C++ Initialization Story

A Guide Through All Initialization Options and Related

C++ Areas

Bartłomiej Filipek
This book is for sale at http://leanpub.com/cppinitbook

This version was published on 2023-06-19

This is a Leanpub book. Leanpub empowers authors and publishers with the Lean
Publishing process. Lean Publishing is the act of publishing an in-progress ebook using
lightweight tools and many iterations to get reader feedback, pivot until you have the right
book and build traction once you do.

© 2021 - 2023 Bartłomiej Filipek

For Wiola, Mikołaj and Feliks
Contents

About the Book . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . i

Why should you read this book? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . i
Learning objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . i
The structure of the book . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ii
Who is this book for? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iii
Prerequisites . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iii
Reader feedback & errata . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iii
Example code . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iv
Formatting & Layout . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iv

About the Author . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii

Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viii

Revision History . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix

1. Local Variables and Simple Types . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

Starting with simple types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
Setting values to zero . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
Initialization with aggregates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
Default data member initialization . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

2. Classes and Initialization With Constructors . . . . . . . . . . . . . . . . . . . . 11

A simple class type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
Basics of constructors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
Body of a constructor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
Adding constructors to DataPacket . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
Compiler-generated default constructors . . . . . . . . . . . . . . . . . . . . . . . . 24
Explicit constructors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
CONTENTS

Difference between direct and copy initialization . . . . . . . . . . . . . . . . . . . 27

Constructor summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

3. Copy and Move Constructors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

Copy constructor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
Move constructor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
Distinguishing from assignment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
Adding logging to constructors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
Trivial classes and user-declared/user-provided default constructors . . . . . . . . 37

4. Delegating and Inheriting Constructors . . . . . . . . . . . . . . . . . . . . . . . 38

Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
Inheritance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
Inheriting constructors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

5. Destructors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
Basics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
Objects allocated on the heap . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
Destructors and data members . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
Virtual destructors and polymorphism . . . . . . . . . . . . . . . . . . . . . . . . . . 43
Partially created objects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
Use Cases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
A compiler-generated destructor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

6. Initialization and Type Deduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

7. Quiz on Constructors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46

8. Non-Static Data Member Initialization . . . . . . . . . . . . . . . . . . . . . . . . 47

How it works . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
Investigation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
Other forms of NSDMI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
Copy constructor and NSDMI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
Move constructor and NSDMI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
C++14 changes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
C++20 changes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
Limitations of NSDMI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
NSDMI: Advantages and Disadvantages . . . . . . . . . . . . . . . . . . . . . . . . . 54
NSDMI summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
CONTENTS

NSDMI: Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55

9. Containers as Data Members . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

The basics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
Using std::initializer list . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
Example implementation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

10. Non-regular Data Members . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57

Constant non-static data members . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
References as data members . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
Pointers as data members . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
Moveable-only data members . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58

11. Inline Variables in C++17 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59

About static data members . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
Motivation for inline variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
Exercise for inline variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
Global inline variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
Constexpr and inline variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64

12. Aggregates and Designated Initializers in C++20 . . . . . . . . . . . . . . . . . . 65

Aggregates in C++20 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
The basics of Designated Initializers . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
Rules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
Advantages of designated initialization . . . . . . . . . . . . . . . . . . . . . . . . . 68
Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
The summary example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
Compiler support . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69

13. Techniques and Use Cases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70

Using explicit for strong types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
Best way to initialize string data members . . . . . . . . . . . . . . . . . . . . . . 74
The copy and swap idiom . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
Several initialization types in one class . . . . . . . . . . . . . . . . . . . . . . . . . 74
Vector like RAII object . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
Factory with self-registering types . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
CONTENTS

14. The Final Quiz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79

Appendix A - Quiz and Exercises Answers . . . . . . . . . . . . . . . . . . . . . . . . 81

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
About the Book
Initialization in C++ is a hot topic! The internet is full of discussions about best practices,
and there are even funny memes on that subject. The situation is not surprising, as there are
more than a dozen ways to initialize a simple integer value, complex rules for the auto-type
deduction, data members, and object lifetime nuances.
And here comes the book.
Throughout this text, you will learn practical options to initialize various categories of
variables and data members in Modern C++. More specifically, this text teaches multiple
types of initialization, constructors, non-static data member initialization, inline variables,
designated initializers, and more. Additionally, you’ll see the changes and new techniques
from C++11 to C++20 and lots of examples to round out your understanding.
The plan is to explain most (if not all) parts of initialization, learn lots of excellent C++
techniques, and see what happens under the hood.

Why should you read this book?

With Modern C++ (since C++11), we have many new features to streamline work and
simplify our code. One area of improvement is initialization. Modern C++ added new
initialization rules, trying to make it easy while keeping old behavior and compatibility
(mainly from the C language). Sometimes the rules might seem confusing and complex,
though, and even the ISO committee might need to correct some things along the way. The
book will help you navigate through those principles and understand this topic better. What’s
more, initialization is just one aspect of this text. You’ll learn all related topics around classes,
constructors, destructors, object lifetime, or even how the compiler processes data at start-up.

Learning objectives
The goal is to equip you with the following knowledge:
• Explain rules about object initialization, including regular variables, data members,
and non-local objects.
About the Book ii

• How to implement special member functions (constructors, destructors, copy/move

operations) and when they are helpful.
• How to efficiently initialize non-static data members using C++11 features like non-
static data member initialization, inheriting, and delegating constructors.
• How to streamline working with static variables and static data members with inline
variables from C++17.
• How to work with container-like members, non-copyable data members (like const
data members) or move-able only data members, or even lambdas.
• What is an aggregate, and how to create such objects with designated initializers from
C++20.

The structure of the book

The book contains 14 chapters in the following structure:
• Chapters 1 to 5 create a foundation for the rest of the book. They cover basic
initialization rules, constructors, destructors, and the basics of data members.
• Chapter 6 describes type deduction that can be used to declare objects: auto, decltype,
Almost Always Auto rule, structured bindings.
• Chapter 7 is a quiz with 10 questions from the first “part” of the book.
• Chapter 8 describes Non-static Data Member Initialization (NSDMI), a powerful
feature from C++11 that improves how we work with data members. At the end of
the chapter, you can solve a few exercises.
• Chapter 9 discusses how to initialize container-like data members.
• Chapter 10 contains information about non-regular data members and how to handle
them in a class. You’ll learn about const data members, unique_ptr as a data member,
and references.
• Chapter 11 describes static non-local variables, static objects, various storage duration
options, and inline variables from C++17 and constinit from C++20.
• Chapter 12 moves to C++20 and describes Designated Initializers, a handy feature
based on similar thing from the C language.
• Chapter 13 shows various techniques like passing strings into constructors, strong
typing, CRTP class counter, Copy and Swap idiom, and more.
• Chapter 14 is the final quiz with questions from the whole book.
About the Book iii

And there are two appendices:

• Appendix A - a handy guide about rules for compiler-generated special member
functions.
• Appendix B - answers to quizzes and exercises.

Who is this book for?

The book is intended for beginner/intermediate C++ programmers who want to learn various
aspects of initialization in Modern C++ (from C++11 to C++20).
You should know at least some of the basics of creating and using custom classes.
This text is also helpful for experienced programmers who know older C++ standards and
want to move into C++17/C++20.

Prerequisites
• You should have basic knowledge of C++ expressions and primitive types.
• You should be able to implement an elementary class with several data members. Know
how to create and manipulate objects of such a class in a basic way.

Reader feedback & errata

If you spot an error, a typo, a grammar mistake, or anything else (especially logical issues!)
that should be corrected, please send your feedback to bartek@cppstories.com or submit an
issue at github.com/fenbf/cppinitbook_public/issues¹.
Here’s the errata with the list of fixes:
www.cppstories.com/p/cppinitbook/²
Your feedback matters! Writing an honest review can help with the book promotion and the
quality of my further work. The book has a dedicated page at GoodReads. Please share your
feedback at:
C++ Initialization Story by Bartłomiej Filipek @Goodreads³.
Or write a review at Amazon if you get this book in print form.
¹https://github.com/fenbf/cppinitbook_public/issues
²https://www.cppstories.com/p/cppinitbook/
³https://www.goodreads.com/book/show/62606823-c-initialization-story
About the Book iv

Example code
You can find source code of all examples in this separate GitHub public repository.
https://github.com/fenbf/cppinitbook_public/tree/main/examples
You can browse individual files or download the whole branch:
https://github.com/fenbf/cppinitbook_public/archive/refs/heads/main.zip

Code license

The code for the book is available under the MIT License model.

Formatting & Layout

Code samples are presented in a monospaced font, similar to the following example:
Title Of the Example

#include <iostream>

int main() {
const std::string text { "Hello World" };
std::cout << text << '\n';
}

Or shorter snippets (without a title and sometimes include statements):

int foo() {
return std::clamp(100, 1000, 1001);
}

When available, you’ll also see a link to online compilers where you can play with the code.
For example:
About the Book v

Example title. Run @Compiler Explorer

#include <iostream>

int main() {
std::cout << "Hello World!";
}

You can click on the link in the title, and then it should open the website of a given online
compiler (in the above case, it’s Compiler Explorer). You can compile the sample, see the
output, and experiment with the code directly in your browser. Here’s a basic overview of
Compiler Explorer:

A Compiler Explorer layout used in the book

Snippets of longer programs might be shortened to present only the core mechanics. They
may lack some #include statements or have a “compressed” line. Click on the Online
Compiler link to see the full version of the program or see them in the public repository.

Recommendation for Compiler Explorer and C++ Reference

When executing the examples on Compiler Explorer, you may select a term (keyword, class,
function, container or other) and then right-click (or equivalent). A context menu will appear
and you can select Search on CppReference, which will take you to the C++ Reference
documentation of the corresponding browser language, if such C++ Reference version exists⁴.
⁴Thanks to Javier Estrada for suggesting this cool tip!
About the Book vi

Syntax highlighting limitations

The current version of the book might show some limitations regarding syntax highlighting.
For example:

• The first method of a class is not highlighted - First method of class not highlighted in
C++ · Issue #791⁵.
• Template method is not highlighted C++ lexer doesn’t recognize function if return type
is templated · Issue #1138⁶.
• Modern C++ attributes are sometimes not appropriately recognized.

Other issues for C++ and Pygments: C++ Issues · github/pygments/pygments⁷.

Special sections

Throughout the book, you can also see the following sections:

This is an Information Box with extra notes related to the current section.

This is a Warning Box with potential risks and threats related to a given topic.

This is a Quote Box. In the book, it’s often used to quote the C++ Standard.

⁵https://github.com/pygments/pygments/issues/791
⁶https://github.com/pygments/pygments/issues/1138
⁷https://github.com/pygments/pygments/issues?q=is%3Aissue+is%3Aopen+C%2B%2B
About the Author
Bartłomiej (Bartek) Filipek is a C++ software developer from the beautiful city of Cracow
in Southern Poland. He started his professional career in 2007 and in 2010 he graduated from
Jagiellonian University with a Masters Degree in Computer Science.
Bartek currently works at Xara⁸, where he develops features for advanced document editors.
He also has experience with desktop graphics applications, game development, large-scale
systems for aviation, writing graphics drivers and even biofeedback. In the past, Bartek
has also taught programming (mostly game and graphics programming courses) at local
universities in Cracow.
Since 2011 Bartek has been regularly blogging at cppstories.com⁹ (started as bfilipek.com¹⁰).
The blog focuses on core C++ and getting up-to-date with the C++ Standards. He’s also a
co-organiser of the C++ User Group in Cracow¹¹. You can hear Bartek in one @CppCast
episode¹² where he talks about C++17, blogging and text processing.
Since October 2018, Bartek has been a C++ Expert for the Polish National Body which works
directly with ISO/IEC JTC 1/SC 22 (C++ Standardisation Committee). Bartek was awarded
his first MVP title for the years 2019/2020 by Microsoft.
In his spare time, he loves collecting and assembling Lego models with his son.
Bartek is the author of C++17 In Detail¹³ and C++ Lambda Story¹⁴.
⁸http://www.xara.com/
⁹https://www.cppstories.com]
¹⁰https://www.bfilipek.com
¹¹https://www.meetup.com/C-User-Group-Cracow/
¹²http://cppcast.com/2018/04/bartlomiej-filipek/
¹³https://leanpub.com/cpp17indetail
¹⁴https://leanpub.com/cpplambda
Acknowledgements
This book wouldn’t be possible without valuable input from many C++ experts and friends.
I especially would like to thank to the following people:

• JFT (John Taylor),

• Mariusz Jaskółka,
• Florin Chertes (see his profile at LinkedIn¹⁵),
• Konrad Jaśkowiec (see his profile at LinkedIn¹⁶),
• Professor Boguslaw Cyganek (see his profile at AGH university page¹⁷),
• Dawid Pilarski (see his blog at panicsoftware.com¹⁸),
• Javier Estrada (see his blog at javierestrada.blog¹⁹),
• Jonathan Boccara (from fluentcpp.com²⁰),
• Andreas Fertig (see his blog at andreasfertig.blog²¹),
• Peter Sommerlad (see his website and training info at sommerlad.ch²²),
• Timur Doumler (see his website at timur.audio²³ and his Twitter²⁴),
• Michael Goldshteyn, Software Architect.

They spent a lot of time on finding even little things that could be improved and extended.
Last but not least, I got a lot of valuable feedback from my blog readers, Patreon Discord
Server (See C++Stories @Patreon²⁵), and discussions at C++ Polska²⁶. Thank you all!
With all of the help from those kind people, the book quality got better and better!
¹⁵https://www.linkedin.com/in/florin-ioan-chertes-41b6845/
¹⁶https://pl.linkedin.com/in/konrad-ja%C5%9Bkowiec-84585159
¹⁷https://home.agh.edu.pl/~cyganek/
¹⁸https://blog.panicsoftware.com/
¹⁹https://javierestrada.blog/
²⁰https://www.fluentcpp.com/
²¹https://andreasfertig.blog/
²²https://sommerlad.ch/
²³https://timur.audio/
²⁴https://twitter.com/timur_audio
²⁵https://www.patreon.com/cppstories
²⁶https://cpp-polska.pl/
Revision History
• 20th June 2022 - The first public version! Missing parts: some sections in 10. Containers
as Data Members, some sections in 11. Non-regular Data Members.
• 22nd June 2022 - new sections on NSDMI, direct init and parens, more about inheriting
constructors, link to GoodReads, wording, hotfixes.
• 24th June 2022 - updated the “copy and move constructor” chapter, typos and small
wording improvements.
• 16th July 2022 - Containers as Data Members chapter rewritten, noexcept consistency
and noexcept move operations advantages in the move constructor section, wording,
fixes, layout.
• 13th September 2022 - changed title to “C++ Initialization Story”, adapted book
structure, rewritten “Non-local objects” chapter (previously only on inline variables),
new extracted chapter on Techniques, new section on CRTP.
• 18th November 2022 - heavily updated and completed “Non-regular data members”
chapter, constinit and thread_local sections in the “Non-local objects” chapter,
filled the “implicit conversion” section in the Constructors chapter.
• 23rd December 2022 - content completed! Added Deduction chapter, filled missing
sections in the Techniques chapter. Layout improvements, a few more questions,
exercises and fixes.
• 4th February 2023 - updated layout, wording improvements, fixes, improve consistency
of examples.
• 27th February 2023 - the first edition in Print!
• 13th April 2023 - the final quiz questions 6th and 8th clarification, small layout fixes.
• 19th June 2023 - a batch of minor fixes and clarifications based on user feedback.
1. Local Variables and Simple Types
Let’s start simple and ask, “what is initialization?” When we go to the definition from
C++Reference¹, we can read:

Initialization of a variable provides its initial value at the time of construction.

We can translate this definition to the following example:

void foo() {
int x = 42;
// ... use 'x' later...
}

Above, we have a function with a local variable x. The variable is declared as integer and
initialized with the value 42. This is not the only way you can assign that initial value. Here
are some more options:

struct Point { int x; int y; }; // declare a custom type

Point createPoint(int x) { return {x, -x}; }
int main() {
int x { 42 }; // list initialization
double y = { 100.0 }; // copy list initialization
auto ptr = std::make_unique<float>(90.5f); // auto type deduction
auto z = createPoint(42); // through a factory function
std::string s (10, 'x'); // calling a constructor
Point p { 10 }; // aggregate initialization
std::array<float, 100> numbers { 1.1f, 2.2f }; // array initialization
// ...
}
¹https://en.cppreference.com/w/cpp/language/initialization
Local Variables and Simple Types 2

You can also come up with many other forms of setting a value. We can also extend the
syntax on class data members, static variables, thread locals, or even dynamic memory
allocations.
In theory, initialization is a simple task: “put a value into a memory location of a newly
created variable”. However, such action relates to many different parts of an application
(local vs. non-local scope) and various places in the memory (like stack vs. heap). That’s
why the syntax or the behavior might be slightly different.
In C++, we have at least the following forms of initialization:

• aggregate initialization
• constant initialization
• default initialization
• direct initialization
• copy initialization
• list initialization
• reference initialization
• value initialization
• zero initialization
• plus related topics like copy elision, static variables, conversion sequences, constructors,
assignment, dynamic memory, storage, and more.

While the list sounds complex, we’ll move through those topics step by step revealing core
concepts. Later we’ll address more advanced examples and see what happens inside the C++
machinery.
While we can explain most cases on integers and other numerical types, it’s best to work
on something more practical. The book starts with some elementary custom types, then
considers various issues we might have with their early implementations. Later the types
will expand, giving us more context and compelling use cases.

Starting with simple types

Defining a class or a struct (a custom type) in C++ allows you to model your problem domain
and solve problems more naturally. Rather than working with a bunch of variables and
functions, it’s best to group them and provide a consistent API (Application Programming
Local Variables and Simple Types 3

Interface). C++ provides a set of built-in types, including boolean, integral, character, and
floating-point. Additionally, you can use objects from the Standard Library, like various
collections, std::string, std::vector, std::map, std::set, and many others. You can
collect these essential components and build your types.
To create a background for our main topic, let’s start with a type representing Car
Information for a car listing app. A system reads the car/truck information from a database
and displays it in the application. For an easy start, the type holds four members: name (a
std::string), production year, number of seats, and engine power.

Below there’s the first version of the code for that CarInfo type:
Ex 1.1. Simple CarInfo structure. Run @Compiler Explorer
#include <iostream>
#include <string>

struct CarInfo {
std::string name;
unsigned year;
unsigned seats;
double power;
};

int main() {
CarInfo firstCar;
firstCar.name = "Renault Megane";
firstCar.year = 2003;
firstCar.seats = 5;
Local Variables and Simple Types 4

firstCar.power = 116;
std::cout << "name: " << firstCar.name << '\n';
std::cout << "year: " << firstCar.year << '\n';
std::cout << "seats: " << firstCar.seats << '\n';
std::cout << "power (hp): " << firstCar.power << '\n';
}

In the above example, we defined a simple structure that holds data for a CarInfo. The code
is super simple, contains some issues, and follows the style of C++03. In the following few
chapters, I’ll guide you through the code and help you understand the problems and how to
eliminate them. We’ll also modernize it to include the latest C++ (up to C++20) features.
First: name, year, seats and power are called non-static data members. Each instance of
the CarInfo class has its own set of those members. In other words, we group variables to
create a representation for models in our problem domain. A user-defined type might also
have static data members, which are data shared between all instances of a given type. For
example, we could imagine a static member variable called numAllCars that would indicate
the total number of cars created in our program. We’ll talk about static data members later
in chapter 11 Static Variables.
Now, let’s investigate the code in detail. The definition and the declaration of the variable
firstCar in the main() function:

CarInfo firstCar;

It is called default initialization and, since our struct is simple, will leave all data members
of built-in types with indeterminate values. Similarly, you can get the same (potentially
buggy effect) for simple types when declared in function (as such variables have automatic
storage duration) ²:

void foo() {
int i; // indeterminate value!
double d; // indeterminate value!
}

The std::string data member name, on the other hand, will have an empty state (an empty
string) because its default constructor will be called. More on that later.
²In contrast, static and thread-local objects will be zero-initialized.
Local Variables and Simple Types 5

Once the object is created and uninitialized, we can access its members and set proper values.
By default, struct has public access to its members (and class has private access). This way,
we can access and change their values directly.

What is “Automatic Storage Duration” ?

All objects in a program have four possible ways to be “stored”: automatic, static,
thread, or dynamic. Automatic means that the storage is allocated at the start of
the scope, like in a function. Most local variables have automatic storage duration
(except those declared as static, extern, or thread_local). We’ll talk about this
more in the separate chapter on non-local objects³.

Setting values to zero

You might feel very unsatisfied that after creating a CarInfo object, most data members
have some indeterminate values. We can fix this and make sure data is at least set to “zero”.
Have a look:
Ex 1.2. Value initialization for CarInfo structure. Run @Compiler Explorer
CarInfo emptyCar{};
std::cout << "name: " << emptyCar.name << '\n';
std::cout << "year: " << emptyCar.year << '\n';
std::cout << "seats: " << emptyCar.seats << '\n';
std::cout << "power (hp): " << emptyCar.power << '\n';

The output:

name:
year: 0
seats: 0
power (hp): 0

The initialization with empty braces {} is called value initialization and by default (for built-
in types and classes with default constructors that are neither user-provided nor deleted),
sets data to “zero” (adapted for different types). This is similar to declaring and defining the
following variables:
³chapterinlinevars
Local Variables and Simple Types 6

int i{}; // i == 0
double d{}; // d == 0.0
std::string s{}; // s is an empty string

int j = {}; // other form of value initialization

std::string str = {}; // ...

This time the storage duration doesn’t matter, and value initialization works the same for
static, dynamic, thread-local, or automatic variables. For types with default constructors
(more on that later), the code will call them and, in the case of string s; will initialize it
to an empty string.

Initialization with aggregates

Our structure is very simple, and for such types, C++ has special rules where we can initialize
their internal values with so-called aggregate initialization. We can use such syntax also for
arrays. Here are some basic examples:
Ex 1.3. Aggregate Initialization basic syntax. Run @Compiler Explorer

// arrays:
int arr[] { 1, 2, 3, 4 };
float numbers[] = { 0.1f, 1.1f, 2.2f, 3.f, 4.f, 5. };
int nums[10] { 1 }; // 1, and then all 0s

// structures:
struct Point { int x; int y; };
struct Line { Point p1; Point p2; };
Line longLone {0, 0, 100, 100};
Line anotherLine = {100}; // rest set to 0
Line shortLine {{-10, -10}, {10, 10}}; // nested

In summary, for the above code:

• Each array element, or non-static class member, in order of array subscript/appear-

ance in the class definition, is copy-initialized from the corresponding clause of the
initializer list.
Local Variables and Simple Types 7

• You can use list initialization for arrays, and when the number of elements is not
provided, the compiler will deduce the count.
• If you pass fewer elements in the initializer list than the number of elements in the
array, the remaining elements will be value initialized. For built-in types, it means the
value of zero.
• For structures, you can use a single initializer list or nested one; the expansion will be
recursive.
• If you provide fewer values than the number of data members in the aggregate, then the
remaining data members (in the declaration order) will be effectively value initialized.
The first bullet point says that each element is copy initialized. We’ll return to this topic and
explain the difference between a copy vs. direct initialization syntax once we know explicit
constructors.
For our structure, we can write the following test code:
Ex 1.4. Aggregate initialization for the CarInfo structure. Run @Compiler Explorer
struct CarInfo {
std::string name;
unsigned year;
unsigned seats;
double power;
};

void printInfo(const CarInfo& c) {

std::cout << c.name << ", "
<< c.year << " year, "
<< c.seats << " seats, "
<< c.power << " hp\n";
}

int main() {
CarInfo firstCar{"Megane", 2003, 5, 116 };
printInfo(firstCar);
CarInfo partial{"unknown"};
printInfo(partial);
CarInfo largeCar{"large car", 1975, 10};
printInfo(largeCar);
}
Local Variables and Simple Types 8

This will output:

Megane, 2003 year, 5 seats, 116 hp

unknown, 0 year, 0 seats, 0 hp
large car, 1975 year, 10 seats, 0 hp

To give you the full picture, as of C++20, here’s the definition of an aggregate type from the
C++ Standard: dcl.init.aggr⁴.

An aggregate is an array or a class type with:

• no user-provided, explicit, or inherited constructors

• no private or protected non-static data members
• no virtual functions, and
• no virtual, private, or protected base classes

Don’t worry if you’re not familiar with all of the cases listed above. We’ll discuss them along
the way and see more aggregates in the further parts. There’s also a dedicated chapter about
Aggregates and Designated Initialization in C++20.

Default data member initialization

What if you want to provide some default value for your data member? With value
initialization, you can get zeros for various types, but sometimes it might not be good enough.
Since C++14, we can leverage Non-static Data Member Initializers (NSDMI), also called
Default Member Initializers, to provide default values for aggregates. Have a look:

⁴https://timsong-cpp.github.io/cppwp/n4868/dcl.init.aggr#:initialization,aggregate
Local Variables and Simple Types 9

Ex 1.5. Default member initialization and aggregates. Run @Compiler Explorer

#include <iostream>
#include <string>

struct CarInfo {
std::string name { "unknown" };
unsigned year { 1920 };
unsigned seats { 4 };
double power { 100. };
};

void printInfo(const CarInfo& c) { /* */ }

int main() {
CarInfo unknown;
printInfo(unknown);
CarInfo zeroed{};
printInfo(zeroed);
CarInfo partial{"large car", 1975};
printInfo(partial);
}

This will print:

unknown, 1920 year, 4 seats, 100 hp

unknown, 1920 year, 4 seats, 100 hp
large car, 1975 year, 4 seats, 100 hp

The syntax is quite intuitive; you can initialize a data member at the place where it’s declared.
This can prevent accidental bugs where your data has some indeterminate value. As you
can see from the example, even if you use default initialization or value initialization, data
members will get values that were provided in the struct declaration. If you give fewer
values in the aggregate initializer, the remaining members will get their defaults from the
declaration.
Technically, in-class member initializers have been available since C++11, but aggregate
types weren’t supported initially. In this section, we’ve only scratched the surface of
this handy technique. See the dedicated chapter for this topic: Non-static data member
initialization chapter.
Local Variables and Simple Types 10

Summary
In this chapter, we covered some simple custom types and looked at ways to initialize their
data members. We went from objects with indeterminate values to zero initialization, and
then we learned about aggregates and techniques to provide default values.
Things to keep in mind:

• Default initialization for objects and variables yields indeterminate values for built-
in types or default-initialize complex types (like std::string and set it to an empty
string). That’s why it’s essential to be sure your objects and simple variables are
always initialized.
• Value initialization like int x{}; for built-in types effectively yields zero initialization
for them so that they will be zero (in their type).
• With value initialization CarInfo car{}; all data members will be zero-initialized
(for built-in types) or default initialized for complex types.
• Aggregates are simple types or arrays with all public data members; we can initialize
them with an aggregate initialization syntax.
• Thanks to the in-class member initializer feature, you can provide default values for
your data members.

What’s next?
While simple types are handy, in C++, we often need to build large objects where data
members depend on each other or have invariants. In such cases, it’s best to hide them
behind member functions and give access to them under certain conditions. That’s why in
the next chapter, we’ll look at class’s and constructors. We’ll also expand the knowledge
that we got so far.
2. Classes and Initialization With
Constructors
In the previous chapter, you’ve seen that C++ might treat simple structures with all public
data members as an aggregate class. Still, aggregates might not be enough if we want better
data encapsulation and a more complex class API. For full flexibility in C++, we can leverage
constructors that are special member functions invoked when an object is created.

A simple class type

As a background example, let’s create a type that will hold some elementary network data. To
complicate things, we’d like to compute a basic checksum for the data part. Such a checksum
might be handy for checking if the data was transferred correctly across the Internet (read
more @Wikipedia¹).
Ex 2.1. Simple DataPacket class. Run @Compiler Explorer

#include <iostream>
#include <numeric>

size_t calcCheckSum(const std::string& s) {

return std::accumulate(s.begin(), s.end(), static_cast<size_t>(0));
}

class DataPacket {
private:
std::string data_;
size_t checkSum_;
size_t serverId_;

public:
const std::string& getData() const { return data_; }
¹https://en.wikipedia.org/wiki/Checksum
Classes and Initialization With Constructors 12

void setData(const std::string& data) {

data_ = data;
checkSum_ = calcCheckSum(data);
}
size_t getCheckSum() const { return checkSum_; }
size_t getServerId() const { return serverId_; }
void setServerId(size_t serverId) { serverId_ = serverId; }
};

The class above contains three non-static data members: data_, checkSum_ and serverID_.
I’m using the underscore suffix to indicate private data members, a common practice in
many codebases. See Google C++ Style Guide².
To keep things simple, I implemented the calcCheckSum function in terms of
std::accumulate(), which is an algorithm from the C++ Standard Library. This
code starts from 0 (we can use 0UZ since C++23 instead of explicit static_cast) and adds
numerical values of letters from the input std::string. For example, for "HELLO", we’ll
get the following computations:

Calculating simple checksum for a string

DataPacket has so-called getters and setters - functions that return or change a particular
data member. For example getData() returns the data_ data member, while setData(...)
allows to change it.
One important topic is that getters usually have const applied at the end. This means that
a given member function is constant and cannot change the value of the members (unless
they are mutable). If you have a const object, you can only call its const member functions.
Applying const might improve program design as it’s usually easier to reason about the state
²https://google.github.io/styleguide/cppguide.html#Variable_Names
Classes and Initialization With Constructors 13

of const instances. For more information, see this C++ core guideline: Con.2: By default,
make member functions const³.

Member functions might also have noexcept specifier applied. However, this
topic is outside the scope of the book and won’t be covered. You can find more
@C++Reference - noexcept specifier⁴.

Here’s the continuation of the example where we create and use the object of the DataPacket
class:
Ex 2.2. Simple DataPacket class, continuation. Run @Compiler Explorer

int main() {
DataPacket packet;
packet.setData("Programming World");
std::cout << packet.getCheckSum() << '\n';
}

The code doesn’t access data members directly but calls member functions to operate on the
object and change its properties.
You can notice public and private parts in the class declaration. The order of those sections
is just a coding convention and they group elements together based on their access modifier.
In short, a member under the public keyword can be accessed from the outside (like calling
a member function or accessing a data member). On the other hand, members under the
private section cannot be accessed from ⁵. In C++, you can also add protected to your class
declaration, which means that member functions or fields are not accessible outside. Still,
they are accessible to all inherited classes (assuming public inheritance, members become
private outside, but public to derived types, see more about different inheritance options
@C++Reference⁶).
For example, in the main() function above, I cannot write:

³https://isocpp.github.io/CppCoreGuidelines/CppCoreGuidelines#con2-by-default-make-member-functions-const
⁴https://en.cppreference.com/w/cpp/language/noexcept_spec
⁵Unless accessed by friend functions or classes.
⁶https://en.cppreference.com/w/cpp/language/access
Classes and Initialization With Constructors 14

DataPacket packet;
packet.serverId = 10; // error: 'size_t DataPacket::serverId'
// is private within this context

The only difference between class and struct in C++ is that class has private
as the default access modifier and private inheritance, while struct has both
specified as public. Some C++ guidelines, for example, Google Style Guide see
this link⁷, suggest using struct only for smaller, “passive” types, with only public
data members. The C++ Core Guidelines also recommend using class if any
member is not public; see C++ Core Guidelines - C.8⁸.

Since our class doesn’t have any user-defined constructors (more on them in the next section),
we can also use value initialization syntax to set values to zero or default values:
Ex 2.3. Value initialization for the DataPacket class. Run @Compiler Explorer

int main() {
DataPacket packet{};
std::cout << "data: " << packet.getData() << '\n';
std::cout << "checkSum: " << packet.getCheckSum() << '\n';
std::cout << "serverId: " << packet.getServerId() << '\n';
}

This will generate the following output:

data:
checkSum: 0
serverId: 0

However, the main difference now is that because we moved the data members to the private
section, the class is not an aggregate. That’s why we cannot use aggregate initialization to
set all values at once. To fix this, we need to look at constructors. And that is the plan for
further sections.
⁷https://google.github.io/styleguide/cppguide.html#Structs_vs._Classes
⁸https://isocpp.github.io/CppCoreGuidelines/CppCoreGuidelines#c8-use-class-rather-than-struct-if-any-member-is-non-
public
Classes and Initialization With Constructors 15

Basics of constructors
A constructor is a special member function without a name, but we declare it using the
enclosing class name. You cannot invoke a constructor like other member functions. Instead,
the compiler calls it when an object of its class is being initialized. It has the following basic
syntax:

class/struct ClassName {
// ...
/*explicit*/ ClassName(parameter-list) = default/=delete
: base-class-initializer
, member-init
{ /*body*/ }
// ...
};

A constructor has the following parts:

• constructor has no name, but we define it using the name of the class,
• optional explicit - keyword to block implicit conversions on a given class type,
• ClassName - the name of the given class type (they have to match),
• parameter-list- a list of parameters, as in a regular function, might be empty
• optional = default/=delete specifies if a constructor should be deleted (not present)
or defaulted by the compiler,
• : - indicates the start of the member/base initialization list, required when
base-class-initializer or member-init lists are present,
• optional base-class-initializer - a list of base classes’ constructors that we
explicitly want to call,
• optional member-init - a list of data members where we can directly initialize them,
• {/*body*/} - a function body.

You can also apply noexcept, [[attributes]], constexpr, consteval on a

constructor, but the full explanation of those additional properties goes beyond
the scope of the book. Read more at C++Reference - Constructors and member
initializer lists⁹.
⁹https://en.cppreference.com/w/cpp/language/constructor
Classes and Initialization With Constructors 16

Let’s have a look at one snippet:

class Product {
public:
Product() : id_{-1}, name_{"none"} { } // a default constructor
explicit Product(int id, const std::string& name)
: id_{id}, name_{name} { }

private:
int id_;
std::string name_;
};

The above example shows a class Product with two constructors. The first one is called a
default constructor; it has no arguments. The second one takes two arguments. As you can
notice, C++ allows multiple constructors that look like overloaded functions (they differ by
the number or types of arguments). Each constructor also has a regular function body where
you can execute some code; in our case, they are both empty for now. I also applied the
explicit keyword on the second constructor; we’ll talk about it later.

The primary function of constructors is to perform some actions at the start of a lifetime of
an object. Usually, it means data member initialization, resource allocation (opening a file, a
socket, memory allocation), or even doing some special logic (like logging).
In our case, constructors touch only data members inside a special section of constructors
called member initializer list: like, id_{-1}, name_{"none"}. Inside this initializer list, we
can also call constructors of base classes (if any). Later, we’ll address inheritance in the
Inheritance section.
The member initializer list is more efficient than using the body of a constructor. Sometimes
it’s even the only option to initialize the value, as with types that are not assignable. See the
following alternative:
Classes and Initialization With Constructors 17

class Product {
public:
Product() { id_ = 0; name_ = "none"; }

private:
int id_;
std::string name_;
};

The code will yield the same values for data members as in the previous example, but the data
members are set in two steps rather than one. With the member initializer list data members
are set directly, same as calling: int id_ { 0 } or std::string name_ {"none"}. On the
other hand, if we use assignment in the constructor body, it requires two steps:

// step 1: default init:

int id_; // indeterminate value!
std::string name_; // default ctor called
// step 2: assignment:
id_ = 0;
name_ = "none";

While this might not be a big issue for built-in simple types like int , you’ll need some more
CPU cycles for larger objects like strings.
There’s also one important aspect about the initializer list: the order of initialization. This is
covered in The C++ Specification: 11.10.3 Classes¹⁰:

Non-static data members are initialized in the order they were declared in the class
definition (regardless of the order of the mem-initializers).

When I write:

¹⁰https://timsong-cpp.github.io/cppwp/n4868/class.base.init#13.3
Classes and Initialization With Constructors 18

class Product {
public:
Product() : name_{"none"}, id_{-1} { }

private:
int id_;
std::string name_;
};

The values will be set correctly, but the order will differ from what we think. A compiler
might show us a warning in this case. Here’s the warning from GCC compiled with -Wall
option (experiment @Compiler Explorer¹¹):

<source>: In constructor 'Product::Product()':

<source>:15:17: warning: 'Product::name_' will be initialized after [-Wreorder]
15 | std::string name_;
| ^~~~~
<source>:14:9: warning: 'int Product::id_' [-Wreorder]
14 | int id_;
| ^~~

The initialization order might be critical when you imply some dependency on the values.
For example, we can write the following artificial sample:

struct S {
int x;
int y;
int z;

S(): x{0}, y{1}, z{x+y} { }

// S(): y{0}, z{0}, x{z+y}, { }
};

In the above example, the first constructor initializes x and y and then uses those values to
initialize z. This is complicated and might be hard to read, but it works correctly. On the
other hand, in the second (commented out) constructor, the order of initialization will create
¹¹https://godbolt.org/z/jE77169qd
Classes and Initialization With Constructors 19

an undefined behavior for initializing x, as z and y won’t be initialized yet. It’s best to avoid
such dependencies to minimize the risk of bugs.
Let’s see how a constructor works by creating some objects of the Product class:

Product none;

In the first example, we created the none object, which is default constructed. The compiler
will call our default constructor; thus, the data members will be initialized to id_ = -1 and
name_ = "none".

Product car(10, "car");

The example uses the form of direct initialization which calls the constructor with two
arguments. After the call data members will be: id_ = 10 and name_ = "car".
And the last example:

Product tvSet{100, "tv set" };

This time we also called a constructor with two arguments, but the syntax is called * direct
list initialization* - "{}". Please notice that I also used this form of initialization inside the
initializer list in constructors.
Here’s the complete example:
Ex 2.4. Constructors for the Product class. Run @Compiler Explorer

#include <iostream>
#include <string>

class Product {
public:
Product() : id_{-1}, name_{"none"} { } // a default constructor
explicit Product(int id, const std::string& name)
: id_{id}, name_{name} { }

int Id() const { return id_; }

std::string Name() const { return name_; }
Classes and Initialization With Constructors 20

private:
int id_;
std::string name_;
};

int main() {
Product none;
std::cout << none.Id() << ", " << none.Name() << '\n';

Product car(10, "super car");

std::cout << car.Id() << ", " << car.Name() << '\n';

Product tvSet{77, "tv set" };

std::cout << tvSet.Id() << ", " << tvSet.Name() << '\n';
}

You might also scratch your head and ask why I declared the name parameter as const
std::string& rather than just std::string&. First, we don’t want to modify this parameter
in the constructor’s body. What’s more, const T& - const references can bind to “temporary”
objects like a string literal "super car". Without a const reference, we would have to pass
some named string object. Alternatively, we can pass the name by value and perform a
“move operation” on that argument. Further in the book, I’ll address this topic in detail, see
chapter: A Use Case - Best Way to Initialize string Data Members.

More on uniform initialization

Content available in the full version of the book.

Body of a constructor
After the member initializer list, each constructor has a regular function body, { ... },
where you can perform additional steps to modify variables or call other functions. The only
difference between a regular function and a constructor is that a constructor cannot return
any values. Typically, a constructor throws an exception to report an error.
Here’s a small example that shows how to add some logging into a constructor body and
throw an exception on error:
Classes and Initialization With Constructors 21

Ex 2.7. Logging in a constructor. Run @Compiler Explorer

#include <iostream>
#include <stdexcept> // for std::invalid_argument

constexpr int LOWEST_ID_VALUE = -100;

class Product {
public:
explicit Product(int id, const std::string& name)
: id_{id}, name_{name}
{
std::cout << "Product(): " << id_ << ", " << name_ << '\n';
if (id_ < LOWEST_ID_VALUE)
throw std::invalid_argument{"id lower than LOWEST_ID_VALUE!"};
}

std::string Name() const { return name_; }

private:
int id_;
std::string name_;
};

int main() {
try {
Product car(10, "car");
std::cout << car.Name() << " created\n";
Product box(-101, "box");
std::cout << box.Name() << " created\n";
}
catch (const std::exception& ex) {
std::cout << "Error - " << ex.what() << '\n';
}
}

The above example shows a constructor that performs logging and basic parameter checking.
It uses a LOWEST_ID_VALUE, a global constant marked with the constexpr keyword (the
second time we used this keyword).
Classes and Initialization With Constructors 22

The constexpr specifier has been available since C++11 and guarantees that a
value is available at compile time for constant expressions. For example, you
can use such a variable to set the number of elements in a C-style array. It’s
often perceived as a “type-safe macro definition”. The keyword applies to all
built-in trivial types like integral values, floating-point, or even character literals
(but not std::string); there’s also a way to declare custom constexpr-ready
types. You can also create a function to be constexpr and possibly evaluate it at
compile-time; however, we won’t cover such functions in this book. See more at
C++Reference - constexpr¹².

If you run this program, you can see the following output:

Product(): 10, car

car created
Product(): -101, box
Error - id cannot be lower than LOWEST_ID_VALUE!

Please notice that while two constructors were called, we can see that only the first one
succeeded. Since the constructor for box threw an exception, this object is not treated as
fully created. More on that later, when we’ll talk about destructors.

Adding constructors to DataPacket

After the introduction, we can start adding constructors to our DataPacket class.
Ex 2.8. Adding constructors. Run @Compiler Explorer

class DataPacket {
std::string data_;
size_t checkSum_;
size_t serverId_;

public:
DataPacket()
: data_{}
, checkSum_{0}
, serverId_{0}
¹²https://en.cppreference.com/w/cpp/language/constexpr
Classes and Initialization With Constructors 23

{ }

explicit DataPacket(const std::string& data, size_t serverId)

: data_{data}
, checkSum_{calcCheckSum(data)}
, serverId_{serverId}
{ }

const std::string& getData() const { return data_; }

void setData(const std::string& data) {
data_ = data;
checkSum_ = calcCheckSum(data);
}
size_t getCheckSum() const { return checkSum_; }

void setServerId(size_t id) { serverId_ = id; }

size_t getServerId() const { return serverId_; }
};

And here’s the demo code that creates some objects:

Ex 2.9. Adding constructors, Demo. Run @Compiler Explorer

void printInfo(const DataPacket& packet) {

std::cout << "data: " << packet.getData() << '\n';
std::cout << "checkSum: " << packet.getCheckSum() << '\n';
std::cout << "serverId: " << packet.getServerId() << '\n';
}

int main() {
DataPacket empty;
printInfo(empty);
DataPacket zeroed{};
printInfo(zeroed);
DataPacket packet{"Hello World", 101};
printInfo(packet);
DataPacket reply{"Hi, how are you?", 404};
printInfo(reply);
}
Classes and Initialization With Constructors 24

The output:

data:
checkSum: 0
serverId: 0
data:
checkSum: 0
serverId: 0
data: Hello World
checkSum: 1052
serverId: 101
data: Hi, how are you?
checkSum: 1375
serverId: 404

In the above example, we used two constructors:

• The first one is a default constructor and initializes data members to default values. It
will be called for default and value initialization.
• The second constructor takes several arguments and matches them with data members.
This constructor makes it easy to pass parameters all at once (previously, we needed
to call setters). This one takes two parameters, but we can initialize as many data
members as we need. For example, the constructors ensure the checkSum_ variable
matches data_. Since those two members are related, thanks to constructors and the
setData member function, we keep the relation safe.

We can also use default member initializers inside a class, but we’ll address that in detail in
a separate chapter.

Compiler-generated default constructors

While C++ allows you to implement various constructors, it can make your life easier by
automatically declaring and defining an implicit default constructor.
In other words, if you write a class type with no default constructor:
Classes and Initialization With Constructors 25

class Example {
public:
std::string Name() const { return name_; }

private:
std::string name_;
};

Then the compiler will create an implicit empty constructor:

inline Example() noexcept { }

A simple rule is that if a class has no user-declared constructors, the compiler will create a
default one if possible.
Have a look:
Ex 2.10. Implicit default constructor. Run @Compiler Explorer

struct Value {
int x;
};

struct CtorValue {
CtorValue(int v): x{v} { }
int x;
};

int main() {
Value v; // fine, default constructor available
// CtorValue y; // error! no default ctor available
CtorValue z{10}; // using custom ctor
}

As you can see above, the compiler will create an implicit default constructor for the Value
class (since it has no other constructors), but it won’t generate a default constructor for the
CtorValue class. Also, notice that Value::x will have an indeterminate value as a default
constructor is empty and won’t set any value for x.
Classes and Initialization With Constructors 26

Default constructors only default-initialize data members, so in the case of built-in

types, it means indeterminate values!

You can control the creation of such a default constructor using two keywords, default
and delete. In short, default tells the compiler to use the default implementation, while
delete blocks the implementation.

Ex 2.11. Default and Delete Constructors. Run @Compiler Explorer

struct Value {
Value() = default;

int x;
};

struct CtorValue {
CtorValue() = default;
CtorValue(int v): x{v} { }
int x;
};

struct DeletedValue {
DeletedValue() = delete;
DeletedValue(int v): x{v} { }
int x;
};

int main() {
Value v; // fine, default constructor available
CtorValue y; // ok now, default ctor available
CtorValue z{10}; // using custom ctor
// DeletedValue w; // err, deleted ctor!
DeletedValue u{10}; // using custom ctor
}

In the above example, you can see that we declare Value() = default; this tells the
compiler to create an empty (doing nothing) implementation. Also, in the CtorValue class,
we also use the same technique, and, as you can notice, the default construction works now.
Classes and Initialization With Constructors 27

The third class has = delete as its default constructor, and you’ll get an error if you want
to create an object of this class using its default constructor.
The implicit default constructor won’t be created if your type has data members that are
not default-constructible or inherits from a type that is not default-constructible. That
includes references, const data members, unions, and others. See the complete list here
@C++Reference¹³.

You may also ask what’s the difference between Value() = default and
Value() { } they both are “empty”. Still, according to the C++ Standard the
second constructor is considered user-declared or user-provided and has some
consequences in the type characteristics. We’ll cover that later once we cover copy
constructors in the section: Trivial classes and user-declared/user-provided default
constructors.

Explicit constructors
Content available in the full version of the book.

Difference between direct and copy initialization

Content available in the full version of the book.

Even more
Content available in the full version of the book.

Constructor summary
This chapter was probably the longest, as we had to prepare the background for the rest
of the book. Once you know the basics of how data members can be initialized through
constructors, we can move further and explore various new C++ features and examples.
Now, it’s essential to summarize two other types of constructors: copy and move. Read on
to the next chapter.
¹³https://en.cppreference.com/w/cpp/language/default_constructor#Deleted_implicitly-declared_default_constructor
3. Copy and Move Constructors
Regular constructors allow you to invoke some logic and initialize data members when an
object is created from a list of arguments. But C++ also has two special constructor types
that let you control a situation when an object is created using an instance of the same class
type. Those constructors are called copy and move constructors. Let’s have a look.

Copy constructor
A copy constructor is a special member function taking an object of the same type as the
first argument, usually by const reference.

ClassName(const ClassName&);

Technically it might have other parameters, but they all have to have default values assigned.
It’s used and called when you create an object using a variable of the same type, to be precise,
when you use copy initialization.

Product base { 42, "base product" }; // an initial object

// various forms of initialization, where a copy constructor is called

Product other { base };
Product another(base);
Product oneMore = base;
Product arr[] = { base, other, oneMore };

Implementing a copy constructor might be necessary when your class has data members
that shouldn’t be shallow copied, like pointers, resource ids (like file handles), etc.

A canonical implementation of a copy constructor

Implementing a copy constructor is straightforward and very similar to regular constructors.
The only difference is that you have a single parameter which is a (const) reference to an
object of that same type.
For the Product class, we can write the following:
Copy and Move Constructors 29

class Product {
public:
explicit Product(int id, const std::string& name)
: id_{id}, name_{name}
{
std::cout << "Product(): " << id_ << ", " << name_ << '\n';
}

// copy constructor
Product(const Product& other)
: id_{other.id_}, name_{other.name_}
{ }

private:
int id_;
std::string name_;
};

As you can see, the copy constructor uses the member initialization list to copy the data from
other. Please notice that there’s no need to use public getters, as we have access to all private
data members. The compiler requires you to use a reference, so writing Product(Product
other) won’t be treated as a copy constructor.

A copy constructor can also take a non-const argument like Product(Product&

other). However, such a constructor might modify the other object and might
be hard to reason about the code. It might be better to use move semantics and
move constructors when you want to “steal” the guts of some other object.

Here’s another example where logging is enabled:

Copy and Move Constructors 30

Ex 3.1. An example of a logging copy constructor. Run @Compiler Explorer

#include <iostream>
#include <string>

class Product {
public:
explicit Product(int id, const std::string& name)
: id_{id}, name_{name}
{
std::cout << "Product(): " << id_ << ", " << name_ << '\n';
}

Product(const Product& other)

: id_{other.id_}, name_{other.name_}
{
std::cout << "Product(copy): " << id_ << ", " << name_ << '\n';
}

const std::string& Name() const { return name_; }

private:
int id_;
std::string name_;
};

int main() {
Product base { 42, "base product" }; // an initial object
std::cout << base.Name() << " created\n";
std::cout << "Product other { base };\n";
Product other { base };
std::cout << "Product another(base);\n";
Product another(base);
std::cout << "Product oneMore = base;\n";
Product oneMore = base;
std::cout << "Product arr[] = { base, other, oneMore };\n";
Product arr[] = { base, other, oneMore };
}
Copy and Move Constructors 31

If you run the code, you should see the following output:

Product(): 42, base product

base product created
Product other { base };
Product(copy): 42, base product
Product another(base);
Product(copy): 42, base product
Product oneMore = base;
Product(copy): 42, base product
Product arr[] = { base, other, oneMore };
Product(copy): 42, base product
Product(copy): 42, base product
Product(copy): 42, base product

In the first line, we construct base product, and then use it to copy-construct all other
instances.

Copy constructors can be marked with explicit, but this is not a common
practice and might prevent copy initialization.

A compiler-generated copy constructor

Content available in the full version of the book.

Move constructor
Move constructors take rvalue references of the same type.

ClassName(ClassName&&);

In short, rvalue references are temporary objects, usually appearing on the right-hand side
of an expression and which value is about to expire.
For example:
Copy and Move Constructors 32

std::string hello { "Hello"}; // lvalue, a regular object

std::string world { "World"}; // lvalue
std::string msg = hello + world;

Above, the expression hello + world creates a temporary object. It doesn’t have a name,
and we cannot access it easily. Such temporary objects will end their lifetime immediately
after the expression completes (unless it’s assigned to a const or rvalue reference¹), so we
can steal resources from them safely. It doesn’t make sense in the case of built-in types like
integers or floats, as we need to copy values anyway. But in the case of strings or memory
buffers, we can avoid data copy and just reassign the pointers.
Move constructors are a way to support the case with initialization from temporary objects.
In many cases, they are an optimization over regular copy constructor calls. Additionally,
they can also be used to pass “ownership” of the resource, for example, with smart pointers.
You can mark a regular object as expiring with the std::move function when you have a
regular object with a name. This tells the compiler that the object’s value is no longer needed,
so it’s safe to “steal” resources from it.
Have a look at this example:
Ex 3.3. Move Constructor. Run @Compiler Explorer

#include <iostream>
#include <string>

class Product {
public:
explicit Product(int id, const std::string& name)
: id_{id}, name_{name}
{
std::cout << "Product(): " << id_ << ", " << name_ << '\n';
}

Product(Product&& other)
: id_{other.id_}, name_{std::move(other.name_)}
{
std::cout << "Product(move): " << id_ << ", " << name_ << '\n';
}
¹The lifetime of a temporary object may be extended by binding to a const lvalue reference or to an rvalue reference. See more
at https://en.cppreference.com/w/cpp/language/lifetime.
Copy and Move Constructors 33

const std::string& name() const { return name_; }

private:
int id_;
std::string name_;
};

int main() {
Product tvSet {100, "tv set"};
std::cout << tvSet.name() << " created...\n";
Product setV2 { std::move(tvSet) };
std::cout << setV2.name() << " created...\n";
std::cout << "old value: " << tvSet.name() << '\n';
}

When you run the code, you can see the following output:

Product(): 100, tv set

tv set created...
Product(move): 100, tv set
tv set created...
old value:

As you can see, we create the first object, and then mark it as expiring. This gives a chance
for the compiler to call the move constructor.

Product(Product&& other)
: id_(other.id_), name_(std::move(other.name_))

The above implementation is similar, but we need to pay attention to details. Since id_ is just
an integer, all we can do is copy the value. We cannot perform any optimizations here. As
for the name_ member, we can initialize it with std::move(other.name_). We encounter
the first problem, other.name_ is a name, so not a temporary (a temporary has no name);
we can not move (take, steal) its contents. That is why we tell the compiler to interpret it as
temporary by using the expression std::move(other.name_). This will invoke the move
constructor for std::string, and, potentially, “steal” the buffer from other.name_.
Copy and Move Constructors 34

The move constructor must ensure that the other object is left in an unspecified but valid
state. In our case, we can see it in the last line of the output. The line old value: ends with
nothing, so the string was simply cleared.

Move constructors can be marked with explicit, but it’s not a common practice
and might affect generic code that relies on implicit move constructors (like
standard algorithms).

noexcept and move constructors

Content available in the full version of the book.

A compiler-generated move constructor

Content available in the full version of the book.

Distinguishing from assignment

Content available in the full version of the book.

Adding logging to constructors

As an exercise, let’s add logging to our DataPacket class and see when each constructor is
called:
 Copy and Move Constructors 35

Ex 3.6. Logging in the DataPacket class. Run @Compiler Explorer

1 class DataPacket {
2 std::string data_;
3 size_t checkSum_;
4 size_t serverId_;
5
6 public:
7 DataPacket()
8 : data_{}
9 , checkSum_{0}
10 , serverId_{0}
11 { }
12
13 explicit DataPacket(const std::string& data, size_t serverId)
14 : data_{data}
15 , checkSum_{calcCheckSum(data)}
16 , serverId_{serverId}
17 {
18 std::cout << "Ctor for \"" << data_ << "\"\n";
19 }
20
21 DataPacket(const DataPacket& other)
22 : data_{other.data_}
23 , checkSum_{other.checkSum_}
24 , serverId_{other.serverId_}
25 {
26 std::cout << "Copy ctor for \"" << data_ << "\"\n";
27 }
28
29 DataPacket(DataPacket&& other)
30 : data_{std::move(other.data_)} // move string member...
31 , checkSum_{other.checkSum_} // no need to move built-in types...
32 , serverId_{other.serverId_}
33 {
34 other.checkSum_ = 0; // leave this in a proper state
35 std::cout << "Move ctor for \"" << data_ << "\"\n";
36 }
37
 Copy and Move Constructors 36

38 DataPacket& operator=(const DataPacket& other) {

39 if (this != &other) {
40 data_ = other.data_;
41 checkSum_ = other.checkSum_;
42 serverId_ = other.serverId_;
43 std::cout << "Assignment for \"" << data_ << "\"\n";
44 }
45 return *this;
46 }
47
48 DataPacket& operator=(DataPacket&& other) {
49 if (this != &other) {
50 data_ = std::move(other.data_);
51 checkSum_ = other.checkSum_;
52 other.checkSum_ = 0; // leave this in a proper state
53 serverId_ = other.serverId_;
54 std::cout << "Move Assignment for \"" << data_ << "\"\n";
55 }
56 return *this;
57 }
58
59 // getters/setters
60 };

And here’s the main() function:

Ex 3.6. Logging in the DataPacket class, the main function. Run @Compiler Explorer

1 int main() {
2 DataPacket firstMsg {"first msg", 101 };
3 DataPacket copyMsg { firstMsg };
4
5 DataPacket secondMsg { "second msg", 202 };
6 copyMsg = secondMsg;
7
8 DataPacket movedMsg { std::move(secondMsg)};
9 // now we stole the data, so it should be empty...
10 std::cout << "secondMsg's data after move ctor): \""
11 << secondMsg.getData() << "\", sum: "
 Copy and Move Constructors 37

12 << secondMsg.getCheckSum() << '\n';

13
14 movedMsg = std::move(firstMsg);
15
16 // now we stole the name, so it should be empty...
17 std::cout << "firstMsg's data after move ctor): \""
18 << firstMsg.getData() << "\", sum: "
19 << firstMsg.getCheckSum() << '\n';
20 }

When you run the example, you should see the following output:

Ctor for "first msg"

Copy ctor for "first msg"
Ctor for "second msg"
Assignment for "second msg"
Move ctor for "second msg"
secondMsg's data after move ctor): "", sum: 0
Move Assignment for "first msg"
firstMsg's data after move ctor): "", sum: 0

The example creates several DataPacket objects, and with each creation, you can see that
the compiler invokes the appropriate constructor or an assignment operator. For instance,
in line 3, we need a copy constructor call. On the other hand, line 5 shows an assignment
(copyMsg already exists). In the last section of main(), lines 8 and 14, there are calls to
std::move(), which marks secondMsg and firstMsg as an rvalue reference, from which
the contents could be moved. This means that the object is unimportant later, and we can
“steal” from it. In this case, the compiler will call a move constructor or move assignment
operator.

Trivial classes and user-declared/user-provided

default constructors
Content available in the full version of the book.
 4. Delegating and Inheriting
Constructors
Content available in the full version of the book.

Limitations
Content available in the full version of the book.

Inheritance
Content available in the full version of the book.

Inheriting constructors
In our previous example with DebugPropertyInfo we didn’t have any new data members,
only some new member functions. The code showed a single constructor called the base class
constructor. Since C++11, you can tell the compiler to “reuse” the code:
Ex 4.4. Inheriting constructors. Run @Compiler Explorer

1 class DebugDataPacket : public DataPacket {

2 public:
3 using DataPacket::DataPacket;
4
5 void DebugPrint(std::ostream& os) {
6 os << getData() << ", " << getCheckSum() << '\n';
7 }
8 };
9
10 int main() {
 Delegating and Inheriting Constructors 39

11 DebugDataPacket hello{"hello!", 404};

12 hello.DebugPrint(std::cout);
13 }

Consider line 3 - using DataPacket::DataPacket;. This tells the compiler that it can
use all constructors from the base class, ignoring access modifiers. It means that all public
constructors are visible and can be called, but the protected will still be protected in that
context. Still, if you want to limit the access to constructors, you must explicitly write
constructors for DebugDataPacket.
We completed all information about constructors, but it’s good to mention one more thing:
destructors. See in the next chapter.
5. Destructors
While constructors are responsible for various situations where an object is created, C++
also offers a way to handle object destruction. C++ doesn’t provide any form of garbage
collection available in many popular programming languages, but thanks to precise lifetime
specification, you can be confident when your object will be destroyed.
Each class has a special member function called a destructor. If you don’t write one, the
compiler prepares a default implementation. A destructor is called when an object ends its
lifetime. In most cases, it means that an object goes out of the scope (for stack-allocated
variables), or when a delete operator is called (for heap-allocated variables). Additionally,
when you have a user-defined class, it will automatically call destructors for its data
members. For more information about lifetime, see a good summary at C++Reference page¹.

Basics

Before we move on, it would be good to expand our terminology. So far I mentioned “object”
to refer to entities of some type and relied on our “intuition” on how to access such entities.
But the C++ Standard defines an object in the following terms (simplified, based on C++
Draft - intro.object²):

The constructs in a C++ program create, destroy, refer to, access, and manipulate objects.
An object is created by a definition, by a new-expression, by an operation that implicitly
creates objects, or when a temporary object is created. An object occupies a region
of storage in its period of construction, throughout its lifetime, and in its period of
destruction.

And continuing:

¹https://en.cppreference.com/w/cpp/language/lifetime
²https://timsong-cpp.github.io/cppwp/n4868/intro.object#1
Destructors 41

• An object can have a name,

• An object has a storage duration which influences its lifetime,
• An object has a type,
• Objects can contain other objects, called subobjects. A subobject can be a member
subobject, a base class subobject, or an array element.

Here’s a basic scenario for a destructor that handles a case where the lifetime of an object
ends:
Ex 5.1. A logging destructor. Run @Compiler Explorer

#include <iostream>
#include <string>

class Product {
public:
explicit Product(const char* name, unsigned id)
: name_(name)
, id_(id)
{
std::cout << name << ", id " << id << '\n';
}

~Product() {
std::cout << name_ << " destructor...\n";
}

std::string Name() const { return name_; }

unsigned Id() const { return id_; }

private:
std::string name_;
unsigned id_;
};

The example contains the following special member function:

Destructors 42

~Product() {
std::cout << name_ << " destructor...\n";
}

The syntax is unique as it has no parameters and has the ∼ prefix. You can also have only
one destructor in a class. What’s more, a destructor doesn’t return any value.
Now, let’s create two objects of that type:
Ex 5.1. A logging destructor, continuation. Run @Compiler Explorer

int main() {
{
Product tvset("TV Set", 123);
}
{
Product car("Mustang", 999);
}
}

In our case, the constructor and the destructor is used to perform the logging. When you run
the example, you’ll see the following output:

TV Set, id 123
TV Set destructor...
Mustang, id 999
Mustang destructor...

I specifically enclosed objects (created on the stack) in separate scopes so that their lifetime
ends when their scope ends. On the other hand, if we have code:

int main() {
Product tvset("TV Set", 123);
Product car("Mustang", 999);
}

Then both tvset and car share the same lifetime scope so that we can expect the following
output:
Destructors 43

TV Set, id 123
Mustang, id 999
Mustang destructor...
TV Set destructor..

As you can see, the destructors are called in the reverse order of how they were created. It’s
because the stack is a LIFO structure (Last In First Out). tvset was created first and added
to the stack, then car is added. When the function goes out of the scope, the stack is cleared,
taking elements in the reverse order. So car is deleted first, and then tvset. This is illustrated
by the following diagram:

Adding and removing objects from the stack.

Objects allocated on the heap

Content available in the full version of the book.

Destructors and data members

Content available in the full version of the book.

Virtual destructors and polymorphism

Content available in the full version of the book.
Destructors 44

Partially created objects

Content available in the full version of the book.

Use Cases
The primary use case for destructors is when you need to release resources allocated in a
constructor. For example, you allocate some memory when the object is created, and then
the memory must be released to avoid memory leaks. Similarly, you can open a file or a
database connection, and then you must ensure the file or the connection is closed when the
object goes out of scope. Fortunately, in Modern C++, there are fewer and fewer places where
you need custom destructors. For example, when your data members are standard containers
(like std::vector<int>, or std::map<std::string, int>) in your classes, then you can
rely on default destructors to do the job. Standard containers like std::vector<int> might
allocate memory buffers, but they also manage that buffer and release it properly, so you
don’t need to take any action when using them in a class.

A compiler-generated destructor
Content available in the full version of the book.
6. Initialization and Type
Deduction
Content available in the full version of the book.
7. Quiz on Constructors
Congratulations!
You’ve just completed the section on the basics and constructors.
Here’s a quick quiz. Try answering the following questions, and then we will continue our
journey :)

1. Can a constructor have a different name than the class name?

1. Yes
2. No
3. Yes, but it can be only named self()

2. What operations are called in the following code? Pick one option.

std::string s { "Hello World" };

std::string other = s;

1. A constructor is called for s. Then, as assignment operation is called for other.

2. A constructor is called for s, and then a copy constructor is called to create other.
3. A constructor is called for s, and then another regular constructor is called for other.

More questions available in the full version of the book.

8. Non-Static Data Member
Initialization
You’ve learned a lot of techniques related to constructors! You can initialize data members
in various constructors, delegate them to reuse code, and inherit them from base classes. Yet,
we can still improve on assigning default values for data members. I mentioned this feature
in the first chapter, where we gave default values for aggregates. We can do the same for
classes. And in this chapter, we’ll look at the full syntax and options related to this feature.
Please have a look at the example below:
Ex 8.1. NSDMI Basics. Run @CompilerExplorer
class DataPacket {
std::string data_;
size_t checkSum_ { 0 };
size_t serverId_ { 0 };

public:
DataPacket() = default;

DataPacket(const std::string& data, size_t serverId)

: data_{data}
, checkSum_{calcCheckSum(data)}
, serverId_{serverId}
{ }

// getters and setters...

};

As you can see, the variables are assigned their default values individually in their place of
declaration. There’s no need to set values inside a constructor. It’s much better than using
a default constructor because it combines declaration and initialization code. This way, it’s
harder to leave data members uninitialized!
Let’s explore this handy feature of Modern C++ in detail.
Non-Static Data Member Initialization 48

How it works

This section shows how the compiler “expands” the code to initialize data members.
For a simple declaration:

struct SimpleType {
int field { 0 };
};

The code has to behave similarly as you’d define a constructor ¹:

struct SimpleType {
SimpleType() : field(0) { }

int field;
};

Experiment with the basic code below:

Ex 8.2. Basic Non-static data member initialization. Run @Compiler Explorer

#include <iostream>

struct SimpleType {
int field { 0 };
};

int main() {
SimpleType st;
std::cout << "st.field is " << st.field << '\n';
}

As a small exercise, you can experiment with the above sample and assign different values
to the field data member.
¹Technically, those types will be different as the version without the constructor will be considered an aggregate type, but for
the purpose of the discussion, it’s not essential now.
Non-Static Data Member Initialization 49

Investigation

With some “machinery,” we can see when the compiler performs the initialization.
Let’s consider the following type:

struct SimpleType {
int a { initA() };
std::string b { initB() };

// ...
};

The implementation of initA() and initB() functions have side effects, and they log extra
messages:

int initA() {
std::cout << "initA() called\n";
return 1;
}

std::string initB() {
std::cout << "initB() called\n";
return "Hello";
}

This allows us to see when the code is called.

Experiments
Now, we can experiment and write some additional constructors:
Non-Static Data Member Initialization 50

struct SimpleType {
int a { initA() };
std::string b { initB() };

SimpleType() { }
SimpleType(int x) : a(x) { }
};

Next, we can run our test and see the results.

Ex 8.3. Calling init functions. Live code @Compiler Explorer

#include <iostream>
#include <string>

int initA() {
std::cout << "initA() called\n";
return 1;
}

std::string initB() {
std::cout << "initB() called\n";
return "Hello";
}

struct SimpleType {
int a { initA() };
std::string b { initB() };

SimpleType() { }
SimpleType(int x) : a(x) { }
};

int main() {
std::cout << "SimpleType t0\n";
SimpleType t0;
std::cout << "SimpleType t1(10)\n";
SimpleType t1(10);
}
Non-Static Data Member Initialization 51

After running the code, we can see the following output:

SimpleType t1
initA() called
initB() called
SimpleType t1(10)
initB() called

You can observe the following:

t0 is default-initialized; therefore, both fields are initialized with their default values. In other
words, the compiler calls {initA()} and {initB{}}. Please notice that they are initialized
in the order they appear in the class/struct declaration.
In the second case, for t1, only one value is default initialized, and the other comes from the
constructor parameter.
As you might already guess, the compiler initializes the fields as if the fields were initialized
in a “member initialization list”. Therefore, they get the default values before the construc-
tor’s body is invoked.
In other words, the compiler “conceptually” expands the code:

struct SimpleType {
int a { initA() };
std::string b { initB() };

SimpleType() { }
SimpleType(int x) : a(x) { }
};

Into:
Non-Static Data Member Initialization 52

struct SimpleType {
int a;
std::string b;

SimpleType() : a(initA()), b(initB()) { }

SimpleType(int x) : a(x), b(initB()) { }
};

We can also visualize it using the following diagram:

Other forms of NSDMI

Content available in the full version of the book.

Copy constructor and NSDMI

Content available in the full version of the book.

Move constructor and NSDMI

Content available in the full version of the book.
Non-Static Data Member Initialization 53

C++14 changes
Originally, in C++11, if you used default member initialization, your class couldn’t be an
aggregate type:

struct Point { float x = 1.0f; float y = 2.0f; };

// won't compile in C++11

Point myPt { 10.0f, 11.0f };

The above code won’t work when compiling with the C++11 flag because you cannot
aggregate-initialize our Point structure. It’s not an aggregate.
Fortunately, C++14 provides a solution to this problem, and that’s this line:

Point myPt { 10.0f, 11.0f};

The code works as expected now. You can see and play with the full code below:
Ex 10.1. Aggregates and NSDMI in C++14. Run @CompilerExplorer
#include <iostream>

struct Point { float x = 1.0f; float y = 2.0f; };

int main()
{
Point myPt { 10.0f };
std::cout << myPt.x << ", " << myPt.y << '\n';
}

C++20 changes
Content available in the full version of the book.

Limitations of NSDMI
Content available in the full version of the book.
Non-Static Data Member Initialization 54

NSDMI: Advantages and Disadvantages

Let’s summarize non-static data member initialization.

Advantages of NSDMI
Content available in the full version of the book.

Any negative sides of NSDMI?

Content available in the full version of the book.

NSDMI summary
Before C++11, the best way to initialize data members was through a member initialization
list inside a constructor. Thanks to C++11, we can now initialize data members in the place
where we declare them, and the initialization happens just before the constructor body kicks
in.
In the chapter, we covered the syntax, how it works with various types of constructors and
the limitations. You also saw changes made in C++14 (aggregate classes) and missing bitfield
initialization fixed in C++20.
The C++ Core Guidelines advise using NSDMI in at least two sections.
C++ Core Guidelines - C.48²:

C.48 Prefer in-class initializers to member initializers in constructors for constant initial-
izers:
Reason: Makes it explicit that the same value is expected to be used in all constructors.
Avoids repetition. Avoids maintenance problems. It leads to the shortest and most efficient
code.

And in C++ Core Guidelines - C.45³

²https://isocpp.github.io/CppCoreGuidelines/CppCoreGuidelines#c48-prefer-in-class-initializers-to-member-initializers-in-
constructors-for-constant-initializers
³https://isocpp.github.io/CppCoreGuidelines/CppCoreGuidelines#c45-dont-define-a-default-constructor-that-only-
initializes-data-members-use-in-class-member-initializers-instead
Non-Static Data Member Initialization 55

C.45Don’t define a default constructor that only initializes data members; use in-class
member initializers instead
Reason: Using in-class member initializers lets the compiler generate the function for you.
The compiler-generated function can be more efficient.

If you like to read more about NSDMI, I highly recommend reading the book
“Embracing Modern C++ Safely”, chapter 2, page 318. There’s a whole section on
advanced cases for this powerful C++ feature.

NSDMI: Exercises
Check your skills with two coding exercises.

The first exercise

Content available in the full version of the book.

The second exercise

Content available in the full version of the book.

9. Containers as Data Members
CarInfo, DataPacket, and Product types used relatively simple data members like integers,
doubles, or strings. While std::string is, in fact, a container (of characters), we tend to use
it as an elementary type. In this section, I’d like to discuss more complex data members like
arrays, vectors, or maps.

The basics
Content available in the full version of the book.

Using std::initializer list

(*) this section will be added in the future.

Example implementation
Content available in the full version of the book.
10. Non-regular Data Members
Thus far, we spoke about mutable non-static data members like integers, doubles, or
strings. Such objects are regular, meaning they are copyable, default constructible, and
equally comparable. In C++20 there’s even a concept for that purpose: std::regular, see
@C++Reference¹.
However, you can also have other categories of objects in a class. For example, a custom
type might contain constant data members, pointers, references, or moveable only fields like
unique pointers or mutexes. For such members, we have immediate issues with default copy
constructors (the compiler won’t create them).
In this chapter, we’ll shed some light on such cases.

Constant non-static data members

Content available in the full version of the book.

References as data members

(*) this section will be added in the future.

Pointers as data members

(*) this section will be added in the future.

Moveable-only data members

(*) this section will be added in the future.
¹https://en.cppreference.com/w/cpp/concepts/regular
Non-regular Data Members 58

Summary
Having discussed other categories of non-static data members, we can now examine static
data members. How to use them in Modern C++? See the next chapter.
11. Inline Variables in C++17
In this chapter, you’ll see how to enhance and simplify code using inline variables from
C++17.

About static data members

In general, each and every instance (object) of a class has non-static data members as its
own data fields; each instance is separate from the other. If we consider a type (a class)
representing a Fruit and it has a data member named “mass”, then each particular instance
of that Fruit class has a “mass” member belonging to it. If we have 10 Fruit objects, the “mass”
data member is replicated ten times. On the other hand, each type can also have static data
members that are not bound to any instance of the class. In the case of our Fruit class, we can
specify a so-called static variable named “default mass”, accessible to each Fruit instance, but
it wouldn’t be part of any instance. In other words, it’s like a global variable in the namespace
of the Fruit type.
Consider the following example:
Ex 12.1. Simple static Data Member. Run @Compiler Explorer

#include <iostream>

struct Value {
int x;

static int y;
};

int Value::y = 0; // definition

int main() {
Value v { 10 };
std::cout << "sizeof(int): " << sizeof(int) << '\n';
std::cout << "sizeof(Value): " << sizeof(Value) << '\n';
Inline Variables in C++17 60

std::cout << "v.x: " << v.x << '\n';

Value::y = 10;
std::cout << "Value::y: " << Value::y << '\n';
}

When you run this program, you’ll see the following output:

sizeof(int): 4
sizeof(Value): 4
v.x: 10
Value::y: 10

static int y declared in the scope of the Value class created a variable that is not part of
any Value type instance. You can see that it doesn’t contribute to the size of the whole class.
It’s the same as the size of the int type.
In the further sections, let’s consider a more practical use case for such class members.

Motivation for inline variables

In C++11/14, if you wanted to add a static data member to a class, you needed to declare
it and define it later in one compilation unit. In the example from the previous section, we
defined it in the same compilation unit as the main() function. Commonly, such variables
are defined in the corresponding implementation file.
For example:
Ex 12.2. Static data member, multiple files. Run @Wandbox

// a header file:
struct OtherType {
static int classCounter;

// ...
};

// implementation, cpp file

int OtherType::classCounter = 0;
Inline Variables in C++17 61

This time we also used Wandbox online compiler - as it’s easy to create and compile multiple
files:

As you can see above, classCounter is an int, and you have to write it twice: in a header
file and then in the CPP file.

The only exception to this rule (even before C++11) is a static constant integral variable that
Inline Variables in C++17 62

you can declare and initialize in one place:

class MyType {
static const int ImportantValue = 42;
};

You do not have to define ImportantValue in a CPP file.

Fortunately, C++17 gave us inline variables, which means we can define a static inline
variable inside a class without defining them in a CPP file.
Ex 12.3. Static inline member. Run @Wandbox

// a header file, C++17:

struct OtherType {
static inline int classCounter = 0;

// ...
};

The compiler (and the linker) guarantees that there’s precisely one definition of this static
variable for all translation units that include the class declaration. Inline variables remain
static class variables, so they will be initialized before the main() function is called.
Inline Variables in C++17 63

This feature makes it much easier to develop header-only libraries because there’s no need
to create CPP files for static variables or use hacks to keep them in a header file (for example,
by creating static member functions with static variables inside).
See the example below:

// CountedType.h
struct CountedType {
static inline int classCounter = 0;

// simple counting... only ctor and dtor implemented...

CountedType() { ++classCounter; }
~CountedType() { --classCounter; }
};

And the main() function:

Ex 12.4. Static inline member. Run @Wandbox

#include <iostream>
#include "CountedType.h"

int main() {
{
CountedType c0;
CountedType c1;
std::cout << CountedType::classCounter << '\n';
}
std::cout << CountedType::classCounter << '\n';
}

The code above declares classCounter inside CountedType, which is a static data member.
The class is defined in a separate header file. Thanks to C++17, we can declare the variable
as inline. Then, there’s no need to write a corresponding definition later. Without inline,
the code wouldn’t compile.
Later, in the main() function, the example creates two objects of CountedType. The static
variable is incremented when there’s a call to the constructor. When an object is destroyed,
the variable is decremented. We can output this value and see the current count of objects.
Inline Variables in C++17 64

Exercise for inline variables

Content available in the full version of the book.

Global inline variables

Content available in the full version of the book.

Constexpr and inline variables

Content available in the full version of the book.
12. Aggregates and Designated
Initializers in C++20
Across the book, you’ve seen a lot of cases for intuitively simple structures with all public
data members. Such types, along with arrays, are called Aggregates. In this chapter, we’ll
look at some C++20 changes and new ways to initialize such objects.

Aggregates in C++20
To sum up, as of C++20, here’s the definition of an aggregate type from the C++ Standard:
dcl.init.aggr¹.

An aggregate is an array or a class type with:

• no user-provided, explicit, or inherited constructors

• no private or protected non-static data members
• no virtual functions, and
• no virtual, private, or protected base classes

Here are some examples of aggregates:

¹https://timsong-cpp.github.io/cppwp/n4868/dcl.init.aggr#:initialization,aggregate
Aggregates and Designated Initializers in C++20 66

Ex 13.1. Aggregate classes, several examples. Run @Compiler Explorer

struct Base { int x {42}; };

struct Derived : Base { int y; };

struct Param {
std::string name;
int val;
void Parse(); // member functions allowed
};

int main() {
Derived d {100, 1000};
std::cout << "d.x " << d.x << ", d.y " << d.y << '\n';
Derived d2 { 1 };
std::cout << "d2.x " << d2.x << ", d2.y " << d2.y << '\n';
Param p {"value", 10};
std::cout << "p.name " << p.name << ", p.val " << p.val << '\n';

double arr[] { 1.1, 2.2, 3.3, 4.4};

std::cout << "arr[0] " << arr[0] << '\n';
Param params[] {{"val", 10}, {"name", 42}};
std::cout << "params[0].name " << params[0].name << '\n';
}

In C++20, in some limited cases, you can also use parens X(args...) to initialize an
aggregate:

// C++20 and parens:

Point pt (1, 2);
// Point pt = (1, 2); // doesn't work

double params[] (9.81, 3.14, 1.44);

// double paramsDeduced[] = (9.81, 3.14, 1.44); // won't deduce
int arrX[10] (1, 2, 3, 4); // rest is 0

Such improvement helps, especially in a generic template code where you want to work with
various types of objects. For example, the following code wasn’t possible until C++20:
Aggregates and Designated Initializers in C++20 67

Ex 13.2. Aggregates and parens for make_unique. Run @Compiler Explorer

struct Point { int x; int y; };

int main() {
auto ptr = std::make_unique<Point>(10, 20);
}

make_unique takes a variable number of arguments and passes them to a constructor.

This function uses parens to call the constructor. Since aggregates has no user-declared
constructors, then such syntax generated errors. With the C++20 change, the code works fine
now. Suppose you want to dig more into this topic. In that case, I highly recommend reading
C++20’s parenthesized aggregate initialization has some downsides – Arthur O’Dwyer²,
which discusses pros and cons of this new initialization syntax.

The basics of Designated Initializers

The C++20 Standard also gives us another handy way to initialize data members. The new
feature is called designated initializers, which might be familiar to C programmers.
As of C++20, to initialize an aggregate object, you can write the following:

Type obj = { .designator = val, .designator { val2 }, ... };

For example:

struct Point { double x; double y; };

Point p { .x = 10.0, .y = 20.0 };

Designator points to a name of a non-static data member from our class, like .x or .y.
One of the main reasons to use this new kind of initialization is to increase readability.
Compare the following initialization forms:

²https://quuxplusone.github.io/blog/2022/06/03/aggregate-parens-init-considered-kinda-bad/
Aggregates and Designated Initializers in C++20 68

struct Date {
int year;
int month;
int day;
};

// new
Date inFutureCpp20 { .year = 2050, .month = 4, .day = 10 };
// old
Date inFutureOld { 2050, 4, 10 };

In the case of the Date class, it might be unclear what the order of days/month or month/days
is. With designated initializers (inFutureCpp20), it’s very easy to see the order of data
members.

Rules
Content available in the full version of the book.

Advantages of designated initialization

• Readability: A designator points to the specific data member, so it’s impossible to make
mistakes here.
• Flexibility: You can skip some data members and rely on default values for others.
• Compatibility with C: In C99, it’s popular to use a similar form of initialization
(although even more relaxed). With the C++20 feature, it’s possible to have very similar
code and share it.
• Standardization: Some compilers, like GCC or Clang, already had some extensions for
this feature, so it’s a natural step to enable it in all compilers.

Examples
Content available in the full version of the book.
Aggregates and Designated Initializers in C++20 69

Summary
As you can see, designated initializers are handy and usually more readable way of
initializing aggregate types. The new technique is also common in other programming
languages, like C or Python, so having it in C++ makes the programming experience even
better.

The summary example

Content available in the full version of the book.

Compiler support
Here’s a table with support for the features we discussed:
Content available in the full version of the book.
13. Techniques and Use Cases
Across the book, we’ve touched on many different topics, sometimes only in a theoretical
way. In this chapter, however, I grouped many of those features and demonstrated their
benefits in several practical use cases.
You’ll learn about the following aspects:

• Strong types and the explicit keyword,

• Initializing string data members,
• Copy and Swap Idiom as a potential simplification of copy and move operations,
• CRTP,
• Creating a simple resource manager (RAII) class.
• Factory With Self-Registering Types
• And more!

Let’s start.

Using explicit for strong types

If you recall the first chapter, I used double to indicate horsepower (hp) inside the CarInfo
structure. However, we might quickly encounter a problem where we forget about the unit
and treat it as Watts instead. Can we somehow limit such problematic cases?
The answer is positive, and the main idea is to wrap the data member double power in a
separate class type with explicit constructors. That it will be harder to misuse it, such an
approach is called Strong Typing.
Have a look at two similar wrapper types:
Techniques and Use Cases 71

Ex 13.1. Strong types and area units classes. Run @Compiler Explorer

constexpr double ToWattsRatio { 745.699872 };

class HorsePower;

class WattPower {
public:
WattPower() = default;
explicit WattPower(double p) : power_{p} { }
explicit WattPower(const HorsePower& h);

double getValue() const { return power_; }

private:
double power_ {0.};
};

class HorsePower {
public:
HorsePower() = default;
explicit HorsePower(double p) : power_{p} { }
explicit HorsePower(const WattPower& w);

double getValue() const { return power_; }

private:
double power_ {0.};
};

As you can see, we have two types that use explicit constructors to initialize their private
data members. To create an object, you have to write the correct type name explicitly, and
thus it should limit the chance of mistakes.
And here is the implementation of the converting constructors as well as stream operators
for easy output:
Techniques and Use Cases 72

Ex 13.2. Strong Types and area units, implementation. Run @Compiler Explorer

constexpr double ToWattsRatio { 745.699872 };

class HorsePower;

class WattPower {
/* as before */
};

class HorsePower {
/* as before */
};

WattPower::WattPower(const HorsePower& h)
: power_{h.getValue()*ToWattsRatio}
{ }

HorsePower::HorsePower(const WattPower& w)
: power_{w.getValue()/ToWattsRatio}
{ }

std::ostream& operator<<(std::ostream& os, const WattPower& w) {

os << w.getValue() << "W";
return os;
}

std::ostream& operator<<(std::ostream& os, const HorsePower& h) {

os << h.getValue() << "hp";
return os;
}

The interface allows us to convert between various units safely.

Techniques and Use Cases 73

//HorsePower hp = 10.; // not possible, copy initialization

HorsePower hp{ 10. }; // fine
WattPower w { 1. }; // fine
WattPower watts { hp }; // fine, performs the proper conversion for us!

Additionally, we have the output support that writes out the proper unit name.
We can use the solution now:

void printInfo(const CarInfo& c) {

std::cout << c.name << ", "
<< c.year << " year, "
<< c.seats << " seats, "
<< c.power << '\n';
}

int main() {
CarInfo firstCar{"Megane", 2003, 5, HorsePower{116}};
printInfo(firstCar);
CarInfo superCar{"Ferrari", 2022, 2, HorsePower{300}};
printInfo(superCar);
superCar.power = HorsePower{WattPower{500000}};
printInfo(superCar);
}

And we’ll get the following output:

Megane, 2003 year, 5 seats, 116hp

Ferrari, 2022 year, 2 seats, 300hp
Ferrari, 2022 year, 2 seats, 670.511hp

While I had to be more explicit and write the types, the code can be safer as it’s harder to
type something accidentally.

In C++11, you can also leverage user-defined literals to allow easier creation of
objects. Especially useful for units, string, numerical types, time, and dates. For
example, We could create a named literal _m2 and then write 50.0_m2 to create an
instance rather than SqMeters{50.2}. See more at C++Reference - User-defined
literals¹.
¹https://en.cppreference.com/w/cpp/language/user_literal
Techniques and Use Cases 74

For more information about Strong Types, I highly recommend reading many
articles on the Fluent C++ blog. For example, start with this one: Strong types for
strong interfaces - Fluent C++².

Best way to initialize string data members

Content available in the full version of the book.

The copy and swap idiom

Content available in the full version of the book. ## CRTP class counter {#sectioncrtp}
Content available in the full version of the book.

Several initialization types in one class

As the demo of various initialization techniques, I’d like to show code that creates N random
“application windows.”
Here are the core points of the demo:

• A Window class contains basic parameters like name (on the title bar), width, height,
and some flags (bits per pixel, visibility).
• The demo selects a random number X and will try to generate X Window objects.
• Each object will have a random name composed of predefined words and a random
size.
• The application prints each window using std::cout.
• As an additional check, an InstanceCounter class counts the number of Window
objects. We can use this helper to verify the correctness of the demo.

Here’s the first part that defines the Flags object:

²https://www.fluentcpp.com/2016/12/08/strong-types-for-strong-interfaces/
Techniques and Use Cases 75

Ex 13.5. The Flags type. Run @Compiler Explorer

struct Flags {
unsigned bppMode_ : 4 { 0 }; // bits per pixel
unsigned visible_ : 1 { 1 };
unsigned extData : 2 { 0 };
};

Here’s the main class:

Ex 13.5. The Window type. Run @Compiler Explorer

class Window : public InstanceCounter<Window> {

static constexpr unsigned default_width { 1028 };
static constexpr unsigned default_height { 768 };
static constexpr unsigned default_bpp { 8 };

unsigned width_ { default_width };

unsigned height_ { default_height };
Flags flags_ {.bppMode_ { default_bpp } };
std::string title_ { "Default Window" };

public:
Window() = default;
explicit Window(std::string title) : title_(std::move(title)) { }
Window(std::string title, unsigned w, unsigned h) :
width_(w), height_(h), title_(std::move(title)) {}

friend std::ostream& operator<<(std::ostream& os, const Window& w) {

os << w.title_ << ": " << w.width_ << "x" << w.height_;
return os;
}
};

The Window class uses several features discussed in the book:

• NSDMI to initialize data members,

• designated initializers from C++20, combined with NSDMI for the flags_ data
member,
Techniques and Use Cases 76

• Custom constructors that offer several options to initialize the data members,
• We inherit from InstanceCounter, so each constructor invocation for the Window
will also invoke the appropriate constructor in InstanceCounter. Similarly, the
InstanceCounter destructor will be nicely called from the implicit default destructor
of the Window class.

And now the final demo code:

Ex 13.5. The Window type. Run @Compiler Explorer

void WindowDemo() {
std::random_device rd;
std::mt19937 gen(rd());
std::uniform_int_distribution<> distrib(0, 20);

const int windowCount = std::uniform_int_distribution<>(2, 10)(gen);

std::cout << "Generating " << windowCount << " random Windows\n";

const std::array adjs { "regular ", "empty ", "blue ", "super " };
const std::array nouns { "app", "tool", "console", "game" };
const std::array sizes { 1080u, 1920u, 768u, 320u, 640u, 3840u, 800u };

std::vector<Window> windows;
for (int i = 0; i < windowCount; ++i) {
auto r = distrib(gen);
auto r2 = distrib(gen);
auto name = std::string { adjs[(r + i) % adjs.size()] } +
nouns[r2 % nouns.size()];
Window w{name, sizes[r2 % sizes.size()],
sizes[r % sizes.size()]};
windows.push_back(w);
}

for (const auto& w : windows)

std::cout << w << '\n';

std::cout << "Created " << Window::GetInstanceCounter() << " Windows\n";

}

int main() {
Techniques and Use Cases 77

WindowDemo();

if (Window::GetInstanceCounter() != 0) {
std::cout << Window::GetInstanceCounter()
<< " Windows are still alive!\n";
}
}

Here’s the possible output:

Generating 8 random Windows

super tool: 320x320
regular tool: 320x640
super game: 1080x768
super game: 640x1080
regular tool: 1920x3840
empty tool: 1920x3840
blue game: 320x768
empty console: 320x320
Created 8 Windows

In WindowDemo, the code declares some basic data and generates a random number. Later, in
the main loop, we generate random numbers to pick values from adjs, nouns, and sizes
arrays. Once the data is ready, I can create a Window object and place it in the std::vector.
To show the creation of the Window object, I used push_back on a vector, but we can optimize
it and call emplace_back, which doesn’t need a temporary object:

windows.emplace_back(name, sizes[r2 % sizes.size()], sizes[r % sizes.size()]);

Later there’s another loop that prints all windows.

In the code, I didn’t have to specify the full type for std::array<Type, Count>
as the compiler could deduce everything for me! Thanks to Class Type Argument
Deduction (CTAD) and Deduction guides from C++17, the compiler can help us
save some typing. See more @C++Reference - deduction guides for array³.
³https://en.cppreference.com/w/cpp/container/array/deduction_guides
Techniques and Use Cases 78

The code uses InstanceCounter as a bonus debugging facility to ensure we have the correct
number of active objects. When WindowDemo() finishes, all instances should be removed, and
we can double-check it inside main().

Vector like RAII object

(*) this section will be added in the future.

Factory with self-registering types

(*) this section will be added in the future.

Summary
(*) this section will be added in the future.
14. The Final Quiz
Check your knowledge from this mini-book!

1. Which C++ Standard did add in-class default member initializers?

1. C++98
2. C++11
3. C++14
4. C++17

2. Can you use auto type deduction for non-static data members?

1. Yes, since C++11

2. No
3. Yes, since C++20

3. Do you need to define a static inline data member in a separate cpp

file?

1. No, the definition happens at the same place where a static inline member is declared.
2. Yes, the compiler needs the definition in a cpp file.
3. Yes, the compiler needs a definition in all translation units that use this variable.

4. Can a static inline variable be non-constant?

1. Yes, it’s just a regular variable.

2. No, inline variables must be constant.
The Final Quiz 80

5. Consider the following code:

struct S {
int a { 10 };
int b { 42 };
};

What’s the output of the following line?

S s { 1 };
std::cout << s.a << ", " << s.b;

1. 1, 0
2. 10, 42
3. 1, 42

6. Consider the following code:

class C {
C(int x) : a(x) { }

int a { 10 };
int b { 42 };
};

C c(0);

More questions available in the full version of the book.

Appendix A - Quiz and Exercises
Answers
Content available in the full version of the book.
References
Related materials and links about data member initialization in C++:
Proposals for C++ features:
• N2756¹ - Non-static data member initializers for C++11,
• P0683² - Default Bit Field Initializer for C++20,
• P0386³ - Inline Variables C++17,
• P0329⁴ - Designated Initializers C++20,
• P0960⁵ and P1975⁶ - Aggregate initialization from a parenthesized list for C++20.
Valuable resources for C++:
• C++ Standard Draft⁷ - N4868 (October 2020 pre-virtual-plenary working draft/C++20
plus editorial changes),
• C++ compiler support - C++Reference⁸ - a list of features and their compiler support
since C++11,
• C++ Core Guidelines⁹ - a community-edited and open guideline for C++ style, lead by
Bjarne Stroustrup and Herb Sutter.
Books:
• “Embracing Modern C++ Safely”¹⁰ by J. Lakos, V. Romeo , R. Khlebnikov, A. Meredith,
a wonderful and very detailed book about latest C++ features, from C++11 till C++14
in the 1st edition.
• “Effective Modern C++: 42 Specific Ways to Improve Your Use of C++11 and C++14”¹¹
by Scott Meyers
¹https://wg21.link/N2756
²https://wg21.link/P0683
³https://wg21.link/P0386
⁴https://wg21.link/P0329
⁵https://wg21.link/p0960
⁶https://wg21.link/p1975
⁷https://timsong-cpp.github.io/cppwp/n4868/
⁸https://en.cppreference.com/w/cpp/compiler_support
⁹https://isocpp.github.io/CppCoreGuidelines/CppCoreGuidelines
¹⁰https://amzn.to/3PywHTg
¹¹https://amzn.to/3t5tmS4
References 83

Presentations:
• Core C++ 2019: Initialisation in modern C++¹² by Timur Doumler,
• CppCon 2018: “The Nightmare of Initialization in C”¹³ by Nicolai Josuttis,
• CppCon 2021: Back To Basics: The Special Member Functions¹⁴ by Klaus Iglberger,
• ACCU 2022: What Classes We Design and How¹⁵ - by Peter Sommerlad,
• CppCon 2018 “The Bits Between the Bits: How We Get to main()”¹⁶ - by Matt Godbolt.
Articles and other links:
• Non-Static Data Members Initialization - C++ Stories¹⁷ - initial source for the book,
• What happens to your static variables at the start of the program? - C++ Stories¹⁸,
• Always Almost Auto Style¹⁹ by Herb Sutter,
• C++ Core Guidelines - C51²⁰ and C52²¹ - about delegating and inheriting constructors,
• Modern C++ Features - Inherited and Delegating Constructors²² by Arne Mertz,
• Trivial, standard-layout, POD, and literal types²³ at Microsoft Docs,
• Modern C++ Features - Uniform Initialization and initializer_list²⁴ by Arne Mertz,
• The cost of std::initializer_list²⁵ by Andrzej Krzemieński,
• Objects, their lifetimes and pointers²⁶ by Dawid Pilarski,
• Tutorial: When to Write Which Special Member²⁷ by Jonathan Müller,
• The implication of const or reference member variables in C++²⁸ by Lesley Lai,
• Brace initialization of user-defined types²⁹ by Glennan Carnie³⁰.
¹²https://www.youtube.com/watch?v=v0jM4wm1zYA
¹³https://www.youtube.com/watch?v=7DTlWPgX6zs
¹⁴https://www.youtube.com/watch?v=9BM5LAvNtus
¹⁵https://www.youtube.com/watch?v=fzsBZicBe88
¹⁶https://www.youtube.com/watch?v=dOfucXtyEsU
¹⁷https://www.cppstories.com/2015/02/non-static-data-members-initialization/
¹⁸https://www.cppstories.com/2018/02/staticvars/
¹⁹https://herbsutter.com/2013/08/12/gotw-94-solution-aaa-style-almost-always-auto/
²⁰https://isocpp.github.io/CppCoreGuidelines/CppCoreGuidelines#c51-use-delegating-constructors-to-represent-common-
actions-for-all-constructors-of-a-class
²¹https://isocpp.github.io/CppCoreGuidelines/CppCoreGuidelines#c52-use-inheriting-constructors-to-import-constructors-
into-a-derived-class-that-does-not-need-further-explicit-initialization
²²https://arne-mertz.de/2015/08/new-c-features-inherited-and-delegating-constructors/
²³https://docs.microsoft.com/en-us/cpp/cpp/trivial-standard-layout-and-pod-types?view=msvc-170
²⁴https://arne-mertz.de/2015/07/new-c-features-uniform-initialization-and-initializer_list/
²⁵https://akrzemi1.wordpress.com/2016/07/07/the-cost-of-stdinitializer_list/
²⁶https://blog.panicsoftware.com/objects-their-lifetimes-and-pointers/
²⁷https://www.foonathan.net/2019/02/special-member-functions/
²⁸https://lesleylai.info/en/const-and-reference-member-variables/
²⁹https://blog.feabhas.com/2019/04/brace-initialization-of-user-defined-types/
³⁰https://blog.feabhas.com/author/glennan/