6 }
7
8 /**
9 * Find the node by its value.
12 */
13 find(value) {
14 return this.root.find(value);
15 }
16
17 /**
18 * Find the node with min value.
20 */
21 findMin() {
22 return this.root.findMin();
23 }
24
25 /**
26 * Find the node with max value.
28 */
29 findMax() {
30 return this.root.findMax();
31 }
32
33 /**
34 * Insert the new node in a tree.
38 */
39 insert(value, data = null) {
40 return this.root.insert(value, data);
41 }
42
43 /**
44 * Check if tree contains the node with specific value.
47 */
48 contains(value) {
49 return this.root.contains(value);
50 }
51
52 /**
53 * Remove the node from a tree by its value.
56 */
57 remove(value) {
58 return this.root.remove(value);
59 }
60
61 /**
62 * Convert tree to string.
64 */
65 toString() {
66 return this.root.toString();
67 }
68 }
Access Search Insertion Deletion Comment

O(log(n)) O(log(n)) O(log(n)) O(log(n)) For balanced trees
We have O(log(n)) time here because every time we search/insert/remove anything from a tree we
do the number of operations that are proportional to the height of a tree which, in case of a balanced
tree, is described as log(n).
Since every search/insert/remove operation is being done iteratively and on every iteration we’re
cutting approximately half of the tree off. This gives us O(log(n)) time.
Problems Examples
• Convert a Sorted List to a Binary Search Tree²⁷

• Unique Binary Search Trees²⁸
Quiz
Q1: Does the root node in a tree have a parent node?
Q2: Is the node in a binary tree data structure allowed to have only one right child?
Q3: What type of binary search tree will have faster lookup: balanced or unbalanced?
References
BST on Wikipedia https://en.wikipedia.org/wiki/Binary_search_tree²⁹
BST on YouTube https://www.youtube.com/watch?v=wcIRPqTR3Kc³⁰
BST visualization https://www.cs.usfca.edu/∼galles/visualization/BST.html³¹
BST example and test cases https://github.com/trekhleb/javascript-algorithms/tree/master/src/data-
structures/tree/binary-search-tree³²
²⁷https://leetcode.com/problems/convert-sorted-list-to-binary-search-tree/
²⁸https://leetcode.com/problems/unique-binary-search-trees/
²⁹https://en.wikipedia.org/wiki/Binary_search_tree
³⁰https://www.youtube.com/watch?v=wcIRPqTR3Kc
³¹https://www.cs.usfca.edu/~galles/visualization/BST.html
³²https://github.com/trekhleb/javascript-algorithms/tree/master/src/data-structures/tree/binary-search-tree
Binary Heap
A binary heap is a binary tree which also satisfies two additional constraints:
• Shape property: a binary heap is a complete binary tree; that is, all levels of the tree, except
possibly the last one (deepest) are completely filled, and, if the last level of the tree is not
complete, the nodes of that level are filled from left to right.
• Heap property: if P is a parent node of child node C, then the value of P is either greater than
or equal to (in a max heap) or less than or equal to (in a min heap) the key of C.
Take a look at the following example of a binary min-heap below. Since 1 is the smallest number
in the heap, it is placed at the root of the heap. All children of the root are bigger than the root. The
same works for the children: the left child of the root is smaller than its children and so on. The tree
itself is a complete binary tree because all of its levels are full (except the last one that is being filled
from left to right).
Binary Heap 82
Here is an example of a binary max-heap. 100 is the largest of all the numbers in the heap so it is
placed at the root of the heap. All children of the root are smaller than the root. The same works
for the children: the left child of the root is smaller than its children and so on. The tree itself is also
complete binary tree because all its levels are full (except the last one that is being filled from left to
right).
Binary Heap 83
Application
One of the direct applications of the heap is the heap-sort algorithm (one of the best sorting methods
being in-place and with no quadratic worst-case scenarios). The binary heap was introduced by J.
W. J. Williams in 1964, as a data structure particularly used for the heap-sort algorithm.
You might have noticed from the illustrations above that heaps allows us to have a quick access to
the min/max values of a set of values. This property of a heap makes it very efficient when it comes
to implementation of an abstract data type called a priority queue. In a priority queue, an element
with a high priority is served before an element with a low priority. By storing the priorities of the
elements in a heap we’ll be able to have a quick access to the next most important element in a
queue.
Heaps are also used by graph algorithms such as Prim’s minimal-spanning-tree algorithm and
Dijkstra’s shortest-path algorithm.
Binary Heap 84
Basic Operations
Here are the basic operations normally performed on a heap:
• Peek - returns the minimum (for min-heap) or maximum (for max-heap) element of a heap
without extracting it.
• Poll - extracts the minimum (for min-heap) or maximum (for max-heap) element out of a heap.
• Add - adds new element to a heap.
• Remove - removes the element from the heap by its value.
Later in this chapter we will get to exact implementation of each of these operations and you will
see the details. But what is worth mentioning here is that for operations that modify the heap (like
“poll”, “add” or “remove”) we first try to keep the shape property of the heap. And then we try to
modify the heap binary tree structure in such a way that heap property is also met. So there are two
more common heap operations: heapify up and heapify down. These operations are internal and
are not used directly. The only purpose of these two operations is to preserve the heap property after
heap modifications.
Usage Example
Before we move on to the actual implementation let’s imagine we already have a MinHeap class
implemented and let’s use it. This will help us to better understand the meaning of each class method
that we’re going to implement.
07-heap/example.js
1 // Import MinHeap class.

2 import { MinHeap } from './MinHeap';
3
4 // Create min-heap instance.
5 const minHeap = new MinHeap();
6
7 minHeap.isEmpty(); // => true
8
9 // Add new element to min-heap.
10 minHeap.add(5);
11 minHeap.isEmpty(); // => false
12 minHeap.peek(); // => 3
13 minHeap.toString(); // => '5'
14
15 // Add more elements.
16 minHeap.add(3);
Binary Heap 85
17 // Peek operation will always return the minimum element of the heap.
19 minHeap.toString(); // => '3,5'
20
21 minHeap.add(10);
23 minHeap.toString(); // => '3,5,10'
24
25 // Extract the minimum value out of a heap.
26 minHeap.poll(); // => 3
28 minHeap.toString(); // => '5,10'
29
30 // Remove element from a heap.
31 minHeap.remove(10);
33 minHeap.toString(); // => '5'
Implementation
Storing Heap as Array

We will construct a heap as a complete binary tree. This means that it is a binary tree where every
level, except possibly the last one, is completely filled, and all nodes are as far left as possible.
Elements of a heap are commonly stored in array. Any binary tree can be stored in an array, but
because a binary heap is always a complete binary tree, it can be stored compactly (without any
empty cells). No space is required for pointers; instead, the parent and children of each node can be
found by arithmetic on array indices.
Binary Heap 86
After saving the heap as an array we may easily calculate every node’s children and its parent
indices. Let n be the number of elements in the heap and i be an arbitrary valid index of the array
storing the heap. If the tree root is at index 0, with valid indices 0 through n − 1, then each element
a at index i has:
• children at indices 2i + 1 and 2i + 2,

• its parent at index floor((i − 1) � 2).
The floor() function mentioned above is a function that takes as input a real number x and gives
as output the greatest integer less than or equal to x. For example floor(5.4) is equal to 5.
For example:
• The root node (with i = 0) has left child at index 1 (because 2 * 0 + 1 = 1) and right child at
index 2 (because 2 * 0 + 2 = 2).
• The left child of the root node is placed at index 1 (as we’ve just found out) and it has its left
child at index 3 (because 2 * 1 + 1 = 3) and its right child being placed at index 4 (because 2
* 1 + 2 = 4).
• The parent node for the node with index i = 4 has index 1 (because floor(4 - 1 / 2) =
floor(1.5) = 1)
Storing Objects in Heap

Sometimes it may be required to store not only plain numbers but also objects in a heap. For example
we might want to store user objects like {name: 'Paul', priority: 5} and {name: 'Jane',
priority: 3} and others in a heap so that the next call of poll() method would return us the user
with higher priority (in our example is {name: 'Paul', priority: 5}). In order to achieve that we
need to provide a custom way of comparing objects in a heap. Therefore we’re going to implement a
universal Comparator utility class that will provide a common interface for object comparison. After
that we will use Comparator utility to implement a MinHeap class.
Class Comparator
Let’s declare the class.
utils/comparator/Comparator.js
1 export default class Comparator {

Binary Heap 87
constructor
The Comparator constructor will accept the compare function callback as an argument. This function
accepts two values for comparison. It will return 0 if two objects are equal, -1 if the first object is
smaller than the second one and 1 if the first object is greater than the first one.
The Comparator class itself will not know anything about the comparison logic. It will just provide
a common interface (the methods that we’re going to implement below) to work with.
compareFunction is an optional argument for the Comparator constructor. If the compareFunction is
not provided we fall back to defaultCompareFunction() function.
16 /**
17 * @param {function(a: *, b: *)} [compareFunction] - It may be custom compare func\
18 tion that, let's
19 * say may compare custom objects together.
20 */
21 constructor(compareFunction) {
22 this.compare = compareFunction || Comparator.defaultCompareFunction;
23 }
defaultCompareFunction()
This is default fallback comparison function that treats two arguments a and b as a string or a number
and applies common comparison operators to them like ===, < and >.
2 /**
3 * Default comparison function. It just assumes that "a" and "b" are strings or nu\
4 mbers.
5 * @param {(string|number)} a
6 * @param {(string|number)} b
7 * @returns {number}
8 */
9 static defaultCompareFunction(a, b) {
10 if (a === b) {
11 return 0;
12 }
13
14 return a < b ? -1 : 1;
15 }
Binary Heap 88
equal()
This function checks if two variables are equal.

24 /**
25 * Checks if two variables are equal.
26 * @param {*} a
27 * @param {*} b
29 */
30 equal(a, b) {
31 return this.compare(a, b) === 0;
32 }
lessThan()
This function checks if variable a is less than b.

34 /**
35 * Checks if variable "a" is less than "b".
36 * @param {*} a
37 * @param {*} b
39 */
40 lessThan(a, b) {
41 return this.compare(a, b) < 0;
42 }
greaterThan()
This function checks if variable a is greater than b.

Binary Heap 89
44 /**
45 * Checks if variable "a" is greater than "b".
46 * @param {*} a
47 * @param {*} b
49 */
50 greaterThan(a, b) {
51 return this.compare(a, b) > 0;
52 }
lessThanOrEqual()
This function checks if variable a is less than or equal to b.

54 /**
55 * Checks if variable "a" is less than or equal to "b".
56 * @param {*} a
57 * @param {*} b
59 */
60 lessThanOrEqual(a, b) {
61 return this.lessThan(a, b) || this.equal(a, b);
62 }
greaterThanOrEqual()
This function checks if variable a is greater than or equal to b.

64 /**
65 * Checks if variable "a" is greater than or equal to "b".
66 * @param {*} a
67 * @param {*} b
69 */
70 greaterThanOrEqual(a, b) {
71 return this.greaterThan(a, b) || this.equal(a, b);
72 }
Binary Heap 90
Class MinHeap
Let’s move on and implement the MinHeap class.
Let’s import dependencies first. We will use Comparator instance in class constructor in order to
allow class consumers to change the way the heap elements are compared with each other.
4-stack/Stack.js
// Import dependencies.
import Comparator from '../utils/comparator/Comparator';
constructor
Let’s declare the MinHeap class:

4-stack/Stack.js
7 export class MinHeap {
In class constructor we need to create an array that will hold all heap data. To improve the
performance of has() operation we also create the heapElements map that will hold all heap
elements and will allow us to check if the element exists in the heap in O(1) time.
4-stack/Stack.js
8 /**
9 * @constructs MinHeap
10 * @param {Function} [comparatorFunction]
11 */
12 constructor(comparatorFunction) {
13 // Array representation of the heap.
14 this.heapContainer = [];
15
16 // A map of heap elements for fast lookup.
17 this.heapElements = new Map();
18
19 // Allow class consumers to change the way the heap elements are compared with e\
20 ach other.
21 // This is particularly useful when we want to save objects in a heap.
22 this.compare = new Comparator(comparatorFunction);
23 }
Binary Heap 91
Helper Methods
Before we move on and implement main heap operations let’s implement several helper methods
that will help us traverse the heap.
getLeftChildIndex
This helper method will return the index of the left child by the index of the parent. If i is the parent
index then the formula for the left child index is 2 * i + 1.
4-stack/Stack.js
24 /**
25 * @param {number} parentIndex
27 */
28 getLeftChildIndex(parentIndex) {
29 return (2 * parentIndex) + 1;
30 }
getRightChildIndex
This helper method will return the index of the right child by the index of the parent. If i is the
parent index then the formula for the right child index is 2 * i + 2.
4-stack/Stack.js
32 /**
35 */
36 getRightChildIndex(parentIndex) {
37 return (2 * parentIndex) + 2;
38 }
getParentIndex
This helper method will return the index of the parent node by the index of the child node. If i is
the child index then the formula for the parent node index is floor((i - 1) / 2).
Binary Heap 92
4-stack/Stack.js
40 /**
41 * @param {number} childIndex
43 */
44 getParentIndex(childIndex) {
45 return Math.floor((childIndex - 1) / 2);
46 }
hasParent
This helper method indicates whether the node with index childIndex has a parent. If the parent
node index we calculate using this formula: floor((i - 1) / 2) is positive then we return true.
Otherwise, the method returns false (the current node is the root node).
4-stack/Stack.js
48 /**
51 */
52 hasParent(childIndex) {
53 return this.getParentIndex(childIndex) >= 0;
54 }
has
This helper method checks if the item has exists in the heap.
4-stack/Stack.js
123 /**
124 * @param {*} item
126 */
127 has(item) {
128 return !!this.heapElements.get(item);
129 }
hasLeftChild
This helper method indicates whether the node with index parentIndex has a left child. If the child
node index is greater than the heapContainer array length then the current node doesn’t have a left
child.
Binary Heap 93
4-stack/Stack.js
56 /**
59 */
60 hasLeftChild(parentIndex) {
61 return this.getLeftChildIndex(parentIndex) < this.heapContainer.length;
62 }
hasRightChild
This helper method indicates whether the node with index parentIndex has a right child. If the child
node index is greater than the heapContainer array length then it means that current node doesn’t
have right child.
4-stack/Stack.js
64 /**
67 */
68 hasRightChild(parentIndex) {
69 return this.getRightChildIndex(parentIndex) < this.heapContainer.length;
70 }
leftChild
This helper method returns the left child of the node.

4-stack/Stack.js
72 /**
74 * @return {*}
75 */
76 leftChild(parentIndex) {
77 return this.heapContainer[this.getLeftChildIndex(parentIndex)];
78 }
rightChild
This helper method returns the right child of the node.

Binary Heap 94
4-stack/Stack.js
80 /**
82 * @return {*}
83 */
84 rightChild(parentIndex) {
85 return this.heapContainer[this.getRightChildIndex(parentIndex)];
86 }
parent
This helper method returns the parent of the child node.

4-stack/Stack.js
88 /**
90 * @return {*}
91 */
92 parent(childIndex) {
93 return this.heapContainer[this.getParentIndex(childIndex)];
94 }
swap
This helper method swaps two elements of the heapContainer array.

4-stack/Stack.js
96 /**
97 * @param {number} indexOne
98 * @param {number} indexTwo
99 */
100 swap(indexOne, indexTwo) {
101 const tmp = this.heapContainer[indexTwo];
102 this.heapContainer[indexTwo] = this.heapContainer[indexOne];
103 this.heapContainer[indexOne] = tmp;
104 }
isEmpty
This helper method checks if we have at least one element in the heapContainer array.
Binary Heap 95
4-stack/Stack.js
286 /**
288 */
289 isEmpty() {
290 return !this.heapContainer.length;
291 }
find
This helper method finds the positions of the elements in a heap by the value. It is possible that heap
will have duplicate values and thus this method returns an array of positions instead of just one
number (the position of the element we’re searching for). The equality of two elements of a heap is
being checked by using comparator() function.
4-stack/Stack.js
106 /**
108 * @param {Comparator} [comparator]
109 * @return {Number[]}
110 */
111 find(item, comparator = this.compare) {
112 const foundItemIndices = [];
113
114 for (let itemIndex = 0; itemIndex < this.heapContainer.length; itemIndex += 1) {
115 if (comparator.equal(item, this.heapContainer[itemIndex])) {
116 foundItemIndices.push(itemIndex);
117 }
118 }
119
120 return foundItemIndices;
121 }
toString
This helper method returns a string representation of the heap by just calling toString() method
of heapContainer array.
Binary Heap 96
4-stack/Stack.js
293 /**
295 */
296 toString() {
297 return this.heapContainer.toString();
298 }
Internal Methods that Preserve Heap Property
The methods in this section: heapifyUp and heapifyDown are responsible for keeping the heap-
property after any heap modification. The heap modification methods that use heapifyUp and
heapifyDown – peek,poll,add,remove – are explained in the next section.
heapifyUp
This method makes sure that heap-property of the heap is satisfied after new element has been added
to the min-heap. It is possible that new element may be smaller than its parent. In order to fix this
situation we take the last element (last in array or the bottom left in a tree) in the heap container
and lift it up until it is in the correct order with respect to its parent element.
We also will provide a possibility to do the “heapify-up” operation for any element of the heap by
adding arbitrary parameter customStartIndex to the function. We need this for remove() method.
4-stack/Stack.js
131 /**
132 * @param {number} [customStartIndex]
133 */
134 heapifyUp(customStartIndex) {
135 // Take the last element (last in array or the bottom left in a tree)
136 // in the heap container and lift it up until it is in the correct
137 // order with respect to its parent element.
138 let currentIndex = customStartIndex || this.heapContainer.length - 1;
139
140 while (
Binary Heap 97
141 this.hasParent(currentIndex)
142 && this.compare.greaterThan(this.parent(currentIndex), this.heapContainer[curr\
143 entIndex])
144 ) {
145 this.swap(currentIndex, this.getParentIndex(currentIndex));
146 currentIndex = this.getParentIndex(currentIndex);
147 }
148 }
heapifyDown
This method makes sure that heap-property of the heap is satisfied after the root of the heap has
been polled out. The common behaviour in this case is to replace the root of the heap with the last
element of the heapContainer. But the new head element may be greater than its children. In this
case we need to compare the parent element to its children and swap parent with the appropriate
child (smallest child for min-heap, largest child for max-heap). We need to do the same for next
children after the swap.
We also will provide a possibility to do the “heapify-down” operation for any element of the heap by
adding arbitrary parameter customStartIndex to the function. We need this for remove() method.
4-stack/Stack.js
149 /**
150 * @param {number} [customStartIndex]
151 */
152 heapifyDown(customStartIndex = 0) {
153 // Compare the parent element to its children and swap parent with the appropria\
154 te
155 // child (smallest child for MinHeap, largest child for MaxHeap).
156 // Do the same for next children after swap.
157 let currentIndex = customStartIndex;
158 let nextIndex = null;
159
160 while (this.hasLeftChild(currentIndex)) {
161 if (
162 this.hasRightChild(currentIndex)
Binary Heap 98
163 && this.compare.lessThanOrEqual(this.rightChild(currentIndex), this.leftChil\

164 d(currentIndex))
165 ) {
166 nextIndex = this.getRightChildIndex(currentIndex);
167 } else {
168 nextIndex = this.getLeftChildIndex(currentIndex);
169 }
170
171 if (this.compare.lessThanOrEqual(
172 this.heapContainer[currentIndex],
173 this.heapContainer[nextIndex],
174 )) {
175 break;
176 }
177
178 this.swap(currentIndex, nextIndex);
179 currentIndex = nextIndex;
180 }
181 }
Main Heap Methods
Now that we have a bunch of helper methods implemented let’s move on and add main methods of
the heap.
peek
This method will return the head of the heap without actually removing it. It is useful when we just
need to check the minimum element of the min-heap. The head of the heap is a very first element
of the heapContainer array.
4-stack/Stack.js
181 /**
182 * @return {*}
183 */
184 peek() {
185 // Check if heap is not empty.
186 if (this.heapContainer.length === 0) {
187 // If heap is empty then there is nothing to peek.
188 return null;
189 }
190
Binary Heap 99
191 // Return the first element of heap array (the head).

192 return this.heapContainer[0];
193 }
poll
This method will remove and return the head of the heap. It is useful when we need to fetch the
minimum element of the min-heap. The head of the heap is a very first element of the heapContainer
array.
After we remove the head of the heap we need to move the last element of the heapContainer to
the head. This operation will preserve the shape property of the heap (the tree of the heap must be
a complete binary tree).
After moving the last element of the heap to its head we’ve preserved the shape property. But it
is possible that now the heap property is violated and the head of the min-heap is bigger than its
children. To fix that we call heapifyDown() method that in turn will preserve the heap property. We
will implement it shortly.
4-stack/Stack.js
195 /**
196 * @return {*}
197 */
198 poll() {
200 // If heap is empty then there is nothing to poll.
201 return null;
202 }
203
205 // If the heap consist of the one element then just extract it.
206 return this.heapContainer.pop();
207 }
208
Binary Heap 100
209 // Remember the value we want to return.

210 const item = this.heapContainer[0];
211
212 // Move the last element from the end to the head to preserve the "shape propert\
213 y".
214 this.heapContainer[0] = this.heapContainer.pop();
215
216 // Heapify the heap down in order to preserve the "heap-property".
217 this.heapifyDown();
218
219 return item;
220 }
add
This method adds a new element to the min-heap. In order to preserve the shape-property of the
heap, the new element is added to the end of the heapContainer array. After adding the new element,
it is possible that the newly added element is smaller than its parent. In this case, the heap-property
is violated. To fix that, we call the heapifyUp() method that will lift the new element up in a tree
structure so the heap-property is satisfied.
Binary Heap 101
4-stack/Stack.js
221 /**
223 * @return {MinHeap}
224 */
225 add(item) {
226 // Add new item to the end of the heap (shape property is preserved).
227 this.heapContainer.push(item);
228
229 // Add current item to the map of heap elements for fast access.
230 this.heapElements.set(item, item);
231
232 // Make sure that the heap-property is preserved by moving the
233 // element up in case if it smaller than its parent.
234 this.heapifyUp();
235
236 return this;
237 }
Binary Heap 102
remove
This method removes the element(s) from the heap by value. In order to remove the element(s) we
need to do the following:
• Find all elements in the heap with the specified value (it is possible that we might have
duplicates and all of them need to be removed)
• Go through all the elements that needs to be removed. – Move the last element of a heap to
the vacant (removed) position. – For the element that we’ve just moved we need to get its new
parent and children to decide whether heap property is not violated. – If the heap property
is violated we need to call heapifyUp() (if parent is greater than the new element value) or
heapifyDown() (if the element value is greater than the value of any of its children) method.
The function will return this in order to be able to chain the commands like heap.remove(10).add(5)
just for convenience.
4-stack/Stack.js
239 /**
241 * @param {Comparator} [comparator]
242 * @return {MinHeap}
243 */
244 remove(item, comparator = this.compare) {
245 // Find number of items to remove.
246 const numberOfItemsToRemove = this.find(item, comparator).length;
247
248 // Remove current item from the map of heap elements.
249 this.heapElements.delete(item);
250
251 for (let iteration = 0; iteration < numberOfItemsToRemove; iteration += 1) {
252 // We need to find item index to remove each time after removal since
253 // indices are being changed after each heapify process.
254 const indexToRemove = this.find(item, comparator).pop();
255
256 // If we need to remove last child in the heap then just remove it.
257 // There is no need to heapify the heap afterwards.
258 if (indexToRemove === (this.heapContainer.length - 1)) {
259 this.heapContainer.pop();
260 } else {
261 // Move last element in heap to the vacant (removed) position.
262 this.heapContainer[indexToRemove] = this.heapContainer.pop();
263
264 // Get parent.
Binary Heap 103
265 const parentItem = this.parent(indexToRemove);

266
267 // If there is no parent or parent is in correct order with the node
268 // we're going to delete then heapify down. Otherwise heapify up.
269 if (
270 this.hasLeftChild(indexToRemove)
271 && (
272 !parentItem
273 || parentItem <= this.heapContainer[indexToRemove]
274 )
275 ) {
276 this.heapifyDown(indexToRemove);
277 } else {
278 this.heapifyUp(indexToRemove);
279 }
280 }
281 }
282
283 return this;
284 }
Complexities
Peek Poll Add Remove

O(1) O(log(n)) O(log(n)) O(log(n))
Polling, adding and removing from queue have O(log(n)) time complexity because every time we
do these operations we need to re-heapify the tree of a heap. This would make us access all the
elements in a branch of a tree in worst case.
Peeking has O(1) time complexity because we’re just accessing the first element (the root) of the
heap array.
Problems Examples
• Check Completeness of a Binary Tree³³

• Binary Tree Inserter³⁴
³³https://leetcode.com/problems/check-completeness-of-a-binary-tree/
³⁴https://leetcode.com/problems/complete-binary-tree-inserter/
Binary Heap 104
Quiz
Q1: What are two constraints that each binary heap must satisfy?
Q2: When we do poll operation on a heap and extract the head then what value must be put on the
place of extracted one?
Q3: What is the time complexity of poll() operation in a MinHeap?
References
Heap on Wikipedia <a href=”https://en.wikipedia.org/wiki/Heap(data_structure)”>https://en.wikipedia.org/wiki/He
structure)</a>
Heap on YouTube https://www.youtube.com/watch?v=t0Cq6tVNRBA³⁵
Heap example and test cases https://github.com/trekhleb/javascript-algorithms/tree/master/src/-
data-structures/heap³⁶
³⁵https://www.youtube.com/watch?v=t0Cq6tVNRBA
³⁶https://github.com/trekhleb/javascript-algorithms/tree/master/src/data-structures/heap
Priority Queue
Priority queue is a data type which is like a regular queue or stack data structure, but where
additionally each element has a “priority” associated with it. In a priority queue, an element with
high priority is served before an element with low priority (or vice versa depending on the queue
setup).
While priority queues are often implemented with heaps, they are conceptually distinct from heaps.
A priority queue is an abstract concept like “a list” or “a map”; just as a list can be implemented with
a linked list or an array, a priority queue can be implemented with a heap or other methods such as
an unordered array.
The idea of using Heap data structure to implement priority queue is based on the heap property of
always returning the lowest (in case of min heap) or highest (in case of max heap) elements when
polling data from the heap. Once we change the heap behavior to respect the priority of each item
instead of their values we will get a priority queue.
On the example below you may see the order of polling the elements from priority queue. The poll()
operation fetches the next most prioritized element from the queue and reduces the queue by one
element.
Normally the priority queue is implemented in a way that item with higher value of the priority
property is being polled from the queue first. But alternative way of implementing priority queue
is to treat lower numbers as more prioritised so that the item with priority 0 will be pulled before
the item with priority 8. It doesn’t matter that much which way we take. But since we already have
MinHeap class implemented in one of the previous chapters let’s implement the reversed version of
priority queue.
Priority Queue 106
Application
Priority queue is being used as a part of Dijkstra’s algorithm (finding the shortest paths in a graph),
Prim’s algorithm (finding the minimum spanning tree of a connected and undirected graph), sorting
algorithms and others.
Basic Operations
Here are the basic operations normally performed on a priority queue:
• Poll - extracts the next most important item from the queue according to its priority (in our
case we will fetch the items with lower values of priorities first).
• Peek - returns the next prioritized item from the queue without extracting it.
• Add - adds new element to the queue with specific priority.
• Remove - removes the element from the queue by its value.
• Change Priority - changes the priority of the item in priority queue.
• Has Value - checks if item with specific value exists in a queue.
Usage Example
Before we move on with actual PriorityQueue class implementation let’s imagine that we already
have it implemented and try to use it. This will help us to understand the interface of working the
PriorityQueue class.
Just for example we may do a simple trick and sort the array of cities in population increasing order
using PriorityQueue.
Priority Queue 107
07-priority-queue/example.js
1 // Import dependencies.
2 import { PriorityQueue } from './PriorityQueue';
3
4 // Let's init an empty priority queue.
5 const priorityQueue = new PriorityQueue();
6
7 // Let's init an array of US cities with population specified in millions.
8 const notSortedCities = [
9 { name: 'New York', population: 8.6 },
10 { name: 'Chicago', population: 2.7 },
11 { name: 'San Francisco', population: 0.84 },
12 { name: 'Houston', population: 2 },
13 ];
14
15 // Let's add all cities to our priority queue treating population as a priority.
16 notSortedCities.forEach(city => priorityQueue.add(city, city.population));
17
18 // Now let's fetch cities from priority queue one by one and putting them into the
19 // sortedCities array. We're expecting the cities bing polled in population increasi\
20 ng order.
21 const sortedCities = [];
22 while (priorityQueue.peek()) {
23 // While we can peek something from the queue it would mean that queue is not empt\
24 y yet.
25 sortedCities.push(priorityQueue.poll());
26 }
27
28 // We're expecting the sortedCities to be truly sorted in population increasing orde\
29 r.
31 console.log(sortedCities);
32
33 /*
34 The output would be:
35
36 [
37 { name: 'San Francisco', population: 0.84 },
38 { name: 'Houston', population: 2 },
39 { name: 'Chicago', population: 2.7 },
40 { name: 'New York', population: 8.6 },
41 ]
42 */
Priority Queue 108
Implementation
We’re going to build PriorityQueue class based on MinHeap class as it was explained above. The
MinHeap class supports custom comparator functions. So we will use Comparator instance to alter
the comparison algorithm in MinHeap so that it would compare items priorities instead of values. For
more details about MinHeap and Comparator implementation please address the “Heap” chapter of
this book.
Let’s import dependencies that we’re going to use.
07-priority-queue/PriorityQueue.js
import { MinHeap } from '../07-heap/MinHeap';
Let’s declare the PriorityQueue class and extend it from MinHeap class.
7 export class PriorityQueue extends MinHeap {
constructor()
In constructor we will call the parent constructor of MinHeap class first by calling super().
Next we will create a map of priorities this.priorities. This map we store all priority queue items
priority values mapped by items themselves. For example this.priorities may look like {itemA:
10, itemB: 14}, where itemaA and itemB are items keys and 10 and 14 are items priorities.
Also we will create a custom this.compare() function. This is the main place when we alter MinHeap
behavior so that MinHeap methods respect items priorities instead of items values. You may see here
we’re providing a custom this.comparePriority() function as a main comparison function. This
function will be explained below.
Priority Queue 109
34 constructor() {
35 // Call MinHip constructor first.
36 super();
37
38 // Setup priorities map.
39 this.priorities = new Map();
40
41 // Use custom comparator for heap elements that will take element priority
42 // instead of element value into account.
43 this.compare = new Comparator(this.comparePriority.bind(this));
44 }
comparePriority()
This function compares two items according to their priorities. To do so it gets priorities of item a
and b from this.priorities map. If both priorities are equal it returns 0, if a is less important than
b then it returns -1 (items are not in correct order), otherwise it returns 1 (items are in correct order).
8 /**
9 * Compares priorities of two items.
10 * @param {*} a
11 * @param {*} b
13 */
14 comparePriority(a, b) {
15 if (this.priorities.get(a) === this.priorities.get(b)) {
16 return 0;
17 }
18 return this.priorities.get(a) < this.priorities.get(b) ? -1 : 1;
19 }
compareValue()
This method does the same as comparePriority() method but instead of comparing priorities it
compares items themselves.
Priority Queue 110
21 /**
22 * Compares values of two items.
23 * @param {*} a
24 * @param {*} b
26 */
27 compareValue(a, b) {
28 if (a === b) {
29 return 0;
30 }
31 return a < b ? -1 : 1;
32 }
add()
This method adds items to the priority queue according to their priority value. It firsts sets up new
priority value to the this.priorities map and than calls the super.add() method to add the item
to the heap.
46 /**
47 * Add item to the priority queue.
48 * @param {*} item - item we're going to add to the queue.
49 * @param {number} [priority] - items priority.
50 * @return {PriorityQueue}
51 */
52 add(item, priority = 0) {
53 this.priorities.set(item, priority);
54 super.add(item);
55 return this;
56 }
remove()
This method removes the item from the priority queue. It accepts two parameters, the item that is
going to be removed and optional customFindingComparator callback that allows to customize the
way the item will be searched in a heap. This is useful when we want to store objects in priority
queue and we want to implement custom logic of finding them (i.e. finding objects my name property
in it).
Priority Queue 111
58 /**
59 * Remove item from priority queue.
60 * @param {*} item - item we're going to remove.
61 * @param {Comparator} [customFindingComparator] - custom function for finding the\
62 item to remove
64 */
65 remove(item, customFindingComparator) {
66 super.remove(item, customFindingComparator);
67 this.priorities.delete(item);
68 return this;
69 }
changePriority()
This method changes priority of the item in a queue. Since on every priority change the heap needs
to be rebuild in order for the item to take a proper place in a heap we remove the item from the
heap first and then add it back but with updated priority. We could alternatively implement heap
rebuilding without removing the item but for simplicity reasons let’s stick with this approach.
70 /**
71 * Change priority of the item in a queue.
72 * @param {*} item - item we're going to re-prioritize.
73 * @param {number} priority - new item's priority.
75 */
76 changePriority(item, priority) {
77 this.remove(item, new Comparator(this.compareValue));
78 this.add(item, priority);
79 return this;
80 }
Complexities
Since PriorityQueue is based on MinHeap class and its main methods just wraps the main methods
of MinHeap class with a altering the comparator function we may assume that time complexity
of PriorityQueue class are the same as of MinHeap class. Also notice that all operations with
this.priorities map are done in O(1) time.
Priority Queue 112

Problems Examples
• Queue Reconstruction by Height³⁷

• Orderly Queue³⁸
Quiz
Q1: If we would want to implement a queue that will return the items with higher value of priority
first, what we would need to change in PriorityQueue class implementation?
Q2: What is the time complexity of poll() operation in PriorityQueue?
References
Priority Queue on Wikipedia https://en.wikipedia.org/wiki/Priority_queue³⁹
Priority Queue on YouTube https://www.youtube.com/watch?v=wptevk0bshY⁴⁰
Priority Queue example and test cases https://github.com/trekhleb/javascript-algorithms/tree/mas-
ter/src/data-structures/priority-queue⁴¹
³⁷https://leetcode.com/problems/queue-reconstruction-by-height/
³⁸https://leetcode.com/problems/orderly-queue/
³⁹https://en.wikipedia.org/wiki/Priority_queue
⁴⁰https://www.youtube.com/watch?v=wptevk0bshY
⁴¹https://github.com/trekhleb/javascript-algorithms/tree/master/src/data-structures/priority-queue
Graphs
A Graph is a collection of vertices and edges that connect all or some of those vertices together.
A Graph may be directed or undirected. In a directed graph each edge has a direction (like a
one-way street). You may access an end-node from a start-node but not the other way around. In
undirected graphs the edges don’t have a specific direction (like a two-way street) and thus every
two vertices that are connected together have access to each other.
What is a vertex?
A vertex is a basic unit used in graphs to represent data. The terms “node” and “vertex” are
used interchangeably to describe how data is represented in graphs. Vertices (the plural
form of vertex) are connected by edges.
Another way of illustrating the difference between directed and undirected graphs is to think about
social networks. If each vertex is a “user”, then in a directed graph each edge would a “follow”
relationship – you may follow someone but the doesn’t mean the other person follows you back.
Whereas in an undirected graph the edge would be like a “friend” relationship – if someone is in
your friend list then it automatically means that you’re in that person’s friend list too.
In the example above we have three different graphs that all have vertices and edges. All of the
graphs have a set of vertices {A, B, C, D} and all of the graphs have a set of edges. The undirected
graph has a set of vertices: {AB, BC, AC, CD} . The directed graph also has a set of vertices: {AB, BC,
CB, AC, CD} Notice that for directed graph to connect vertices B and C together in both directions
there are two edges BC and CB in the set.
If two vertices are connected to each other by one edge, we call these vertices adjacent or neighbors.
In the undirected graph above, for example, the vertices A and B are neighbors as well as the vertices
A and C. But the vertices B and D are not neighbors since these vertices are connected via more than
one edge (B is connected to C which is connected to D).
Graphs 114
The sequence of edges between two vertices is called a path. In the directed graph above there is no
path from D to A but there is a path from A and D and the path is the set of edges: {AC, CD}.
All vertices in a graph do not need to be connected. When there are vertices that are unreachable
we call this type of graph a disconnected graph. If every pair of vertices in a graph has an edge, it
is called a connected graph.
Graphs can also have cycles. If a graph doesn’t have cycles it is called acyclic graph.
A weighted graph is a graph where each edge has a weight associated with it. Depending on
the context, the edge weight may represent different concepts like “distance”, “traffic”, “cost”,
“resistance” and so on.
Application
• Maps applications use graphs to compute shortest paths between vertices (addresses, cities,
locations).
• Social networks use graphs for friends suggestion algorithms.
• Package managers use graphs to calculate correct order of installing packages based on their
dependencies.
• Search engines use graphs to analyze page relevance. In graphs used by search engines, the
web-page is a vertex and the link from this web-page to another is directed edge.
• Network providers use graphs to analyze network traffic and security. In this case each vertex
has an IP address and every edge represent the network packets that flow between different IP
addresses.
Graph Representation
Graphs are normally being represented as adjacency lists or adjacency matrices. Let’s take the graphs
from the beginning of this chapter as an example.
Adjacency Matrix
The graph with n vertices may also be represented as an adjacency matrix of size n Ã— n of binary
values (0 or 1). In such matrix the value 1 at position (i, j) means that there is an edge between
vertices i and j. Otherwise if the value is 0 there is no edge between those vertices.
Graphs 115
Undirected graph adjacency matrix example

For undirected graphs the adjacency matrix will be symmetrical.
A B C D
A 0 1 1 0
B 1 0 1 0
C 1 1 0 1
D 0 0 1 0
Directed graph adjacency matrix example

For directed graph the adjacency matrix may not be symmetrical.
A B C D
A 0 1 1 0
B 0 0 1 0
C 0 1 0 1
D 0 0 0 0
Directed weighted graph adjacency matrix example

In case of the weighted graphs the weights of the edges may be used instead of boolean 0/1 values.
A B C D
A 0 5 3 0
B 0 0 1 0
C 0 4 0 8
D 0 0 0 0
Adjacency List
A popular way to represent a graph is to have a Node (or Vertex) entity stored as an object which
contains a list of all the adjacent Nodes.
Graphs 116
Sometimes, instead of the list of adjacent Nodes the list of Edge entities may be used. The Edge entity
stores startNode, endNode and weight properties. This approach may be more convenient when we
need to store some meta-data for the edges like weight.
Graphs 117
class Node {
constructor() {
/** @type {Edge[]} */
this.edges = [];
}
}
class Edge {
constructor() {
/** @type {Node} */
this.startVertex = null;
/** @type {Node} */
this.endVertex = null;
this.weight = null;
}
}
class Graph {
constructor() {
/** @type {Node[]} */
this.nodes = [];
}
}
We also need to have a separate Graph class here to support graphs which are not connected (they
contain outstanding nodes).
In this chapter we will implement graphs using adjacency list approach.
Adjacency Matrix and Adjacency List Comparison

Graph related algorithms may be implemented using adjacency matrix or adjacency list interchange-
ably. The difference is that for some algorithms one way of implementing the graphs may be more
efficient than other depending on the operations that are going to be performed on the graph.
The main issue with adjacency matrices is their space complexity of O(n2). For sparse
graphs a lot of matrix space will be wasted (filled with 0). In case of adjacency list the space
complexity will be proportional to the number of vertices and edges.
In an adjacency list we can iterate over the neighbors of the vertex. In an adjacency matrix we need
to iterate over all the vertices in the graph to get all the neighbors of the vertex. Adjacency matrices
have fast lookups to check for presence or absence of a specific edge.
In most real-world problems we’re dealing with sparse and large graphs, which are better suited for
adjacency list representations that we’re going to implement in this chapter.
Graphs 118
Basic Operations
The most common and basic graph operations are:
• Add/delete vertex to the graph.

• Add/delete edge (connection) between two vertices.
• Get all neighbors of the vertex.
• Check if two vertices are connected.
• Find a vertex by its value.
Usage Example
Before we implement the JavaScript version of our graph let’s imagine that we already have Graph,
GraphVertex and GraphEdge classes and describe how we’re going to use them.
Let’s say we want to create a really simple social network model. We want to be able to register new
users, to create “friends” relationship and to see the list of friends of specific user.
08-graph/example.js
2 import Graph from './Graph';
3 import GraphVertex from './GraphVertex';
4 import GraphEdge from './GraphEdge';
5
6 // Create a network.
7 const network = new Graph();
8
9 // Create users.
10 const bill = new GraphVertex('Bill');
11 const mary = new GraphVertex('Mary');
12 const john = new GraphVertex('John');
13 const jane = new GraphVertex('Jane');
14
15 // Register users in our network.
16 network
17 .addVertex(bill)
18 .addVertex(mary)
19 .addVertex(john)
20 .addVertex(jane);
21
22 // Check if users have been registered successfully.
Graphs 119
23 network.getVertexByKey('Bill'); // -> bill

24 network.getVertexByKey('Mary'); // -> mary
25
26 // Establish friendship connections.
27 network
28 .addEdge(new GraphEdge(bill, mary))
29 .addEdge(new GraphEdge(john, jane))
30 .addEdge(new GraphEdge(jane, mary));
31
32 // Check if specific users are friends.
33 network.findEdge(bill, mary); // -> GraphEdge entity
34 network.findEdge(john, jane); // -> GraphEdge entity
35 network.findEdge(bill, john); // -> null
36
37 // Get all friends of specific user.
38 // eslint-disable-next-line no-unused-expressions
39 mary.getNeighbors().length; // -> 2
40 mary.getNeighbors(); // -> [bill, jane]
Implementation
Let’s split the implementation into three parts. First, let’s implement GraphEdge and GraphVertex
classes, then wrap them into the Graph class.
GraphEdge
08-graph/GraphEdge.js
1 export default class GraphEdge {
constructor
The constructor method will accept startVertex and endVertex along with the weight parameters.
The edge instance will represent a connection between startVertex and endVertex. For a directed
graph the order of the parameters (startVertex and endVertex) are important because if you
remember from the introduction, the edges in a directed graph are a “one way street” – they can
only go one direction. For an undirected graph, the order of the parameters are not important.
Graphs 120
2 /**
3 * GraphEdge constructor.
4 * @param {GraphVertex} startVertex
5 * @param {GraphVertex} endVertex
6 * @param {number} [weight=1]
7 */
8 constructor(startVertex, endVertex, weight = 0) {
9 this.startVertex = startVertex;
10 this.endVertex = endVertex;
11 this.weight = weight;
12 }
getKey()
This method will return unique edge key which will consist of start and end vertices keys. This key
will be used in the Graph class later to quickly look up the edges of the graph We will implement
getKey() method for GraphVertex instance below. The key value will look something like A_B which
would mean that this edge connects vertices A and B together.
14 /**
15 * Get string representation of the edge key.
17 */
18 getKey() {
19 const startVertexKey = this.startVertex.getKey();
20 const endVertexKey = this.endVertex.getKey();
21
22 return `${startVertexKey}_${endVertexKey}`;
23 }
toString()
This method will return a string representation of the edge. For simplicity’s sake let’s return the edge
key here.
Graphs 121
25 /**
26 * Convert edge to string.
28 */
29 toString() {
30 return this.getKey();
31 }
GraphVertex
We will use the LinkedList class from the previous chapters to store all adjacent edges of the vertex.
Let’s import our dependencies first.
08-graph/GraphVertex.js
Let’s now create a class.

4 export default class GraphVertex {
constructor
The constructor method will accept the value of the vertex. Also in this method we will create an
empty edges using our LinkedList class that will store all adjacent edges.
5 /**
6 * Graph vertex constructor.
8 */
9 constructor(value) {
11 this.edges = new LinkedList();
12 }
addEdge()
This method adds new edge to the current vertex. It does it by simple appending to the edge
linked list. The method returns this pointer to method to allow for chaining other methods (e.g.,
vertex.addEdge(edgeAB).addEdge(edgeBA)).
Graphs 122
14 /**
15 * Add new edge to the vertex.
16 * @param {GraphEdge} edge
17 * @returns {GraphVertex}
18 */
19 addEdge(edge) {
20 this.edges.append(edge);
21
22 return this;
23 }
deleteEdge()
This method deletes the edge from the LinkedList object.

25 /**
26 * Delete vertex edge.
28 */
29 deleteEdge(edge) {
30 this.edges.delete(edge);
31 }
getEdges()
This method returns an array of all adjacent edges of the vertex. To do so it converts the edge’s
LinkedList into array of linked list vertices and then goes through each lined list vertex using map()
function and extracts the values from them. Those values are our vertex edge entities.
Graphs 123
57 /**
58 * Get all vertex edges as an array.
59 * @return {GraphEdge[]}
60 */
61 getEdges() {
62 return this.edges.toArray().map(linkedListNode => linkedListNode.value);
63 }
hasEdge()
This method checks whether the vertex has a specific edge instance. To do so, it uses the LinkedList
find() method. We pass in a function as an argument to the find() method. Inside the function,
we look for the correct edge by doing a strict comparison of the objects using ===. If the edge has
been found, we convert it to boolean by doing double conversion of the object to boolean using !!.
65 /**
66 * Check if vertex has a specific edge connected to it.
67 * @param {GraphEdge} requiredEdge
68 * @returns {boolean}
69 */
70 hasEdge(requiredEdge) {
71 const edgeNode = this.edges.find({
72 callback: edge => edge === requiredEdge,
73 });
74
75 return !!edgeNode;
76 }
getNeighbors()
This methods returns an array of all adjacent vertices of the current vertex. To do so we need to:
• go through all adjacent edges,

• for each edge we need to check the startVertex and the endVertex and return the vertex which
is not equal to the current one.
We need to do so because we can’t guarantee that the current vertex is stored as startVertex in the
edge. For example consider the situation when we have vertices A and B and the edge AB. In this
case the edge AB will be added to both vertices A and B. For vertex A it will be a startVertex in
edge AB but for vertex B it will be the endVertex. Both of these cases need to be handled.
Graphs 124
33 /**
34 * Get the list of vertex neighbors.
35 * @returns {GraphVertex[]}
36 */
37 getNeighbors() {
38 const edges = this.edges.toArray();
39
40 /** @param {LinkedListNode} node */
41 const neighborsConverter = (node) => {
42 const edge = node.value;
43 const neighbor = edge.startVertex === this ? edge.endVertex : edge.startVertex;
44
45 // Add edge property to the neighbor so that consumers
46 // of this method could access edge property like weight instantly.
47 neighbor.edge = edge;
48
49 return neighbor;
50 };
51
52 // Return either start or end vertex.
53 // For undirected graphs it is possible that current vertex will be the end one.
54 return edges.map(neighborsConverter);
55 }
hasNeighbor()
This method checks if current vertex has a specific neighbor. To do so it uses the linked list find()
method to find the edge which connects the current vertex to the specified neighbor vertex. Then
the result of find() operation is converted to a boolean value using !! sequence. If there is an edge
from current vertex to the specified neighbor vertex this method will return true. Otherwise it will
return false.
Graphs 125
78 /**
79 * Check if specific vertex is a neighbor of the current vertex.
80 * @param {GraphVertex} vertex
81 * @returns {boolean}
82 */
83 hasNeighbor(vertex) {
84 const edgeToNeighbor = this.edges.find({
85 callback: edge => edge.startVertex === vertex || edge.endVertex === vertex,
86 });
87
88 return !!edgeToNeighbor;
89 }
findEdge()
This method finds the edge that connects current vertex to the specified vertex. To do so it uses the
LinkedList find() method with edgeFinder() the callback. In edgeFinder() callback we’re trying
to find a vertex which has startVertex or endVertex being set to the specified vertex object from
findEdge() function arguments. If the edge is found this method will return its value. Otherwise it
will return null.
91 /**
92 * Find edge that connects current vertex to the specified vertex.
93 * @param {GraphVertex} vertex
94 * @returns {(GraphEdge|null)}
95 */
96 findEdge(vertex) {
97 const edgeFinder = (edge) => {
98 return edge.startVertex === vertex || edge.endVertex === vertex;
99 };
100
101 const edge = this.edges.find({ callback: edgeFinder });
102
103 return edge ? edge.value : null;
104 }
deleteAllEdges()
This method deletes all edges that are connected to the current vertex. It simply goes through all the
edges and executes the deleteEdge() method for every edge.
Graphs 126
114 /**
115 * Delete all edges, connected to the vertex.
116 * @return {GraphVertex}
117 */
118 deleteAllEdges() {
119 this.getEdges().forEach(edge => this.deleteEdge(edge));
120
121 return this;
122 }
getKey()
This method returns a vertex’s key. We use the vertex’s value as a unique key. We will need this key
in the Graph class later to store all vertices in a map to quickly lookup vertices.
106 /**
107 * Get vertex string key.
108 * @returns {string}
109 */
110 getKey() {
111 return this.value;
112 }
toString()
This method creates a string representation of the vertex. We print out the value of the vertex.
124 /**
125 * Get string representation of graph vertex.
126 * @returns {string}
127 */
128 toString() {
129 return `${this.value}`;
130 }
Graph
Finally, we can create the Graph class. This class is needed to support disconnected graphs where we
may have several unconnected vertices.
Graphs 127
08-graph/Graph.js
1 export default class Graph {
constructor
To support directed and undirected graphs the constructor method will accept isDirected flag. We
also create empty this.vertices and this.edges maps (objects) that we will use for storing all
vertices and edges that are added to the graph.
08-graph/Graph.js
2 /**
3 * Graph constructor.
4 * @param {boolean} isDirected
5 */
6 constructor(isDirected = false) {
7 this.vertices = {};
8 this.edges = {};
9 this.isDirected = isDirected;
10 }
addVertex()
This method adds a vertex instance to the graph. It simply creates a unique key in the this.vertices
object and stores a newVertex instance in it.
08-graph/Graph.js
12 /**
13 * Add new vertex to the graph.
14 * @param {GraphVertex} newVertex
15 * @returns {Graph}
16 */
17 addVertex(newVertex) {
18 this.vertices[newVertex.getKey()] = newVertex;
19
20 return this;
21 }
getVertexByKey()
This method returns the vertex by its key from the this.vertices object.
Graphs 128
08-graph/Graph.js
23 /**
24 * Get vertex by its key.
25 * @param {string} vertexKey
26 * @returns GraphVertex
27 */
28 getVertexByKey(vertexKey) {
29 return this.vertices[vertexKey];
30 }
getAllVertices()
This method returns all graph vertices. To do so it converts this.vertices object into array by using
Object.values() method.
08-graph/Graph.js
32 /**
33 * Get the list of all graph vertices.
34 * @return {GraphVertex[]}
35 */
36 getAllVertices() {
37 return Object.values(this.vertices);
38 }
getAllEdges()
This method returns the list of all graph edges. To do so it uses Object.values() method that
converts this.edges object into array.
08-graph/Graph.js
40 /**
41 * Get the list of all graph edges.
42 * @return {GraphEdge[]}
43 */
44 getAllEdges() {
45 return Object.values(this.edges);
46 }
Graphs 129
addEdge()
This method adds specific edge instance to the graph. There are a few things we need to take care
of when adding a new edge.
First we need to check whether some (or all) of the vertices in the edge have been already added to the
graph or not. If some (or all) the vertices from the edge have not been added before to this.vertex
object we need to add them. To check vertices existence we use getVertexByKey() method. If it
returns undefined it means we need to create a vertex by using addVertex() method.
Second we need to check whether the edge has been already added to the this.edges object. In case
if edge has been already added we throw an error. Otherwise we add the edge to the this.edges
object.
And finally if graph is directed then we need to add this edge to startVertex instance only.
Otherwise this edge is being added for both startVertex and endVertex instances.
08-graph/Graph.js
48 /**
49 * Add new edge to the graph.
51 * @returns {Graph}
52 */
53 addEdge(edge) {
54 // Try to find and end start vertices.
55 let startVertex = this.getVertexByKey(edge.startVertex.getKey());
56 let endVertex = this.getVertexByKey(edge.endVertex.getKey());
57
58 // Insert start vertex if it wasn't inserted.
59 if (!startVertex) {
60 this.addVertex(edge.startVertex);
61 startVertex = this.getVertexByKey(edge.startVertex.getKey());
62 }
63
64 // Insert end vertex if it wasn't inserted.
65 if (!endVertex) {
66 this.addVertex(edge.endVertex);
67 endVertex = this.getVertexByKey(edge.endVertex.getKey());
68 }
69
70 // Check if edge has been already added.
71 if (this.edges[edge.getKey()]) {
72 throw new Error('Edge has already been added before');
73 } else {
74 this.edges[edge.getKey()] = edge;
Graphs 130
75 }
76
77 // Add edge to the vertices.
78 if (this.isDirected) {
79 // If graph IS directed then add the edge only to start vertex.
80 startVertex.addEdge(edge);
81 } else {
82 // If graph ISN'T directed then add the edge to both vertices.
83 startVertex.addEdge(edge);
84 endVertex.addEdge(edge);
85 }
86
87 return this;
88 }
deleteEdge()
This method deletes specific edge from the graph. We need to delete the edge from the this.edges
object first. And then we need to delete it also from the relevant vertices (startVertex and
endVertex).
08-graph/Graph.js
90 /**
91 * Delete specific edge from the graph.
93 */
94 deleteEdge(edge) {
95 // Delete edge from the list of edges.
96 if (this.edges[edge.getKey()]) {
97 delete this.edges[edge.getKey()];
98 } else {
99 throw new Error('Edge not found in graph');
100 }
101
102 // Try to find and end start vertices and delete edge from them.
103 const startVertex = this.getVertexByKey(edge.startVertex.getKey());
104 const endVertex = this.getVertexByKey(edge.endVertex.getKey());
105
106 startVertex.deleteEdge(edge);
107 endVertex.deleteEdge(edge);
108 }
Graphs 131
findEdge()
This method finds an edge by startVertex and endVertex instances. It basically checks whether two
vertices in the graph are connected. To do so it first tries to find a startVertex in the list of vertices.
If there is no startVertex then method returns null. If the startVertex is found then it tries to find
an edge that connects startVertex with endVertex by using the findEdge() method.
08-graph/Graph.js
110 /**
111 * Find the edge by start and end vertices.
112 * @param {GraphVertex} startVertex
113 * @param {GraphVertex} endVertex
114 * @return {(GraphEdge|null)}
115 */
116 findEdge(startVertex, endVertex) {
117 const vertex = this.getVertexByKey(startVertex.getKey());
118
119 if (!vertex) {
120 return null;
121 }
122
123 return vertex.findEdge(endVertex);
124 }
toString()
This method returns a string representation of the graph. To do so, it converts this.vertices into
array of graph vertex instances. And then it converts this array into string.
The string will look something like 'A,B' in case if graph consists of two vertices with keys 'A' and
'B'.
08-graph/Graph.js
126 /**
127 * Convert graph to string.
129 */
130 toString() {
131 return Object.keys(this.vertices).toString();
132 }

Graphs 132
Implementation Add Vertex Remove Vertex Add Edge Remove Edge

Adjacency list O(1) O(E+V) O(1) O(E)
Adjacency O(V2) O(V2) O(1) O(1)
matrix
Where: V - number of vertices in graph E - number of edges in graph
Problems Examples
Here are some graph related problems that you might encounter during the interview:
• Find the Town Judge⁴²

• Clone Graph⁴³
• Course Schedule⁴⁴
Quiz
Q1: Is a tree data structure a type of graph?
Q2: Can we call a set of vertices {A, B} and an empty set of edges {} a graph?
Q3: What ways of representing the graphs in code do you know?
References
Graph on Wikipedia <a href=”https://en.wikipedia.org/wiki/Graph(abstract_data_type)”>https://en.wikipedia.org/w
data_type)</a>
Graph on YouTube https://www.youtube.com/watch?v=gXgEDyodOJU⁴⁵
Graph example and test cases https://github.com/trekhleb/javascript-algorithms/tree/master/src/-
data-structures/graph⁴⁶
⁴²https://leetcode.com/problems/find-the-town-judge/
⁴³https://leetcode.com/problems/clone-graph/
⁴⁴https://leetcode.com/problems/course-schedule/
⁴⁵https://www.youtube.com/watch?v=gXgEDyodOJU
⁴⁶https://github.com/trekhleb/javascript-algorithms/tree/master/src/data-structures/graph
Bit Manipulation
Intro
In computers, a bit (which is short for “binary digit”) is the most basic representation of data for a
computer. A binary digit can be one of two values: 0 or 1.
We use a combination of 1s and 0s to represent numbers with bits. For example, the number 10 in
decimal is represented like this in binary:
1010
How do we get this number? If we want to convert an integer into binary, we divide the number by
two and then the use the remainder to determine whether we use a 1 or a 0 for that position:
10/2 = 5 with remainder 0 5/2 = 2 with remainer 1 2/2 = 1 with remainder 0 1/2 = 0 with remainder 1
Then we display our number in reverse order that we calculated it - 1010 and that gets us the binary
representation of 10.
JavaScript uses 64 bits to represent numbers. So in the case of the number 10 in binary, it would be
1010 with 60 0s in front of it:
0000000000000000000000000000000000000000000000000000000000001010
There are a few operations we can use when working with binary numbers and we will use each of
these operations in our code examples. The table below explains each of these operations:
Operator Name Description

———– ———- ———-
& AND Sets each bit to 1 if both bits are 1
OR Sets each bit to 1 if one of two bits is 1
^ XOR Sets each bit to 1 if only one of two bits is 1
∼ NOT Inverts all the bits
<< Zero fill left shift Shifts left by pushing zeros in from the
right and let the leftmost bits fall off
>> Signed right shift Shifts right by pushing copies of the leftmost bit in
from the left, and let the rightmost bits fall off
>>> Zero fill right shift Shifts right by pushing zeros in from the left, and
let the rightmost bits fall off
Any time one of these types of operators is used in JavaScript, JavaScript converts the numbers to
32 bits and then returns the result as a 64 bit signed number.
Bit Manipulation 134
What is a signed number?
Binary numbers are just 1s and 0s. There is no way to determine whether a number is a negative
number or a positive number as there is with decimal numbers. A signed binary number is one that
uses the leftmost digit (also known as the “most significant bit” or MSB) to determine whether the
number is positive or negative. If the leftmost digit is a 1, the number is negative. If the leftmost
digit is a 0, the number is positive.
Applications
When bitwise operations are performed in JavaScript the numbers are converted from 64 bit numbers
to 32 bit signed integers. So for example if you use the NOT operator on 6 like this:
∼6
6 in 64 bits looks like this:

0000000000000000000000000000000000000000000000000000000000000110
When you perform the NOT operation on this number the number gets converted to 32 bits signed
integer:
00000000000000000000000000000110
and the result of the operation looks like this:

1111111111111111111111111111111111111111111111111111111111111001
The result of ∼6 is -7. Why? When you use the NOT operator to flip all the bits, the number that is
returned is a signed number in two’s complement format.
What is two’s complement?
Two’s complement is a format for binary numbers that is a slightly modified version of signed
numbers. The most significant bit is still used to determine whether the number is positive or
negative, with 0 being positive and 1 being negative. To calculate the value of negative numbers
in two’s complement format, you invert all the bits and add 1. The advantage of using two’s
complement format for signed numbers is that there is no “double zero” problem as you can see
in the table below:
Decimal Signed Two’s complement

+7 0111 0111
+6 0110 0110
+5 0101 0101
+4 0100 0100
+3 0011 0011
+2 0010 0010
+1 0001 0001
+0 0000 0000
-0 1000 -
-1 1001 1111
-2 1010 1110
-3 1011 1101
-4 1100 1100
-5 1101 1011
-6 1110 1010
-7 1111 1001
In Java, the bitwise operators work with integers. JavaScript doesn’t have integers. It
only has double precision floating-point numbers. So, the bitwise operators convert their
number operands into integers, do their business, and then convert them back. In most
languages, these operators are very close to the hardware and very fast. In JavaScript,
they are very far from the hardware and very slow. JavaScript is rarely used for doing bit
manipulation. - Douglas Crockford, Javascript: The Good Parts
Although in practice JavaScript isn’t the most ideal programming language for doing bitwise
operations, it is still an important concept to learn. The following section explains all the various
types of bitwise operations and how they are implemented in JavaScript. For simplicity’s sake,
instead of using 64-bit binary numbers, we will use 6-bit numbers to explain the different bitwise
operations. All the bitwise operation use a zero-based index when shifting bits left and right.
Code
Get Bit
To get a relevant bit from a binary number, we first want to shift the relevant bit to the zeroth
position, then use the AND operation to determine if the bit is a 1 or a 0.
For example, the number 5 in binary looks like this:
000101
If we want to find out the value of the bit in the second position, we first shift the bit over 2 places
like this:
shift 1 –> 000010 shift 2 –> 000001

Then once we’ve shifted the bit the correct number of positions, we use the & (AND) operation to
determine if the bit is a 1 or a 0. Remember, when using the & operation, the result of the operation
will only return 1 if both numbers are 1. In this case, when we call our getBit function with the
number 5 and 2:
let result = getBit(5,2)
result is equal to 1.
09-bit-manipulation/getBit.js
1 /**
3 * @param {number} bitPosition - zero based.
5 */
6 export default function getBit(number, bitPosition) {
7 return (number >> bitPosition) & 1;
8 }
Set Bit
To set a bit in a binary number, we first must shift the number 1 to the correct position, then OR
that with the number.
To set the bit in the 3rd position for the number 5 we first take the number 1 and shift the digit in
the 0th position 3 times like this:
000001 <– initial binary representation of 1 000010 <– shift 1 000100 <– shift 2 001000 <– shift 3
Then we perform the | (OR) operation with the number 5 to set the number.
000101 | 001000 = 001101
09-bit-manipulation/setBit.js
1 /**
5 */
6 export default function setBit(number, bitPosition) {
7 return number | (1 << bitPosition);
8 }
Clear Bit
To clear a bit in a particular position, we first create a mask using the number 1, then NOT the mask
and AND it with the number.
For example, if we want to clear the bit in the second position for the number 5, we first take the
number 1 and shift the bits to the second position:
000001 <– initial binary representation of 1 000010 <– shift 1 000100 <– shift 2
Then, we use the ∼ (NOT) operation on this number which gives us this result:
111011
Finally, we use the & operation with the number to clear the bit in the correct position:
000101 & 111011 = 000001
09-bit-manipulation/clearBit.js
1 /**
5 */
6 export default function clearBit(number, bitPosition) {
7 const mask = ~(1 << bitPosition);
8
9 return number & mask;
10 }
Update Bit
This method is a combination of clearBit and setBit methods.

09-bit-manipulation/updateBit.js
1 /**
4 * @param {number} bitValue - 0 or 1.
6 */
7 export default function updateBit(number, bitPosition, bitValue) {
8 // Normalized bit value.
9 const bitValueNormalized = bitValue ? 1 : 0;
10
11 // Init clear mask.

12 const clearMask = ~(1 << bitPosition);
13
14 // Clear bit value and then set it up to required value.
15 return (number & clearMask) | (bitValueNormalized << bitPosition);
16 }
isEven
This method determines if the number provided is even. For odd numbers, the right most digit (or
least significant bit) is 1 and for even numbers the least significant bit is 0.
Number: 5 = 000101
isEven: false
Number: 4 = 000100
isEven: true
By performing the bitwise AND operation with 1 we can determine if a number is even or odd.
000101 & 000001 === 000001 (odd) 000100 & 000001 === 000000 (even)
09-bit-manipulation/isEven.js
1 /**
4 */
5 export default function isEven(number) {
6 return (number & 1) === 0;
7 }
isPositive
This method determines if the number is positive. A number is positive if the most significant bit
(the leftmost bit) is 0 and the function returns true. If the most significant bit is 1, the number is
negative and the function returns false.
Number: 1 = 000001
isPositive: true
Number: -1 = 100001
isPositive: false
To get the most significant bit, we shift the leftmost bit 31 times and then AND the number with 1
to determine if it’s a 1 or a 0.
shift 1 –> 100000000000000000000000000000001 shift 2 –> 010000000000000000000000000000000 ….
shift 31 –> 00000000000000000000000000000001
00000000000000000000000000000001 & 000001 = 000001 <– Negative number
09-bit-manipulation/isPositive.js
1 /**
2 * @param {number} number - 32-bit integer.
4 */
5 export default function isPositive(number) {
6 // Zero is neither a positive nor a negative number.
7 if (number === 0) {
8 return false;
9 }
10
11 // The most significant 32nd bit can be used to determine whether the number is po\
12 sitive.
13 return ((number >> 31) & 1) === 0;
14 }
Multiply By Two
This method shifts the original number by one bit to the left. By shifting all bits to the left, this
multiplies the number by two.
000101 <– Initial binary represenation of 5 001010 <– shift 1
001010 = 10
09-bit-manipulation/multiplyByTwo.js
1 /**
4 */
5 export default function multiplyByTwo(number) {
6 return number << 1;
7 }
Divide By Two
To divide a binary number by two, we shift the original number by one bit to the right. By shifting
all bits to the right, this divides the number by 2.
Initial binary represenation of 5 –> 000101 shift 1 –> 000010
000010 = 2
09-bit-manipulation/divideByTwo.js
1 /**
4 */
5 export default function divideByTwo(number) {
6 return number >> 1;
7 }
Switch Sign
This method switches positive numbers to negative and negative numbers to positive. To do so it
uses the “Twos Complement” approach by inverting all of the bits of the number and adding 1 to it.
1 1101 -3
2 1110 -2
3 1111 -1
4 0000 0
5 0001 1
6 0010 2
7 0011 3
Switching 5 to -5:
000101 = 5
Invert bits:
111010
Add 1:
111011
And then switching -5 back to 5:

111011 = -5
Invert bits:
000100
Add 1:
000101
09-bit-manipulation/switchSign.js
1 /**
2 * Switch the sign of the number using "Twos Complement" approach.
5 */
6 export default function switchSign(number) {
7 return ~number + 1;
8 }
Multiply Two Signed Numbers
This method multiplies two signed integer numbers using bitwise operators. This method is based
on the following facts:
a * b can be written in the below formats:

0 if a is zero or b is zero or both a and b are zeroes
2a * (b/2) if b is even
2a * (b - 1)/2 + a if b is odd and positive
2a * (b + 1)/2 - a if b is odd and negative
The advantage of this approach is that in each recursive step one of the operands reduces to half its
original value. Hence, the run time complexity is O(log(b) where b is the operand that reduces to
half on each recursive step.
09-bit-manipulation/multiply.js
1 import multiplyByTwo from './multiplyByTwo';
2 import divideByTwo from './divideByTwo';
3 import isEven from './isEven';
4 import isPositive from './isPositive';
5
6 /**
7 * Multiply two signed numbers using bitwise operations.
8 *
9 * If a is zero or b is zero or if both a and b are zeros:
10 * multiply(a, b) = 0
11 *
12 * If b is even:
13 * multiply(a, b) = multiply(2a, b/2)
14 *
15 * If b is odd and b is positive:
16 * multiply(a, b) = multiply(2a, (b-1)/2) + a
17 *
18 * If b is odd and b is negative:
19 * multiply(a, b) = multiply(2a, (b+1)/2) - a
20 *
21 * Time complexity: O(log b)
22 *
23 * @param {number} a
24 * @param {number} b
26 */
27 export default function multiply(a, b) {
28 // If a is zero or b is zero or if both a and b are zeros then the production is a\
29 lso zero.
30 if (b === 0 || a === 0) {
31 return 0;
32 }
33
34 // Otherwise we will have four different cases that are described above.
35 const multiplyByOddPositive = () => multiply(multiplyByTwo(a), divideByTwo(b - 1))\
36 + a;
37 const multiplyByOddNegative = () => multiply(multiplyByTwo(a), divideByTwo(b + 1))\
38 - a;
39
40 const multiplyByEven = () => multiply(multiplyByTwo(a), divideByTwo(b));
41 const multiplyByOdd = () => (isPositive(b) ? multiplyByOddPositive() : multiplyByO\
42 ddNegative());
43
44 return isEven(b) ? multiplyByEven() : multiplyByOdd();
45 }
Multiply Two Unsigned Numbers
This method multiplies two integer numbers using bitwise operators. This method is based on that
“Every number can be denoted as the sum of powers of 2”.
The main idea of bitwise multiplication is that every number may be split to the sum of powers of
two:
19 = 2^4 + 2^1 + 2^0
Then multiplying number x by 19 is equivalent of:
x * 19 = x * 2^4 + x * 2^1 + x * 2^0
Now we need to remember that x * 2^4 is equivalent of shifting x left by 4 bits (x << 4).
09-bit-manipulation/multiplyUnsigned.js
1 /**
2 * Multiply to unsigned numbers using bitwise operator.
3 *
4 * The main idea of bitwise multiplication is that every number may be split
5 * to the sum of powers of two:
6 *
7 * I.e. 19 = 2^4 + 2^1 + 2^0
8 *
9 * Then multiplying number x by 19 is equivalent of:
10 *
11 * x * 19 = x * 2^4 + x * 2^1 + x * 2^0
12 *
13 * Now we need to remember that (x * 2^4) is equivalent of shifting x left by 4 bits\
14 (x << 4).
15 *
16 * @param {number} number1
17 * @param {number} number2
19 */
20 export default function multiplyUnsigned(number1, number2) {
21 let result = 0;
22
23 // Let's treat number2 as a multiplier for the number1.
24 let multiplier = number2;
25
26 // Multiplier current bit index.
27 let bitIndex = 0;
28
29 // Go through all bits of number2.
30 while (multiplier !== 0) {
31 // Check if current multiplier bit is set.
32 if (multiplier & 1) {
33 // In case if multiplier's bit at position bitIndex is set
34 // it would mean that we need to multiply number1 by the power
35 // of bit with index bitIndex and then add it to the result.
36 result += (number1 << bitIndex);
37 }
38
39 bitIndex += 1;
40 multiplier >>= 1;
41 }
42
43 return result;
44 }
Count Set Bits
This method counts the number of set bits (bits that are 1) in a number using bitwise operators.
While the number is still greater than 0, we shift the number right one bit at a time and AND the
number with 1 to determine if the bit is set or not.
Initial binary representation of 5 –> 000101 Set bit count: 1 shift 1 –> 000010 Set bit count:1 shift 2
–> 000001 Set bit count:2 shift 3 –> 000000 Set bit count:2
Number is now 0, so terminate while loop. Set bit count is 2.
09-bit-manipulation/countSetBits.js
1 /**
2 * @param {number} originalNumber
4 */
5 export default function countSetBits(originalNumber) {
6 let setBitsCount = 0;
7 let number = originalNumber;
8
9 while (number) {
10 // Add last bit of the number to the sum of set bits.
11 setBitsCount += number & 1;
12
13 // Shift number right by one bit to investigate other bits.
14 number >>= 1;
15 }
16
17 return setBitsCount;
18 }
Count the Bit Difference
This methods determines the number of bits that are different between two numbers. By using the
XOR operation, we can determine the number of bits that are different between the two numbers,
and then use the countBit function to get a total number.
1 000101 ^ 000001 = 000100

2 Count of Bits to be Flipped: 1
09-bit-manipulation/bitsDiff.js
1 import countSetBits from './countSetBits';

2
3 /**
4 * Counts the number of bits that need to be change in order
5 * to convert numberA to numberB.
6 *
7 * @param {number} numberA
8 * @param {number} numberB
10 */
11 export default function bitsDiff(numberA, numberB) {

12 return countSetBits(numberA ^ numberB);
13 }
Count the Number of Significant Bits of a Number
To calculate the number of valuable bits we need to shift 1 one bit left each time and see if the shifted
number is bigger than the input number.
5 = 000101 Initial binary represenation of 1 –> 000001 Bit count: 0 000010 <– shift 1, 2 < 5 000100 <–
shift 2, 4 < 5 001000 <– shift 3, 8 > 5
Count of valuable bits is: 3 When we shift 1 four times it is bigger than 5.
09-bit-manipulation/bitLength.js
1 /**
2 * Return the number of bits used in the binary representation of the number.
3 *
6 */
7 export default function bitLength(number) {
8 let bitsCounter = 0;
9
10 while ((1 << bitsCounter) <= number) {
11 bitsCounter += 1;
12 }
13
14 return bitsCounter;
15 }
Problems Examples
• Reverse Bits⁴⁷
• Bitwise AND of Numbers Range⁴⁸
⁴⁷https://leetcode.com/problems/reverse-bits/
⁴⁸https://leetcode.com/problems/bitwise-and-of-numbers-range/
Quiz
Q1: What does -5 look like in two’s complement binary format?
Q2: If we clear the bit in position 0 for the binary representation for the number 5, what is the result?
References
Bit Manipulation on YouTube https://www.youtube.com/watch?v=NLKQEOgBAnw⁴⁹
Negative Numbers in binary on YouTube https://www.youtube.com/watch?v=4qH4unVtJkE⁵⁰
Bit Hacks on stanford.edu https://graphics.stanford.edu/∼seander/bithacks.html⁵¹
Bit manipulation examples and test cases https://github.com/trekhleb/javascript-algorithms/tree/-
master/src/algorithms/math/bits⁵²
⁴⁹https://www.youtube.com/watch?v=NLKQEOgBAnw
⁵⁰https://www.youtube.com/watch?v=4qH4unVtJkE
⁵¹https://graphics.stanford.edu/~seander/bithacks.html
⁵²https://github.com/trekhleb/javascript-algorithms/tree/master/src/algorithms/math/bits
Factorial
Intro
In mathematics, the factorial of a non-negative integer n, denoted by n!, is the product of all positive
integers less than or equal to n. For example:
1 5! = 5 * 4 * 3 * 2 * 1 = 120
n n!
0 1
1 1
2 2
3 6
4 24
5 120
6 720
7 5 040
8 40 320
9 362 880
10 3 628 800
11 39 916 800
12 479 001 600
13 6 227 020 800
14 87 178 291 200
15 1 307 674 368 000
Applications
The factorial function is useful for calculating distinct sets of objects, for example say we have
three people in a group – Sally, Susan and Sarah – and we want to determine the number of unique
ways these people can stand in line.
You could write it out:
Sally, Susan, Sarah Sally, Sarah, Susan Susan, Sarah, Sally Susan, Sally, Sarah Sarah, Sally, Susan
Sarah, Susan, Sally
Factorial 149
Writing this list down takes a relatively short amount of time, but what if we had 10 or 20 people?
We’d be writing all possible combinations for a long time. By calculating the factorial of the number
of people we have in a group, we aren’t required to write each combination down.
Recursion
Solving the factorial of a number using programming is a problem commonly used to demonstrate
the concept of recursion. In the code below you’ll see we have two different approaches to solving
this problem – by using recursion and a non-recursive approach.
In many of the chapters in this book, we will be using recursion to solve our problems. Recursion
is frequently used in computer science and is an important (albeit somewhat confusing) concept to
understand. The basic idea is we have a function and within the function definition, we call the
function again. We must also include a base case within our function, which forces our function to
stop calling itself. Without a base case, our function would run indefinitely, until our system runs
out of memory.
Code
Recursive implementation
The recursive implementation to calculate the factorial of a number is only one line in our function,
but there is a lot to unpack in that function, so let’s walk through it step by step.
To implement the factorial, we use a ternary operator to determine whether or not to continue calling
our function.
10-factorial/factorialRecursive.js
1 /**
4 */
5 export default function factorialRecursive(number) {
6 return number > 1 ? number * factorialRecursive(number - 1) : 1;
7 }
A reminder on ternary operators
A ternary operator in JavaScript is a way to simplify the number of lines we have to write.
Instead of an if/else statement:
Factorial 150
if(someTrueValue){
return someValue;
}else{
return aDifferentValue;
}
We can consolidate this logic into one line like this:
return someTrueValue ? someValue : aDifferentValue;
The first part of the ternary operator is a true/false expression, the second part (between
the ? and :) is the value that returns if the expression is true and the final part is what is
returned when the expression is false.
For our factorial function, the first part of our ternary operator (the true/false expression) determines
whether the number we are using is greater than 1.
number > 1
If the number we are calculating is greater than 1, then the second part of our ternary operator gets
executed. This is where our function calls itself with the number that is one less than the current
number. We multiply the result of that function call with the current number.
? number * factorialRecursive(number - 1)
Finally, we get to the last part of our ternary operator, which is our base case. When we get to 1,
we return 1.
If we wanted to call our function with the number 5, what would that look like “under the hood”?
return factorialRecursive(5);
Is 5 greater than 1? Yes return 5 * factorialRecursive(4)

Is 1 greater than 1? No return 1 <– This is our base case. We’re done calling our function and can
return the values from each previous call to our function:
return 2 * 1 return 3 * 2 (result from previous call 2 * 1) return 4 * 6 (result from previous call 3 * 2)
return 5 * 24 (result from previous call 4 * 6)
Final result: 120
Factorial 151
Non-recursive implementation
For the non-recursive implementation, we first create a variable to store our result and initialize it
to 1:
10-factorial/factorial.js
5 export default function factorial(number) {
6 let result = 1;
Next, we create a for loop that starts at 2, and continues looping until it reaches the number we are
calculating.
7 for (let i = 2; i <= number; i += 1) {
Inside the for loop we multiply i by our result:

8 for (let i = 2; i <= number; i += 1) {
9 result *= i;
10 }
Finally, when we break out of the for loop we return the result.
11 return result;
Problems Examples
• Clumsy Factorial⁵³
• Factorial Trailing Zeroes⁵⁴
Quiz
Q1: In the recursive implementation of 6!, how many times will the factorialRecursive function be
called?
⁵³https://leetcode.com/problems/clumsy-factorial/
⁵⁴https://leetcode.com/problems/factorial-trailing-zeroes/
Factorial 152
References
Factorial on Wikipedia https://en.wikipedia.org/wiki/Factorial⁵⁵
Factorial implementation examples and test cases https://github.com/trekhleb/javascript-algorithm-
s/tree/master/src/algorithms/math/factorial⁵⁶
⁵⁵https://en.wikipedia.org/wiki/Factorial
⁵⁶https://github.com/trekhleb/javascript-algorithms/tree/master/src/algorithms/math/factorial
Fibonacci Number
In mathematics, the Fibonacci numbers are the numbers in the following integer sequence, called
the Fibonacci sequence, and characterized by the fact that every number after the first two is the
sum of the two previous ones:
0, 1, 1, 2, 3, 5, 8, 13, 21, 34, 55, 89, 144, ...
A tiling with squares whose side lengths are successive Fibonacci numbers
Fibonacci
The Fibonacci spiral: an approximation of the golden spiral created by drawing circular arcs
connecting the opposite corners of squares in the Fibonacci tiling;[4] this one uses squares of sizes
1, 1, 2, 3, 5, 8, 13 and 21.
Fibonacci Number 154
Fibonacci Spiral
Applications
The Fibonacci sequence is found all throughout nature–flowers, pinecones, even insects and humans
all have elements of the Fibonacci sequence. For example, a pinecone’s seed pods are arranaged in
a Fibonacci sequence and similarly the seeds in a sunflower are arranged in a Fibonacci sequence.
Code
Creating an array of Fibonacci numbers

To create an array of N Fibonacci numbers, we must first create an array with one value in it, the
number 1.
11-fibonacci/fibonacci.js
7 export default function fibonacci(n) {

8 const fibSequence = [1];
Then we create two variables, one to hold the current number we are working with and one to hold
the previous number.
10 let currentValue = 1;
11 let previousValue = 0;
If we only want the first number in the Fibonacci sequence (1) then we return the array immediately:
13 if (n === 1) {
14 return fibSequence;
15 }
If we want more than the first number in the Fibonacci sequence, we create an iterationsCounter
variable to keep track of the numbers we’ve added to our array. Since we already included 1 in our
array, we set the iterationsCounter to n-1:
16 let iterationsCounter = n - 1;
Then we create a while loop which continues until iterationsCounter reaches 0. We add up the
current value and the previously calculated value and put that into the array.
18 while (iterationsCounter) {
19 currentValue += previousValue;
20 previousValue = currentValue - previousValue;
21
22 fibSequence.push(currentValue);
23
24 iterationsCounter -= 1;
25 }
Finally, when we’ve broken out of the while loop, we return our array.
28 return fibSequence;
29 }
Calculating the Nth Fibonacci number

Calculating the Nth Fibonacci number is very similar to creating an array of Fibonacci numbers.
Instead of placing the numbers in an array, when we reach Nth number we just return it.
11-fibonacci/fibonacciNth.js
7 export default function fibonacciNth(n) {

8 let currentValue = 1;
9 let previousValue = 0;
10
11 if (n === 1) {
12 return 1;
13 }
14
15 let iterationsCounter = n - 1;
16
17 while (iterationsCounter) {
18 currentValue += previousValue;
19 previousValue = currentValue - previousValue;
20
21 iterationsCounter -= 1;
22 }
Calculating the Nth Fibonacci number (using Binet’s formula)

Binet’s formula is a mathematical way to calculate the Nth Fibonacci number without having to use
a for or while loop.
First, we calculate the square root of 5 and save that into a variable:
11-fibonacci/fibonacciNthClosedForm.js
16 // Calculate √5 to re-use it in further formulas.

17 const sqrt5 = Math.sqrt(5);
Next we calculate the value of Ï† (â‰ˆ 1.61803) and save that into a variable
18 // Calculate φ constant (≈ 1.61803).

19 const phi = (1 + sqrt5) / 2;
Finally, we use these two values to calculate the Nth Fibonacci number. You’ll notice that we are
using the exponentiation operator (**) introduced in EcmaScript 2017 instead of Math.pow().
21 // Calculate fibonacci number using Binet's formula.

22 return Math.floor((phi ** position) / sqrt5 + 0.5);
Problems Examples
• Length of Longest Fibonacci Sequence⁵⁷

• Split Array into Fibonacci Sequence⁵⁸
References
Fibonacci sequence on Wikipedia https://en.wikipedia.org/wiki/Fibonacci_number⁵⁹
Fibonacci sequence on YouTube https://www.youtube.com/watch?v=T8xgfVzef_E⁶⁰
Fibonacci sequence implementation examples and test cases https://github.com/trekhleb/javascript-
algorithms/tree/master/src/algorithms/math/fibonacci⁶¹
⁵⁷https://leetcode.com/problems/length-of-longest-fibonacci-subsequence/
⁵⁸https://leetcode.com/problems/split-array-into-fibonacci-sequence/
⁵⁹https://en.wikipedia.org/wiki/Fibonacci_number
⁶⁰https://www.youtube.com/watch?v=T8xgfVzef_E
⁶¹https://github.com/trekhleb/javascript-algorithms/tree/master/src/algorithms/math/fibonacci
Primality Test
A prime number (or a prime) is a natural number greater than 1 that cannot be formed by
multiplying two smaller natural numbers.
A natural number is any number greater than or equal to 0 that is not a fraction or a
decimal
A natural number greater than 1 that is not prime is called a composite number. For example, 5 is
prime because the only ways of writing it as a product, 1 Ã— 5 or 5 Ã— 1, involve 5 itself. However,
6 is a composite number because it is the product of two numbers (2 Ã— 3) that are both smaller
than 6.
Primality Test 159
A primality test is an algorithm for determining whether an input number is prime.

Primality Test 160
Applications
Among other fields of mathematics, primality tests are used in cryptography.
Code
To determine if a number is a prime or not, we’ll be using the modulo operator, also known as the
remainder operator. This operator calculates the remainder of two numbers when they are divided.
First, we start out by ensuring the number is not a fraction or decimal by using the modulo operator.
If the remainder of the number divided by 1 is not 0, we know we have a fraction or decimal and
it’s not a prime.
12-primality-test/trialDivision.js
5 export default function trialDivision(number) {

6 // Prime numbers can only be integers
7 if (number % 1 !== 0) {
8 return false;
Next, we check if the number is less than 1, if it is, it’s not a prime number
11 if (number <= 1) {
12 // Any number less than or equal to 1 is not a prime
13 return false;
14 }
If the number is less than or equal to 3, then the number is a prime number (2 & 3 are prime numbers)
16 if (number <= 3) {
17 // 2 and 3 are prime numbers
18 return true;
19 }
If the number is divisible by 2, then it is an even number and there are no even prime numbers (they
all can be divided by 2)
Primality Test 161
21 // If the number is divisble by 2, then it is not a prime number
22 if (number % 2 === 0) {
23 return false;
24 }
Finally, if we’ve made it this far in the code, we can actually do some calculations:
We calculate the square root of the number we are testing and determine if any numbers up to
the square root of the number are divisible. If any of the numbers divided by the number we are
checking result in a remainder of 0, we know it is not a prime number.
26 // If there is no dividers up to square root of n then there is no higher dividers\
27 as well.
28 const dividerLimit = Math.sqrt(number);
29 for (let divider = 3; divider <= dividerLimit; divider += 2) {
30 if (number % divider === 0) {
31 return false;
32 }
33 }
34
35 return true;
Why do we use the square root of the number?

We use the square root of the number because any number greater than the square root of a number
will have a corresponding number smaller than the square root that will be used to calculate the
number.
For example, the square root of 36 is 6.
1 * 36 = 36 2 * 18 = 36 3 * 12 = 36 4 * 9 = 36 5 * 7.2 = 36 6 * 6 = 36 7 * 5.15 = 36 <– The second operand
is less than the square root of 36; we don’t need to check numbers higher than the square root.
Problems Examples
• Prime Palindrome⁶²
• Count Primes⁶³
⁶²https://leetcode.com/problems/prime-palindrome/
⁶³https://leetcode.com/problems/count-primes/
Primality Test 162
Quiz
Q1: Is 32 a prime number? Follow the logic from the code above and find out
Q2: Is 31 a prime number? Follow the logic from the code above and find out
References
Prime numbers on Wikipedia https://en.wikipedia.org/wiki/Prime_number⁶⁴
Primality test on Wikipedia https://en.wikipedia.org/wiki/Primality_test⁶⁵
Primality test example and test cases https://github.com/trekhleb/javascript-algorithms/tree/mas-
ter/src/algorithms/math/primality-test⁶⁶
⁶⁴https://en.wikipedia.org/wiki/Prime_number
⁶⁵https://en.wikipedia.org/wiki/Primality_test
⁶⁶https://github.com/trekhleb/javascript-algorithms/tree/master/src/algorithms/math/primality-test
Is a power of two
The Task
Given a positive integer, write a function to find if it is a power of two or not.
Naive solution
In the naive solution, we first check that the number is greater than 1, if it is less than 1, it is not a
power of 2.
14-is-power-of-two/isPowerOfTwo.js
5 export default function isPowerOfTwo(number) {

6 // 1 (2^0) is the smallest power of two.
7 if (number < 1) {
8 return false;
9 }
Then, we divde the number by 2 and check that the remainder is 0. If the remainder is not 0, then
the number is not a power of 2. If we continue dividing and reach 1, then we know the number is a
power of 2.
14-is-power-of-two/isPowerOfTwo.js
11 // Let's find out if we can divide the number by two

12 // many times without remainder.
13 let dividedNumber = number;
14 while (dividedNumber !== 1) {
15 if (dividedNumber % 2 !== 0) {
16 // For every case when remainder isn't zero we can say that this number
17 // couldn't be a result of power of two.
18 return false;
19 }
20
21 dividedNumber /= 2;
Is a power of two 164
22 }
23
24 return true;
25 }
Bitwise solution
Powers of two in binary form always have just one bit. The only exception is with a signed integer
(e.g. an 8-bit signed integer with a value of -128 looks like: 10000000)
1 1: 0001
2 2: 0010
3 4: 0100
4 8: 1000
To check if a number is a power of two, we first subtract 1 from the number.

For example, the number 8 looks like this in binary:
1 1000
8 - 1 in binary looks like this:
1 0111
Any power of two minus 1 is all 1s except for the most significant bit. After subtracting 1 from the
number, we AND it with the number. If the result of the AND operation is 0, we know it is a power
of 2. If the result of the AND operation is 1, then it is not a power of two.
1 1000
2 & 0111
3 ----
4 0000
14-is-power-of-two/isPowerOfTwoBitwise.js
5 export default function isPowerOfTwoBitwise(number) {

6 // 1 (2^0) is the smallest power of two.
7 if (number < 1) {
8 return false;
9 }
10
11 /*
12 * Powers of two in binary look like this:
13 * 1: 0001
14 * 2: 0010
15 * 4: 0100
16 * 8: 1000
17 *
18 * Note that there is always exactly 1 bit set. The only exception is with a signe\
19 d integer.
20 * e.g. An 8-bit signed integer with a value of -128 looks like:
21 * 10000000
22 *
23 * So after checking that the number is greater than zero, we can use a clever lit\
24 tle bit
25 * hack to test that one and only one bit is set.
26 */
27 return (number & (number - 1)) === 0;
Problems Examples
• Power of Four⁶⁷
• Reordered Power of Two⁶⁸
Quiz
Q1: Is 32 a power of 2? Use the naive solution to find out
Q2: Is 31 a power of 2? Use the bitwise solution to find out
⁶⁷https://leetcode.com/problems/power-of-four/
⁶⁸https://leetcode.com/problems/reordered-power-of-2/
References
Bitwise solution on GeeksForGeeks https://www.geeksforgeeks.org/program-to-find-whether-a-no-
is-power-of-two/⁶⁹
Bitwise solution on Stanford http://www.graphics.stanford.edu/∼seander/bithacks.html#DetermineIfPowerOf2⁷⁰
Binary number subtraction on YouTube https://www.youtube.com/watch?v=S9LJknZTyos⁷¹
Bitwise solution example and test cases https://github.com/trekhleb/javascript-algorithms/tree/mas-
ter/src/algorithms/math/is-power-of-two⁷²
⁶⁹https://www.geeksforgeeks.org/program-to-find-whether-a-no-is-power-of-two/
⁷⁰http://www.graphics.stanford.edu/~seander/bithacks.html#DetermineIfPowerOf2
⁷¹https://www.youtube.com/watch?v=S9LJknZTyos
⁷²https://github.com/trekhleb/javascript-algorithms/tree/master/src/algorithms/math/is-power-of-two
Search. Linear Search.
The Task
Find the position of a specific element in a list (or array).
The list is not sorted, it may contain strings, numbers or objects. The distribution of the element in
a list is unknown, we can’t say if it sorted or not.
The Algorithm
A linear search algorithm is a method for finding an element within a list. It sequentially checks
each element of the list until a match is found or the whole list has been searched.
Search. Linear Search. 168
Application
If we need to find one value in an unsorted array, the linear search algorithm is a good option.
If many values need to be found in the same array, then pre-processing the array – like using a
sorting algorithm first – and using another search algorithm, like a binary search algorithm is a
better approach.
Usage Example
Let’s say we want to implement a linear search for the phone book array that contains a list of users
(objects with name and phone).
To make our linear search a little bit more useful we will make it work not only with numbers and
strings but also with objects. This option will allow us to be more flexible when searching the array.
15-linear-search/example.js
2 import linearSearch from './linearSearch';
3
4 // Create people objects.
5 const tim = { name: 'Tim', phone: '+111111111112' };
6 const jane = { name: 'Jane', phone: '+111111111111' };
7 const bill = { name: 'Bill', phone: '+111111111113' };
8 const janeNamesake = { name: 'Jane', phone: '+111111111114' };
9 const peter = { name: 'Peter', phone: '+111111111115' };
10
11 // Add people to the phone book.
12 const phoneBook = [tim, jane, bill, janeNamesake];
13
14 // Create custom comparator function that can compare two person
15 // and decide whether they are equal or not.
16 const personComparator = (person1, person2) => {
17 return person1.name === person2.name ? 0 : -1;
18 };
19
20 // Let's try to find Jane in the phone book using linear search.
21 // Notice that we have two persons with the same name Jane in the phone book and
22 // our comparision function takes only the person name into account so we are
23 // expecting to have two indices in the output.
24
26 linearSearch(phoneBook, jane, personComparator); // -> [1, 3]
27
28 // Now let's try to find Peter. He is not in the phone book so we expect an empty ar\
29 ray.
30
32 linearSearch(phoneBook, peter, personComparator); // -> []
Implementation
In the example above we use a custom comparison function. To allow us to use this comparison
function, we need to split our linear search algorithm into two parts. First, we will implement
a universal Comparator utility class that will provide a common interface for object comparison.
And second, we will move on and implement the linearSearch() function itself that will use the
Comparator utility class internally.
Comparator

1 export default class Comparator {
constructor
The Comparator constructor will accept the compare function callback as an argument. This function
accepts two values for comparison. It will return 0 if two objects are equal, -1 if the first object is
smaller than the second one and 1 if the first object is greater than the first one.
The Comparator class itself will not know anything about the comparison logic. It will just provide
a common interface (the methods that we’re going to implement below) to work with.
compareFunction is an optional argument for the Comparator constructor. If the compareFunction is
not provided we fall back to defaultCompareFunction() function.
16 /**
17 * @param {function(a: *, b: *)} [compareFunction] - It may be custom compare func\
18 tion that, let's
19 * say may compare custom objects together.
20 */
21 constructor(compareFunction) {
22 this.compare = compareFunction || Comparator.defaultCompareFunction;
23 }
defaultCompareFunction()
This is default fallback comparison function that treats two arguments a and b as a string or a number
and applies common comparison operators to them like ===, < and >.
2 /**
3 * Default comparison function. It just assumes that "a" and "b" are strings or nu\
4 mbers.
5 * @param {(string|number)} a
6 * @param {(string|number)} b
7 * @returns {number}
8 */
9 static defaultCompareFunction(a, b) {
10 if (a === b) {
11 return 0;
12 }
13
14 return a < b ? -1 : 1;
15 }
equal()
This function checks if two variables are equal.

24 /**
25 * Checks if two variables are equal.
26 * @param {*} a
27 * @param {*} b
29 */
30 equal(a, b) {
31 return this.compare(a, b) === 0;
32 }
lessThan()
This function checks if variable a is less than b.

34 /**
35 * Checks if variable "a" is less than "b".
36 * @param {*} a
37 * @param {*} b
39 */
40 lessThan(a, b) {
41 return this.compare(a, b) < 0;
42 }
greaterThan()
This function checks if variable a is greater than b.

44 /**
45 * Checks if variable "a" is greater than "b".
46 * @param {*} a
47 * @param {*} b
49 */
50 greaterThan(a, b) {
51 return this.compare(a, b) > 0;
52 }
lessThanOrEqual()
This function checks if variable a is less than or equal to b.

54 /**
55 * Checks if variable "a" is less than or equal to "b".
56 * @param {*} a
57 * @param {*} b
59 */
60 lessThanOrEqual(a, b) {
61 return this.lessThan(a, b) || this.equal(a, b);
62 }
greaterThanOrEqual()
This function checks if variable a is greater than or equal to b.

64 /**
65 * Checks if variable "a" is greater than or equal to "b".
66 * @param {*} a
67 * @param {*} b
69 */
70 greaterThanOrEqual(a, b) {
71 return this.greaterThan(a, b) || this.equal(a, b);
72 }
linearSearch()
Now that we’ve implemented our Comparator utility class let’s implement the linearSearch()
function.
The linearSearch() function compares each element of the array to the element we are trying to
find. The comparison is done using the Comparator class. Once the element is found, its position is
recorded in the foundIndices array.
15-linear-search/linearSearch.js
11 export default function linearSearch(array, seekElement, comparatorCallback) {

12 const comparator = new Comparator(comparatorCallback);
13 const foundIndices = [];
14
15 array.forEach((element, index) => {
16 if (comparator.equal(element, seekElement)) {
17 foundIndices.push(index);
18 }
19 });
20
21 return foundIndices;
22 }
Complexity
Space Worst Case Best Case Average Performance

Performance Performance
O(1) O(n) O(1) O(n)
The algorithm iteratively goes through the list. It doesn’t use additional data structures and thus the
space complexity is O(1).
The worst-case scenario for performance is when the element we are looking for does not exist in
the array. In this case, the algorithm needs to traverse all list elements in this case. As a result, we
have O(n) time complexity for the worst-case scenario.
In the best-case scenario, the element we’re looking for is placed at the very beginning of the list.
In this case, time complexity would be O(1). In our implementation, the complexity would still be
O(n) because we made an assumption that the array might have duplicates and we want to find the
position of all the duplicates if there are any.
In the average scenario if we know that the probability of finding every element in the array is
equal, then the time complexity would be O(n/2). If we don’t know that the probability of finding
every element in the array is equal, then the average case is O(n) time complexity.
Problems Examples
• Peak Index in a Mountain Array⁷³

• Find First and Last Position of Element in Sorted Array⁷⁴
Quiz
Q1: What time complexity of the linear search algorithm?
Q2: Is it true that we should apply linear search algorithm for huge list?
References
Linear search on Wikipedia https://en.wikipedia.org/wiki/Linear_search⁷⁵
Linear search example and test cases https://github.com/trekhleb/javascript-algorithms/tree/mas-
ter/src/algorithms/search/linear-search⁷⁶
⁷³https://leetcode.com/problems/peak-index-in-a-mountain-array/
⁷⁴https://leetcode.com/problems/find-first-and-last-position-of-element-in-sorted-array/
⁷⁵https://en.wikipedia.org/wiki/Linear_search
⁷⁶https://github.com/trekhleb/javascript-algorithms/tree/master/src/algorithms/search/linear-search
Search. Binary Search.
The Task
Find the position of a specific element in a sorted array.
The Algorithm
Binary search is an efficient algorithm for finding the item in a sorted array. It works by doing the
following steps:
1. Get the middle element of an array (it splits array by two halves).
2. Check if the middle element is equal to the element we’re searching for.
3. If the middle element is equal to the seek element then we’ve found it. Just return the current
element index.
4. If the middle element is bigger than the seek element then abandon the right half of the array.
Otherwise, abandon the left half of the array. We can do so because we know that the array is
sorted.
5. Treat half of the array that is left as a new array and go to step 1 again.
The algorithm ends up by either finding the position of the seek element or by returning null or -1
if the right or left half consists of just one element that is not equal to the seek element.
Take a look at the illustration below that shows the flow of finding number 8 in a sorted array.
Search. Binary Search. 176
Algorithm Complexities
Time Complexity
In the example above can see that after every comparison iteration we abandon the half of the
array. For example, after the first comparison (is 12 bigger than 8?) we make the decision to go left
(to abandon the elements with indices starting from 4 to 8). We do so because the middle element 12
is larger than 8. It’s not possible to find 8 in the right half of the array since the array is sorted. We
repeat this process over and over again until we find 8. This is what makes binary search efficient.
Its time complexity is O(log(n)), where n is the number of elements in the array.
To illustrate the fact that time complexity of this algorithm is O(log(n)), let’s think about the worst
case when the element we are searching for doesn’t exist in the array. In this case, how many time
do we split the array by two halves until the next half has just one element? In our example above
we have an array of 9 elements:
Math.floor(9 / 2); // -> 4

Math.floor(4 / 2); // -> 2
Math.floor(2 / 2); // -> 1
The number of possible divisions is proportional to the power of two that will give us the total
number of elements in the array (or at least close to the total number of elements). We can write
this statement using this formula:
2x = n x = log(n)
And x here is our time complexity.
To grasp the idea of how efficient O(log(n)) time is for binary search let’s compare it to
the O(n) time complexity of linear search. The Tycho-2 star catalog contains information
about the brightest 2,539,913 stars in our galaxy. Let’s imagine that we want to find a star
by its name in that catalog. In the worst case scenario for linear search we would need
to do 2,539,913 iterations. In the worst case scenario for binary search and if all stars
are sorted by their names it would take only 22 iterations (log2(2539913) =
21.27).
Space Complexity
The algorithm may be implemented in different ways using iterative approach or recursive approach
(allocating additional memory for each recursive call). Depending on the implementation that space
complexity may vary. In this chapter we will use simple iterative approach without creating array
copies. And thus the space complexity, in this case, would be O(1).
Application
• Binary search is used for doing fast search operation in sorted arrays.
• Binary search approach is used for Binary Search Tree implementation.
• Binary search approach may be used to quickly find a bug commit in a commit history. Just
split commit history by halves. If a bug already exists in the “middle” commit then move left to
the earlier commits and split the left half. Otherwise, if the bug is not present then move right
to the later commits.
• Binary search approach may be used to quickly find a buggy line of code in the program.
Just comment the half of the code and see if the bug is still presented. If yes then move up,
otherwise move down.
Usage Example
Earlier in this chapter, we’ve discussed how fast it would be to search for a star by its name in
Tycho-2 catalog in case if it was sorted by name. Instead of linear 2,539,913 iterations, it would give
us only 22 iterations to find the star.
Let’s imagine that we already have this sorted catalog of stars in our script and let’s see how we
would want to apply binary search to it.
15-binary-search/example.js
2 import binarySearch from './binarySearch';
3
4 // Let's create a really simple and small demo version of our sorted stars catalog.
5 const sortedArrayOfStars = [
6 { name: 'Alpha Centauri A', position: {} },
7 { name: 'Alpha Centauri B', position: {} },
8 { name: 'Betelgeuse', position: {} },
9 { name: 'Polaris', position: {} },
10 { name: 'Rigel', position: {} },
11 { name: 'Sirius', position: {} },
12 // And 2.5 millions more records here :)
13 ];
14
15 // Our custom object comparator for binary search function.
16 // We will use string comparison here. For example 'Polaris' is smaller than 'Rigel'.
17 const comparator = (star1, star2) => {
18 if (star1.name === star2.name) return 0;
19 return star1.name < star2.name ? -1 : 1;
20 };
21
22 // Now let's search for the stars.
23 binarySearch(sortedArrayOfStars, { name: 'Not Existing Name' }, comparator); // -> -1
24 binarySearch(sortedArrayOfStars, { name: 'Alpha Centauri A' }, comparator); // -> 0
25 binarySearch(sortedArrayOfStars, { name: 'Alpha Centauri B' }, comparator); // -> 1
26 binarySearch(sortedArrayOfStars, { name: 'Polaris' }, comparator); // -> 3
Implementation
In the example above we’ve been using a custom comparison function that compared star objects
with each other. To make it possible we will re-use the Comparator utility class that we’ve already
implemented in “Linear Search” chapter. Please address the “Linear Search” chapter to get the details
of Comparator class implementation. This is a really simple class that provides a common interface
for object comparison.
So let’s import dependencies first.
16-binary-search/binarySearch.js
Now we may move on to binary search function implementation.

16-binary-search/binarySearch.js
4 /**
5 * Binary search implementation.
6 *
7 * @param {*[]} sortedArray
8 * @param {*} seekElement
9 * @param {function(a, b)} [comparatorCallback]
11 */
12 export default function binarySearch(sortedArray, seekElement, comparatorCallback) {
13 // Let's create comparator from the comparatorCallback function.
14 // Comparator object will give us common comparison methods like equal() and lessT\
15 hen().
17
18 // These two indices will contain current array (sub-array) boundaries.
19 let startIndex = 0;
20 let endIndex = sortedArray.length - 1;
21
22 // Let's continue to split array until boundaries are collapsed
23 // and there is nothing to split anymore.
24 while (startIndex <= endIndex) {
25 // Let's calculate the index of the middle element.
26 const middleIndex = startIndex + Math.floor((endIndex - startIndex) / 2);
27
28 // If we've found the element just return its position.
29 if (comparator.equal(sortedArray[middleIndex], seekElement)) {
30 return middleIndex;
31 }
32
33 // Decide which half to choose for seeking next: left or right one.
34 if (comparator.lessThan(sortedArray[middleIndex], seekElement)) {
35 // Go to the right half of the array.
36 startIndex = middleIndex + 1;
37 } else {
38 // Go to the left half of the array.
39 endIndex = middleIndex - 1;
40 }
41 }
42
43 // Return -1 if we have not found anything.
44 return -1;
45 }
Problems Examples
Here are some binary search related problems that you might encounter during the interview:
• Search Insert Position⁷⁷

• Divide Two Integers⁷⁸
• Search in Rotated Sorted Array⁷⁹
Quiz
Q1: Can we apply binary search to not sorted array?
Q2: What is the time complexity of binary search algorithm?
References
Binary search on Wikipedia https://en.wikipedia.org/wiki/Binary_search_algorithm⁸⁰
Binary search on YouTube https://www.youtube.com/watch?v=P3YID7liBug⁸¹
Binary search example and test cases https://github.com/trekhleb/javascript-algorithms/tree/mas-
ter/src/algorithms/search/binary-search⁸²
⁷⁷https://leetcode.com/problems/search-insert-position/
⁷⁸https://leetcode.com/problems/divide-two-integers/
⁷⁹https://leetcode.com/problems/search-in-rotated-sorted-array/
⁸⁰https://en.wikipedia.org/wiki/Binary_search_algorithm
⁸¹https://www.youtube.com/watch?v=P3YID7liBug
⁸²https://github.com/trekhleb/javascript-algorithms/tree/master/src/algorithms/search/binary-search
Sets. Cartesian Product.
Sets
The Cartesian Product algorithm uses sets to calculate the sets of multiple elements. When working
with sets, there are some basic terms we need to understand first.
In mathematics, a set is a collection of distinct objects, considered as an object in its own right. For
example numbers 2, 1, 0 and 14 may form a set of numbers {2, 1, 0, 14}. Strings 'apple', 'banana'
and 'grape' may form a set of strings/fruits {'apple', 'banana', 'grape'}. We can use arrays in
JavaScript to create sets.
const setOfNumbers = [2, 1, 0, 14];

const setOfStrings = ['apple', 'banana', 'grape'];
The objects that form a set are called set elements.
Cartesian Product
A cartesian product is a mathematical operation that returns a set (or product set or product) from
multiple sets. For two sets A and B the product set A×B will consist of all possible ordered pairs (a,b)
where a is an element of A set and b is an element of B.
For example the power set of two sets A = {x, y, z} and B = {1, 2, 3} would be the following
set:
{(x,1), (x,2), (x,3), (y,1), (y,2), (y,3), (z,1), (z,2), (z,3)}
The illustration below shows an example of how the power set of two sets is formed:
Sets. Cartesian Product. 182
The number of the elements in a product set A×B of two sets A and B may be calculated as a
multiplication of lengths of sets A and B as |A|×|B|. In example above we had two sets of length 3.
The product of those sets contains 3 * 3 = 9 elements.
It is also possible to create a cartesian product of more than just two sets. For example we have a set
of sweater colors COLORS = {blue, red, yellow}, set of neck forms NECK_FORMS = {round-neck,
V−neck, polo−neck} and a set of sizes SIZE = {XS, S, M, L, XL, XXL} then the cartesian
product of these three sets would look like {(blue,round-neck,XS), (blue,round-neck,S),
(blue,round-neck,M), ..., (yellow,polo−neck,XXL)}.
Applications
A deck of cards is an example of a cartesian product being applied to the set of card ranks {A, K,
Q, J, 10, 9, 8, 7, 6, 5, 4, 3, 2} and the set of card suits {�, �, �, �}. The product of these
two sets forms the standard 52-card deck {(A,�), (A,�), (A,�), ..., (2,�)}.
A two-dimensional coordinate system is also a product set of two sets R×R with R denoting the set
of real numbers.
In general, the cartesian product may be used to generate all possible pairs (sub-sets) of different
entities. We may think of generating pairs of (model, color) properties for a price list of cars or
(neck_form, size, color) sub-sets of properties for a sweater catalog and so on.
Usage Example
Let’s use our cartesianProduct() function to generate a deck of cards.
17-cartesian-product/example.js
2 import cartesianProduct from './cartesianProduct';
3
4 // Init the sets of card ranks and suits.
5 const cardRanks = ['A', 'K', 'Q', 'J', '10', '9', '8', '7', '6', '5', '4', '3', '2'];
6 const cardSuits = ['�', '�', '�', '�'];
7
8 // Generate a deck of cards.
9 const cardDeck = cartesianProduct(cardRanks, cardSuits);
10
11 // Check the deck.
13 console.log(cardDeck); // -> [['A', '�'], ['A', '�'], ..., ['2', '�']]
Implementation
In order to implement cartesianProduct() function, first we need to check the input. If the input
is incorrect and some input sets are empty the function will return null. Then we generate all pairs
of elements from two input sets setA and setB by using two nested for loops.
17-cartesian-product/cartesianProduct.js
1 /**
2 * Generates Cartesian Product of two sets.
3 * @param {*[]} setA
4 * @param {*[]} setB
5 * @return {*[]}
6 */
7 export default function cartesianProduct(setA, setB) {
8 // Check if input sets are not empty.
9 // Otherwise return null since we can't generate Cartesian Product out of them.
10 if (!setA || !setB || !setA.length || !setB.length) {
11 return null;
12 }
13
14 // Init product set.
15 const product = [];

16
17 // Now, let's go through all elements of a first and second set and form all possi\
18 ble pairs.
19 for (let indexA = 0; indexA < setA.length; indexA += 1) {
20 for (let indexB = 0; indexB < setB.length; indexB += 1) {
21 // Add current product pair to the product set.
22 product.push([setA[indexA], setB[indexB]]);
23 }
24 }
25
26 // Return cartesian product set.
27 return product;
28 }
Complexities
If we’re generating the product set of two sets A and B then two nested for loops will give us |A| *
|B| number of loops, where |A| is number of elements in set A and |B| is number of elements in set
|B|. So the time complexity is O(|A| * |B|).
The additional space complexity is also O(|A| * |B|) since we’re allocating a new array that will
hold all cartesian product elements.
Problems Examples
• Letter Combinations of a Phone Number⁸³

• Subsets II⁸⁴
Quiz
Q1: How many elements does a Cartesian Product of two sets A = ['a', 'b'] and B = ['c', 'd']
have?
Q2: Is it possible to generate Cartesian Product for four sets?
Q3: What is the time complexity of a cartesian product algorithm?
⁸³https://leetcode.com/problems/letter-combinations-of-a-phone-number/
⁸⁴https://leetcode.com/problems/subsets-ii/
References
Cartesian product on Wikipedia https://en.wikipedia.org/wiki/Cartesian_product⁸⁵
Cartesian product example and test cases https://github.com/trekhleb/javascript-algorithms/tree/-
master/src/algorithms/sets/cartesian-product⁸⁶
⁸⁵https://en.wikipedia.org/wiki/Cartesian_product
⁸⁶https://github.com/trekhleb/javascript-algorithms/tree/master/src/algorithms/sets/cartesian-product
Sets. Power Set.
In mathematics, the power set of a set S is a set of all subsets of S, including the empty set and set
S itself. The power set of a set S is denoted as P(S). The order of the elements in subsets doesn’t
matter, meaning that the set {a, b} is the same as {b, a}.
For example for a set of {a, b, c}:
• the empty set is {} (which may also be denoted as âˆ… or as the null set),
• one-element subsets are: {a}, {b}, abd {c},
• two-element subsets are: {a, b}, {a, c} and {b, c},
• and the original set {a, b, c} is also counted as a subset.
Summing all up we can say that the power set of a set {a, b, c} is a following set:
P({a, b, c}) = {
{},
{a}, {b}, {c},
{a, b}, {a, c}, {b, c},
{a, b, c}
}
The image below illustrates the process of generating a power set of a set {x, y, z} graphically.
Sets. Power Set. 187
Total Number of Subsets

For the set S with n members, the total number of subsets in a power set P(S) is written as |P(S)|
and is equal to |P(S)| = 2n.
For example, say we have a set S = {a, b, c} with 3 members then the total number of subsets is
23 = 8.
Why does the power set have a power of two members? The power set is binary in nature. Let’s
write down a table of 8 binary numbers. Each binary number in this case will consist of three bits
and each bit corresponds to an element in the set S. If the bit is set to 1, then the element is included
in the subset. If the bit is set to 0, the element will not be in the subset. Take a look at the picture
and the table below to get a better understanding of this idea.
Number abc Subset

0 000 {}
1 001 {c}
2 010 {b}
3 011 {c, b}
4 100 {a}
5 101 {a, c}
6 110 {a, b}
7 111 {a, b, c}
So the bit combinations starting from 0 and up to 2n are describe all possible subsets of
a power set. Thus each power set will have 2n subsets in total, where n is the number
of elements in the original set.
Usage Example
Before we move on to the power set algorithm implementation, let’s imagine that we already have a
bwPowerSet() function implemented and let’s use it to generate all possible ingredient combinations
for a fruit salad. This will help us understand what types of inputs and outputs we need for the
function.
18-power-set/example.js
2 import bwPowerSet from './bwPowerSet';
3
4 // Set up ingredients.
5 const ingredients = ['banana', 'orange', 'apple'];
6
7 // Generate all possible salad mixes out of our ingredients.
8 const saladMixes = bwPowerSet(ingredients);
9
11 console.log(saladMixes);
12
13 /*
14 The output will be:
15
16 [
17 [],
18 ['banana'],
19 ['orange'],
20 ['banana', 'orange'],
21 ['apple'],
22 ['banana', 'apple'],
23 ['orange', 'apple'],
24 ['banana', 'orange', 'apple'],
25 ]
26 */
Implementation
We will implement a power set generator using two approaches: the bitwise approach that we’ve
discussed above and a backtracking approach that will be described in the next section.
Bitwise Approach
In the bitwise approach, we will use the fact that the binary representation of the numbers up to
2n (where n is the size of a set) will represent bit masks for our original set that will
produce subsets.
Take a look at the implementation below:
18-power-set/bwPowerSet.js
1 /**
2 * Find power-set of a set using BITWISE approach.
3 *
4 * @param {*[]} originalSet
5 * @return {*[][]}
6 */
7 export default function bwPowerSet(originalSet) {
8 const subSets = [];
9
10 // We will have 2^n possible combinations (where n is a length of original set).
11 // It is because for every element of original set we will decide whether to inclu\
12 de
13 // it or not (2 options for each set element).
14 const numberOfCombinations = 2 ** originalSet.length;
15
16 // Each number in binary representation in a range from 0 to 2^n does exactly what\
17 we need:
18 // it shows by its bits (0 or 1) whether to include related element from the set o\
19 r not.
20 // For example, for the set {1, 2, 3} the binary number of 0b010 would mean that w\
21 e need to
22 // include only "2" to the current set.
23 for (let combinationIndex = 0; combinationIndex < numberOfCombinations; combinatio\
24 nIndex += 1) {
25 const subSet = [];
26
27 for (let setElementIndex = 0; setElementIndex < originalSet.length; setElementIn\
28 dex += 1) {
29 // Decide whether we need to include current element into the subset or not.
30 if (combinationIndex & (1 << setElementIndex)) {
31 subSet.push(originalSet[setElementIndex]);
32 }
33 }
34
35 // Add current subset to the list of all subsets.
36 subSets.push(subSet);
37 }
38
39 return subSets;
40 }
Our function accepts only one argument: originalSet which is an array. Then it creates a subSets
array that will hold all generated subsets of a power set.
The following line calculates the number of combinations. This number also is the maximum number
in a range of numbers that we use as bit masks. To raise the number to the power we use ** operator.
const numberOfCombinations = 2 ** originalSet.length;
The first for loop goes over every number in our range [0, numberOfCombinations]. For each
iteration, it creates a new power set subSet array that will hold all elements of the current subset
that corresponds to the current combinationIndex.
In the internal for loop, we iterate over the original set elements to see whether we need to include
each particular element or not. This is decided by checking the appropriate bit in combinationIndex.
If the bit at setElementIndex position is set then we need to include the element with the index
setElementIndex to the subSet. Otherwise, we’re ignoring it.
The bit checking is happens by using the following construction:
if (combinationIndex & (1 << setElementIndex)) {

// Add the element to the subset.
}
Here we’re shifting 1 to the left by setElementIndex positions (i.e. if setElementIndex is 4 then 0b1
<< 4 will give us 0b1000). And then we’re doing logical and operation to check if appropriate bit in
combinationIndex is set (i.e. if combinationIndex is 0b0100 and we’re doing 0b0100 & 0b1000 this
will give us 0 indicating the fact that the 4th bit of combinationIndex is not set).
Backtracking Approach
In the backtracking approaches we try to generate all possible solutions, but each time we generate
next solution we test if it satisfies all conditions (or try to make it satisfy all our conditions), and only
then continue generating subsequent solutions. Otherwise, backtrack, and go on a different path of
finding a solution.
On every iteration in our example, we’re going to add one more valid element. Validity in our case
is defined by the rule that we don’t want to have duplicates in subsets. Thus, on every iteration, we
may add any elements from the original set except the ones that have been already added.
This approach may be illustrated using the following decision tree.
This decision tree helps us to decide what elements we may include at this particular moment. And
if there are no options, let’s go one step back (backtrack) and try another valid option.
So, for example, let’s trace the decision tree in depth-first manner:
1. At the top of the tree, we have only one option and it is to “add empty set”. We may now
memorize empty set {} as a “power set” subset and move on trying to add more elements to
the empty set.
2. The next valid element that we may add to our current subset is a. So memorize subset {a} as
a valid “power set” subset and move on trying to add something new to the {a}.
3. At this moment we may add b since it was not yet added to the current subset. So let’s memorize
subset {a, b} as a valid “power set” subset and let’s move on trying to add more valid elements
to {a, b}.
4. We may add c now since it has not been added so far. Let’s memorize {a, b, c} as a valid
subset and move on.
5. At this moment we don’t have any other elements to add to {a, b, c} since we will have
duplicates. Backtrack now! Get rid of lastly added element (do one step to the top in our decision
tree). Now the current subset is {a, b}.
6. At this moment we still don’t have any other element to add since c element has been already
added before. Backtrack again! Move one step to the top of our decision tree. The current subset
now is {a}.
7. At this moment the b element has been already added at step 3 so we can’t add it now. But we
can add c now. So we may memorize the current subset {a, c} as a valid “power set” subset.
8. And so on…
Now let’s try to implement this approach in code:

18-power-set/bwPowerSet.js
1 /**
2 * Find power-set of a set using BACKTRACKING approach.
3 *
4 * @param {*[]} originalSet - Original set of elements we're forming power-set of.
5 * @param {*[][]} allSubsets - All subsets that have been formed so far (empty
6 * subset is included by default).
7 * @param {*[]} currentSubSet - Current subset that we're forming at the moment.
8 * @param {number} startAt - The position of in original set we're starting to form \
9 current subset.
10 * @return {*[][]} - All subsets of original set.
11 */
12 export default function btPowerSet(
13 originalSet,
14 allSubsets = [[]],
15 currentSubSet = [],
16 startAt = 0,
17 ) {
18 // Let's iterate over originalSet elements that may be added to the subset
19 // without having duplicates. The value of startAt prevents adding the duplicates.
20 for (let position = startAt; position < originalSet.length; position += 1) {
21 // Let's push current element to the subset
22 currentSubSet.push(originalSet[position]);
23
24 // Current subset is already valid so let's memorize it.
25 // We do array destruction here to save the clone of the currentSubSet.
26 // We need to save a clone since the original currentSubSet is going to be
27 // mutated in further recursive calls.
28 allSubsets.push([...currentSubSet]);
29
30 // Let's try to generate all other subsets for the current subset.
31 // We're increasing the position by one to avoid duplicates in subset.
32 btPowerSet(originalSet, allSubsets, currentSubSet, position + 1);
33
34 // BACKTRACK. Exclude last element from the subset and try the next valid one.
35 currentSubSet.pop();
36 }
37
38 // Return all subsets of a set.
39 return allSubsets;
40 }
Complexities
Bitwise Approach
In the bitwise approach, we’re iterating over 2n numbers and for each number we’re
iterating over n-sized original set to decide what elements to include in the current subset. Thus the
time complexity of this approach is O(n * 2n), where n is a number of elements in
the original set.
Backtracking Approach
In the backtracking approach, we’re iterating over a decision tree which has 2n nodes.
Thus the time complexity of this approach is O(2n), where n is a number of elements
in the original set.
Problems Examples
• Subsets⁸⁷
• Subsets II⁸⁸
Quiz
Q1: Is empty set a part of power set?
Q2: How many elements does power set of a set {a, b, c, d} have?
⁸⁷https://leetcode.com/problems/subsets/
⁸⁸https://leetcode.com/problems/subsets-ii/description/
References
Power set on Wikipedia https://en.wikipedia.org/wiki/Power_set⁸⁹
Power set on “Math is Fun” https://www.mathsisfun.com/sets/power-set.html⁹⁰
Power set example and test cases https://github.com/trekhleb/javascript-algorithms/tree/master/sr-
c/algorithms/sets/power-set⁹¹
⁸⁹https://en.wikipedia.org/wiki/Power_set
⁹⁰https://www.mathsisfun.com/sets/power-set.html
⁹¹https://github.com/trekhleb/javascript-algorithms/tree/master/src/algorithms/sets/power-set
Sets. Permutations.
Let’s say we have a collection or set of something (collection of numbers, letters, fruits, coins etc.)
and we need to pick items from a collection to form another collection. For example, imagine that
you’re picking lottery numbers and from the collection of available numbers {0, 1, 2, 3, 4, 5,
6, 7, 8, 9} you pick {4, 5, 9}. Or you’re picking the fruits from collections of available fruits
{orange, apple, banana, grape} to make a fruit salad out of {apple, banana}. Or you’re trying to
guess the lock password and you’re choosing 3 numbers from the collection {0, 1, 2, 3, 4, 5, 6,
7, 8, 9} to guess the correct password by forming sub-collections like {1, 1, 2}, {1, 1, 3}, {1,
1, 4}, …. In all these cases you’re creating one collection out from the other one by following some
rules. And these rules define whether your new collection is a permutation or a combination.
If the order of the elements in your new collection matters then you’re dealing with permutations.
In the case of the lock password the set of {1, 1, 2} is not the same as set of {2, 1, 1}). If the
order doesn’t matter, then you’re making a combination. For example, in the combination of fruits
{apple, banana} to make a salad, it doesn’t matter if you pick apple or banana first.
A permutation arranges all the members of a set into some sequence or order, or if the
set is already ordered, rearranging (reordering) its elements.
Your new collection may or may not contain duplicates (or repetitions). For example in the lock
password you’re allowed to use duplicate numbers but when you’re dealing with collection of race
winners {Black, Smith, Brown} you’re not allowed to make duplicates because it doesn’t make
sense to have the same person to be in two positions at the same time.
Sets. Permutations. 197
In this chapter we will focus on implementing permutations. In the next chapter we will deal with
combinations.
Permutations With Repetitions

Let’s take a simple example and form all permutations of a set S = {A, B} with repetitions. When
we say repititions, we mean the same letter can be used in different positions. The permutation set
P with permutations size of 2 in this case will look like the following:
S = {A, B}
P = {
{A, A},
{A, B},
{B, A},
{B, B},
}
Let’s generalize this example and calculate the number of sub-sets in P. Let’s say we have an original
set S of size n and we need to generate all possible permutations with repetitions (set P) of size r
for it. So the set P will consist of subsets of length r. For every position of a subset in P we will have
an option of n possible elements. In the example above, for each permutation position we choose
from two possible elements A and B. This would mean we will have n options for the first position in
permutations, n options again for the second position and so on up until rth position:
$P(n, r) = n \times n \times \dotsc n = n^{r}$
In our example above n = 2 and r = 2 so we have 4 permutations.
Permutations Without Repetitions

Let’s take another simple example and form all permutations of a set S = {A, B, C} without
repetitions. This means the same letter can’t be used in multiple positions. We must use one unique
letter per position. The permutation set P with permutations size of 3 in this case will look like the
following:
S = {A, B, C}
P = {
{A, B, C},
{A, C, B},
{B, A, C},
{B, C, A},
{C, A, B},
{C, B, A},
}
Let’s also generalize this example and calculate the number of sub-sets in P. Let’s say we have an
original set S of size n and we need to generate all possible permutations without repetitions (set P)
of size r for it. So the set P will consist of subsets of length r. For every position of a subset in P we
will need to reduce the number of possible elements each time. In example above (when we had n =
3 and r = 3) for the first permutation position we choose from three possible elements A, B and C.
But then for the second position we choose only from the rest of the elements that were absent in
the first position:
3×2×1=6
In other words in case if n and r are equal the number of permutations may be calculated as follows:
(n) × (n - 1) × (n - 2) × … × 1
This is actually the formula for the factorial of n! Let’s revert the order of multiplication so that this
equation would look a little bit more traditional and recognizable:
1 × 2 × 3 × … × (n - 1) × (n) = n!
In our example above, we had n = 3 so that 3! = 6 and we had 6 possible permutations without
repetitions.
But what happens if r is smaller than n? Let’s say we have a set of 20 elements and we need generate
permutations of size 3 out of them. In this case we will have 20 × 19 × 18 = 6840 permutations.
To generalize this calculation we may apply a trick to our first formula with factorial.
20 × 19 × 18 × 17 × 16 × ...
---------------------------- = 20 × 19 × 18
17 × 16 × ...
The formula for this calculation looks like this:
$P(n, r) = \frac{n!}{(n - r)!}$
If we have n = 20 and r = 3 then 20! / (20 - 3)! = 20! / 17! = 6840 as we’ve calculated above
as 20 × 19 × 18 = 6840. If n and r are equal than the formula falls back to n! / 0! = n! / 1 = n!.
Application
Permutations and combinations are useful when solving problems relevant to probability. Since we
can generate all possible options for specific situation (or at list to calculate the number of possible
options) we can also conclude what the chances are that each option will occur. This might be
simple cases like predicting the probability of lottery winning but also it may go beyond that to
more scientific areas.
In computer science permutations are used for analyzing sorting algorithms, in quantum physics
for describing states of particles and in biology for describing RNA sequences.
Usage Example
Before we move on with permutation implementation let’s imagine that we already have a
permutateWithoutRepetitions() and permutateWithRepetitions() functions implemented and
let’s try to use it. By using permutateWithRepetitions() we will try to generate all possible
passwords to the lock. And by using permutateWithoutRepetitions() we will try to generate all
possible variants of how the race of three racers might be ended (let’s say that the 0th
position in array would mean that the racer came first).
19-permutations/example.js
2 import permutateWithoutRepetitions from './permutateWithoutRepetitions';
3 import permutateWithRepetitions from './permutateWithRepetitions';
4
5 // PERMUTATIONS WITH REPETITIONS EXAMPLE.
6 // Let's generate all possible passwords combinations using permutations with repeti\
7 tions.
8 const possiblePasswordSymbols = ['A', 'B', 'C'];
9 const passwordLength = 3;
10 const allPossiblePasswords = permutateWithRepetitions(possiblePasswordSymbols, passw\

11 ordLength);
12
14 console.log(allPossiblePasswords);
15 /*
17 [
18 ['A', 'A', 'A'],
19 ['A', 'A', 'B'],
20 ['A', 'A', 'C'],
21 ['A', 'B', 'A'],
22 ['A', 'B', 'B'],
23 ['A', 'B', 'C'],
24 ['A', 'C', 'A'],
25 ['A', 'C', 'B'],
26 ['A', 'C', 'C'],
27 ['B', 'A', 'A'],
28 ['B', 'A', 'B'],
29 ['B', 'A', 'C'],
30 ['B', 'B', 'A'],
31 ['B', 'B', 'B'],
32 ['B', 'B', 'C'],
33 ['B', 'C', 'A'],
34 ['B', 'C', 'B'],
35 ['B', 'C', 'C'],
36 ['C', 'A', 'A'],
37 ['C', 'A', 'B'],
38 ['C', 'A', 'C'],
39 ['C', 'B', 'A'],
40 ['C', 'B', 'B'],
41 ['C', 'B', 'C'],
42 ['C', 'C', 'A'],
43 ['C', 'C', 'B'],
44 ['C', 'C', 'C'],
45 ]
46 */
47
48 // PERMUTATIONS WITHOUT REPETITIONS EXAMPLE.
49 // Now let's generate all possible racing results for three racers.
50 const racers = ['John', 'Bill', 'Jane'];
51 const racingResults = permutateWithoutRepetitions(racers);
52
54 console.log(racingResults);
55 /*
57 [
58 ['John', 'Bill', 'Jane'],
59 ['Bill', 'John', 'Jane'],
60 ['Bill', 'Jane', 'John'],
61 ['John', 'Jane', 'Bill'],
62 ['Jane', 'John', 'Bill'],
63 ['Jane', 'Bill', 'John'],
64 ]
65 */
Implementation
Permutations With Repetitions

We will solve this problem recursively. For example, if we need to find all permutations with
repetitions for the set {A, B, C} of size 3 we do the following:
• remember the first element A of the set,

• find all permutations of size r = 1 for a set S = {A, B, C},
• concatenate A with all smaller permutations from the previous step,
• remember the second element B of the set,
• find all permutations of size r = 1 for a set S = {A, B, C},
• concatenate B with all smaller permutations from the previous step,
• and so on…
For permutations of size 1, each element is a permutation. So for the set {A, B, C} the permutations
of size 1 will be {A}, {B} and {C}. This will be our base case for our recursive function.
Take a look at JavaScript implementation of this algorithm below.

19-permutations/permutateWithRepetitions.js
1 /**
2 * @param {*[]} permutationOptions
3 * @param {number} permutationLength
4 * @return {*[]}
5 */
6 export default function permutateWithRepetitions(
7 permutationOptions,
8 permutationLength = permutationOptions.length,
9 ) {
10 // If permutation length is equal to 1 than every element of the permutationOptions
11 // is a permutation subset.
12 if (permutationLength === 1) {
13 return permutationOptions.map(permutationOption => [permutationOption]);
14 }
15
16 // Init permutations array.
17 const permutations = [];
18
19 // Get permutations af smaller size that made of all permutation options.
20 const smallerPermutations = permutateWithRepetitions(
21 permutationOptions,
22 permutationLength - 1,
23 );
24
25 // Recursively go through all options and join it to the smaller permutations.
26 permutationOptions.forEach((currentOption) => {
27 // Concatenate current options to smaller permutations.
28 smallerPermutations.forEach((smallerPermutation) => {
29 permutations.push([currentOption].concat(smallerPermutation));
30 });
31 });
32
33 // Return permutations.
34 return permutations;
35 }
We’re using concat() function to attach two arrays together.
Permutations Without Repetitions

We will solve this problem recursively. It means that we generate permutations of a bigger size out
of permutations of smaller sizes. For example, if we need to generate all permutations of size r = 3
without repetitions for a set S = {A, B, C, D} we do the following:
• extract the first element A out of the set,

• find all permutations of size r = 2 for a set S = {B, C, D},
• concatenate A with all smaller permutations from the previous step,
• extract the second element B out of the set,
• find all permutations of size r = 2 for a set S = {C, D},
• concatenate B with all smaller permutations from the previous step,
• and so on…
If we will continue applying this logic recursively by ejecting one element on each iteration and
forming the permutations out of the set of elements that are left we’ll eventually end up forming
permutations of one element. The permutation of one element will be the element itself. This will be
our base case for recursion since we don’t need further recursion calls to calculate the permutations.
Take a look at JavaScript implementation of this algorithm below.

19-permutations/permutateWithoutRepetitions.js
1 /**
2 * @param {*[]} permutationOptions
3 * @return {*[]}
4 */
5 export default function permutateWithoutRepetitions(permutationOptions) {
6 // If we have only one element to permutate then this element is already a permuta\
7 tion of itself.
8 if (permutationOptions.length === 1) {
9 return [permutationOptions];
10 }
11
12 // Init permutations array.
13 const permutations = [];
14
15 // Get all permutations for permutationOptions excluding the first element.
16 // By doing this we're excluding the first element from all further smaller permu\
17 tations.
18 const smallerPermutations = permutateWithoutRepetitions(permutationOptions.slice(1\
19 ));
20
21 // Insert first option into every possible position of every smaller permutation.
22 const firstOption = permutationOptions[0];
23 for (let permIndex = 0; permIndex < smallerPermutations.length; permIndex += 1) {
24 const smallerPermutation = smallerPermutations[permIndex];
25
26 // Insert first option into every possible position of smallerPermutation.
27 for (let positionIndex = 0; positionIndex <= smallerPermutation.length; position\
28 Index += 1) {
29 const permutationPrefix = smallerPermutation.slice(0, positionIndex);
30 const permutationSuffix = smallerPermutation.slice(positionIndex);
31 permutations.push(permutationPrefix.concat([firstOption], permutationSuffix));
32 }
33 }
34
35 // Return all permutations.
36 return permutations;
37 }
We’re using concat() function to attach two arrays together.
Problems Examples
• Clone Graph⁹²
• Is Graph Bipartite?⁹³
• Number of Islands⁹⁴
Quiz
Q1: Does the order of the elements matter when permuting the set?
Q2: If we will need to generate all permutations without repetitions of size 2 from the set of size 4,
how many permutations will we have?
⁹²https://leetcode.com/problems/clone-graph/
⁹³https://leetcode.com/problems/is-graph-bipartite/
⁹⁴https://leetcode.com/problems/number-of-islands/
References
Permutations on Wikipedia https://en.wikipedia.org/wiki/Permutation⁹⁵
Permutations on “Math is Fun” https://www.mathsisfun.com/combinatorics/combinations-permu-
tations.html⁹⁶
Permutations on Medium https://medium.com/@trekhleb/permutations-combinations-algorithms-
cheat-sheet-68c14879aba5⁹⁷
Permutations example and test cases https://github.com/trekhleb/javascript-algorithms/tree/mas-
ter/src/algorithms/sets/permutations⁹⁸
⁹⁵https://en.wikipedia.org/wiki/Permutation
⁹⁶https://www.mathsisfun.com/combinatorics/combinations-permutations.html
⁹⁷https://medium.com/@trekhleb/permutations-combinations-algorithms-cheat-sheet-68c14879aba5
⁹⁸https://github.com/trekhleb/javascript-algorithms/tree/master/src/algorithms/sets/permutations
Sets. Combinations.
In mathematics, a combination is a selection of items from a collection, such that (unlike

permutations) the order of selection does not matter. More formally, a r-combination of a set S
of size n is a subset of r distinct elements of S.
For example, let’s say that we have a set of 5 (n = 5) web developers {Bill, John, Kate, Mike,
Julia}. When you need to select 3 (r = 3) web developers out of 5 to work on the new project it
doesn’t matter which order you select them. The only thing that matters is whether the developer
has been selected or not. In a result you may end up with combination {John, Kate, Julia} which
is actually the same as {Kate, John, Julia} because the order doesn’t matter. This is an example
of a combination without repetition because you can’t select the same developer twice (the set
{Bill, Bill} is not possible and doesn’t make sense).
Another example of a combination is a change set that you generate out of the coins set in your
pocket. Let’s say you have a set of coins {1, 1, 5, 5, 5, 10, 10, 10} and you need to give a
change of 27 coins. You have a set of 3 unique elements {1, 5, 10} (n = 3) and you’re allowed to
pick the same element several times. You may end up with the combination {10, 10, 5, 1, 1} (r
= 5) which is the same as {1, 10, 10, 1, 5} because the order doesn’t matter. This is an example
of a combination with repetition because you may include the same element into a combination
several times (the set of coins {1, 1, 1} is possible and it makes sense).
Combinations Without Repetitions

Let’s take a simple example and form combinations without repetitions of size r = 2 out of a set S
= {A, B, C} (n = 3). The combination set C in this case will look like the following:
Sets. Combinations. 208
S = {A, B, C}
r = 2
C = {
{A, B},
{A, C},
{B, C},
}
Now, let’s try to generalize this example and find out the formula of how many combinations without
repetitions (of size r) are there for the set of size n. We may do it by following these two steps:
• Assume that the order does matter (like with permutations) and count the number of
permutations of size r without repetitions.
• Afterwards assume that the order does not matter and eliminate all duplicated combinations.
From the chapter on permutations we already know that the number of permutations without
repetitions of size r for a set of size n may be calculated as $P(n, r) = \frac{n!}{(n - r)!}$. If we’re
dealing with the permutations of size r = 2 for the set S = {A, B, C} we will have $P(n, r) =
\frac{n!}{(n - r)!} = \frac{3!}{(3 - 2)!} = 3! = 6$ permutations:
S = {A, B, C}
r = 2
P = {
{A, B},
{B, A},
{A, C},
{C, A},
{B, C},
{C, B},
}
This set P is a permutation. What if we’ll ignore the ordering and will treat subsets of a set P as
combinations? We will find out that P has duplicate combinations.
P = {
{A, B},
{B, A}, <-- duplicate combination
{A, C},
{C, A}, <-- duplicate combination
{B, C},
{C, B}, <-- duplicate combination
}
Now we need to figure out how many duplicates we have and eliminate them from our calculations.
To do this, we need to figure out how many ways are there to generate all permutations without
repetitions of size r = 2 for the sets of size n = 2. For example the set {A, B} has two permutations
without repetitions and they are {A, B} and {B, A}.
From the chapter on permutations, we know that there are $r!$ permutations that exist in this case
(when r = n). And this is because we have r options for the first element of the permutations to
choose from. Then we have r - 1 options for the second element of the permutations and so on. In
result we will have $r \times (r-1) \times (r-2) \times \dots \times 1 = r!$ permutations.
Finally, in order find out how many combinations without permutations we have we need to divide
the number of permutations without repetitions by the number of duplicated combinations.
$C(n, r) = \frac{n!}{(n - r)!} \times \frac{1}{r!} = \frac{n!}{r!(n - r)!}$
So according to this formula the number of combinations without repetitions of size r = 2 for the
set S = {A, B, C} equals to $C(n, r) = \frac{3!}{2!(3-2)!} = \frac{3!}{2!} = \frac{6}{2} = 3$.
Combinations With Repetitions

Let’s take a simple example and form combinations with repetitions of size r = 2 out of a set S =
{A, B, C} (n = 3). The combination set C in this case will look like the following:
S = {A, B, C}
r = 2
C = {
{A, A},
{A, B},
{A, C},
{B, B},
{B, C},
{C, C},
}
It turned out that the number of combinations with repetitions equals to number of combinations
without repetitions but for imaginary set S of length r + (n - 1). The proof of this fact is out of
scope of this article but the final formula would look like the following:
$C(n, r) = \frac{(r + n - 1)!}{r!(n - 1)!}$
Application
Permutations and combinations are useful when solving the problems relevant to probability. Since
we may generate all possible options for specific situation (or at list to calculate the number of
possible options) we may also conclude what are the chances for each options to occur. This might
be simple cases like predicting the probability of lottery winning but also it may go beyond that to
more scientific areas.
Usage Example
Before we move on with combination implementation let’s imagine that we already have a
combineWithoutRepetitions() and combineWithRepetitions() functions implemented and let’s try
to use it. By using combineWithoutRepetitions() we will try to generate all possible web-developers
teams compositions out of the list of available developers. And by using combineWithRepetitions()
we will try to generate all possible ice cream scoops combinations.
20-combinations/example.js
2 import combineWithoutRepetitions from './combineWithoutRepetitions';
3 import combineWithRepetitions from './combineWithRepetitions';
4
5 // Combination WITHOUT repetitions.
6 // Let's generate all possible teams compositions that may work on the next projects.
7 const teamSize = 3;
8 const candidates = ['Bill', 'John', 'Kate', 'Jane', 'Mike'];
9 const possibleTeams = combineWithoutRepetitions(candidates, teamSize);
10
12 console.log(possibleTeams);
13 /*
15 [
16 ['Bill', 'John', 'Kate'],
17 ['Bill', 'John', 'Jane'],
18 ['Bill', 'John', 'Mike'],
19 ['Bill', 'Kate', 'Jane'],
20 ['Bill', 'Kate', 'Mike'],
21 ['Bill', 'Jane', 'Mike'],
22 ['John', 'Kate', 'Jane'],
23 ['John', 'Kate', 'Mike'],
24 ['John', 'Jane', 'Mike'],
25 ['Kate', 'Jane', 'Mike'],
26 ]
27 */
28
29 // Combination WITH repetitions.
30 // Let's generate all possible combinations of ice cream scoops.
31 const iceCreamFlavours = ['banana', 'mint', 'pistachio', 'vanilla'];
32 const numberOfScoops = 3;
33 const scoopCombinations = combineWithRepetitions(iceCreamFlavours, numberOfScoops);
34
36 console.log(scoopCombinations);
37 /*
39 [
40 ['banana', 'banana', 'banana'],
41 ['banana', 'banana', 'mint'],
42 ['banana', 'banana', 'pistachio'],
43 ['banana', 'banana', 'vanilla'],

44 ['banana', 'mint', 'mint'],
45 ['banana', 'mint', 'pistachio'],
46 ['banana', 'mint', 'vanilla'],
47 ['banana', 'pistachio', 'pistachio'],
48 ['banana', 'pistachio', 'vanilla'],
49 ['banana', 'vanilla', 'vanilla'],
50 ['mint', 'mint', 'mint'],
51 ['mint', 'mint', 'pistachio'],
52 ['mint', 'mint', 'vanilla'],
53 ['mint', 'pistachio', 'pistachio'],
54 ['mint', 'pistachio', 'vanilla'],
55 ['mint', 'vanilla', 'vanilla'],
56 ['pistachio', 'pistachio', 'pistachio'],
57 ['pistachio', 'pistachio', 'vanilla'],
58 ['pistachio', 'vanilla', 'vanilla'],
59 ['vanilla', 'vanilla', 'vanilla'],
60 ]
61 */
Implementation
Combinations Without Repetitions

This problem may be solved recursively. It means that we generate combinations of a bigger size
out of combinations of smaller sizes. For example, if we need to generate all combinations of size r
= 3 without repetitions for a set S = {A, B, C, D} we do the following:
• extract the first element A out of the set,

• find all combinations of size r = 2 for a set S = {B, C, D},
• concatenate A with all smaller combinations from the previous step,
• extract the second element B out of the set,
• find all combinations of size r = 2 for a set S = {C, D},
• concatenate B with all smaller combinations from the previous step,
• and so on…
The base case for our recursion would be the case when we need to generate combinations of size r
= 1 for a set. In this case every element of a set is a combination we’re looking for, thus no further
recursion calls are needed.
Take a look at full recursion calls tree below to get better understanding of the algorithm.
Here is an example of how this algorithm may be implemented in JavaScript.

20-combinations/combineWithoutRepetitions.js
1 /**
2 * @param {*[]} comboOptions - original set that we will take elements from.
3 * @param {number} comboLength - the length of combinations we're going to make.
4 * @return {*[]}
5 */
6 export default function combineWithoutRepetitions(comboOptions, comboLength) {
7 // If the length of combinations is 1 then each element of original set is a combi\
8 nation.
9 if (comboLength === 1) {
10 return comboOptions.map(comboOption => [comboOption]);
11 }
12
13 // Init combinations array.
14 const combos = [];
15
16 // Extract characters one by one and concatenate them to combinations of smaller s\
17 ize.
18 // We need to extract characters since we don't want to have duplicates.
19 comboOptions.forEach((currentOption, optionIndex) => {
20 // Get all smaller combinations WITHOUT the current element from original set.
21 const smallerCombos = combineWithoutRepetitions(

22 comboOptions.slice(optionIndex + 1),
23 comboLength - 1,
24 );
25
26 // Concatenate current element (option) to all smaller combinations.
27 smallerCombos.forEach((smallerCombo) => {
28 combos.push([currentOption].concat(smallerCombo));
29 });
30 });
31
32 // Return all combinations.
33 return combos;
34 }
Combinations With Repetitions

This problem may be solved recursively. It means that we may generate combinations of bigger size
out of combinations of smaller sizes. For example, if we need to generate all combinations of size r
= 2 with repetitions for a set S = {A, B, C} we do the following:
• remember the first element A of the set,

• find all combinations of size r = 1 for a set S = {A, B, C},
• concatenate A with all smaller combinations from the previous step,
• remember the second element B of the set,
• find all combinations of size r = 1 for a set S = {A, B, C},
• concatenate B with all smaller combinations from the previous step,
• and so on…
The base case for our recursion would be the case when we need to generate combinations of size r
= 1 for a set. In this case every element of a set is a combination we’re looking for, thus no further
recursion calls are needed.
The main difference from the algorithm for combinations without repetitions here is that instead
of extracting the elements out of a set we just remember them. We don’t need extractions since
repetitions are allowed.
Take a look at full recursion calls tree below to get better understanding of the algorithm.
Here is an example of how this algorithm may be implemented in JavaScript.

20-combinations/combineWithRepetitions.js
1 /**
2 * @param {*[]} comboOptions - original set that we will take elements from.
3 * @param {number} comboLength - the length of combinations we're going to make.
4 * @return {*[]}
5 */
6 export default function combineWithRepetitions(comboOptions, comboLength) {
7 // If the length of combinations is 1 then each element of original set is a combi\
8 nation.
9 if (comboLength === 1) {
10 return comboOptions.map(comboOption => [comboOption]);
11 }
12
13 // Init combinations array.
14 const combos = [];
15
16 // Go through every character of original set and concatenate it to combinations
17 // of smaller size.
18 comboOptions.forEach((currentOption, optionIndex) => {
19 // Get all smaller combinations WITH the current element from original set.
20 const smallerCombos = combineWithRepetitions(
21 comboOptions.slice(optionIndex),
22 comboLength - 1,
23 );
24
25 // Concatenate current element (option) to all smaller combinations.

26 smallerCombos.forEach((smallerCombo) => {
27 combos.push([currentOption].concat(smallerCombo));
28 });
29 });
30
31 // Return all combinations.
32 return combos;
33 }
Problems Examples
• Combinations⁹⁹
• Combination Sum III¹⁰⁰
• Combination Sum IV¹⁰¹
Quiz
Q1: Does the order of the elements matter when forming a combination?
Q2: If we will need to generate all combinations without repetitions of size 2 from the set of size 4,
how many combinations will we have?
References
Combinations on Wikipedia https://en.wikipedia.org/wiki/Combination¹⁰²
Combinations on “Math is Fun” https://www.mathsisfun.com/combinatorics/combinations-permu-
tations.html¹⁰³
Combinations on Medium https://medium.com/@trekhleb/permutations-combinations-algorithms-
cheat-sheet-68c14879aba5¹⁰⁴
Combinations example and test cases https://github.com/trekhleb/javascript-algorithms/tree/mas-
ter/src/algorithms/sets/combinations¹⁰⁵
⁹⁹https://leetcode.com/problems/combinations/
¹⁰⁰https://leetcode.com/problems/combination-sum-iii/
¹⁰¹https://leetcode.com/problems/combination-sum-iv/
¹⁰²https://en.wikipedia.org/wiki/Combination
¹⁰³https://www.mathsisfun.com/combinatorics/combinations-permutations.html
¹⁰⁴https://medium.com/@trekhleb/permutations-combinations-algorithms-cheat-sheet-68c14879aba5
¹⁰⁵https://github.com/trekhleb/javascript-algorithms/tree/master/src/algorithms/sets/combinations
Sorting: Quicksort
The Task
Sort an unordered array of values in increasing order. For example, an unsorted array [10, 3, 0,
15, 10, 18, 2, 7, 11] sorted in increasing order would look like this: [0, 2, 3, 7, 10, 10, 11,
15, 18].
In addition to sorting numbers, we’d like to also be able to sort an array of objects. In order to
accomplish this, in our algorithm we will implement a comparison callback function. For example,
it should be possible to sort an array [{name: "Bill", age: 23}, {name: "Jane", age: 18}] by
age in increasing order to get an array [{name: "Jane", age: 18}, {name: "Bill", age: 23}].
In our description of the task, we only mention sorting by increasing order. Once you understand
how the quicksort algorithm works though, you’ll be able to change the algorithm to support sorting
by increasing order.
The Algorithm
Quicksort is an O(n * log(n)) efficient and commonly used sorting algorithm. Quicksort is a
comparison sort, meaning that it can sort items of any type for which a “less-than” relation is defined.
Quicksort is a divide and conquer algorithm. It means that it splits the original problem into smaller
sub-problems and tries to solve them first. Then it combines the solutions to the smaller problems
to come up with the final solution. In our case, we will split an array into two parts and then sort
smaller arrays first before merging them into one final sorted array.
The algorithm consists of three recursive (repeating over sub-problems) steps:
• Pivot selection: pick an element, called a pivot, from the array (usually the first, middle or the
last one is taken).
• Partitioning: split an array into three sub-arrays: left (with all the elements that are smaller
than the pivot element), center (with all the elements that are equal to pivot element) and
right (with all the elements that are greater than the pivot element).
• Recurse: apply two previous steps to smaller arrays left and right by choosing the new pivot
element for them with further partitioning. We don’t need to apply those steps to the center
array because it is already sorted and contains all equal elements.
Sorting: Quicksort 218
The base case for this algorithm is to sort an array with just one element. With one element in the
array, no sorting is necessary so we return the value.
The recursive steps of the Quicksort algorithm are illustrated in the diagram below:
Usage Example
Before we move on with quickSort() function implementation let’s imagine that we already have
one implemented and let’s use it for sorting a list of usernames in O(n * log(n)) time. This will help
us to understand what parameters the function needs and what the result will be when the algorithm
is finished. In this example, we will use a comparator function to sort the array. This allows us to
use custom comparison logic when sorting the array of objects.
21-quicksort/example.js
2 import quickSort from './quickSort';
3
4 // Let's sort user list by age in O(n * log(n)) time.
5
6 // Init users list.
7 const notSortedUserList = [
8 { age: 18, name: 'Bill' },
9 { age: 20, name: 'Kate' },
10 { age: 20, name: 'Tom' },
11 { age: 24, name: 'Cary' },
12 { age: 37, name: 'Mike' },
13 { age: 42, name: 'Ben' },
14 { age: 50, name: 'Jane' },
15 { age: 60, name: 'Julia' },
16 ];
17
18 // Create comparator function that will compare two user objects.
19 const userComparator = (user1, user2) => {
20 if (user1.age === user2.age) {
21 return 0;
22 }
23
24 return user1.age > user2.age ? 1 : -1;
25 };
26
27 // Sort user list.
28 const sortedUserList = quickSort(notSortedUserList, userComparator);
29
31 console.log(sortedUserList);
32 /*
34 [
35 { age: 18, name: 'Bill' },
36 { age: 20, name: 'Kate' },
37 { age: 20, name: 'Tom' },
38 { age: 24, name: 'Cary' },
39 { age: 37, name: 'Mike' },

40 { age: 42, name: 'Ben' },
41 { age: 50, name: 'Jane' },
42 { age: 60, name: 'Julia' },
43 ]
44 */
Implementation
There are several ways to implement the quicksort algorithm. They differ in two ways: how the pivot
point is chosen (first element, middle element, last element) and the way the partitioning happens
(in-place, not in-place). You will find the different implementations in the JavaScript Algorithms
repository¹⁰⁶.
In this chapter, we will implement quicksort by picking the first element as the pivot point and we
will do “not in-place” partitioning to make the algorithm implementation more clear and readable.
Since we want our quickSort() function to be able to sort not only numbers but objects as
well, we will re-use the Comparator utility class that we’ve already implemented in the “Linear
Search” chapter. Please address “Linear Search” chapter to get the details of Comparator class
implementation. This is a simple class that provides a common interface for object comparison with
equal(), lessThan() and greaterThan() methods.
21-quicksort/quickSort.js
4 /**
5 * @param {*[]} originalArray - array to be sorted.
6 * @param {function} comparatorCallback - function that compares two elements
7 * @return {*[]} - sorted array
8 */
9 export default function quickSort(originalArray, comparatorCallback = null) {
10 // Let's create comparator from the comparatorCallback function.
11 // Comparator object will give us common comparison methods like equal() and lessT\
12 hen().
14
15 // Clone original array to prevent it from modification.
16 // We don't do in-place sorting in this example and thus we don't want side effect\
17 s.
18 const array = [...originalArray];
19
20 // If array has less than or equal to one elements then it is already sorted.
¹⁰⁶https://github.com/trekhleb/javascript-algorithms
21 // This is a base case for our recursion.

22 if (array.length <= 1) {
23 return array;
24 }
25
26 // Init left and right arrays.
27 const leftArray = [];
28 const rightArray = [];
29
30 // Take the first element of array as a pivot and init the center array.
31 const pivotElement = array.shift();
32 const centerArray = [pivotElement];
33
34 // Split all array elements between left, center and right arrays.
35 // Since we're extracting elements out of array we may just check array length to \
36 stop the loop.
37 while (array.length) {
38 // Extract first element out from the unsorted array.
39 const currentElement = array.shift();
40
41 // Compare extracted element to the pivot to decide what sub-array
42 // it belongs to (left, center or right).
43 if (comparator.equal(currentElement, pivotElement)) {
44 centerArray.push(currentElement);
45 } else if (comparator.lessThan(currentElement, pivotElement)) {
46 leftArray.push(currentElement);
47 } else {
48 rightArray.push(currentElement);
49 }
50 }
51
52 // Sort left and right arrays recursively.
53 const leftArraySorted = quickSort(leftArray, comparatorCallback);
54 const rightArraySorted = quickSort(rightArray, comparatorCallback);
55
56 // Let's now join sorted left array with center array and with sorted right array.
57 return leftArraySorted.concat(centerArray, rightArraySorted);
58 }
Complexities
Best Average Worst

O(n * log(n)) O(n * log(n)) O(n2)
The worst case scenario with O(n2) time complexity might happen if we have an array
that is already sorted in increasing order and we always pick the first element of the array as the
pivot point. In this case, after the partitioning step, all the elements will be in the right array leaving
the left array empty. This situation will continue for all further recursive partitions and we will end
up with (n - 1) recursive steps in which we will do full traversal over sub-arrays that will result
in O(n2) time complexity.
The best case scenario with O(n * log(n)) time complexity will happen when the pivot point will
divide an array into two almost equal left and right partitions. In this case, we will have log(n)
levels in the quicksort diagram (see the illustration above) with full sub-arrays traversal which gives
us O(n * log(n)) time complexity.
Problems Examples
• Sort List¹⁰⁷
• Kth Largest Element in an Array¹⁰⁸
Quiz
Q1: On average, what is the time complexity of the Quicksort algorithm?
Q2: How can we change our quickSort() function to make it sort the elements in decreasing order?
References
Quicksort on Wikipedia https://en.wikipedia.org/wiki/Quicksort¹⁰⁹
Quicksort on YouTube https://www.youtube.com/watch?v=SLauY6PpjW4¹¹⁰
Quicksort example and test cases https://github.com/trekhleb/javascript-algorithms/tree/master/sr-
c/algorithms/sorting/quick-sort¹¹¹
¹⁰⁷https://leetcode.com/problems/sort-list/
¹⁰⁸https://leetcode.com/problems/kth-largest-element-in-an-array/
¹⁰⁹https://en.wikipedia.org/wiki/Quicksort
¹¹⁰https://www.youtube.com/watch?v=SLauY6PpjW4
¹¹¹https://github.com/trekhleb/javascript-algorithms/tree/master/src/algorithms/sorting/quick-sort
Trees. Depth-First Search.
The Task
Imagine that we have a tree of nodes and we want to:
• check whether a specific node exists in a tree (i.e. perform a search operation for node “E”),
• check whether two nodes are connected (i.e. to find a path between node “A” and “E”),
• do nodes inventory by visiting all the nodes of a tree (i.e. to convert a tree to an array).
Trees. Depth-First Search. 224
To accomplish all these tasks we need to utilize a tree traversal algorithm. In other words, we need
to write some code that will allow us to visit every node of a tree.
In this chapter, we will traverse a binary tree (a tree in which every node may have at most two
children). Once you understand the idea of a depth-first tree traversal for a binary tree you’ll be
able to extend it for trees with more than two possible children for each node.
The tree that we will traverse is an abstraction of some real-world data. It may be a binary tree
of subsequent “yes/no” questions and you might want to find out what “yes/no” answers will lead
you to the final suggestion/conclusion that is located on one of the leaves of a tree. You may also
think about chess game moves that are presented as a tree where the children of the node are all the
possible moves. Each move will lead you to the next node and the next node will have another set
of possible moves. And you might want to find out which move will lead you to the “winning” leaf
of a tree.
The Algorithm
One of the ways to solve the tasks mentioned above is a depth-first search algorithm.
Depth-first search (DFS) is an algorithm for traversing a tree or graph data structures. The idea of
the algorithm is to start at some specific node (normally the root node) and explore as far as possible
along each branch before backtracking.
The word backtracking means that when we’re moving deeper along the branch of a tree and there
are no more nodes along the current path to explore, then we move back up the tree to find the next
closest unexplored sub-branch (node) and explore it.
The illustration below shows the sequence of binary tree traversal using a depth-first search
algorithm.
Think about this algorithm as if you are walking through a maze. As you walk, you come upon
intersections in the maze – these are like the nodes of our tree. Each intersection has three options:
take a right turn, take a left turn or go back to the previous intersection (backtracking). Let’s say
we’ve came up with a strategy to try the right turns first. So we choose to go right and mark the
right turn with a paintbrush as “visited” to avoid this turn in the future. If we come to a dead end
in the maze (a leaf node), we backtrack until we reach an intersection where we’ve already gone
right. Then we go to the left and mark with our paintbrush the left turn as “visited”. Once we came
to the intersection with both turns being marked as “visited” we need to go back (backtrack) to the
previous intersection and find the first unvisited turn. And so on until we reach the exit of the maze.
This algorithm may be implemented recursively using a stack (see the Stack chapter for an in-depth
explanation on this data structure). The idea of using the stack for depth-first search is the following:
1. Pick the starting node and put all its children into the stack.
2. Pop the node from the stack to select the next node to visit and push all its children into the
stack.
3. Go to step 2 (repeat this process until the stack is empty).
Usage Example
Before moving on with the actual implementation of the depth-first search algorithm let’s imagine
that we already have a depthFirstSearch() function implemented. This function allows us to
traverse binary tree nodes that are implemented using simple BinaryTreeNode class. By doing that
we’ll a get better understanding of what arguments we want to pass to our depthFirstSearch()
function and what output we expect from it.
In this usage example, we will use a callback function as the second argument to our depthFirstSearch()
function. This callback will be called every time the function visits a node in the tree. This allows
us to add custom logic when we visit each node.
22-tree-depth-first-search/example.js
2 import { depthFirstSearch } from './depthFirstSearch';
3 import { BinaryTreeNode } from '../06-binary-search-tree/BinaryTreeNode';
4
5 // Create tree nodes.
6 const nodeA = new BinaryTreeNode('A');
7 const nodeB = new BinaryTreeNode('B');
8 const nodeC = new BinaryTreeNode('C');
9 const nodeD = new BinaryTreeNode('D');
10 const nodeE = new BinaryTreeNode('E');
11
12 // Form a binary tree out of created nodes.
13 nodeA
14 .setLeft(nodeB)
15 .setRight(nodeC);
16
17 nodeB
18 .setLeft(nodeD)
19 .setRight(nodeE);
20
21 // Init the array that will contain all traversed nodes of th tree.
22 const traversedNodes = [];

23
24 // Create visiting node callback.
25 const visitNodeCallback = (visitedNode) => {
26 // Once we visit the node let's add it to the list of traversed nodes.
27 traversedNodes.push(visitedNode);
28 };
29
30 // Perform depth-first tree traversal.
31 depthFirstSearch(nodeA, visitNodeCallback);
32
34 console.log(traversedNodes);
35 /*
37 [nodeA, nodeB, nodeD, nodeE, nodeC]
38 */
Implementation
In order to traverse our tree, we must first create a BinaryTreeNode class that will represent
a node of the binary tree. In the class constructor, we will assign a value to the node. Our
setLeft() and setRight() methods allow us to attach left and right nodes to it. For usability
reasons we’ll return this from setLeft() and setRight() methods to be able to chain them like
nodeA.setLeft(nodeB).setRight(nodeC).
06-binary-search-tree/BinaryTreeNode.js
1 export class BinaryTreeNode {

2 /**
3 * @param {*} [value]
4 */
5 constructor(value = null) {
6 this.left = null;
9 }
10
11 /**
12 * @param {BinaryTreeNode} node
13 */
14 setLeft(node) {
16 return this;
17 }
18
19 /**
21 */
22 setRight(node) {
24 return this;
25 }
26 }
Once we’ve implemented the BinaryTreeNode class we can then implement the depthFirstSearch()
function that will traverse BinaryTreeNode instances. This function will accept the starting node as
a first parameter and visitNodeCallback() as a second parameter. visitNodeCallback() will be
called each time we meet the new node on our DFS-path.
We will implement the depthFirstSearch() function using recursion. It means that we will call
depthFirstSearch() function for the left and right child of the current node if they exist.
22-tree-depth-first-search/depthFirstSearch.js
1 /**
2 * Perform depth-first search (DFS) traversal of the rootNode.
3 *
4 * @param {BinaryTreeNode} node - The starting node to be traversed.
5 * @param {callback} [visitNodeCallback] - Visiting node callback.
6 */
7 export function depthFirstSearch(node, visitNodeCallback = () => {}) {
8 // Call the visiting node callback.
9 visitNodeCallback(node);
10
11 // Traverse left branch.
12 if (node.left) {
13 depthFirstSearch(node.left, visitNodeCallback);
14 }
15
16 // Traverse right branch.
17 if (node.right) {
18 depthFirstSearch(node.right, visitNodeCallback);
19 }
20 }
Previously, it was mentioned that the depth-first search algorithm may be implemented using
the stack data structure. But looking at the code of depthFirstSearch() function you might be
wondering why we didn’t use the Stack class implemented in the stack chapter. The reason for that
is that we’ve used a stack but in a slightly different form. Instead of using the Stack class directly
we’ve used the call stack of the depthFirstSearch() function instead. Since we’re dealing with
recursion calls here and calling the function out from itself the first function call can’t be resolved
unless subsequent function calls are resolved. Therefore, all unresolved function calls are placed into
the function call stack.
Complexities
Time Complexity
The depth-first search algorithm that was implemented above visits every node of a tree exactly
once. If the tree we are traversing has |N| number of nodes then the function depthFirstSearch()
will be called recursively |N| number of times (once for every tree node). Therefore we may say that
the time complexity of tree traversal using a depth-first approach is O(|N|), where |N| is the total
number of nodes in a tree.
Auxiliary Space Complexity

In the worst-case scenario, we will have an unbalanced tree that will look like a linked list (each node
of the tree has one left (or only one right) child). Since we used the recursive approach to implement
the depthFirstSearch() function it will make the function call stack grow with every new recursive
call (for every new node). Thus, in the worst-case scenario, we will need O(|N|) auxiliary memory
space for our recursive function call stack.
Problems Examples
• Same Tree¹¹²
• Maximum Depth of Binary Tree¹¹³
• Validate Binary Search Tree¹¹⁴
Quiz
Q1: What is the time complexity of the depth-first search algorithm?
Q2: Can we avoid recursion while implementing the depthFirstSearch() function?
References
Depth-first search (DFS) on Wikipedia https://en.wikipedia.org/wiki/Depth-first_search¹¹⁵
Tree DFS example and test cases https://github.com/trekhleb/javascript-algorithms/tree/master/sr-
c/algorithms/tree/depth-first-search¹¹⁶
¹¹²https://leetcode.com/problems/same-tree/
¹¹³https://leetcode.com/problems/maximum-depth-of-binary-tree/
¹¹⁴https://leetcode.com/problems/validate-binary-search-tree/
¹¹⁵https://en.wikipedia.org/wiki/Depth-first_search
¹¹⁶https://github.com/trekhleb/javascript-algorithms/tree/master/src/algorithms/tree/depth-first-search
Trees. Breadth-First Search.
The Task
Imagine that we have a tree of nodes and we want to:
• check whether a specific node exists in a tree (i.e. perform search operation for node “E”),
• do a nodes inventory by visiting all the nodes of a tree (i.e. to convert a tree to an array).
Trees. Breadth-First Search. 232
To accomplish all these tasks we need to utilize a tree traversal algorithm. In other words, we need
to write some code that will allow us to visit every node of a tree.
In this chapter, we will traverse a binary tree (a tree in which every node may have at most two
children). Once you understand the idea of a depth-first tree traversal for a binary tree you’ll be
able to extend it for trees with more than two possible children for each node.
The Algorithm
One of the ways to solve the tasks mentioned above is breadth-first search algorithm.
Breadth-first search (BFS) is an algorithm for traversing or searching tree or graph data structures.
It starts at the tree root, and explores all of the neighbor nodes at the present depth prior to moving
on to the nodes at the next depth level.
Breadth-first search (BFS) and depth-first search (DFS) algorithms are very similar in that they are
both traversing a tree, but they do it in two different ways. BFS goes wide by exploring all the nodes
of a tree on a certain level and only then moves one level deeper and explores nodes on the next
level. DFS goes deep by exploring all the nodes of a tree by going down to the tree leaves and only
then backtracks and tries another branch until it reaches leaf nodes.
The illustration below shows the sequence of binary tree traversal using breadth-first search
algorithm.
This algorithm may be implemented using queue (see one of the previous chapters of this book about
the Queue data structure). The idea of using a queue to implement breadth-first search algorithm is
the following:
1. Pick the starting node and put it to the nodes queue.

2. Pop the next node from the queue and add all its children to the queue to visit them later.
3. Go to step 2 (repeat this process until the queue is empty).
Usage Example
Before moving on with actual implementation of breadth-first search algorithm let’s imagine that
we already have a breadthFirstSearch() function implemented. This function allows us to traverse
binary tree nodes that are implemented using simple BinaryTreeNode class. By doing that we’ll get
better understanding of what arguments we want to pass to our breadthFirstSearch() function

and what output we expect from it.
In this usage example, we will use a callback function as the second argument to our breadtFirstSearch()
function. This callback will be called every time the function visits a node in the tree. This allows
us to add custom logic when we visit each node.
23-tree-breadth-first-search/example.js
2 import { BinaryTreeNode } from '../06-binary-search-tree/BinaryTreeNode';
3 import { breadthFirstSearch } from './breadthFirstSearch';
4
5 // Create tree nodes.
6 const nodeA = new BinaryTreeNode('A');
7 const nodeB = new BinaryTreeNode('B');
8 const nodeC = new BinaryTreeNode('C');
9 const nodeD = new BinaryTreeNode('D');
10 const nodeE = new BinaryTreeNode('E');
11
12 // Form a binary tree out of created nodes.
13 nodeA
14 .setLeft(nodeB)
15 .setRight(nodeC);
16
17 nodeB
18 .setLeft(nodeD)
19 .setRight(nodeE);
20
21 // Init the array that will contain all traversed nodes of th tree.
22 const traversedNodes = [];
23
24 // Create visiting node callback.
25 const visitNodeCallback = (visitedNode) => {
26 // Once we visit the node let's add it to the list of traversed nodes.
27 traversedNodes.push(visitedNode);
28 };
29
30 // Perform breadth-first tree traversal.
31 breadthFirstSearch(nodeA, visitNodeCallback);
32
33 // Check that traversal happened in correct order.
34 expect(traversedNodes).toEqual([
35 nodeA, nodeB, nodeC, nodeD, nodeE,
36 ]);
37
39 console.log(traversedNodes);
40 /*
42 [nodeA, nodeB, nodeC, nodeD, nodeE]
43 */
Implementation
Since we’re going to traverse binary tree let’s create a simple BinaryTreeNode class that will represent
the node of binary tree. Using class constructor we will assign the value to the node. And also by
using setLeft() and setRight() methods we’ll be able to attach left and right nodes to it. For
usability reasons we’ll return this from setLeft() and setRight() methods to be able to chain
them like nodeA.setLeft(nodeB).setRight(nodeC).
06-binary-search-tree/BinaryTreeNode.js
1 export class BinaryTreeNode {

2 /**
3 * @param {*} [value]
4 */
5 constructor(value = null) {
6 this.left = null;
9 }
10
11 /**
13 */
14 setLeft(node) {
16 return this;
17 }
18
19 /**
21 */
22 setRight(node) {
24 return this;
25 }
26 }
Since the breadth-first traversal uses queue to store all the nodes that are supposed to be visited let’s
import a Queue class first.
23-tree-breadth-first-search/breadthFirstSearch.js
import { Queue } from '../03-queue/Queue';
Once we’ve implemented the BinaryTreeNode class and imported Queue as a dependency we may
proceed with implementation of breadthFirstSearch() function that will traverse BinaryTreeNode
instances. This function will accept the starting rootNode as a first parameter and visitNodeCallback()
as a second parameter. visitNodeCallback() will be called each time we meet the new node on our
BFS-path.
What we need to do inside the breadthFirstSearch() function is to:
1. create an empty queue instance nodeQueue,

2. add the rootNode (the node from which we’ll start BFS traversal) to the queue instance
nodeQueue,
3. pop the next node in the queue and assign it to currentNode variable,
4. call the visitNodeCallback() callback to notify watchers/subscribers about visiting the new
node,
5. add all currentNode children to the queue to visit them later,
6. go to step 3 until the nodeQueue is empty.
Here is how this function implementation looks:

23-tree-breadth-first-search/breadthFirstSearch.js
4 /**
5 * Perform breadth-first search (BFS) traversal of the rootNode.
6 *
7 * @param {BinaryTreeNode} rootNode - The starting node to be traversed.
8 * @param {callback} [visitNodeCallback] - Visiting node callback.
9 */
10 export function breadthFirstSearch(rootNode, visitNodeCallback = () => {}) {
11 // Do initial queue setup.
12 // We need to add a rootNode to the queue first to start the
13 // traversal process from it.
14 const nodeQueue = new Queue();
15 nodeQueue.enqueue(rootNode);
16
17 // Visit all the nodes of the queue until the queue is empty.
18 while (!nodeQueue.isEmpty()) {
19 // Fetch the next node to visit.
20 const currentNode = nodeQueue.dequeue();
21
22 // Call the visiting node callback.
23 visitNodeCallback(currentNode);
24
25 // Add left node to the traversal queue.
26 if (currentNode.left) {
27 nodeQueue.enqueue(currentNode.left);
28 }
29
30 // Add right node to the traversal queue.
31 if (currentNode.right) {
32 nodeQueue.enqueue(currentNode.right);
33 }
34 }
35 }
Complexities
Time Complexity
The breadth-first search algorithm that was implemented above visits every node of a tree exactly
once. This is achieved by using a Queue instance. Since a tree can’t have circular connections (i.e.
when two nodes share the same child) we can say that we put every node to the queue only once.
And then extract each node from the queue only once by visiting it. Therefore we may say that the
time complexity of tree traversal using breadth-first approach is O(|N|), where |N| is total number
of nodes in a tree.

In worst-case scenario we will have unbalanced tree that will look like a linked list (in case if each
node of a tree has only left (or only right) child). Since we use queue to store the list of nodes that
are going to be visited this queue in worst-case scenario may contain all the nodes of a tree. Thus,
in worst-case scenario, we will need O(|N|) auxiliary memory space for our queue instance.
Problems Examples
• Permutation Sequence¹¹⁷
• Letter Case Permutation¹¹⁸
• Next Permutation¹¹⁹
Quiz
Q1: What is time complexity of the breadth-first search algorithm?
Q2: How the function depthFirstSearch() may be adjusted in order to support more than two
children for tree node?
References
Breadth-first search (BFS) on Wikipedia https://en.wikipedia.org/wiki/Breadth-first_search¹²⁰
Tree BFS example and test cases https://github.com/trekhleb/javascript-algorithms/tree/master/sr-
c/algorithms/tree/breadth-first-search¹²¹
¹¹⁷https://leetcode.com/problems/permutation-sequence/
¹¹⁸https://leetcode.com/problems/letter-case-permutation/
¹¹⁹https://leetcode.com/problems/next-permutation/
¹²⁰https://en.wikipedia.org/wiki/Breadth-first_search
¹²¹https://github.com/trekhleb/javascript-algorithms/tree/master/src/algorithms/tree/breadth-first-search
Graphs. Depth-First Search.
The Task
Imagine that we have a graph and we want to:
• check whether specific vertex exists in a graph (i.e. perform search operation for vertex “John”),
• check whether two nodes are connected (i.e. to find connection path between vertex “Bill” and
“Tom”),
• do vertices inventory by visiting all the vertices of a graph (i.e. to convert a graph to an array),
• etc (for more usecases see Application section below).
To accomplish all these tasks we need to to utilize a graph traversal algorithm. In other words, we
need to write an algorithm that will visit all the vertices of a graph that can be reached from a
specific vertex.
Graphs. Depth-First Search. 240
The Algorithm
One of the ways to solve the tasks mentioned above is depth-first search algorithm.
Depth-first search (DFS) is an algorithm for traversing or searching a tree or graph data structures.
The idea of the algorithm is to start at some specific node (or vertex) and explore as far as possible
along each branch before backtracking.
Here, the word backtracking means that when we’re moving deeper along the branch of a graph
and there are no more nodes along the current path to explore, then we’re moving backwards to
find another closest unexplored sub-branch (vertex) and then explore it.
The illustration below shows the sequence of graph traversal using depth-first search algorithm.
To get better understanding of depth-first search algorithm you can think of the process of walking
through the maze. Graph is a maze, graph edges are paths and graph vertices are intersections of
maze paths. When you get to an intersection, you mark the path you intend to travel to avoid taking
that same path in the future. Once you come to the intersection where all paths are marked as
“visited”, you go back to the previous intersection (backtracking) and choose the next path that has
not been visited yet. This process is repeated until you reach the exit of the maze.
This algorithm may be implemented recursively using stack (see one of the previous chapters of
this book about the Stack data structure). The idea of using the stack for depth-first search is the
following:
1. Pick the starting vertex and add it to the stack.

2. Pop the next vertex from the stack and “visit” it (i.e. call a callback or do some operations over
the vertex).
3. Push all unvisited neighbors of current vertex into the stack.
4. Go to step 2 (repeat this process until the stack is empty).
A Hash map data structure may also be used for efficient checking if the vertex has been already
visited or not. The vertex must have a unique key value that is used as a key for a hash map.
Application
Depth-first search algorithm serves as a building block for other graph related algorithms such as:
• finding connected components,

• topological sorting,
• finding the bridges of a graph,
• detecting cycle in a graph,
• path Finding,
• solving puzzles with only one solution (such as mazes),
• maze generation may use a randomized depth-first search
• etc.
Usage Example
Before implementing the depthFirstSearch() function let’s see how we’re going to use it and what
parameters we’re going to pass into it. The depthFirstSearch() function needs to know what graph
to traverse and from starting vertex to begin a traversal process. Also we want to pass a callback
to the function that will be called every time a new vertex is being visited. Having all that in mind
let’s create really simple social network using Graph class from the “Graphs” chapter of this book
and then let’s traverse it.
Here is a structure of the network we’re going to implement.
Let’s start from Bill and traverse all users that connected to Bill directly or via another users. First
we will create a graph (analogy of social network). Next we will create graph vertices (analogy of
registering the users in the network). Then we will create graph edges between graph nodes (analogy
of adding users to friend lists). Then we will traverse all users starting from Bill.
24-graph-depth-first-search/example.js
2 import Graph from '../08-graph/Graph';
3 import GraphVertex from '../08-graph/GraphVertex';
4 import GraphEdge from '../08-graph/GraphEdge';
5 import { depthFirstSearch } from './depthFirstSearch';
6
7 // Create a demo-version of our overly-simplified social network.
8 const socialNetwork = new Graph();
9
10 // Let's register several users in our network.
12 const alice = new GraphVertex('Alice');
14 const kate = new GraphVertex('Kate');
15 const ann = new GraphVertex('Ann');
16 const tom = new GraphVertex('Tom');
17 const sam = new GraphVertex('Sam');
18
19 // Now let's establish friendship connections between the users of our network.
20 socialNetwork
21 .addEdge(new GraphEdge(bill, alice))
22 .addEdge(new GraphEdge(bill, john))
23 .addEdge(new GraphEdge(bill, kate))
24 .addEdge(new GraphEdge(alice, ann))
25 .addEdge(new GraphEdge(ann, sam))
26 .addEdge(new GraphEdge(john, ann))
27 .addEdge(new GraphEdge(kate, tom));
28
29 // Now let's traverse the network in depth-first manner staring from Bill
30 // and add all users we will encounter to the userVisits array.
31 const userVisits = [];
32 depthFirstSearch(socialNetwork, bill, (user) => {
33 userVisits.push(user);
34 });
35
36 // Now let's see in what order the users have been traversed.
38 console.log(userVisits);
39 /*
41 [bill, alice, ann, sam, john, kate, tom]
42 */
Implementation
Function arguments
As described above the depthFirstSearch() function will accept three parameters:
• graph - graph that is going to be traversed (instance of Graph class),

• startVertex - vertex that we will use as a starting point (instance of GraphVertex class),
• enterVertexCallback - callback that will be called on every vertex visit.
Visited vertices memorization

When we’ve been implemented the depth-first search algorithm for trees we didn’t care about the
case when the same node could be visited twice during the traversal from different parents. It is
because the tree node by definition can’t have two or more parents. Otherwise it would become a
graph. So we have only one way to enter the node in a tree. But when we’re dealing with graphs
we need to remember that the same graph vertex may be visited many times from many neighbor
vertices. This way the same vertex may be added to our traversal stack many times. That makes
the whole traversal process inefficient and may even cause an endless loop. That’s why we need to
mark all visited vertices while traversing a graph to avoid visiting the same vertex over and over
again. We will use visitedVertices object for that. The key value of visited graph vertices will be
the keys of visitedVertices object. This will help us efficiently check if the node has been already
visited in O(1) time.
Recursive closure
Inside our depthFirstSearch() function let’s implement a recursive depthFirstSearchRecursive()
function. This function will contain the main traversal logic. The depthFirstSearchRecursive()
function is a closure, which means that it has an access to all variables and parameters of the
depthFirstSearch() function. Therefore we may pass only one argument to it and it will be a vertex
that is currently going to be traversed (currentVertex parameter).
What depthFirstSearchRecursive() does is simply calling the enterVertexCallback() function,
getting all unvisited neighbors of currentVertex, marking them as visited and calling itself again
for every unvisited neighbor of the currentVertex.
Implementation example
24-graph-depth-first-search/depthFirstSearch.js
1 /**
2 * Traverse the graph in depth-first manner.
3 *
4 * @param {Graph} graph - Graph that is going to be traversed.
5 * @param {GraphVertex} startVertex - Vertex that we will use as a starting point.
6 * @param {function} enterVertexCallback - Callback that will be called on every ver\
7 tex visit.
8 */
9 export function depthFirstSearch(graph, startVertex, enterVertexCallback) {
10 // In order to prevent the visiting of the same vertex twice
11 // we need to memorize all visited vertices.
12 const visitedVertices = {};
13
14 // Since we're going to visit startVertex first let's add it to the visited list.
15 visitedVertices[startVertex.getKey()] = true;
16
17 /**
18 * Recursive implementation of depth-first search.
19 *
20 * @param {GraphVertex} currentVertex - Graph vertex that is currently being trave\
21 rsed.
22 */
23 const depthFirstSearchRecursive = (currentVertex) => {
24 // Call the callback to notify subscribers about the entering a new vertex.
25 enterVertexCallback(currentVertex);
26
27 // Iterate over every neighbor of currently visited vertex.
28 currentVertex
29 .getNeighbors()
30 .forEach((nextVertex) => {
31 // In case if the neighbor vertex was not visited before let's visit it.
32 if (!visitedVertices[nextVertex.getKey()]) {
33 // Memorize current neighbor to avoid visiting it again in the feature.
34 visitedVertices[nextVertex.getKey()] = true;
35 // Visit the next neighbor vertex.
36 depthFirstSearchRecursive(nextVertex);
37 }
38 });
39 };
40
41 // Start graph traversal by calling recursive function.
42 depthFirstSearchRecursive(startVertex);
43 }
Usage of Stack
It was mentioned above the depth-first search algorithm may be implemented using Stack data-
structure. We may use Stack for implementing DFS algorithm explicitly or implicitly.
For explicit stack usage we need to import a Stack class implementation from “Stack” chapter of the
book (or any other stack implementation including a plain JavaScript array) and use stack instance
for storing all the vertices we’re going to traverse. While stack is not empty we need to pop() the
last added vertex from it, get all unvisited neighbors of the vertex, push() them to the stack and
repeat. We don’t need recursion for explicit stack usage, the while loop may be a good fit instead.
But in depthFirstSearch() function above we’ve used stack implicitly using recursion. Even though
we didn’t import the Stack class explicitly we were still dealing with the stack in form of recursive
function call stack. Every time the depthFirstSearchRecursive() function calls itself the JavaScript
engine push current function variables (state) into the memory stack.
Complexities
Time Complexity
Time complexity for depthFirstSearch() function is mostly defined by time complexity of its inter-
nal recursive function depthFirstSearchRecursive(). Because all operations outside of depthFirstSearchRecursive
are constant time operations. Inside the depthFirstSearchRecursive() function we do the follow-
ing:
• Getting all neighbors of the current vertex. To do that we need to iterate over all edges that are
connected to the current vertex. Thus the number of edges will impact the performance of this
step.
• Iterating over all unvisited neighbors of the current vertex. The more vertices we would have,
the more iterations will be needed for this step.
Taking all that into account we may conclude that time complexity of graph depth-first search
algorithm is O(|V| + |E|), where |V| is total number of vertices and |E| is total number of edges
in the graph.

In current implementation of depthFirstSearch() function we’re using recursion. Every next
recursion call will increase the function call stack. In worst-case scenario when graph would look
like linked list the function depthFirstSearchRecursive() will increase the call stack proportionally
to the number of vertices |V| in the graph. Additionally we’re keeping track of visited vertices using
the visitedVertices map. This map will require the space to store all |V| number of vertices.
Taking all that into account we may conclude that auxiliary space complexity of the current DFS
implementation is O(|V|), where |V| is total number of vertices in the graph.
Problems Examples
• Clone Graph¹²²
• Is Graph Bipartite?¹²³
• Number of Islands¹²⁴
Quiz
Q1: What is time complexity of the depth-first search algorithm for graphs?
Q2: What data-structure fits best for implementing depth-first search algorithm for graphs: Stack or
Queue?
References
Depth-first search (DFS) on Wikipedia https://en.wikipedia.org/wiki/Depth-first_search¹²⁵
Depth-first search visualization https://www.cs.usfca.edu/∼galles/visualization/DFS.html¹²⁶
Graph DFS example and test cases https://github.com/trekhleb/javascript-algorithms/tree/master/s-
rc/algorithms/graph/depth-first-search¹²⁷
¹²²https://leetcode.com/problems/clone-graph/
¹²³https://leetcode.com/problems/is-graph-bipartite/
¹²⁴https://leetcode.com/problems/number-of-islands/
¹²⁵https://en.wikipedia.org/wiki/Depth-first_search
¹²⁶https://www.cs.usfca.edu/~galles/visualization/DFS.html
¹²⁷https://github.com/trekhleb/javascript-algorithms/tree/master/src/algorithms/graph/depth-first-search
Graphs. Breadth-First Search.
The Task
Imagine that we have a graph and we want to:
• check whether specific vertex exists in a graph (i.e. perform search operation for vertex “John”),
• do vertices inventory by visiting all the vertices of a graph (i.e. to convert a graph to an array).
• etc (for more usecases see Application section below).
To accomplish all these tasks we need to find an algorithm of graph traversal. In other words we
need to find an algorithm of visiting all the vertices of a graph that can be reached from some specific
vertex.
The Algorithm
One of the ways to solve the tasks mentioned above is breadth-first search algorithm.
Graphs. Breadth-First Search. 249
Breadth-first search (BFS) is an algorithm for traversing or searching tree or graph data structures.
It starts at some arbitrary vertex of a graph, and explores all of the neighbor vertices at the present
depth prior to moving on to the nodes at the next depth level.
It uses the opposite strategy as depth-first search, which instead explores the highest-depth nodes
first before being forced to backtrack and expand shallower nodes.
The illustration below shows the sequence of graph traversal using breadth-first search algorithm.
This algorithm may be implemented using queue (see one of the previous chapters of this book
about the Queue data structure). Queue data structure will give us desired FIFO (First-In-First-Out)
sequence of graph vertex visiting so that we would go wide instead of deep. FIFO sequence basically
means that once we added all neighbors of current vertex to the Queue we then will go from neighbor
to neighbor first and visit the first neighbor, then the second neighbor and so on before moving
deeper to the next level of neighbors’ children. This is the opposite to Stack’s LIFO (Last-In-First-
Out) approach that we’ve used for implementing depth-first traversal where we were visiting the
last node that has been added to the stack every time. And by doing that we were moving not from
the neighbor to neighbor but from the parent to child first.
The idea of using the queue for breadth-first search is the following:
1. Pick the starting vertex and put it to the visiting queue.

2. Fetch next vertex from the queue and “visit” it (i.e. call a callback or do some operations over
the vertex).
3. Add all unvisited neighbors of current vertex to the queue to visit them later.
4. Go to step 2 (repeat this process until the queue is empty).
Hash map data structure may be used for efficient checking if the vertex has been already visited
or not. The vertex must have a unique key value that is used as a key for a hash map.
Application
Breadth-first search algorithm is being used to find a shortest path between to nodes in graph. And
since the graph can represent real-world data such as roads networks, social networks, computer
networks there are many BFS algorithm applications may be found:
• GPS Navigation - to find a shortest path (combination of edges in the map graph) from your
location (the vertex in the map graph) to specific city (another vertex in the map graph).
• Computer Networks - traverse Peer-to-Peer (P2P) network by P2P client (i.e. Torrent client)
to find the “closest” hosts with required file.
• Social Networks - to find people with a given distance from other person.
• Web Crawlers - to traverse the content of the web-site by following the links from the home-
page and then from internal pages. By using BFS it is possible to setup a desired depth of
web-site crawling.
Breadth-first search algorithm also serves as a building block for other graph related algorithms
such as Cheney’s algorithm (copying garbage collection), Cuthillâ€“McKee mesh numbering,
Fordâ€“Fulkerson method for computing the maximum flow in a flow network and others.
Usage Example
Before implementing the breadthFirstSearch() function let’s see how we’re going to use it and
what parameters we’re going to pass into it. The breadthFirstSearch() function needs to know
what graph to traverse and from starting vertex to begin a traversal process. Also we want to pass a
callback to the function that will be called every time a new vertex is being visited. Having all that
in mind let’s create really simple social network using Graph class from the “Graphs” chapter of this
book and then let’s traverse it.
Here is a structure of the network we’re going to implement.
Let’s start from Bill and traverse all users that connected to Bill directly or via another users. First
we will create a graph (analogy of social network). Next we will create graph vertices (analogy of
registering the users in the network). Then we will create graph edges between graph nodes (analogy
of adding users to friend lists). Then we will traverse all users starting from Bill.
25-graph-breadth-first-search/example.js
5 import { breadthFirstSearch } from './breadthFirstSearch';
6
7 // Create a demo-version of our overly-simplified social network.
8 const socialNetwork = new Graph();
9
10 // Let's register several users in our network.
12 const alice = new GraphVertex('Alice');
14 const kate = new GraphVertex('Kate');
15 const ann = new GraphVertex('Ann');
16 const tom = new GraphVertex('Tom');
17 const sam = new GraphVertex('Sam');
18
19 // Now let's establish friendship connections between the users of our network.
20 socialNetwork
21 .addEdge(new GraphEdge(bill, alice))
22 .addEdge(new GraphEdge(bill, john))
23 .addEdge(new GraphEdge(bill, kate))
24 .addEdge(new GraphEdge(alice, ann))
25 .addEdge(new GraphEdge(ann, sam))
26 .addEdge(new GraphEdge(john, ann))
27 .addEdge(new GraphEdge(kate, tom));
28
29 // Now let's traverse the network in breadth-first manner staring from Bill
30 // and add all users we will encounter to the userVisits array.
31 const userVisits = [];
32 breadthFirstSearch(socialNetwork, bill, (user) => {
33 userVisits.push(user);
34 });
35
36 // Now let's see in what order the users have been traversed.
38 console.log(userVisits);
39 /*
41 [bill, alice, john, kate, ann, tom, sam]
42 */
Implementation
Function arguments
As described above the breadthFirstSearch() function will accept three parameters:
• graph - graph that is going to be traversed (instance of Graph class),

• startVertex - vertex that we will use as a starting point (instance of GraphVertex class),
• enterVertexCallback - callback that will be called on every vertex visit.
Visited vertices memorization

When we’ve been implemented the breadth-first search algorithm for trees we didn’t care about the
case when the same node could be visited twice during the traversal from different parents. It is
because the tree node by definition can’t have two or more parents. Otherwise it would become a
graph. So we have only one way to enter the node in a tree. But when we’re dealing with graphs
we need to remember that the same graph vertex may be visited many times from many neighbor
vertices. This way the same vertex may be added to our traversal stack many times. That makes
the whole traversal process inefficient and may even cause an endless loop. That’s why we need to
mark all visited vertices while traversing a graph to avoid visiting the same vertex over and over
again. We will use visitedVertices object for that. The key value of visited graph vertices will be
the keys of visitedVertices object. This will help us efficiently check if the node has been already
visited in O(1) time.
Usage of Queue
As it was mentioned above in “The Algorithm” section we’re going to implement breadth-first
traversal using queues to store all the nodes that are supposed to be visited. We may use JavaScript
array as a simple Queue with unshift() method for adding elements to the queue and pop() method
for fetching elements from the queue. But since we’ve already implemented the Queue class earlier
in this book we’ll use it here.
Then using while loop we’re visiting all the vertices from the queue until the queue is empty. Every
time we visit a new vertex we call a enterVertexCallback() and add all unvisited neighbors of the
currentVertex to the vertexQueue.
25-graph-breadth-first-search/breadthFirstSearch.js
import { Queue } from '../03-queue/Queue';
Implementation example
We will start implementing breadthFirstSearch() with initial setup for the vertexQueue that will
hold all the vertices we’re going to visit and for the visitedVertices map that will hold keys of all
visited vertices to prevent them from visiting twice. Since we’re going to start graph traversal from
startVertex it should be added to both vertexQueue and visitedVertices first.
25-graph-breadth-first-search/breadthFirstSearch.js
4 /**
5 * Traverse the graph in breadth-first manner.
6 *
7 * @param {Graph} graph - Graph that is going to be traversed.
8 * @param {GraphVertex} startVertex - Vertex that we will use as a starting point.
9 * @param {function} enterVertexCallback - Callback that will be called on every ver\
10 tex visit.
11 */
12 export function breadthFirstSearch(graph, startVertex, enterVertexCallback) {
13 // Init vertex queue. Whenever we will meet a new vertex it will be
14 // added to the Queue for further exploration.
15 const vertexQueue = new Queue();
16
17 // Add startVertex to the queue since we're going to visit it first.
18 vertexQueue.enqueue(startVertex);
19
20 // In order to prevent the visiting of the same vertex twice
21 // we need to memorize all visited vertices.
23
24 // Since we're going to visit startVertex first let's add it to the visited list.
25 visitedVertices[startVertex.getKey()] = true;
26
27 // Traverse all vertices from the queue while it is not empty.
28 while (!vertexQueue.isEmpty()) {
29 // Get the next vertex from the queue.
30 const currentVertex = vertexQueue.dequeue();
31
32 // Call the callback to notify subscribers about the entering a new vertex.
33 enterVertexCallback(currentVertex);
34
35 // Add all neighbors to the queue for future traversals.
36 currentVertex.getNeighbors().forEach((nextVertex) => {
37 if (!visitedVertices[nextVertex.getKey()]) {
38 // Memorize current neighbor to avoid visiting it again in the feature.
39 visitedVertices[nextVertex.getKey()] = true;
40
41 // Add nextVertex to the queue for further visits.
42 vertexQueue.enqueue(nextVertex);
43 }
44 });
45 }
46 }
Complexities
Time Complexity
Time complexity of breadthFirstSearch() function is mostly defined by the while loop since all
lines outside of the while loop perform in O(1) time. Remember that by using visitedVertices
hash map we’re preventing the same vertex to be added to the vertexQueue more than once.
Thus we may assume that we will enter the while loop exactly |V| times, where |V| is a
number of connected vertices in the graph that we’re traversing. We also need to take into
account the fact that we’re iterating over every vertex edge to fetch vertex neighbors by calling
currentVertex.getNeighbors(). So our function will also visit all |E| number of graph edges. The
last thing we need to mention here is that inside the while loop we have Queue related enqueue()
and dequeue() operations that are being performed in O(1) time since we’re dealing with linked-list
based implementation of the Queue (for more details please take a look the chapter on Queue data
structure).
Having all that in mind we may conclude that time complexity of breadthFirstSearch() function
is O(|V| + |E|), where |E| is a number of graph edges and |V| is a number of vertices in the graph.

In current implementation of the breadthFirstSearch() function we use the following additional
variables:
• vertexQueue queue - keeps all |V| number of graph vertices,

• visitedVertices hash map - keeps all |V| number of graph vertex keys,
• currentVertex variable - keeps only one current vertex instance.
Since vertexQueue and visitedVertices variables are the most space consuming ones we may
conclude that auxiliary space complexity of the breadthFirstSearch() function is O(|V|), where
|V| is a number of vertices in the graph.
Problems Examples
• Clone Graph¹²⁸
• Is Graph Bipartite?¹²⁹
• Surrounded Regions¹³⁰
Quiz
Q1: What is time complexity of the breadth-first search algorithm for graphs?
Q2: What data-structure fits best for implementing breadth-first search algorithm for graphs: Stack
or Queue?
References
Breadth-first search (BFS) on Wikipedia https://en.wikipedia.org/wiki/Breadth-first_search¹³¹
Breadth-first search visualization https://www.cs.usfca.edu/∼galles/visualization/BFS.html¹³²
Graph BFS example and test cases https://github.com/trekhleb/javascript-algorithms/tree/master/s-
rc/algorithms/graph/breadth-first-search¹³³
¹²⁸https://leetcode.com/problems/clone-graph/
¹²⁹https://leetcode.com/problems/is-graph-bipartite/
¹³⁰https://leetcode.com/problems/surrounded-regions/
¹³¹https://en.wikipedia.org/wiki/Breadth-first_search
¹³²https://www.cs.usfca.edu/~galles/visualization/BFS.html
¹³³https://github.com/trekhleb/javascript-algorithms/tree/master/src/algorithms/graph/breadth-first-search
Dijkstra’s Algorithm
• Difficulty: hard
The Task
Given a graph and a source vertex in the graph, find shortest paths from source to all vertices in the
given graph.
Example
A good example of a graph is a map with various cities (vertices) and roads that connect them
(edges). If graph is directed it means the roads connecting two cities are one way roads. If the graph
is undirected then all roads are two way roads. Also a graph may be weighted. The weight of the
edge may be treated as the distance between two spots. The higher the weight the longer the distance
between two cities.
Perhaps we want to find the shortest distance to every location on the map from the Home vertex.
With this information we can, for example, find the shortest path from Home to the Office.
Dijkstra’s Algorithm 258
The Algorithm
Algorithm Definition
Dijkstra’s algorithm is an algorithm for finding the shortest paths between nodes in a graph,
which may represent, for example, road networks. It was conceived by computer scientist Edsger
W. Dijkstra in 1956.
Dijkstra’s original variant found the shortest path between two nodes, but a more common variant
finds shortest paths from the source node to all other nodes in the graph.
Algorithm Steps
The algorithm has the following steps:
1. Assign all vertices a tentative distance value: set it to 0 for our start vertex (since we’ve already
reached it) and Infinity for all other vertices (since at this point they seem unreachable to us).
2. Create a set of visited vertices. We will use it to avoid endless loops while traversing the
vertices.
3. Create a vertex visiting queue and put the starting vertex to this queue. We will use priority
queue with distance to the vertex as a priority factor (priority is explained in one of the previous
chapters). The shorter the vertex distance to the starting vertex the higher its priority.
4. Poll the next closest vertex from the queue (the vertex with smaller distance from the source).
Priority queue will allow us to do it.
5. Go through every unvisited neighbor of the current vertex. For each neighbor calculate
distanceToTheCurrentVertex + distanceFromCurrentVertexToNeighbor. If this distance is
less than neighbor’s tentative distance then replace it with this new value.
6. When we’re done considering all the neighbors of the current vertex, mark current vertex as
visited.
7. Go to step 4 and repeat it until the vertex visiting queue is empty.
Priority Queue Usage in the Algorithm

In Dijkstra’s algorithm every time we poll the next vertex to visit (step 4) it must be a node that was
not visited before and that has the shortest distance to the starting vertex.
But how can we do this type of polling based on distance?
If we use the Queue class that was implemented in one of the previous chapters the Queue will respect
FIFO (First-In-First-Out) ordering but it won’t take distance into account. For example, if we add
a vertex with distance 20 and then we add a vertex with distance 10 to the queue then the poll()
operation will return us the vertex with distance 20 since it was put to the queue first – not the
vertex with shorter distance 10. So using a Queue won’t work.
Another approach may be to store vertex visiting queue as an array and every time we need to poll
the next vertex from it we may traverse entire array and find/return the next vertex with shortest
distance. This approach works but in O(n) time.
Can we make polling faster?
Yes, it is possible by using priority queue. In this case the time complexity of polling will be O(log(n))
which make the algorithm work faster.
You may find full description of PriorityQueue class with implementation example in “Priority
Queue” chapter of this book.
Step by Step Example

Let’s take the graph above and try to find the shortest paths to all vertices from Home vertex applying
the Dijkstra’s algorithm.
Step 1
Init all vertices with tentative distances. The startingVertex (in our case it is Home vertex) will have
tentative distance 0 meaning that there are no effort required to get to Home since we’re already there.
All other vertices have their tentative distance being set to Infinity since we don’t know how far
are they from Home and we don’t even know if they are reachable at all.
Let’s also add Home vertex to the vertices priority queue to start traversing the graph from it.
Step 2
Poll the next vertex from priority queue. Remember that every time we poll the vertex from priority
queue it will return us the vertex with shortest distance to the startingVertex. In our case there is
only one node in priority queue and it is Home vertex. So the current vertex is Home.
Get all neighbors of Home and set the distance to them using formula distanceToHome + distanceFromHomeToNeighbor
In our case for vertex A we’re setting the distance 0 + 5 = 5, for vertex D we’re setting the distance
0 + 8 = 8. This means that for now the shortest distance ot A is 5 and the shortest distance to D is 8.
Push vertices A and D to the priority queue.

Once we’ve traversed all neighbors of Home vertex we may mark it as visited (you’ll see that Home
node is marked as black one on the next illustration, that means it is visited).
Step 3
Get the next unvisited vertex from our vertex priority queue that is closer to the Home. In our case
we have only two vertices in a queue and they are A and D. Since A vertex is closer to the Home vertex
(distance is 5) we visit it first.
Get all neighbors of A that are not visited yet. The neighbors are vertex B and vertex C. Let’s calculate
the shortest path to neighbors using formula distanceToA + distanceFromAtoNeighbor. For vertex
B the distance is calculated as 5 + 9 = 14. For vertex C the distance is calculated as 5 + 3 = 8.
Add new neighbors B and C to the vertex priority queue.

Since all neighbors of A have been visited we mark A as visited (it will become black on the next
illustration).
Step 4
Get the next unvisited vertex from our vertex priority queue that is closer to the Home. In our case
we have only three vertices in a queue and they are D, C and B. Since D and C vertices are closest to
Home (distance is 8) let’s peak the first vertex that were added to the queue among them. It will be
vertex D.
Get all neighbors of D (vertices C and E) and let’s calculate the shortest distances to them. For C we
will have 8 + 4 = 12, for 8 + 6 = 14. Notice that vertex C already has a calculated distance to it. If
we would go to vertex C from the vertex A then the distance to vertex C from Home would be 8. Since
the overall distance to vertex C from Home via vertex D is longer than the one we’ve found before (12
> 8) then we discard the distance we’ve just found (via vertex D) in favor of the shortest path variant
(via vertex A). So the distance for vertex C stays unchanged at this point.
Add new neighbors E to the vertex priority queue. There is no need to add C to the queue since it is
already there.
Mark vertex D as visited.
Step 5
Get the next vertex from vertices priority queue that is closest to Home. In our case it is vertex C.
Get all unvisited neighbors of C (in our case it is B and E). Calculate the shortest distances to B and E
from C. For vertex B the distance will be calculated as 8 + 5 = 13, for E the distance is 8 + 2 = 10.
Notice that we’ve just found a shorter path to B. Previously the distance to B from Home via A was
equal to 14 but now we’ve figured out that the distance from Home to B via C is shorter and it equals
to 13. In this case we need to update the distance to B to 13 and we also need to update the priority
for vertex B in our priority queue.
Add all unvisited neighbors of C to priority queue.
Mark vertex C as visited.
Step 6
Get the next vertex from vertices priority queue that is closest to Home. In our case it is vertex E.
Get all unvisited neighbors of E and update their distances. In our case there is only one unvisited
neighbor Office and the distance to it is 10 + 4 = 14.
Add Office vertex to priority queue.
Mark vertex E as visited.
Step 7
Get the next vertex from vertices priority queue that is closest to Home. In our case it is vertex B.
Get all unvisited neighbors of B and update their distances. In our case there is only one unvisited
neighbor Office and the distance to it is 13 + 7 = 20. Since it is bigger than the distance to Office
that we’ve just found in a previous step (20 > 14) we ignore it.
Mark vertex B as visited.
Step 8
Finally we’ve got to the last node in a queue and this node doesn’t have unvisited neighbors. This
means that we may stop and that we’ve just found a shortest paths to all vertices from Home vertex.
Application
Dijkstra’s algorithm may be applied to the following areas:
• In GIS (Geographic Information Systems) to find the shortest paths between geographical
spots on the map (traffic and distance may be represented as graph edges weights).
• In Computer Networks (eg, a peer-to-peer application), to find the shortest paths between
machine A and B.
Usage Example
Before we will start implementing the dijkstra() function let’s imagine that we already have it
implemented and see how we want it to work and what interface (input parameters and output
values) it should have.
Let’s take a graph from “The Task” section above and try to find the shortest paths to all vertices
from Home vertex and the shortest path from Home to Office in particular.
When calling dijkstra() function we would want to pass the graph we’re going to traverse along
with the starting vertex startVertex.
This function should return us distances object with shortest paths to all vertices from the
startVertex. This object may be a map with vertices keys as a keys and with shortest distances
as values. For example:
{
Home: 0,
A: 4,
Office: 10,
...
}
This would mean that the shortest distance to the node with key Office from startVertex is 10.
The shortest distance to the node with key A from startVertex is 4. The shortest distance to Home
in this case is zero because our startVertex has key Home and we don’t need to do anything to get
there since we’re already there.
But distances map will not give us an understanding of what the shortest path to the Office is. We
know that it will take us 10 distance units (let’s say miles) to get there but we don’t know which
path to take. Therefore we need our dijkstra() function to also return us another object that we
call previousVertices. The idea of this object is that it is also a map with the keys that match the
graph vertices keys but instead of distances this object contains the links to previous vertex that
forms the shortest path. For example:
{
Office: spotE,
E: spotC,
C: spotA,
A: spotHome,
Home: null,
}
In the example above spotE, spotC, spotA, spotHome are instances of GraphVertex class and Office,
E, C, A, Home are just string key of each vertex of the graph.
26-dijkstra/example.js
5 import dijkstra from './dijkstra';
6
7 // Let's create the spots on our imaginary map.
8 const spotHome = new GraphVertex('Home');
9 const spotA = new GraphVertex('A');
10 const spotB = new GraphVertex('B');
11 const spotC = new GraphVertex('C');
12 const spotD = new GraphVertex('D');
13 const spotE = new GraphVertex('E');
14 const spotOffice = new GraphVertex('Office');
15
16 // Now let's connect those spots with the roads of certain length.
17 const roadHomeA = new GraphEdge(spotHome, spotA, 5);
18 const roadHomeD = new GraphEdge(spotHome, spotD, 8);
19 const roadAB = new GraphEdge(spotA, spotB, 9);
20 const roadAC = new GraphEdge(spotA, spotC, 3);
21 const roadCB = new GraphEdge(spotC, spotB, 5);
22 const roadDC = new GraphEdge(spotD, spotC, 4);
23 const roadCE = new GraphEdge(spotC, spotE, 2);
24 const roadDE = new GraphEdge(spotD, spotE, 6);
25 const roadOfficeB = new GraphEdge(spotOffice, spotB, 7);
26 const roadOfficeE = new GraphEdge(spotOffice, spotE, 4);
27
28 // We will create two way roads that means that our graph will be undirected.
29 const isDirected = false;

30 const graph = new Graph(isDirected);
31
32 // Now let's add the spots and roads we've just created to the graph.
33 // It is like putting all the information about the spots and roads on the map.
34 // Remember that we don't need to add both vertices and edges between them.
35 // We may add only edges and vertices will be added to the graph automatically
36 // (for more details on Graph implementation please refer to Graphs chapter).
37 graph
38 .addEdge(roadHomeA)
39 .addEdge(roadHomeD)
40 .addEdge(roadAB)
41 .addEdge(roadAC)
42 .addEdge(roadCB)
43 .addEdge(roadDC)
44 .addEdge(roadCE)
45 .addEdge(roadDE)
46 .addEdge(roadOfficeB)
47 .addEdge(roadOfficeE);
48
49 // Now we're ready to launch the Dijkstra algorithm and figure out what are the shor\
50 test
51 // paths from home to all other nodes and to the office in particular.
52 const { distances, previousVertices } = dijkstra(graph, spotHome);
53
54 /* eslint-disable no-console */
55
56 // Now let's print and see what are the shortest distances to every spot on the map.
57 console.log(distances);
58 /*
60
61 {
62 Home: 0,
63 A: 5,
64 B: 13,
65 C: 8,
66 D: 8,
67 E: 10,
68 Office: 14,
69 }
70
71 Which means that the shortest distance from Home to Office is 14.
72 */
73
74 // And now when we now the shortest distance from Home to the Office
75 // let's figure out what is the shortest path that has a shortest distance.
76 console.log(previousVertices.Office.getKey()); // -> 'E'
77 console.log(previousVertices.E.getKey()); // -> 'C'
78 console.log(previousVertices.C.getKey()); // -> 'A'
79 console.log(previousVertices.A.getKey()); // -> 'Home'
80
81 // So the shortest path from Home to Office is:
82 // Home -> A -> C -> E -> Office
The output of dijkstra() function will allow us find a shortest path from Home to Office as well as
to any other vertex in a graph.
Implementation
Importing Dependencies
Since we’re going to use PriorityQueue class to keep track of vertex visiting sequence let’s import
it as a dependency.
26-dijkstra/dijkstra.js
import { PriorityQueue } from '../07-priority-queue/PriorityQueue';
Describing Types
First let’s create a ShortestPaths type definition that we will use when describing what the
dijkstra() function will return.
/**
* @typedef {Object} ShortestPaths
* @property {Object} distances - shortest distances to all vertices
* @property {Object} previousVertices - shortest paths to all vertices.
*/
Implementing the dijkstra() function

Now we may use this type definition in dijkstra() JSDoc section to describe the input and the
output of the function. This is not mandatory to describe function’s JSDoc but it will help function
consumers to better understand what parameters does function accept and what it returns.
The dijkstra() function will accept the following parameters:
• graph - graph we’re going to traverse (instance of Graph class, see “Graphs” chapter for
implementation details).
• startVertex - graph vertex that we will use as a starting point for traversal (instance of
GraphVertex class, see “Graphs” chapter for implementation details).
The dijkstra() function first inits the helper variables:
• distances - this map is used to store the shortest distances to each vertex from the startVertex.
It may look like {A: 5, D: 8} meaning that the shortest distance from startVertex to A is 5
and to D is 8. at first we set the distance to all vertices to Infinity because we haven’t reached
them yet.
• visitedVertices - this map is used to keep track of visited vertices and to avoid visiting the
same vertex more than once. It may look like {A: true, D: true} meaning that vertices A and
D have been already visited.
• previousVertices - this map will store the information about which vertex was the previous
one to the current one in order to form a shortest path. For example this map may look like {A:
B, B: D, D: StartVertex} meaning that the shortest path to A lays through B, the shortest
path to B lays through D and finally the shortest path to D is a direct path from StartVertex.
This map will allow us to restore the shortest paths to any connected vertex in the graph.
• queue - instance of PriorityQueue that we will use to keep track of which vertex need to be
visited next. Every time we need to make a decision about which vertex to visit next we need
to take unvisited vertex that is closest to the StartVertex.
The while loop of dijkstra() function mimics the algorithm steps that were described in the
“Algorithm Steps” section above:
1. Poll the next closest vertex from the queue (the vertex with smaller distance from the source)
using poll() method of PriorityQueue instance.
2. Go through every unvisited neighbor of the current vertex using getNeighbors() method
of Graph instance. For each neighbor we calculate distanceToNeighborFromCurrent. If this
distance is less than existingDistanceToNeighbor then replace it with this new value and
change the priority of the neighbor in the queue using changePriority() method.
3. Mark current vertex as visited.
4. Go to step 1 and repeat it until the vertex visiting queue is empty. The emptiness of the queue
is being checked by using isEmpty() method of PriorityQueue instance.
Here is an example of how dijstra() function may be implemented.

10 /**
11 * Implementation of Dijkstra algorithm of finding the shortest paths to graph nodes.
12 * @param {Graph} graph - graph we're going to traverse.
13 * @param {GraphVertex} startVertex - traversal start vertex.
14 * @return {ShortestPaths}
15 */
16 export default function dijkstra(graph, startVertex) {
17 // Init helper variables that we will need for Dijkstra algorithm.
18 const distances = {};
20 const previousVertices = {};
21 const queue = new PriorityQueue();
22
23 // Init all distances with infinity assuming that currently we can't reach
24 // any of the vertices except the start one.
25 graph.getAllVertices().forEach((vertex) => {
26 distances[vertex.getKey()] = Infinity;
27 previousVertices[vertex.getKey()] = null;
28 });
29
30 // We are already at the startVertex so the distance to it is zero.
31 distances[startVertex.getKey()] = 0;
32
33 // Init vertices queue.
34 queue.add(startVertex, distances[startVertex.getKey()]);
35
36 // Iterate over the priority queue of vertices until it is empty.
37 while (!queue.isEmpty()) {
38 // Fetch next closest vertex.
39 const currentVertex = queue.poll();
40
41 // Iterate over every unvisited neighbor of the current vertex.
42 currentVertex.getNeighbors().forEach((neighbor) => {
43 if (!visitedVertices[neighbor.getKey()]) {
44 const edgeWight = neighbor.edge.weight;
45 const existingDistanceToNeighbor = distances[neighbor.getKey()];
46 const distanceToNeighborFromCurrent = distances[currentVertex.getKey()] + ed\
47 geWight;
48
49 // If we've found shorter path to the neighbor - update it.
50 if (distanceToNeighborFromCurrent < existingDistanceToNeighbor) {
51 distances[neighbor.getKey()] = distanceToNeighborFromCurrent;
52
53 // Change priority of the neighbor in a queue since it might have became c\
54 loser.
55 if (queue.has(neighbor)) {
56 queue.changePriority(neighbor, distances[neighbor.getKey()]);
57 }
58
59 // Remember previous closest vertex.
60 previousVertices[neighbor.getKey()] = currentVertex;
61 }
62
63 // Add neighbor to the queue for further visiting.
64 if (!queue.has(neighbor)) {
65 queue.add(neighbor, distances[neighbor.getKey()]);
66 }
67 }
68 });
69
70 // Add current vertex to visited ones to avoid visiting it again later.
71 visitedVertices[currentVertex.getKey()] = true;
72 }
73
74 // Return the set of shortest distances and paths to all vertices.
75 return { distances, previousVertices };
76 }
Complexities
Let’s say that the graph we’re going to traverse has |V| vertices and |E| edges.
Time Complexity
In our dijkstra() function we first do initial setup and then iterate over the queue. Let’s analyze
these two steps.
Initial setup (creating queue, populating distances and previousVertices). The complexity of this
steps is O(|V|) since we’re iterating over all vertices in the graph.
Iterating over the queue. The code inside the while loop will be called |V| times since we’re visiting
every graph vertex exactly once because of tracking the visited vertices in visitedVertices map.
So we will use the |V| multiplier of time complexity of the code inside the while loop.
Let’s analyze the steps we make inside the while loop:
• Polling the next vertex from the queue. This operation uses poll() method of MinHeap class
which in turn has O(log(|V|)) time complexity (please see the “Heap” chapter for further
details of how the complexity of poll() method has been calculated).
• Iterating over unvisited neighbors of the current vertex. In worst case scenario when the
graph is fully connected and every vertex has |V| - 1 edges this step will take O(|E|) time.
– Updating vertex distances. This steps involves operations with distances map and thus
they have O(1) time complexity.
– Changing the vertex priority if needed. This is a MinHeap class related operation and it
is done in O(log(|V|)) time because we’re removing and adding element from the heap
tree and calling heapifyUp() and heapifyDown() methods along the way.
– Checking if vertex is already inside the queue. This operation is done using hash map
and thus it takes O(1) time.
– Adding vertex to the queue. This operation takes O(log(|V|)) because the heapifyUp()
method of MinHeap class is being called to rebuild a heap.
– Mark current vertex as visited. This is O(1) operation since we’re dealing with updating
the visitedVertices map hear.
If we will take all the steps complexities that are mentioned above we’ll get the final time complexity
of the dijkstra() function which is equal to O(|E| * |V| * log(|V|)), where |V| is the number
of graph vertices and |E| is the number of graph edges.

We use auxiliary data structures like maps (distances, visitedVertices, previousVertices vari-
ables), priority queue (queue variable) to make dijkstra() function work. All of these data structures
are directly proportional in size to the number of vertices in the graph. Thus we may assume that
auxiliary space complexity is O(|V|).
Problems Examples
• Network Delay Time¹³⁴

• Cheapest Flights Within K Stops¹³⁵
References
Dijkstra’s algorithm on Wikipedia https://en.wikipedia.org/wiki/Dijkstra%27s_algorithm¹³⁶
Dijkstra’s algorithm on YouTube https://www.youtube.com/watch?v=gdmfOwyQlcI¹³⁷
Dijkstra’s algorithm example and test cases https://github.com/trekhleb/javascript-algorithms/tree/-
master/src/algorithms/graph/dijkstra¹³⁸
¹³⁴https://leetcode.com/problems/network-delay-time/
¹³⁵https://leetcode.com/problems/cheapest-flights-within-k-stops/
¹³⁶https://en.wikipedia.org/wiki/Dijkstra%27s_algorithm
¹³⁷https://www.youtube.com/watch?v=gdmfOwyQlcI
¹³⁸https://github.com/trekhleb/javascript-algorithms/tree/master/src/algorithms/graph/dijkstra
Appendix A: Quiz Answers
Algorithms and Their Complexities
• Q1: No, time complexity shows the rate (proportion) of function execution time and function
input
• Q2: O(log(n)) is faster. For example O(1000) = 1000, but O(log(1000)) = 9
• Q3: No it is not true. Sometimes it is a trade off and you make algorithm work faster by
consuming more memory.
Linked List
• Q1: It’s a reference to the next object in the linked list

• Q2: deleteTail removes the last item in the linked list, deleteHead removes the first
Queue
• Q1: A Queue is first-in-first-out

• Q2: Basic operations for Queue are peek(), enqueue() and dequeue()
• Q3: The time complexity is O(1)
Stack
• Q1: A Queue is first-in-last-out

• Q2: Basic operations for STack are peek(), push() and pop()
• Q3: The time complexity is O(1)
Hash Table
• Q1: The main advantage of hash table is fast lookup

• Q2: The time complexity is O(1) (with good hash function and sufficient buckets number)
• Q3: Collision is a situation when we try to save two values with different keys to the same
bucket either because of not ideal hash function or because of lack of available buckets
• Q4: One of the ways of handling the collision is Separate Chaining when we use links list to
store information in each bucket
Binary Search Tree (BST)

Appendix A: Quiz Answers 275
• Q1: No. The root node by definition is the node in a tree that doesn’t have parent.
• Q2: Yes, it is allowed. Nodes in binary tree may have from zero to two children.
• Q3: The balanced binary search tree will have faster O(log(n)) lookup. If the tree is unbalanced
then in worst case (pathological tree) it will act like a linked list with O(n) lookup time
Binary Heap
• Q1: “Shape property” and “Heap property”

• Q2: We need to move the last element of a heap (bottom left element of a tree) to the head and
then do heapifying down operation
• Q3: O(log(n))
Priority Queue
• Q1: We would need to use MaxHeap instead of MinHeap.

• Q2: O(log(n))
Graphs
• Q1: Yes. The tree is a special case of directed connected acyclic graph.
• Q2: Yes. In this case we will have disconnected graph that consists of only two vertices
(vertices).
• Q3: There are two common ways of representing the graph: adjacency list and adjacency matrix
Bit Manipulation
• Q1: 111011
• Q2: 000101 & 111110 = 000100
Factorial
• Q1: 7
Primality Test
• Q1: No, 32/2 is remainder 0

• Q2: Yes, 31/2 is remainder 1. The square root of 31 is 5.56776436283 – 31/3 = 10.3, 31/4 =
7.75, 31/5 = 6.2
Is a power of two
• Q1: Yes, 32/2 = 16, 16/2 = 8, 8/2 = 4, 4/2 = 2, 2/2 = 1

• Q2: No, 01111 & 11110 = 01110
Search. Linear Search.
• Q1: Time complexity is O(n) for worst and average cases.

• Q2: No. Because of O(n) time complexity the common practice is to apply the algorithm for
small sized lists
Search. Binary Search.
• Q1: No. By its nature the binary search works only with sorted array. Otherwise we couldn’t
abandon array halves on each step.
• Q2: Time complexity is O(log(n)). It is caused by the fact that we abandon the half of the array
on each iteration.
Sets. Cartesian Product.
• Q1: It has 4 elements: ['ac', 'ad', 'bc', 'bd']. The formula is |A| * |B|.
• Q2: Yes, it is possible. If we have four sets A, B, C and D then each element of cartesian product
set would consist of four elements (a, b, c, d).
• Q3: The time complexity is O(|A|*|B|).
Sets. Power Set.
• Q1: Yes, empty set is a part of power set.

• Q2: It has 24=16 elements, where 4 is a size of original set {a, b, c, d}.
Sets. Permutations.
• Q1: Yes, the order matters. This is distinguishing characteristic of permutations comparing to
combinations.
• Q2: We will have 4! / (4 - 2)! = 4! / 2! = 24 / 2 = 12 permutations.
Sets. Combinations.
• Q1: No, the order doesn’t matter. This is distinguishing characteristic of combinations compar-
ing to permutations.
• Q2: We will have 4! / (2! * (4 - 2)!) = 4! / (2! * 2!) = 24 / 4 = 6 combinations.
Sorting: Quicksort
• Q1: The average time complexity of Quicksort is O(n * log(n)).

• Q2: We just need to add the elements that are bigger than the pivot element to the left array
(instead of the right one) and the elements that are smaller than the pivot element need to be
added to the right array (instead of the left one).
Trees. Depth-First Search.
• Q1: The time complexity of DFS is O(|N|), where |N| is total number of nodes in a tree.
• Q2: Yes, we can avoid recursion by using the Stack class implemented earlier. In this case every
time we visit the node, we need to put all its children (not just two) on the stack. And then
fetch the next node to traverse from the top of the stack.
Trees. Breadth-First Search.
• Q1: The time complexity of BFS is O(|N|), where |N| is total number of nodes in a tree.
• Q2: Instead of adding just ‘left’ and ‘right’ child to the queue inside the while loop we need to
fetch all children of the node and add all of them to the queue.
Graphs. Depth-First Search.
• Q1: The time complexity of DFS for graphs is O(|V| + |E|), where |V| is total number of
vertices and |E| is total number of edges in the graph.
• Q2: Stack is a best fit for implementing DFS algorithm.
Graphs. Breadth-First Search.
• Q1: The time complexity of BFS for graphs is O(|V| + |E|), where |V| is total number of
vertices and |E| is total number of edges in the graph.
• Q2: Queue is a best fit for implementing BFS
Appendix B: Big O Times Comparison
The chart below contains functions that are commonly used in algorithms analysis, showing the
number of operations versus the number of input elements for each function.
The table below contains the list of some of the most common Big O notations and their performance
comparisons against different sizes of the input data. This will give you the feeling of how different
algorithms complexities (time and memory consumptions) may be.
Appendix B: Big O Times Comparison 279
Big O Notation Computations for Computations for Computations for

10 elements 100 elements 1000 elements
O(1) 1 1 1
O(log(n)) 3 6 9
O(n) 10 100 1000
O(n log(n)) 30 600 9000
O(n²) 100 10000 1000000
O(2ⁿ) 1024 1.26e+29 1.07e+301
O(n!) 3628800 9.3e+157 4.02e+2567
Appendix C: Data Structures
Operations Complexities
Common Data Structures Operations Complexities
Data Structure Access Search Insertion Deletion Comments

Array 1 n n n
Stack n n 1 1
Queue n n 1 1
Linked List n n 1 1
Hash Table - n n n In case of perfect
hash function
costs would be
O(1)
Binary Search n n n n In case of balanced
Tree tree costs would
be O(log(n))
B-Tree log(n) log(n) log(n) log(n)
Red-Black Tree log(n) log(n) log(n) log(n)
AVL Tree log(n) log(n) log(n) log(n)
Bloom Filter - 1 1 - False positives are
possible while
searching
Graph Operations Complexities
Implementation Add Vertex Remove Vertex Add Edge Remove Edge

Adjacency list O(1) O(E+V) O(1) O(E)
Adjacency O(V2) O(V2) O(1) O(1)
matrix
Where: V - number of vertices in graph E - number of edges in graph
Heap Operations Complexities

Appendix D: Array Sorting Algorithms
Complexities
Name Best Average Worst Memory Stable Comments
Bubble n n2
n2 1 Yes
sort
Insertion n n2
n2 1 Yes
sort
Selection n2
n2
n2 1 No
sort
Heap sort n log(n) n log(n) n log(n) 1 No
Merge n log(n) n log(n) n log(n) n Yes
sort
Quick n log(n) n log(n) n2log(n) No Quicksort
sort is usually
done
in-place
with
O(log(n))
stack
space
Shell sort n log(n) depends n (log(n))2
1 No
on gap
sequence
Counting n+r n+r n+r n+r Yes r - biggest
sort number in
array
Radix n*k n*k n*k n+k Yes k - length
sort of longest
key
Changelog
Revision 2 (11-25-2019)
Pre-release revision 2
Revision 1 (10-29-2019)
Initial pre-release version of the book

JavaScript Algorithms

Uploaded by

JavaScript Algorithms

Uploaded by

JavaScript Algorithms

The Web Developer’s Guide to Data Structures and Algorithms

Written by Oleksii Trekhleb and Sophia Shoemaker

Algorithms and Their Complexities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

Binary Search Tree (BST) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

Priority Queue . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105

Basic Operations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106

Bit Manipulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133

Fibonacci Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153

Primality Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158

Problems Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161

Is a power of two . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163

Search. Linear Search. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167

Search. Binary Search. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175

Sets. Cartesian Product. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181

Sets. Power Set. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186

Sets. Permutations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196

Sets. Combinations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207

Sorting: Quicksort . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217

Trees. Depth-First Search. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223

Problems Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229

Trees. Breadth-First Search. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231

Graphs. Depth-First Search. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239

Graphs. Breadth-First Search. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 248

Dijkstra’s Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 257

Appendix A: Quiz Answers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 274

Appendix B: Big O Times Comparison . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 278

Appendix C: Data Structures Operations Complexities . . . . . . . . . . . . . . . . . . . . . . . 280

Appendix D: Array Sorting Algorithms Complexities . . . . . . . . . . . . . . . . . . . . . . . . 281

EARLY RELEASE VERSION

Join Our Discord

Be notified of updates via Twitter

We’d love to hear from you!

Join Our Discord

How To Read This Book

1. Take a WALK to your home.

1. Call a CAB that will drive you home.

Time and Space Complexities

Big O Notation Definition

Big O Notation Explanation

Input: (20 * n) steps

One important rule to understand here:

Big O Notation Computations for Computations for Computations for

Big O Notation Examples

What would be its time and space complexities?

Time complexity: O(1)

Space complexity: O(1)

Time complexity: O(power)

Space complexity: O(1)

O(n) recursive example

Let’s write recursive factorial function:

Time complexity: O(number)

Space complexity: O(number)

Time complexity: O(letters.length^2)

Space complexity: O(letters.length^2)

Space complexity and auxiliary space complexity example

Space complexity = Input + Auxiliary space complexity

Space complexity: O(array.length)

Auxiliary space complexity: O(1)

Space complexity: O(array.length)

Auxiliary space complexity: O(array.length)

Linked list and its common operations

• prepend - add a node to the beginning of the list

Finally, we return the value that was added to our LinkedList

35 // If there is no head yet let's make new node a head.

44 const currentTail = this.tail;

Finally, we set this.tail to be the new LinkedListNode object.

61 let deletedNode = null;

70 let currentNode = this.head;

72 if (currentNode !== null) {

72 if (currentNode !== null) {

Linked list before delete method is called

delete method is called

Linked list after delete method is called

We have two scenarios for which we need to write code:

1. We only have one item in our LinkedList