Game Development Patterns Best Practices

Title Page
Game Development Patterns and Best Practices
Better games, less hassle
John P. Doran
Matt Casanova
BIRMINGHAM - MUMBAI
Game Development Patterns and Best Practices
Copyright 2017 Packt Publishing
All rights reserved. No part of this book may be reproduced, stored in a retrieval system, or
transmitted in any form or by any means, without the prior written permission of the publisher, except
in the case of brief quotations embedded in critical articles or reviews.
Every effort has been made in the preparation of this book to ensure the accuracy of the information
presented. However, the information contained in this book is sold without warranty, either express
or implied. Neither the authors, nor Packt Publishing, and its dealers and distributors will be held
liable for any damages caused or alleged to be caused directly or indirectly by this book.
Packt Publishing has endeavored to provide trademark information about all of the companies and
products mentioned in this book by the appropriate use of capitals. However, Packt Publishing cannot
guarantee the accuracy of this information.
First published: April 2017
Production reference: 1180417
Published by Packt Publishing Ltd.

Livery Place
35 Livery Street
Birmingham
B3 2PB, UK.
ISBN 978-1-78712-783-8
www.packtpub.com
Credits
Authors
Copy Editor
John P. Doran Safis Editing
Matt Casanova
Reviewers
Project Coordinator
Francesco Sapio
Ritika Manoj
Josh Wittner
Commissioning Editor Proofreader
Amarabha Banerjee Safis Editing

Acquisition Editor Indexer
Larissa Pinto Francy Puthiry
Content Development Editor Graphics
Arun Nadar Jason Monteiro
Technical Editor Production Coordinator
Sushant S Nadkar Arvindkumar Gupta

About the Authors
John P. Doran is a technical game designer who has been creating games for over 10 years. He has
worked on an assortment of games in teams consisting of just himself to over 70 people in student,
mod, and professional projects in different roles, from game designer to lead UI programmer. He
previously worked at LucasArts on Star Wars: 1313 as a game designer. He later graduated from
DigiPen Institute of Technology in Redmond, WA, with a Bachelor of Science in game design.
John is currently a part of DigiPen's research and development branch in Singapore in addition to
DigiPen Game Studios. He is also the lead instructor of the DigiPen-Ubisoft Campus Game
Programming Program, instructing graduate-level students in an intensive, advanced-level game
programming curriculum. In addition to that, he also tutors and assists students in various subjects and
gives lectures on C#, C++, Unreal, Unity, game design, and more.
He is the author of Unreal Engine Game Development Cookbook, Building an FPS Game in Unity,
Unity Game Development Blueprints, Getting Started with UDK, UDK Game Development,
and Mastering UDK Game Development, and he co-wrote UDK iOS Game Development Beginner's
Guide, all available from Packt. More information about him can be found at johnpdoran.com.
Matt Casanova has been programming indie games in C++, as well as mobile games and
applications in Java and Objective C, for over ten years. For the last seven years he has assumed the
role of technical director for his students. He specializes in 2D game engines and tools.
Matt was an instructor for two years at Digipen Institute of Technology's Singapore campus, where he
taught The Advanced Certification in 2D Game Development. This was a six month graduate-level
course that specialized in creating 2D game engines. He then spent five years teaching undergraduate
courses such as C++, data structures, and game engine architecture for Digipens degree in Real Time
Interactive Simulation in South Korea. He has helped dozens of student teams make successful game
projects and avoid the pitfalls of bad code. He currently works as lead mobile developer at a
software company in Las Vegas.
I would like to thank my wife TaeGyeong Lee for listening to me for months on end while I talked
about this book. I would also like to the thank Josh Wittner for constantly giving me technical
advice and encouraging me to learn programming in the first place.
About the Reviewers
Francesco Sapio obtained his Computer Science and Control Engineering degree from Sapienza
University of Rome, Italy, a couple of semesters in advance, scoring summa cum laude. He is
currently studying a Master of Science in Engineering in Artificial Intelligence and Robotics at the
same university.
He is a Unity3D and Unreal expert, a skilled game designer, and an experienced user of the major
graphics programs. He developed Game@School (Sapienza University of Rome), an educational
game for high school students to learn the concepts of physics, and Sticker Book (series) (Dataware
Games), a cross-platform series of games for kids. In addition, he worked as consultant for the
(successfully funded by Kickstarter) game Prosperity Italy 1434 (Entertainment Game Apps, Inc),
and for the open online collaborative ideation system titled Innovoice (Sapienza University of
Rome). Moreover, he has been involved in different research projects such as Belief-Driven-
Pathfinding (Sapienza University of Rome), a new technique of pathfinding in videogames that was
presented as a paper at the DiGRA-FDG Conference 2016; and perfekt.ID (Royal Melbourne Institute
of Technology), which included developing a recommendation system for games.
He is an active writer on the topic of game development. Recently, he authored the book Getting
Started with Unity 5.x 2D Game Development (Packt Publishing), which takes your hand and guide
you through the amazing journey of game development; the successful Unity UI Cookbook (Packt
Publishing), which has been translated into other languages, and which teaches readers how to
develop exciting and practical user interfaces for games within Unity; and a short e-guide What do
you need to know about Unity (Packt Publishing). In addition, he co-authored the book Unity 5.x 2D
Game Development Blueprints (Packt Publishing). Furthermore, he has also been a reviewer for the
following books: Game Physics Cookbook (Packt Publishing), Mastering Unity 5.x (Packt
Publishing), Unity 5.x by Example (Packt Publishing), and Unity Game Development Scripting
(Packt Publishing).
Francesco is also a musician and a composer, especially of soundtracks for short films and video
games. For several years, he worked as an actor and dancer, where he was a guest of honor at the
Brancaccio theater in Rome. In addition, he is a very active person, having volunteered as a
children's entertainer at the Associazione Culturale Torraccia in Rome.
Finally, Francesco loves math, philosophy, logic, and puzzle solving, but most of all, creating video
gamesthanks to his passion for game designing and programming.
You can find him at www.francescosapio.com.
I'm deeply thankful to my parents for their infinite patience, enthusiasm, and support
throughout my life. Moreover, I'm thankful to the rest of my family, in particular to my
grandparents, since they have always encouraged me to do better in my life with the Latin
expressions "Ad maiora" and "Per aspera ad astra."
Finally, a huge thanks to all the special people around me whom I love, in particular to my
girlfriend; I'm grateful for all of your help in everything. I do love you.
At work Josh Wittner is currently applying a decade's worth of experience developing real-time
simulations in the video game industry to mixing virtual with actual reality through holographic
applications. At home, he's the father of a wonderful daughter and the husband to a wonderful wife.
I'd like to thank my wife and daughter for helping me stay focused.
www.PacktPub.com
For support files and downloads related to your book, please visit www.PacktPub.com.
Did you know that Packt offers eBook versions of every book published, with PDF and ePub files
available? You can upgrade to the eBook version at www.PacktPub.com and as a print book customer, you
are entitled to a discount on the eBook copy. Get in touch with us at [email protected] for more
details.
At www.PacktPub.com, you can also read a collection of free technical articles, sign up for a range of free
newsletters and receive exclusive discounts and offers on Packt books and eBooks.
https://www.packtpub.com/mapt
Get the most in-demand software skills with Mapt. Mapt gives you full access to all Packt books and
video courses, as well as industry-leading tools to help you plan your personal development and
advance your career.
Why subscribe?
Fully searchable across every book published by Packt
Copy and paste, print, and bookmark content
On demand and accessible via a web browser
Customer Feedback
Thanks for purchasing this Packt book. At Packt, quality is at the heart of our editorial process. To
help us improve, please leave us an honest review on this book's Amazon page at https://www.amazon.com/
dp/1787127834.
If you'd like to join our team of regular reviewers, you can e-mail us at [email protected]. We
award our regular reviewers with free eBooks and videos in exchange for their valuable feedback.
Help us be relentless in improving our products!
Table of Contents
Preface
What this book covers
What you need for this book
Who this book is for
Conventions
Reader feedback
Customer support
Downloading the example code
Downloading the color images of this book
Errata
Piracy
Questions
1. Introduction to Design Patterns
Chapter overview
Your objective
What are design patterns
Why you should plan for change
Understanding UML class diagrams
Relationships between classes
Inheritance
Aggregation
Composition
Implements
Separating the why and the how
Understanding the separation of concerns
An Introduction to interfaces
Polymorphism refresher
Understanding interfaces
The advantages of compartmentalizing code
The structure of the Mach5 engine
Mach5 core engines and systems
The app
The StageManager
The ObjectManager
The graphics engine
Tools and utilities
The problems with using design patterns in games
Setting up the project
Summary
2. One Instance to Rule Them All - Singletons
Chapter overview
Your objective
An overview on class access specifiers
The static keyword
Static keyword inside a function
Static keyword in class definitions
Static as a file global variable
Pros and cons of global variables
What is a Singleton?
Keeping the single in Singleton
Deleting our object correctly
Learning about templates
Templatizing Singletons
Advantages/disadvantages of using only one instance
The Singleton in action - the Application class
Summary
3. Creating Flexibility with the Component Object Model
Chapter overview
Your objectives
Why a monolithic game object is a bad design
The monolithic game object
The problem with object behavior
The benefits of the monolithic game object
Why inheritance hierarchies are inflexible
Organizing the code by what it does, not what it is
Avoiding the Diamond of Death
The Strategy pattern and the Decorator pattern
The Strategy pattern explained
The Decorator pattern explained
The Component Object Model explained
Implementing the Component Object Model
Implementing components
Creating and removing objects and components
Performance concerns
Summary
4. Artificial Intelligence Using the State Pattern
Chapter overview
Your objectives
The State pattern explained
Introduction to State Machines
An overview of enumerations
Acting on states
Issues with conditionals
Expanding on the State Machine
The State pattern in action - the M5StateMachine class
The State pattern in action - StageManager
Issues with FSMs
Summary
5. Decoupling Code via the Factory Method Pattern
Chapter overview
Your objective
The trouble with switch statements
The Dependency Inversion Principle
The Factory method pattern
The Static Factory
The Dynamic Factory
Creating our Stage Builders
The template builder
Creating the Dynamic Factory class
Using the Dynamic Factory
Creating a component and Object Factory
The Templated Factory
Architecting versus over-architecting
Summary
6. Creating Objects with the Prototype Pattern
Your objectives
The trouble with using a factory for game objects
Using builders with objects
Solution - reading from files
The Prototype pattern explained
The virtual constructor
The problem with constructors
The benefits of a virtual constructor
We don't need to know the type
No need to subclass
It's easy to make exact copies
Examples of the clone method in Mach5
The Gfx and collider components
Cloning an object
Choosing a copy constructor
Covariant return types
Loading archetypes from a file
Archetype files
The object manager
Summary
7. Improving Performance with Object Pools
Chapter overview
Your objectives
Why you should care about memory
The Object Pool pattern explained
Implementing a basic object pool
Operator overloading in C++
Building the object pool for Mach5
Issues with object pools
Summary
8. Controlling the UI via the Command Pattern
Chapter overview
Your objectives
How can we control actions through buttons?
Callback functions
Repeated code in the component
The Command pattern explained
Two parameters and beyond
Pointers to member functions
Pointer to member command
The benefits of the command pattern
Treating a function call like an object
Physically decoupling the client and the function call
Temporal decoupling
Undo and redo
Easy UI with commands in Mach5
Using commands
Summary
9. Decoupling Gameplay via the Observer Pattern
Chapter overview
Your objectives
How gameplay creeps into every system
Hardcoding requirements
Polling
The Observer pattern explained
The Subject and Observer
The Player
The Observers
Push versus Pull
Benefits of using the Observer pattern
Problems using the Observer pattern
Dangling references
Overuse
Implementing interfaces
When to notify
Summary
10. Sharing Objects with the Flyweight Pattern
Chapter overview
Your objectives
Introductions to particles
Implementing particles in Mach5
Why memory is still an issue
Introduction to the Flyweight pattern
Transitioning to ParticleSystems
Creating different system types
Developing the ParticleFactory
Using the ParticleFactory
Summary
11. Understanding Graphics and Animation
Chapter overview
Your objectives
Introduction to monitor refresh rates
What is a pixel?
The horizontal and vertical blank
Refresh rate
Double buffering
The back buffer
VSync
Triple buffering
LCD monitors
Time-based movement and animation
Frame-based movement
Time-based movement
Summary
12. Best Practices
Chapter overview
Your objectives
Learning fundamental code quality techniques
Avoid magic numbers
White space
Indentation
Blank lines and spaces
Comments and self-documenting code
Commenting
Learning and understand the uses of the const keyword
Const function parameters
Const classes as parameters
Const member functions
Problems with const
Learning how iteration can improve your game and code design
The game development cycle
Production phase
Prototyping
Playtesting
Conducting a playtest
Iteration
Meeting milestones
Learning when to use scripting in a game
Introduction to assembly
Moving to higher-level programming languages
Introducing the compiler
Introduction to scripting languages
Using interpreters
Just in time compilation
Why use a scripting language?
When to use C++
Compiled versus scripting
Summary
Preface
As your programs get larger, it is critical that you understand how to write code in such a way that
classes and modules can communicate with each other in an intelligent way. Knowing how to write
code that is clean and extensible is critical for the success of medium to large scale projects,
especially if there are multiple programmers involved. We don't want to spend time rewriting another
programmers code because it is easier than figuring out how the original code works. Likewise, we
dont want other programmers to be confused by our code and rewrite solutions to problems that we
already solved.
This book explores the ins and outs of the most common ways to design code so that it can be
understood, reused, maintained, and extended if necessary. These common design patterns will make
communication between classes simple and clean. Whether you are using a commercial game engine
or writing your own from scratch, knowledge of these patterns will make your game project more
likely to succeed.
Each chapter of this book explores one of the most used design patterns for games. Together we
discuss the problems that we are trying to solve and how a specific pattern can be of use to us. Within
each chapter, we also cover the pros and cons of the pattern so you are better equipped to know when
to use it. This book is not a "cook book". Using a pattern in your project isnt as simple as copying
some code that we provide. Instead of learning recipes for code, we are going to learn the
fundamentals of how a game engine should be written.
We will do this by looking at lots of code examples. As we said, these example, cant simply be copy
and pasted into any project. However, by understanding how these examples fit into the specific
project for this book, you can implement them in any project.
What this book covers
Introduction to Design Patterns, introduces the concept of design patterns and how they can
Chapter 1,
benefit our projects. We also dive into learning about the advantages of compartmentalizing our code
and set up a sample project that we can use to build upon in the book.
One Instance to Rule Them All Singletons, discusses the well-known singleton pattern,
Chapter 2,
explaining common arguments for and against having a single instance thats accessible globally
within our game.
Creating Flexibility with the Component Object Model, explores how to create a large
Chapter 3,
amount of different game object types while minimizing code duplication.
Artificial Intelligence Using the State Pattern, shows the usage of states to allow a game
Chapter 4,
object to change its behavior and functionality due to different stimuli in your game. We discuss and
apply how we can use this concept in both player behavior and artificial intelligence.
Decoupling Code via the Factory Method Pattern, explains how we separate our specific
Chapter 5,
game objects from our game engine. This will allow us to reuse all or at least part of our game
engine for future games.
Creating Objects with the Prototype Pattern, covers another way to reduce dependencies
Chapter 6,
within our game engine. We learn how to create objects through their base classes. In addition, we
cover how to read object data from a file.
Improving Performance with Object Pools, covers a way to improve memory costs in our
Chapter 7,
games by reusing objects that weve created dynamically by making use of a manager to control what
is and isnt being used at the moment.
Controlling the UI via the Command Pattern, discusses how to create reusable actions that
Chapter 8,
can be utilized for UI button clicks or input devices such as keyboard or controller.
Decoupling Gameplay via the Observer Pattern, explains multiple ways that classes can
Chapter 9,
communicate and share data without using global variables.
Sharing Objects with the Flyweight Pattern, discusses how we can design our objects to
Chapter 10,
be as lightweight as possible by separating data that can and can't be shared. We use this knowledge
to create a particle system which produces smaller objects developed to simulate things like fire,
explosions, and smoke.
Understanding Graphics and Animation, provides a low-level explanation of how graphics

Chapter 11,
and rendering work so that you are better able to implement it in your game.
Chapter 12, Best Practices, covers a number of topics that will improve your code and game projects in
the future, touching on things like how to improve your code quality, use the const keyword correctly,
how iteration is a great way to improve game and code design, and when you should consider adding
scripting to your game.
What you need for this book
To write code, all you really need is a text editor, but for serious development you'll need to use an
IDE (Integrated Developer Environment). We use Microsoft Visual Studio 2015 and C++ in this book
using a Windows computer but, the concepts taught can be used within any programming language as
well. For those of you that do not have Visual Studio installed, but are using Windows we do go over
how to install the free Visual Studio Community version and set it up with our project in Chapter 1,
Introduction to Design Patterns.
Who this book is for
This book is for anyone who wants to learn how to write better, cleaner code. While design patterns
in general do not require any specific programming language, the examples in this book are written in
C++. Readers will get much more out of this book if they are already familiar with C++ and STL
containers. The project that is included with this book was written and compiled using Microsoft
Visual Studio 2015, so familiarity with that IDE will be very helpful as well.
Conventions
In this book, you will find a number of text styles that distinguish between different kinds of
information. Here are some examples of these styles and an explanation of their meaning.
Code words in text, database table names, folder names, filenames, file extensions, pathnames,
dummy URLs, user input, and Twitter handles are shown as follows: "Im going to update the
ChasePlayerComponent class that already exists in the EngineTest project."
A block of code is set as follows:

class Animal
{
public:
virtual void Speak(void) const //virtual in the base class
{
//Using the Mach 5 console print
M5DEBUG_PRINT("...\n");
}
};
When we wish to draw your attention to a particular part of a code block, the relevant lines or items
are set in bold:
class StaticExamples
{
public:
static float classVariable;
static void StaticFunction()

{
// Note, can only use static variables and functions within
// static function
std::string toDisplay = "\n I can be called anywhere!
classVariable value: " +
std::to_string(classVariable);
printf(toDisplay.c_str());
}
void InFunction()
{
static int enemyCount = 0;
// Increase the value of enemyCount

enemyCount += 10;
std::string toDisplay = "\n Value of enemyCount: " +

std::to_string(enemyCount);
}
};
New terms and important words are shown in bold. Words that you see on the screen, for example,
in menus or dialog boxes, appear in the text like this: "In order to download new modules, we will go
to Files | Settings | Project Name | Project Interpreter."
Warnings or important notes appear in a box like this.

Tips and tricks appear like this.
Reader feedback
Feedback from our readers is always welcome. Let us know what you think about this book-what you
liked or disliked. Reader feedback is important for us as it helps us develop titles that you will really
get the most out of.
To send us general feedback, simply e-mail [email protected], and mention the book's title in the
subject of your message.
If there is a topic that you have expertise in and you are interested in either writing or contributing to
a book, see our author guide at www.packtpub.com/authors.
Customer support
Now that you are the proud owner of a Packt book, we have a number of things to help you to get the
most from your purchase.
Downloading the example code
You can download the example code files for this book from your account at http://www.packtpub.com. If
you purchased this book elsewhere, you can visit http://www.packtpub.com/support and register to have the
files e-mailed directly to you.
You can download the code files by following these steps:
1. Log in or register to our website using your e-mail address and password.
2. Hover the mouse pointer on the SUPPORT tab at the top.
3. Click on Code Downloads & Errata.
4. Enter the name of the book in the Search box.
5. Select the book for which you're looking to download the code files.
6. Choose from the drop-down menu where you purchased this book from.
7. Click on Code Download.
Once the file is downloaded, please make sure that you unzip or extract the folder using the latest
version of:
WinRAR / 7-Zip for Windows

Zipeg / iZip / UnRarX for Mac
7-Zip / PeaZip for Linux
The code bundle for the book is also hosted on GitHub at https://github.com/PacktPublishing/Game-Development-Pat
terns-and-Best-Practices. We also have other code bundles from our rich catalog of books and videos
available at https://github.com/PacktPublishing/. Check them out!
Downloading the color images of this book
We also provide you with a PDF file that has color images of the screenshots/diagrams used in this
book. The color images will help you better understand the changes in the output. You can download
this file from https://www.packtpub.com/sites/default/files/downloads/GameDevelopmentPatternsandBestPractices_ColorImages.pd
f.
Errata
Although we have taken every care to ensure the accuracy of our content, mistakes do happen. If you
find a mistake in one of our books--maybe a mistake in the text or the code--we would be grateful if
you could report this to us. By doing so, you can save other readers from frustration and help us
improve subsequent versions of this book. If you find any errata, please report them by visiting http://w
ww.packtpub.com/submit-errata, selecting your book, clicking on the Errata Submission Form link, and
entering the details of your errata. Once your errata are verified, your submission will be accepted
and the errata will be uploaded to our website or added to any list of existing errata under the Errata
section of that title.
To view the previously submitted errata, go to https://www.packtpub.com/books/content/support and enter the

name of the book in the search field. The required information will appear under the Errata section.
Piracy
Piracy of copyrighted material on the Internet is an ongoing problem across all media. At Packt, we
take the protection of our copyright and licenses very seriously. If you come across any illegal copies
of our works in any form on the Internet, please provide us with the location address or website name
immediately so that we can pursue a remedy.
Please contact us at [email protected] with a link to the suspected pirated material.
We appreciate your help in protecting our authors and our ability to bring you valuable content.
Questions
If you have a problem with any aspect of this book, you can contact us at [email protected], and we
will do our best to address the problem.
Introduction to Design Patterns
You've learned how to program, and you've probably created some simple games at this point, but
now you want to start building something larger. Perhaps you have tried building an interesting
project but you felt like the code was hacked together. Maybe you worked with a team of
programmers and you couldn't see eye-to-eye on how to solve problems. Maybe your code didn't
integrate well, or features were constantly being added that didn't fit with your original design.
Maybe there wasn't a design to begin with. When building larger game projects, it's important that you
break apart your problems, focus on writing quality code, and spend your time solving problems
unique to your game, as opposed to common programming problems that already have a solution. The
old advice don't reinvent the wheel applies to programming as well. One could say that instead of
just being someone that writes code, you now need to think like a game developer or software
engineer.
Knowing how to program is very similar to knowing a language. It's one thing to use a language to
make conversation, but it's quite different if you're trying to create a novel or write poetry. In much the
same way as when programmers are writing code in their game projects, you'll need to pick the right
parts of the language to use at the best time. To organize your code well, as well as to solve problems
that arise time and time again, you'll need to have certain tools. These tools, design patterns, are
exactly what this book is about.
Chapter overview
Over the course of this chapter, we will be discussing the idea of design patterns as well as the
thought processes to be going through when deciding to use them. We will also be setting up our
project and installing everything necessary to work with the Mach5 engine, which was written by one
of the authors.
Your objective
This chapter will be split into a number of topics. It will contain a simple step-by-step process from
beginning to end. Here is the outline of our tasks:
What are design patterns?

Separating the what and how
An introduction to interfaces
The problems with using design patterns in games
Project setup
What are design patterns
Famously documented in the book Design Patterns: Elements of Reusable Object-Oriented Software
by Erich Gamma, John Vlissides, Ralph Johnson, and Richard Helm, also known as the Gang of
Four (GoF for short), design patterns are solutions for common programming problems. More than
that, they are solutions that were designed and redesigned as developers tried to get more flexibility
and reuse from their code. You don't need to have read the Gang of Four's book in order to understand
this book, but after finishing you may wish to read or reread that book to gain additional insights.
As the Gang of Four title suggests, design patterns are reusable, meaning the implemented solution
can be reused in the same project, or a completely new one. As programmers, we want to be as
efficient as possible. We don't want to spend time writing the same code over and over, and we
shouldn't want to spend time solving a problem that already has an answer. An important
programming principle to follow is the DRY principle, Don't Repeat Yourself. By using and reusing
design patterns, we can prevent issues or silly mistakes that would cause problems down the road. In
addition, design patterns can improve the readability of your code not only by breaking apart sections
that you would have put together, but also by using solutions that other developers are (hopefully)
familiar with.
When you understand and use design patterns, you can shorten the length of a discussion with another
developer. It is much easier to tell another programmer that they should implement a factory than to go
into a lengthy discussion involving diagrams and a whiteboard. In the best-case scenario, you both
know about design patters well enough that there doesn't need to be a discussion, because the solution
would be obvious.
Although design patterns are important, they aren't just a library that we can just plug into our game.
Rather, they are a level above libraries. They are methods for solving common problems, but the
details of implementing them is always going to be unique to your project. However, once you have a
good working knowledge of patterns, implementing them is easy and will feel natural. You can apply
them when first designing your project, using them like a blueprint or starting point. You can also use
them to rework old code if you notice that it's becoming jumbled (something we refer to as spaghetti
code). Either way, it is worth studying patterns so your code quality will improve and your
programming toolbox will grow larger.
With this toolbox, the number of ways to solve a problem is limited only by your imagination. It can
sometimes be difficult to think of the best solution right off the bat. It can be difficult to know the best
place or best pattern to use in a given situation. Unfortunately, when implemented in the wrong place,
design patterns can create many problems, such as needlessly increasing the complexity of your
project with little gain. As I mentioned before, software design is similar to writing poetry in that they
are both an art. There will be advantages and disadvantages to the choices you make.
That means in order to use patterns effectively, you first need to know what problem you are trying to
solve in your project. Then you must know the design patterns well enough to understand which one
will help you. Finally, you'll need to know the specific pattern you are using well enough so you can
adapt it to your project and your situation. The goal of this book is to provide you with this in-depth
knowledge so you can always use the correct tool for the job.
There are many design patterns out there, including the foundational patterns from
the Gang of Four book, architectural patterns, and many more. We will only be
touching on the ones that we feel are best used for game development. We feel it is
better to supply you with deep knowledge of a select few patterns than to give you a
primer on every possible pattern out there. If you're interested in learning more
about all of the ones out there, feel free to visit https://en.wikipedia.org/wiki/Software_design_pa
ttern .
In my many years doing game development, one thing that has always been constant is that a project
never ends up 100% the same as it was imagined in the pre-production phase. Features are added and
removed at a moment's notice and things that you think are pivotal to the game experience will get
replaced with something completely different. Many people can be involved in game development,
such as producers, game directors, designers, quality assurance, or even marketing, so we can never
tell who, where, or when those changes will be made to the project.
Since we never can tell what will be changed, it's a good practice to always write your code so it can
be easily modified. This will involve planning ahead much more than you might be used to, and
typically involves either a drawn flowchart, some form of pseudocode, or possibly both. However,
this planning will get you much further much faster than jumping straight to coding.
Sometimes you may be starting a project from scratch; other times you may be joining a game team
and using an existing framework. Either way, it is important to start coding with a plan. Writing code
is called software engineering for a reason. The structure of code is often likened to building or
architecting. However, let's think in smaller terms for now. Let's say you want to build some furniture
from IKEA.
When you buy your furniture, you receive it unassembled with an instruction manual. If you were to
start building it without following the instructions, it is possible you wouldn't finish it at all. Even if
you did eventually finish, you may have assembled things out of order, causing much more work. It's
much better to have a blueprint that shows you every step along the way.
Unfortunately, building a game is not exactly like following an instruction manual for furniture. In the
case of games and software of any kind, the requirements from the client might constantly change. Our
client might be the producer that has an updated timeline for us to follow. It might be our designer that
just thought of a new feature we must have. It might even be our play testers. If they don't think the
game is fun, we shouldn't just keep moving along making a bad game. We need to stop, think about our
design, and try something new.
Having a plan and knowing a project will change seem to be in opposition to each other. How can we
have a plan if we don't know what our end product will be like? The answer is to plan for that
change. That means writing code in such a way that making changes to the design is fast and easy. We
want to write code so that changing the starting location on the second level doesn't force us to edit
code and rebuild all the configurations of the project. Instead, it should be as simple as changing a
text file, or better yet, letting a designer control everything from a tool.
Writing code like this takes work and planning. We need to think about the design of the code and
make sure it can handle change. Oftentimes this planning will involve other programmers. If you are
working on a team, it helps if everyone can understand the goal of each class and how it connects
with every other class. It is important to have some standards in place so others can start on or
continue with the project without you there.
Understanding UML class diagrams
Software developers have their own form of blueprints as well, but they look different from what you
may be used to. In order to create them, developers use a format called Unified Markup Language,
or UML for short. This simple diagramming style was primarily created by Jim Rumbaugh, Grady
Booch, and Ivar Jacobson and has become a standard in software development due to the fact that it
works with any programming language. We will be using them when we need to display details or
concepts to you via diagrams.
Design patterns are usually best explained through the use of class diagrams, as you're able to give a
demonstration of the idea while remaining abstracted. Let's consider the following class:
class Enemy
{
public:
void GetHealth(void) const;
void SetHealth(int);
private:
int currentHealth;
int maxHealth;
};
Converted to UML, it would look something like this:
Basic UML diagrams consist of three boxes that represent classes and the data that they contain. The
top box is the name of the class. Going down, you'll see the properties or variables the class will
have (also referred to as the data members) and then in the bottom box you'll see the functions that it
will have. A plus symbol (+) to the left of the property means that it is going to be public, while a
minus symbol (-) means it'll be private. For functions, you'll see that whatever is to the right of the
colon symbol (:) is the return type of the function. It can also include parentheses, which will show
the input parameters for the functions. Some functions don't need them, so we don't need to place
them. Also, note in this case I did add void as the return type for both functions, but that is optional.
Relationships between classes
Of course, that class was fairly simple. In most programs, we also have multiple classes and they can
relate to each other in different ways. Here's a more in-depth example, showing the relationships
between classes.
Inheritance
First of all, we have inheritance, which shows the IS-A relationship between classes.
class FlyingEnemy: public Enemy
{
public:
void Fly(void);
private:
int flySpeed;
};
When an object inherits from another object, it has all of the methods and fields that are contained in
the parent class, while also adding their own content and features. In this instance, we have a special
FlyingEnemy, which has the ability to fly in addition to all of the functionality of the Enemy class.
In UML, this is normally shown by a solid line with a hollow arrow and looks like the following:
Aggregation
The next idea is aggregation, which is designated by the HAS-A relationship. This is when a single
class contains a collection of instances of other classes that are obtained from somewhere else in
your program. These are considered to have a weak HAS-A relationship as they can exist outside of
the confines of the class.
In this case, I created a new class called CombatEncounter which can have an unlimited number of
enemies that can be added to it. However when using aggregation, those enemies will exist before the
CombatEncounter starts; and when it finishes, they will also still exist. Through code it would look
something like this:
class CombatEncounter
{
public:
void AddEnemy(Enemy* pEnemy);
private:
std::list<Enemy*> enemies;
};
Inside of UML, it would look like this:

Composition
When using composition, this is a strong HAS-A relationship, and this is when a class contains one or
more instances of another class. Unlike aggregation, these instances are not created on their own but,
instead, are created in the constructor of the class and then destroyed by its destructor. Put into
layman's terms, they can't exist separately from the whole.
In this case, we have created some new properties for the Enemy class, adding in combat skills that it
can use, as in the Pokmon series. In this case, for every one enemy, there are four skills that the
enemy will be able to have:
class AttackSkill
{
public:
void UseAttack(void);
private:
int damage;
float cooldown;
};
class Enemy
{
public:
void GetHealth(void) const;
void SetHealth(int);
private:
int currentHealth;
int maxHealth;
AttackSkill skill1;
AttackSkill skill2;
AttackSkill skill3;
AttackSkill skill4;
};
The line in the diagram looks similar to aggregation, aside from the fact that the diamond is filled in:
Implements
Finally, we have implements, which we will talk about in the Introduction to interfaces section.
The advantage to this form of communication is that the ideas presented will work the same way no
matter what programming language you're using and without a specific implementation. That's not to
say that a specific implementation isn't valuable, which is why we will also include the
implementation for problems in code as well.
There's a lot more information out there about UML and there are various different
formats that different people like to use. A nice guide that I found that may be of
interest can be found at https://cppcodetips.wordpress.com/2013/12/23/uml-class-diagram-explained-wit
h-c-samples/.
Separating the why and the how
When creating games, we have many different systems that need to be juggled around in order to
provide the entire game experience. We need to have objects that are drawn to the screen, need to
have realistic physics, react when they hit each other, animate, have gameplay behavior and, on top of
all that, we then need to make sure that it runs well 60 times every second.
Understanding the separation of concerns
Each of these different aspects is a problem of its own, and trying to solve all of these issues at once
would be quite a headache. One of the most important concepts to learn as a developer is the idea of
compartmentalizing problems, and breaking them apart into simpler and simpler pieces until they're
all manageable. In computer science, there is a design principle known as the separation of concerns
which deals with this issue. In this aspect, a concern would be something that will change the code of
a program. Keeping this in mind, we would separate each of these concerns into their own distinct
sections, with as little overlap in functionality as possible. Alternatively, we can make it so that each
section solves a separate concern.
Now when we mention concerns, they are a distinct feature or a distinct section. Keeping that in mind,
it can either be something as high level as an entire class or as low level as a function. By breaking
apart these concerns into self-contained pieces that can work entirely on their own, we gain some
distinct advantages. By separating each system and making it so they do not depend on each other, we
can alter or extend any part of our project with minimal hassle. This concept creates the basis for
almost every single design pattern that we'll be discussing.
By using this separation effectively, we can create code that is flexible, modular, and easy to
understand. It'll also allow us to build the project in a much more iterative way because each class
and function has its own clearly defined purpose. We won't have to worry nearly as much about
adding new features that would break previously written code because the dependencies are on the
existing functional classes, and never the other way around. This means we can to easily expand the
game with things like Downloadable Content (DLC). This might include new game types, additional
players, or new enemies with their own unique artificial intelligence. Finally, we can take things
we've already written and decouple them from the engine so we can use them for future projects,
saving time and development costs.
An Introduction to interfaces
One of the main features of using design patterns is the idea of always programming to an interface
and not to an implementation. In other words, the top of any class hierarchy should have an abstract
class or an interface.
Polymorphism refresher
In Hollywood, lots of actors and actresses take on many different roles when filming movies. They
can be the hero of a story, a villain, or anything else as they inhabit a role. No matter what role
they've gotten, when they are being filmed they are acting even if what they do specifically can be
quite different. This kind of behavior acts similarly to the idea of polymorphism.
Polymorphism is one of the three pillars of an object-oriented language (along with encapsulation and
inheritance). It comes from the words poly meaning many and morph meaning change.
Polymorphism is a way to call different specific class functions in an inheritance hierarchy, even
though our code only uses a single type. That single type, the base class reference, will be changed
many ways depending on the derived type. Continuing with the Hollywood example, we can tell an
actor to act out a role and, based on what they've been cast in, they will do something different.
By using the virtual keyword on a base class function and overriding that function in a derived class,
we can gain the ability to call that derived class function from a base class reference. While it may
seem a bit complex at first, this will seem clearer with an example. For instance, if we have the
following class:
class Animal
{
public:
virtual void Speak(void) const //virtual in the base class
{
//Using the Mach 5 console print
M5DEBUG_PRINT("...\n");
}
};
I could create a derived class with its own method, without modifying the base class in any way. In
addition, we have the ability to replace or override a method within a derived class without affecting
the base class. Let's say I wanted to change this function:
class Cat: public Animal
{
public:
void Speak(void) const //overridden in the derived class
{
M5DEBUG_PRINT("Meow\n");
}
void Purr(void) const //unrelated function

{
M5DEBUG_PRINT("*purr*\n");
}
};
class Dog: public Animal
{
public:
void Speak(void) const //overridden in the derived class
{
M5DEBUG_PRINT("Woof\n");
}
};
Since a derived class can be used anywhere a base class is needed, we can refer to derived classes
using a base class pointer or an array of pointers and call the correct function at runtime. Let's have a
look at the following code:
void SomeFunction(void)
{
const int SIZE = 2;
Cat cat;
Dog dog;
Animal* animals[SIZE] = {&cat, &dog};
for(int i = 0; i < SIZE; ++i)

{
animals[i]->Speak();
}
}
The following is the output of the preceding code:

Meow
Woof
As you can see, even though we have an array of base class pointers, the correct derived class
function is called. If the functions weren't marked as virtual, or if the derived classes didn't override
the correct functions, polymorphism wouldn't work.
Understanding interfaces
An interface implements no functions, but simply declares the methods that the class will support.
Then, all of the derived classes will do the implementation. In this way, the developer will have more
freedom to implement the functions to fit each instance, while having things work correctly due to the
nature of using an object-oriented language.
Interfaces may contain only static final variables, and they may contain only abstract methods, which
means that they cannot be implemented within the class. However, we can have interfaces that inherit
from other interfaces. When creating theses classes, we can implement whatever number of interfaces
we want to. This allows us to make classes become even more polymorphic but, by doing so, we are
agreeing that we will implement each of the functions defined in the interface. Because a class that
implements an interface extends from that base class, we would say that it has an IS-A relationship
with that type.
Now, interfaces have one disadvantage, and that's the fact that they tend to require a lot of coding to
implement each of the different versions as needed, but we will talk about ways to adjust and/or fix
this issue over the course of this book.
In C++, there isn't an official concept of interfaces, but you can simulate the behavior of interfaces by
creating an abstract class.
Here's a simple example of an interface, and an implementation of it:

class Enemy
{
public:
virtual ~Enemy(void) {/*Empty virtual destructor*/}
virtual void DisplayInfo(void) = 0;
virtual void Attack(void) = 0;
virtual void Move(void) = 0;
};
class FakeEnemy: public Enemy

{
public:
virtual void DisplayInfo(void)
{
M5DEBUG_PRINT("I am a FAKE enemy");
}
virtual void Attack(void)

{
M5DEBUG_PRINT("I cannot attack");
}
virtual void Move(void)

{
M5DEBUG_PRINT("I cannot move");
}
};
And here's how it looks in UML:

One important difference between procedural programming (think C-style) and object-oriented
programming is the ability to encapsulate or compartmentalize code. Oftentimes we think of this as
just data hiding: making variables private. In a C-style program, the functions and data are separate,
but it is hard to reuse any one function because it might depend on other functions or other pieces of
data in the program. In object-oriented programming, we are allowed to group the data and function
together into reusable pieces. That means we can (hopefully) take a class or module and place it in a
new project. This also means that since the data is private, a variable can be easily changed as long
as the interface or public methods don't change. These concepts of encapsulation are important, but
they aren't showing us all of the power that this provides us.
The goal of writing object-oriented code is to create objects that are responsible for themselves.
Using a lot of if/else or switch statements within your code can be a symptom of bad design. For
example, if I have three classes that need to read data from a text file, I have the choice of using a
switch statement to read the data differently for each class type, or passing the text file to a class
method and letting the class read the data itself. This is even more powerful when combined with the
power of inheritance and polymorphism.
By making the classes responsible for themselves, the classes can change without breaking other
code, and the other code can change without breaking the classes. We can all imagine how fragile the
code would be if a game was written entirely in the main function. Anything that is added or removed
is likely to break other code. Anytime a new member joined the team, they would need to understand
absolutely every line and every variable in the game before they could be trusted to write anything.
By separating code into functions or classes, we are making the code easier to read, test, debug, and
maintain. Anyone joining the team would of course need to understand some pieces of the code, but it
might not be necessary to understand all of graphics if they are working on game logic or file loading.
Design patterns are solutions to common programming problems flexible enough to handle change.
They do this by compartmentalizing sections of code. This isn't by accident. For the purposes of this
book, the definition of good design is encapsulated, flexible, reusable code. So it should come as no
surprise that these solutions are organized into classes or groups of classes that encapsulate the
changing sections of your code.
The structure of the Mach5 engine
Throughout this book, we will be using design patterns to solve common game programming
problems. The best way to do this is by example, and so we will be examining how these problems
arise and implementing the solutions using the Mach5 engine, a 2D game engine designed in C++ by
Matt Casanova. By looking at the entire source code for a game, we will be able to see how many of
the patterns work together to create powerful and easy-to-use systems.
However, before we can dive into the patterns, we should spend a little time explaining the structure
of the engine. You don't need to understand every line of source code, but it is important to understand
some of the core engine components and how they are used. This way we can better understand the
problems we will be facing and how the solution fits together.
While looking at the diagram, it may seem a little confusing at first, so let's examine each piece of the
engine separately.
Mach5 core engines and systems
The meaning of engine is getting a little blurred these days. Often when people talk of engines they
think of entire game creation tools such as Unreal or Unity. While these are engines, the term didn't
always require a tool. Game engines such as Id Software's Quake Engine or Valve Corporation's
Source engine existed independently of tools, although the latter did have tools including the Hammer
Editor for creating levels.
The term engine is also used to refer to components within the larger code base. This includes things
like a rendering engine, audio engine, or physics engine. Even these can be created completely
separate from a larger code base. Orge 3D is an open source 3D graphics engine, while the Havok
Physics engine is proprietary software created by the Havok company and used in many games.
So, when we talk about the engines or systems of the Mach5 engine, we are simply referring to groups
of related code for performing a specific task.
The app
The M5App or application layer is a class responsible for interfacing with the operating system. Since
we are trying to write clean, reusable code, it is important that we don't mix our game code with any
operating system function calls. If we did this, our game would be difficult to port to another system.
The M5App class is created in WinMain and responsible for creating and destroying every other system.
Anytime our game needs to interact with the OS, including changing resolution, switching to full
screen, or getting input from a device, we will use the M5App class. In our case, the operating system
that we will be using will be Windows.
The StageManager
The M5StageManager class is responsible for controlling the logic of each stage. We consider things such
as the main menu, credits screen, options menu, loading screen, and playable levels to be stages. They
contain behaviors that control the flow of the game. Examples of stage behavior include reading game
object data from files, spawning units after specific time intervals, or switching between menus and
levels.
StageManageris certainly not a standardized name. In other engines, this section of code may be called
the game logic engine; however, most of our game logic will be separated into components so this
name doesn't fit. No matter what it is called, this class will control which objects need to be created
for the current stage, as well as when to switch to the next stage or quit the game altogether.
Even though this uses the name manager instead of engine, it serves as one of the core systems of the
game. This class controls the main game loop and manages the collection of user stages. In order to
make a game, users must derive at least one class from the base M5Stage class and overload the virtual
functions to implement their game logic.
The ObjectManager
The M5ObjectManager is responsible for creating, destroying, updating, and searching for game objects. A
game object is anything visible or invisible in the game. This could include the player, bullets,
enemies, and triggers--the invisible regions in a game that cause events when collided with. The
derived M5Stage classes will use the M5ObjectManager to create the appropriate objects for the stage. They
can also search for specific game objects to update game logic. For example, a stage may search for a
player object. If one doesn't exist, the manager will switch to the game over stage.
As seen in the previous diagram, our game will use components. This means the M5ObjectManager will be
responsible for creating those as well.
The graphics engine
This book isn't about creating a graphics engine but we do need one to draw to the screen. Similar to
how the M5App class encapsulates important OS function calls, our M5Gfx class encapsulates our graphics
API. We want to make sure there is a clear separation between any API calls and our game logic.
This is important so we can port our game to another system. For example, we may want to develop
our game for PC, XBox One, and PlayStation 4. This will mean supporting multiple graphics APIs
since a single API isn't available for all platforms. If our game logic contains API code, then those
files will need to be modified for every platform.
We won't be going deep into the details of how to implement a full graphics engine, but we give an
overview of how graphics works. Think of this as a primer to the world of graphics engines.
This class allows us manipulate and draw textures, as well as control the game camera and find the
visible extents of the world. M5Gfx also manages two arrays of graphics components, one for world
space and one for screen space. The most common use of the screen space components is for creating
User Interface (UI) elements such as buttons.
Tools and utilities
Besides the core engines and systems for a game, every engine should provide some basics tools and
support code. The Mach5 engine includes a few categories for tools:
Debug Tools: This includes debug asserts, message windows, and creating a debug console
Random: Helper functions to create random int or float from min/max values
Math: This includes 2D vectors and 4 x 4 matrices, as well some more general math helper
functions
FileIO: Support for reading and writing .ini files
The problems with using design patterns in
games
Unfortunately, there are also some issues that may come into play from using design patterns exactly
as described. It's often said that the fastest executing code is the code that is never called, and using
design patterns will typically require you to add more code to your project than what you would have
done otherwise. This will have a performance cost as well, as there will likely need to be more
calculations done whenever you're using a part of your engine.
For instance, using some principles will cause some classes that you write to become extremely
bloated with extra code. Design patterns are another form of complexity to add to your project. If the
problem itself is simple, it can be a much better idea to focus on the simpler solutions before going
straight into implementing a design pattern just because you have heard of it.
Sometimes it's better to follow the simple rule of K.I.S.S. and remember that it is the knowledge of
the pattern that holds the most important value, not using the pattern itself.
Setting up the project
Now that we've gotten a good understanding of why we would want to use design patterns, let's get
set up the game engine that we will be using over the course of the book: the Mach5 engine. Now in
order to get started, we will need to download the engine as well as the software needed to run the
project. Perform the following steps:
1. Open up your web browser of choice and visit the following website: https://beta.visualstudio.com/downl
oads/. Once there, move to the Visual Studio Community version on the left and then click on the
Free download option, as shown in the following screenshot:
2. If you get a window asking what to do with the file, go ahead and open it or save and then open
it by clicking on the Run button:
3. From there, wait until the installer pops up, then select Custom, and then click on Next to start
downloading the program:
4. Now once you get to the Features section, uncheck whatever is selected and then open up the
Programming Languages tab and check Visual C++. You may go ahead and remove the other
options, as we will not be using them. Then go ahead and click on the Next button, then Install,
and allow it to make changes to your computer:
You may need to wait a while at this point, so go ahead and get yourself a coffee, and once it's
finished you'll need to restart your computer. After that, go ahead and continue with the
project.
5. Once you've finished installing, you next need to actually install the engine itself. With that in
mind, go over to https://github.com/mattCasanova/Mach5 and from there, click on the Clone or download
section and then click on Download ZIP:
6. Once you're finished with the download, go ahead and unzip the file to a folder of your choice;
then open up the Mach5-master\EngineTest folder, double-click on the EngineTest.sln file, and start up
Visual Studio.
7. You may get a login screen asking you to log in; go ahead and sign up or press the Not now,
maybe later option on the bottom of the screen. You can then pick a color theme; then click Start
Visual Studio.
8. Upon starting, you may get a security warning asking if you'd still like to open this project. This
is displayed from any Visual Studio solution that wasn't made on your machine, so it wants to
make sure that you know where it came from, but in this case the project is perfectly safe. Go
ahead and uncheck the Ask me for every project in this solution option and then select OK:
9. Once it's finished loading, you should finally see the Visual Studio interface, which should look
like this:
Visual Studio is a very powerful tool and, for developers, it can be quite useful to learn all of
the functionality that it has. We'll be discussing the features as we use them, but this book
shouldn't be considered the end-all book on Visual Studio.
If you're interested in learning more about the Visual Studio interface, check out: https
://msdn.microsoft.com/en-us/library/jj620919.aspxa .
10. The engine is built to work on 32-bit processors, so change the x64 dropdown to x86 and then
click the Play button or press F5. It will then ask if you wish to rebuild the project. Go ahead
and say Yes. If all goes well, you should eventually see a debug window and a gameplay
window as well. After a few seconds, it should transition to a simple default project:
You can play around by using the W, A, and D keys to move the character around and the
Spacebar to shoot bullets at enemies. Once you're finished playing around, go ahead and hit
the Esc key to go to the menu, and then click on the Quit button to leave the project and go
back to the editor!
Summary
And there we have it! In this first chapter, you've learned some fundamentals about design patterns
and also got the Mach5 engine running on your computer.
Specifically, we learned that design patterns are solutions for common programming problems. There
are a lot of reasons why we should use them, but in order to use patterns effectively, you first need to
know what problem you are trying to solve and which ones can help you in that instance, which is
what this book intends to teach you.
We learned how game development is always changing and how important it is to have a plan, as
well as an architecture that can support those changes. With that in mind, we learned about various
aspects of coding that will be used in the creation of our architecture.
We dived into learning about the separation of concerns principle and how important it is for us to
separate the what and how; making it so they do not depend on each other allows us to alter or extend
any part of our project with minimal hassle. Afterwards, we explored what interfaces were and how
they are useful in giving us a foundation we can build on. Later, we dived into the Mach5 engine, saw
an example of how compartmentalized code worked, and the advantages of it. We also saw how using
design patterns in games can be a great thing, as well as the problems that they have.
Finally, we downloaded the Mach5 engine ourselves and made sure that it worked correctly. Moving
on, in the next chapter, we will tackle our first design pattern, the Singleton, and see how it can be
useful to us!
One Instance to Rule Them All - Singletons
Now that we've learned what a design pattern is, as well as why we'd want to use them, let's first talk
about a design pattern that most people learn, the Singleton pattern.
The Singleton pattern is probably the most well-known pattern and it is also the one out there that is
most often misused. It definitely has the most controversy surrounding it, so when discussing this
pattern it is as important (or even more important) to know when not to apply it.
Chapter overview
In this chapter, we will explain about the pattern and many arguments for and against it. We will
describe how and why core systems within the Mach5 engine such as the Graphics Engine and Object
Manager are utilized as Singletons. Finally, we will explain a number of different ways to implement
this in C++, along with the pros and cons of each choice.
Your objective
An overview of class access specifiers

Pros and cons of global access
Understanding the static keyword
The advantages and disadvantages of only one instance
The Singleton in action: the Application class
Design decisions
An overview on class access specifiers
When using an object-oriented programming language, one of the most important features included is
the ability to hide data, preventing classes from accessing properties and functions of another class
type by default. By using access specifiers such as public, private, and protected, we can dictate
specifically how the data and/or functions can be accessed from other classes:
class ModifierExamples
{
public int publicInteger;
private void PrivateMethod() {}
protected float protectedNumber;
};
A class can have unlimited variables or functions that are public, private, or protected and can even
control access to entire sections of the class:
class MoreModifierExamples
{
public:
// until we reach another tag, all variables and functions
// will be public
int publicIntegar;
int anotherExample;
private:
// Now, they'll be private
void PrivateFunction() {}
double safeValue;
protected:
// And now... protected
float protectedNumber;
int AlsoProtected() { return 0; }
};
When you place a labelled section with an access modifier's name and a : next to it, until there is
another section label, all of the parts of the class that are listed will use that specific one.
When we use the public access modifier, we are saying that this variable or function can be used or
accessed from anywhere within our program, even outside of the class we've created. Declaring a
variable outside of a function or class, or marking a variable as public and static, is often referred to
as being a global variable. We will be talking about global variables in the next section, but for right
now, let's go over the other access specifiers as well.
When private is used, we are restricting the usage of our variable or function to being allowed only
inside of the class, or from friend functions. By default, all of the variables and functions in a class are
private.
For more information on friend functions, check out http://en.cppreference.com/w/cpp/languag

e/friend .
The third type, protected, is the same as a private type except that it can still be accessed by child (or
derived) classes. This can be quite useful when using inheritance so you can still access those
variables and/or functions.
The static keyword
Another thing that is important to know before diving into the Singleton pattern is what the static
keyword means, as it's something that we will be using the functionality of when building this pattern.
When we use the static keyword, there are three main contexts that it'll be used in:
Inside a function
Inside a class definition
In front of a global variable in a program with multiple files
Static keyword inside a function
The first one, being used inside of a function, basically means that once the variable has been
initialized, it will stay in the computer's memory until the end of the program, keeping the value that it
has through multiple runs of the function. A simple example would be something like this:
#include <string>
{
public:
void InFunction()
{

enemyCount += 10;

}
};
Now if we were to call this, it would look something like the following:
StaticExamples se;
se.InFunction();
se.InFunction();
And when we call it, the following would be displayed:
As you can see, the value continues to exist, and we can access and/or modify its contents as we see
fit in the function. This could be used for a number of things, such as maybe needing to know what
happened the last time that you called this function, or to store any kind of data between any calls. It's
also worth noting that static variables are shared by all instances of the class, and due to that, if we
had two variables of type StaticExamples, they would both display the same enemyCount. We will utilize
the fact that, if an object is created this way, it will always be available later on in this chapter.
Static keyword in class definitions
The second way is by having a variable or function in a class being defined as static. Normally, when
you create an instance of a class, the compiler has to set aside additional memory for each variable
that is contained inside of the class in consecutive blocks of memory. When we declare something as
static, instead of creating a new variable to hold data, a single variable is shared by all of the
instances of the class. In addition, since it's shared by all of the copies, you don't need to have an
instance of the class to call it. Take a look at the following bolded code to create our variable:
{
public:
static float classVariable;
static void StaticFunction()
{
// Note, can only use static variables and functions within
// static function
std::string toDisplay = "\n I can be called anywhere!
classVariable value: " + std::to_string(classVariable);
}
void InFunction()
{

enemyCount += 10;

}
};
Now, in the preceding code we define a variable and a function, but this isn't all the prep work we
need to do. When creating a static variable, you cannot initialize it from within the class, and instead
need to do it in a .cpp file instead of the .h file we could use for the class definition. You'll get errors
if you do not initialize it, so it's a good idea to do that. In our case, it'd look like the following:
// StaticExamples.cpp
float StaticExamples::classVariable = 2.5f;
Note that, when we initialize, we also need to include the type, but we use the ClassName::variableName
template similar to how you define functions in .cpp files. Now that everything's set up, let's see how
we can access them inside our normal code:
StaticExamples::StaticFunction();
StaticExamples::classVariable = 5;
StaticExamples::StaticFunction();
Note that instead of accessing it via creating a variable, we can instead just use the class name
followed by the scope operator (::) and then select which static variable or function we'd like to use.
When we run it, it'll look like this:
As you can see, it works perfectly!
Static as a file global variable
As you may be aware, C++ is a programming language closely related to the C programming
language. C++ was designed to have most of the same functionality that C had and then added more
things to it. C was not object-oriented, and so, when it created the static keyword, it was used to
indicate that source code in other files that are part of your project cannot access the variable, and
that only code inside of your file can use it. This was designed to create class-like behavior in C.
Since we have classes in C++ we don't typically use it, but I felt I should mention it for completeness.
Pros and cons of global variables
To reiterate, a global variable is a variable that is declared outside of a function or class. Doing this
makes our variable accessible in every function, hence us calling it global. When being taught
programming in school, we were often told that global variables are a bad thing or at least, that
modifying global variables in a function is considered to be poor programming practice.
There are numerous reasons why using global variables is a bad idea:
Source code is the easiest to understand when the scope of the elements used is limited. Adding
in global variables that can be read or modified anywhere in the program makes it much harder
to keep track of where things are being done, as well as making it harder to comprehend when
bringing on new developers.
Since a global variable can be modified anywhere, we lose any control over being able to
confirm that the data contained in the variable is valid. For instance, you may only want to
support up to a certain number, but as a global variable this is impossible to stop. Generally, we
advise using getter/setter functions instead for this reason.
Using global variables tightens how coupled our programs are, making it difficult to reuse
aspects of our projects as we need to grab from a lot of different places to make things work.
Grouping things that are connected to each other tends to improve projects.
When working with the linker, if your global variable names are common, you'll often have
issues when compiling your project. Thankfully, you'll get an error and have to fix the issue in
this case. Unfortunately, you may also have an issue where you are trying to use a locally scoped
variable in a project but end up selecting the global version due to mistyping the name or
relying too heavily on intelligence and selecting the first thing you see, which I see students
doing on multiple occasions.
As the size of projects grow, it becomes much harder to do maintenance and/or make changes
to/on global variables, as you may need to modify many parts of your code to have it adjust
correctly.
This isn't to say that global access is entirely bad. There are some reasons why one would consider
using it in their projects:
Not knowing what a local variable is

Not understanding how to create classes
Wanting to save keystrokes
Not wanting to pass around variables all the time to functions
Not knowing where to declare a variable, so making it global means anyone can get it
To simplify our project for components that need to be accessible anywhere within the project
Aside from the last point, those issues are really bad reasons for wanting to use global variables, as
they may save you some time up front, but as your projects get larger and larger it'll be a lot more
difficult to read your code. In addition, once you make something global it's going to be a lot more
difficult to convert it to not be global down the road. Think that, instead of using global variables, you
could instead pass parameters to different functions as needed, making it easier to understand what
each function does and what it needs to work with to facilitate its functionality.
That's not to say that there isn't any time when using a global variable is a reasonable or even a good
idea. When global variables represent components that truly need to be available throughout your
project, the use of global variables simplifies the code of your project, which is similar to what we
are aiming to accomplish.
Norm Matloff also has an article explaining times that he feels like global variables
are necessary when writing code. If you want to hear an alternative take, check out htt
p://heather.cs.ucdavis.edu/~matloff/globals.html.
Basically, always limit your variables to the minimal scope needed for the project and not any more.
This especially comes to mind when you only ever need one of something, but plan to use that one
object with many different things. That's the general idea of the Singleton design pattern and is the
reason why it's important that we understand the general usage before moving onwards.
The Singleton pattern in a nutshell is where you have a class that you can access anywhere within
your project, due to the fact that only one object (instance) of that class is created (instantiated). The
pattern provides a way for programmers to give access to a class's information globally by creating a
single instance of an object in your game.
Whereas there are quite a few issues with using global variables, you can think of a Singleton as an
improved global variable due to the fact that you cannot create more than one. With this in mind, the
Singleton pattern is an attractive choice for classes that only have a unique instance in your game
project, such as your graphics pipeline and input libraries, as having more than one of these in your
projects doesn't make sense.
This single object uses a static variable and static functions to be able to access the object without
having to pass it through all of our code.
In the Mach5 engine, Singletons are used for the application's, input, graphics, and physics engines.
They are also used for the resource manager, object manager, and the game state manager. We will be
taking a much closer look at one of the more foundational ones in the engine, the Application class, later
on in this chapter. But before we get to it, let's dive into how we can actually create one of our very
own.
There are multiple ways to implement the Singleton pattern or to get Singleton-like behavior. We'll go
over some of the commonly seen versions and their pros and cons before moving to our final version,
which is how the Mach5 engine uses it.
One very common way of implementing the functionality of the Singleton pattern would look
something like the following:
Through code, it will look a little something like this:

class Singleton
{
public:
static Singleton * GetInstance()
{
// If the instance does not exist, create one
if (!instance)
{
instance = new Singleton;
}
return instance;
}
private:
static Singleton * instance;
};
In this class, we have a function called GetInstance and a single property called instance. Note that we
are using pointers in this instance, and only allocating memory to create our Singleton if we are
actually using it. The instance property represents the one and only version of our class, hence it
being made static. As it is private though, there is no way for others to access its data unless we give
them access to it. In order to give this access, we created the GetInstance function. This function will
first check whether instance exists and if it doesn't yet, it will dynamically allocate the memory to
create one, set instance to it, and then return the object.
This will only work if instance is properly set to 0 or nullptr when initialized, which
thankfully is the default behavior of static pointers in C++.
Keeping the single in Singleton
As we've mentioned previously, one of the most important parts of the Singleton pattern is the fact that
there is only one of those objects. That causes some issues with the original code that we've written,
namely that with some simple usage of C++ it is quite easy to have more than one of these classes
created by other programmers on your team. First and most obviously, they can just create a Singleton
variable (a variable of type Singleton) like the following:
Singleton singleton;
In addition, as a higher-level programming language, C++ will try to do some things automatically for
you when creating classes to eliminate some of the busy work that would be involved otherwise. One
of these things is automatically creating some functionality between classes to enable you to create or
copy objects of a custom class that we refer to as a constructor and copy constructor. In our case, you
can also create a copy of your current object in the following way:
Singleton instanceCopy(*(Singleton::GetInstance()));
The compiler will also create a default destructor and an assignment operator, moving the data from
one object to the other.
Thankfully, that's a simple enough thing to fix. If we create these functions ourselves (declaring an
explicit version), C++ notes that we want to do something special, so it will not create the defaults.
So to fix our problem, we will just need to add an assignment operator and some constructors that are
private, which you can see in the bold code that we've changed:
class Singleton
{
public:
static Singleton * GetInstance()
{
if (!instance)
{
}
return instance;
}
private:
static Singleton * instance;
// Disable usability of silently generated functions

Singleton();
~Singleton();
Singleton(const Singleton &);
Singleton& operator=(const Singleton&);
};
If you are using C++ 11 or above, it is also possible for us to instead mark the
functions we don't want to use as deleted, which would look like this:
Singleton() = delete;
~Singleton() = delete;
Singleton(const Singleton &) = delete;
Singleton& operator=(const Singleton&) = delete;
For more information on the delete keyword, check out http://www.stroustrup.com/C++11FAQ.
html#default.
Another thing that may possibly be an issue is that instance is a pointer. This is because, as a pointer,
our users have the ability to call delete on it and we want to make sure that the object will always be
available for our users to access. To minimize this issue, we could change our pointer to be a
reference, instead, by changing the function to the following (note the return type and that we use
*instance now on the last line):
static Singleton& GetInstance()

{
if (!instance)
{
}
return *instance;
}
Programmers are used to working with references as aliases for objects that exist somewhere else in
our project. People would be surprised if they ever saw something like:
Singleton& singleton = Singleton::GetInstance();
delete &singleton;
While technically doable, programmers won't expect to ever use delete on the address of a reference.
The nice thing about using references is that, when you need them in code, you know that they exist
because they're managed somewhere else in the code--and you don't need to worry about how they
are used.
Deleting our object correctly
People also are used to looking for memory leaks with pointers and not references, so that perhaps
leaves us with an issue as, in our current code, we allocate memory but don't actually delete it.
Now, technically, we haven't created a memory leak. Memory leaks appear when you allocate data
and lose all of your references to it. Also, modern operating systems take care of deallocating a
process's memory when our project is quit.
That's not to say that it's a good thing though. Depending on what information the Singleton class uses,
we could have references to things that no longer exist at some point.
To have our object delete itself correctly, we need to destroy the Singleton when our game shuts
down. The only issue is we need to make sure that we do it only when we are sure no one will be
using the Singleton afterwards.
However, as we want to talk about best practices, it's much better for us to actually solve this issue
by removing resource leaks whenever we see them. A solution to this very problem was created by
Scott Meyers in his book More Effective C++, which uses some of the features of the compiler,
namely that a static variable located in a function will exist throughout our program's running time.
For instance, let's take the following function:
void SpawnEnemy()
{
static int numberOfEnemies = 0;
++numberOfEnemies;
// Spawn the enemy

}
The numberOfEnemies variable is created and has been initialized before any code in the project has been
executed, most likely when the game was being loaded. Then, once SpawnEnemy is called for the first
time, it will have already been set to 0 (or nullptr). Conveniently, as the object is not allocated
dynamically, the compiler will also create code so that, when the game exists, it will call the
deconstructor for our object automatically.
With that in mind, we can modify our Singleton class to the following:
class Singleton
{
public:
static Singleton & GetInstance()
{
static Singleton instance;
return instance;
}
private:
// Disable usability of silently generated functions
Singleton();
~Singleton();
Singleton(const Singleton &);
Singleton& operator=(const Singleton&);
};
Specifically note the changes we've made to the GetInstance function and the removal of our class
instance variable. This method provides the simplest way to destroy the Singleton class automatically
and it works fine for most purposes.
Another technique to add to your toolbox of programming concepts that we will use in the next section
is the idea of templates. Templates are a way for you to be able to create generic classes that can be
extended to have the same functionality for different datatypes. It's another form of abstraction, letting
you define a base set of behavior for a class without knowing what type of data will be used on it. If
you've used the STL before, you've already been using templates, perhaps without knowing it. That's
why the list class can contain any kind of object.
Here's an example of a simple templated class:

#include <iostream> // std::cout
template <class T>

class TemplateExample
{
public:
// Constructor
TemplateExample();
// Destructor
~TemplateExample();
// Function
T TemplatedFunction(T);
};
In this case, we created our TemplateExample class and it has three functions. The constructor and
deconstructor look normal, but then I have this TemplateFunction function which takes in an object of type
T, and returns an object of type T. This T comes from the first line of our example code with the
template <class T> section of our code. Anywhere that there is a T it will be replaced with whatever
class we want to use this template with.
Now, unlike regular functions, we have to define templated functions within our .h file, so that, when
we need to create an object using this template, it will know what the functions will do. In addition to
this, the syntax is also a bit different:
template <class T> TemplateExample<T>::TemplateExample()
{
printf("\nConstructor!");
}
template <class T> TemplateExample<T>::~TemplateExample()

{
printf("\nDeconstructor!");
}
template <class T> T TemplateExample<T>::TemplatedFunction(T obj)

{
std::cout << "\nValue: " << obj;
return obj;
}
In this example, I'm just printing out text to display when a certain functionality is called, but I also
want to point out the usage of std::cout and that using it will require you to add #include <iostream> to the
top of your file.
We are using the standard library's cout function in this instance, instead of the printf that we have been
using, because cout allows us to feed in obj--no matter what its type is--to display something, which
isn't possible with printf by default.
Once that's finished, we can go ahead and use this inside of our project:
TemplateExample<int> teInt;
teInt.TemplatedFunction(5);
TemplateExample<float> teFloat;
teFloat.TemplatedFunction(2.5);
TemplateExample<std::string> teString;
teString.TemplatedFunction("Testing");
As you can see, this will create three different kinds of TemplateExample class objects using different
types. When we call the TemplatedFunction function, it will print out exactly the way we were hoping:
Later on, when we learn about abstract types, we can use templates with them to handle any kind of
data. In our case right now, we are going to use this functionality to allow us to make as many
Singletons as we'd like!
Now, assuming we get our Singleton working just the way that we want it to, you may wish to create
more Singletons in the future. You could create them all from scratch, but a better thing to do is
instead create a consistent approach, creating templates and inheritance to create a single
implementation that you can use for any class. At the same time, we can also learn about an
alternative way of creating a Singleton class, which will look something like the following:
template <typename T>
class Singleton
{
public:
Singleton()
{
// Set our instance variable when we are created
if (instance == nullptr)
{
instance = static_cast<T*>(this);
}
else
{
// If instance already exists, we have a problem
printf("\nError: Trying to create more than one Singleton");
}
}
// Once destroyed, remove access to instance

virtual ~Singleton()
{
instance = nullptr;
}
// Get a reference to our instance

static T & GetInstance()
{
return *instance;
}
// Creates an instance of our instance

static void CreateInstance()
{
new T();
}
// Deletes the instance, needs to be called or resource leak

static void RemoveInstance()
{
delete instance;
}
private:
// Note, needs to be a declaration
static T * instance;
};
template <typename T> T * Singleton<T>::instance = nullptr;
You'll notice that most of the differences have to do with the class itself. The very first line in our
code above uses the template keyword which tells the compiler that we are creating a template, and
typename T tells the compiler that, when we create a new object using this, the type T will be replaced
with whatever the class we want it to be based on is.
I also want to point out the use of a static cast to convert our Singleton pointer to a T. static_cast is
used in code generally when you want to reverse an implicit conversion. It's important to note that
static_cast performs no runtime checks for if it's correct or not. This should be used if you know that
you refer to an object of a specific type, and thus a check would be unnecessary. In our case, it is safe
because we will be casting from a Singleton object to the type that we've derived from it (T).
Of course, it may be useful to see an example of this being used, so let's create an example of a class
that we could use as a Singleton, perhaps something to manage the high scores for our game:
class HighScoreManager : public Singleton<HighScoreManager>
{
public:
void CheckHighScore(int score);
private:
int highScore;
};
Notice here that, when we declare our HighScoreManager class, we say that it's derived from the Singleton
class and, in turn, we pass the HighScoreManager class to the Singleton template. This pattern is known as
the curiously recurring template pattern.
For more information on the curiously recurring template pattern, check out https://en.
wikipedia.org/wiki/Curiously_recurring_template_pattern .
After defining the class, let's go ahead and add in an example implementation for the function we've
created for this class:
void HighScoreManager::CheckHighScore(int score)
{
std::string toDisplay;
if (highScore < score)

{
highScore = score;
toDisplay = "\nNew High Score: " + std::to_string(score);
}
else
{
toDisplay = "\nCurrent High Score: " + std::to_string(highScore);
}
}
By using the templatized version of our class, we don't need to create the same materials as in the
preceding class. We can just focus on the stuff that is particular to what this class needs to do. In this
case, it's checking our current high score, and setting it to whatever we pass in if we happen to beat it.
Of course, it's great to see our code in action, and in this case I used the SplashStage class, which is
located in the Mach5 EngineTest project, under SpaceShooter/Stages/SplashStage.cpp. To do so, I added the
following bolded lines to the Init function:
void SplashStage::Init(void)
{
//This code will only show in the console if it is active and you
//are in debug mode.
M5DEBUG_PRINT("This is a demo of the different things you can do\n");
M5DEBUG_PRINT("in the Mach 5 Engine. Play with the demo but you must\n");
M5DEBUG_PRINT("also inspect the code and comments.\n\n");
M5DEBUG_PRINT("If you find errors, report to [email protected]");
HighScoreManager::CreateInstance();
HighScoreManager::GetInstance().CheckHighScore(10);
//Create ini reader and starting vars

M5IniFile iniFile;
// etc. etc.
In this case, our instance has been created by us creating a new HighScoreManager. If that is not done, then
our project could potentially crash when calling GetInstance, so it's very important to call it. Then call
our CheckHighScore functions a number of times to verify that the functionality works correctly. Then, in
the Shutdown function, add the following bolded line to make sure the Singleton is removed correctly:
void SplashStage::Shutdown(void)
{
HighScoreManager::RemoveInstance();
M5ObjectManager::DestroyAllObjects();
}
With all of that gone, go ahead, save the file, and run the game. The output will be as follows:
As you can see, our code works correctly!
Note that this has the same disadvantages we discussed with our initial version of the script, with the
fact that we have to manually create the object and remove it; but it takes away a lot of the busy work
when creating a number of Singletons in your project. If you're going to be creating a number of them
in your project, this could be a good method to look into.
Advantages/disadvantages of using only one
instance
There is the possibility that as you continue your project, something that looks at the time to be a thing
that you'll only need one of will suddenly turn into something you need more of down the road. In
games, one of the easiest examples would be that of a player. When starting the game, you may think
you're only going to have one player, but maybe later you decide to add co-op. Depending on what
you did before, that can be a small or huge change to the project.
Finally, one of the more common mistakes we see once programmers learn about Singletons, is to
create managers for everything, and then make the managers all Singletons.
The Singleton in action - the Application class
The Singleton pattern achieves its ability to be accessible anywhere easily by having a special
function that we use to get the Singleton object. When this function is called, we will check whether
that object has been created yet. If it has, then we will simple return a reference to the object. If not,
we will create it, and then return a reference to the newly created object.
Now, in addition to having this way to access it, we also want to block off our user from being able to
create them, so we will need to define our class constructors to be private.
Now that we have an understanding of some implementations of the Singleton, we have one other
version, which is what we actually used within the Mach5 engine.
In Mach5, the only Singletons that are included are aspects of the engine code. The engine code is
designed to work with any game, meaning there is nothing gameplay-specific about it, which means
that it doesn't need to have instances since they're just instructions. Building the engine in this way
makes it much easier in the future to bring this to other games, since it's been removed from anything
that's game-specific.
In this case, let's open up the M5App.h file which is in the EngineTest project under Core/Singletons/App and
take a look at the class itself:
//! Singleton class to Control the Window
class M5App
{
public:
friend class M5StageManager;
/*Call These in Main*/
/*This must be called first, before the game is started*/

static void Init(const M5InitData& initStruct);
/*Call this after you add your stages to start the game*/
static void Update(void);
/*Call this after Update is finished*/
static void Shutdown(void);
/*Call these to control or get info about the application*/
/*Use this to change to fullscreen and back*/

static void SetFullScreen(bool fullScreen);
/*Use this to show and hide the window*/
static void ShowWindow(bool show);
/*Use this to show and hide the default window cursor*/
static void ShowCursor(bool showCursor);
/*Use this to change the resolution of the game*/
static void SetResolution(int width, int height);
/*Returns the width and height of the window (client area)*/
static M5Vec2 GetResolution(void);
private:
static LRESULT CALLBACK M5WinProc(HWND win, UINT msg, WPARAM wp, LPARAM lp);
static void ProcessMessages(void);
};//end M5APP
Now, the Mach5 engine follows the Singleton pattern. However, it is done in a different way from the
others that we've looked at so far. You may notice in the class definition that every single function and
variable that was created was made static.
This provides us with some unique benefits, namely that we don't need to worry about the user
creating multiple versions of the class, because they'll only be restricted to using static properties and
variables that are shared by everything. This means we don't need to worry about all of those fringe
cases we mentioned in the previous examples that we've seen. This is possibly due to the fact that the
Mach5 engine classes have no need to have child classes; there's no need for us to create a pointer or
even call a GetInstance function.
You'll also notice the Init, Update, and Shutdown functions mentioned previously. We mentioned before
that it was a disadvantage to manually have to create and destroy our singleton classes, but there are
some distinct benefits to having this control. In the previous examples we had, the order in which
classes were created was up to the compiler as we couldn't control the order. However, with our
game engine it makes sense to create our Application (M5App) before we start up the graphics library
(M5Gfx) and the only way we can make sure that happens is by telling our engine to do so, which you
can look at if you open up the Main.cpp file and look at the WinMain function, which is what opens first
when we create our project. I've gone ahead and bolded the uses of M5App:
int WINAPI WinMain(HINSTANCE instance,
HINSTANCE /*prev*/,
LPSTR /*commandLine*/,
int /*show*/)
{
/*This should appear at the top of winmain to have windows find memory leaks*/
M5DEBUG_LEAK_CHECKS(-1);
M5InitData initData; /*Declare my InitStruct*/

M5GameData gameData = { 0 }; /*Create my game data initial values*/
M5IniFile iniFile; /*To load my init data from file*/
iniFile.ReadFile("GameData/InitData.ini");
iniFile.SetToSection("InitData");
/*Set up my InitStruct*/
iniFile.GetValue("width", initData.width);
iniFile.GetValue("height", initData.height);
iniFile.GetValue("framesPerSecond", initData.fps);
iniFile.GetValue("fullScreen", initData.fullScreen);
initData.title = "AstroShot";
initData.instance = instance;
/*Information about your specific gamedata */
initData.pGData = &gameData;
initData.gameDataSize = sizeof(M5GameData);
/*Pass InitStruct to Function. This function must be called first!!!*/

M5App::Init(initData);
/*Make sure to add what stage we will start in*/

M5StageManager::SetStartStage(ST_SplashStage);
/*Start running the game*/

M5App::Update();
/*This function must be called after the window has closed!!!*/
M5App::Shutdown();
return 0;
}
Afterwards, we can look at the Init function of M5App and see that it will initialize the other Singletons
in our project:
void M5App::Init(const M5InitData& initData)
{
// ...
// Other init code above...
M5StageManager::Init(initData.pGData, initData.gameDataSize, initData.fps);

M5ObjectManager::Init();
M5Input::Init();
}
By having this control, our users have a much better idea as to the flow and order that things will be
created. But, of course, with that great power comes great responsibility.
The Singleton pattern is used only for single-threaded applications. Should you be
developing a multithreaded game, you'll want to use the Double-Checked Locking
pattern instead, which was created by Doug Schmidt and Tim Harrison. If you're
interested in learning more about it, check out https://en.wikipedia.org/wiki/Double-checked_loc
king .
Summary
In this chapter, we have demystified a lot of programming concepts in a quick refresher. We also
started learning about our first design pattern, the Singleton, which is intended to allow us to always
have access to a class's functions and variables due to the fact that there will only ever be one of
these objects.
We discussed some of the typical downfalls of using the Singleton pattern, such as the possibility that
objects could have multiple copies of them in the future, even if this is unlikely.
We learned about three different kinds of method for creating Singletons, starting off with the
Singleton, then extending it and templating parts of it to create the curiously reoccurring template
pattern, and then we saw a final all-static version of getting the same effect with minimal hassle.
Each of these methods has their own pros and cons, and we hope that you use them effectively, where
they are relevant. Now that we've touched on the design pattern everyone is familiar with, we can
move towards our next challenge: learning about how to deal with logic that is specific to each of our
individual games.
Creating Flexibility with the Component
Object Model
In the last chapter, we saw how the Singleton pattern can help us solve the problem of creating and
using the big core engines of our game. The engine code is designed to work with any game, meaning
there is nothing gameplay-specific about it. So as the game design evolves, we don't need to worry
about changes in game design breaking our engine. The goal when writing code for a graphics or
physics engine is to make it as reusable or game-agnostic as possible. This means that when you are
done making the current game, you should be able to use the code in the next game with very little or
no change. The way to do this is to separate the engine code from anything related to the specific
game.
Game objects, on the other hand, are completely specific to our game. If the game changes, all our
object types will need to change as well. If we are making a platformer and suddenly change to
making a Space Shooter, our graphics and physics engine code probably doesn't need to change.
However, every single game object and behavior will change. While this may be the most extreme
example, the fact is that our game objects are likely to change a lot. So let's look at how we can use
patterns to solve this small, but very important, piece of our game.
Chapter overview
In this chapter, we will be focusing on creating a game object that is flexible enough to adapt as our
game design changes. We will do this by first looking at the two most common ways that new
programmers create a game object, and the problems that arise when we use these approaches. Then
we will discuss two design patterns that can help us solve our problem. Finally, we will arrive at our
solution of creating a reusable, flexible game object. Since we know our game design and game
objects are likely to change, the questions we are going to answer are the following:
Is it possible to write game objects in a reusable way?

How can we decouple our game objects from our core engine code?
If we have a reusable game object, how can we make it flexible enough to use in different games
or account for changes in our game design while the game is being developed?
Along the way, we will discuss a few important design principles that will come up again and again
in this book that will help you write clean and solid code.
Your objectives
Over the course of this chapter, we will be focusing on a lot of important concepts and diving deep
into some interesting code. Some of these concepts are about ways not to implement a game object.
Learning the wrong way is often just as important as learning the right way. Here is an outline of the
topics we will cover and your tasks for this chapter:

Learning and implementing the Strategy pattern and the Decorator pattern
Learning and implementing the Component Object Model
When you break it down to the simplest terms, programming is about solving problems with code.
Someone has an idea for a game or an app, and the problem that needs to be solved is how to
describe that idea logically and correctly to the computer. Day to day, these problems usually come in
the form of integrating code you wrote today with code written earlier by you or another programmer.
When solving these problems, there is a constant struggle between doing things the easy way or doing
them the right way.
The easy way to solve a problem means solving the immediate problem in the fastest way possible.
Examples of this might be hardcoding a number or string literal instead of using a named constant,
copying code instead of writing a function or refactoring code into a base class, or just writing code
without thinking about how it can impact the rest of the code base.
On the other hand, solving a problem the right way means thinking about how the new code will
interact with the old code. It also means thinking about how the new code will interact with future
code if the design changes. The right way doesn't mean that there is only one correct solution to the
problem. There are often many possible ways to reach the same result. The creativity involved in
programming is one of the reasons programming is so much fun.
Veteran programmers know that in the long run, the easy way often turns out to be more difficult. This
is often because a quick fix solves an immediate problem but doesn't consider the changes that will
occur as the project evolves.
The monolithic game object
The easy way to do a game object is to have a single struct that contains all of the data a game object
will need. This seems correct because everything in the game has the same basic data. For example,
we know players and enemies all have a position, scale, and rotation. So our struct will look like
this:
struct GameObject
{
//using vectors from the Mach 5 Engine
M5Vec2 pos;
M5Vec2 scale;
float rotation;
};
This game object works well in theory, but it is too basic. It is true that everything in our game
probably needs a position, scale, and rotation. Even an invisible trigger region needs these
properties. However, as it stands, we can't draw our object: we have no health, and no way to do
damage. So, let's add a few things to make the game object a little more real:
struct Object
{
//using vectors from the Mach 5 Engine
M5Vec2 pos;
M5Vec2 scale;
float rotation;
float damage;
int health;
int textureID; //for drawing
float textureCoords[4]; //for sprite animation
unsignedchar color[4]; //the color of our image
};
Now we have added a few more basic elements to our game object. Most of our game object types
will have health and damage, and we have added a texture ID so we can draw our game object, and
some texture coordinates so we can use a sprite sheet for animation. Finally, we added a color so we
can reuse the same texture and color it differently for different enemies (think about the different
ghosts in Namco's Pacman).
This is not that bad yet but, unfortunately, this is just the beginning. Once we start making a real game
instead of just brainstorming about a basic game object, our struct member count starts to explode.
Imagine we are making a Space Shooter. There are lots of things we will want to add:
The player will have multiple types of weapons that all do different amounts of damage
The player might have access to bombs and missiles that each have an ammo count
The missile needs a target to seek
The bomb needs an explosion radius
There are two super enemies that each have a special ability with a cool-down time
The player and one super enemy both have the ability to use a shield
The UI buttons have some actions associated with clicking them
We have power-ups that add health and add lives
We need to add a lives count to all objects to account for the power up
We should add velocity to objects and do time-based movement instead of just setting the
position directly
We need to add an enumeration for the type of the game object so we can update it properly
Here is what our game object looks like now:

struct GameObject
{
M5Vec2 pos;
M5Vec2 scale;
M5Vec2 vel;
float rotation;
ObjectType type; //Our object type enum
int objectID; //So the missile can target
int lives;
int shieldHealth; //For Player and SuperBomber
int health;
float playerLaserDamage;
float playerIonDamage;
float playerWaveCannonDamage;
float superRaiderDamage;
float superRaiderAbilityDamage;
float superRaiderAbilityCoolDownTime;
float superBomberDamage;
float superBomberAbilityDamage;
float superBomberAbilityCoolDownTime;
int bombCount;
float bombRadius;
int missileCount;
int missileTargetID;
int textureID; //the object image
float textureCoords[4];//for sprite animation
unsigned char color[4]; //the color of our image
Command* command; //The command to do
};
As you can see, this basic method of creating a game object doesn't scale very well. We already have
more than 25 members in our struct and we haven't even talked about adding space stations that can
spawn or repair units. We have only two boss types, we can make a few enemy types by allowing
different enemies to use different player weapons such as the laser or missiles, but we are still
limited.
The major problem with this approach is that, as the game gets bigger, our game object must also get
very big. Some types, such as the player, will use many of these members, but other types, such as a
UI button, will only use a small amount. This means if we have lots of game objects, we are very
likely wasting a lot of memory per object.
The problem with object behavior
So far, we have only considered what members the game object has. We haven't considered how each
object will have its behavior updated. Right now, the game object is just data. Since it has no
functions, it can't update itself. We could easily add an Update function for the game object but, in order
to update each type of object correctly, we would need a switch statement:
//Create our objects
Object gameObjects[MAX_COUNT];
//initialization code here

//...
//Update loop
for(int i = 0; i < objectInUse; ++i)
{
switch(gameObjects[i].type)
{
case OT_PLAYER:
//Update based on input
break;
case OT_SUPER_RAIDER:
//Add intercept code here
break;
case OT_SUPER_BOMBER:
//Add case code here
break;
case OT_MISSILE:
//Add find target and chase code here
break;
case OT_BOMB:
//add grow to max radius code here
break;
default:
M5DEBUG_ASSERT(true, "Incorrect Object Type");
}
}
Again, this approach doesn't scale well. As we add more object types, we need to add even more
cases to our switch statement. Since we only have one struct type, we need to have a switch statement,
whenever we need to do something object-type-specific.
If we are adding behaviors, we will also face the decision of adding data to our object or hardcoding
a value into the switch statement. For example, if our bomb grows in size, how does it grow? We could
hard code scale.x *= 1.1f into our switch statement or we can add member data float bombScaleFactor to
our struct.
In the end, this approach just isn't that flexible. Changing our design is very difficult because there are
switch statements and public members throughout our code. If we were to make a game like this, then
our code base would be a complete mess after only a few months. The worst part would be that once
the game was completed, we wouldn't be able to reuse any code. The game object and all behaviors
would be so gameplay-specific that unless we make a sequel, we would need to remake a brand new
game object.
The benefits of the monolithic game object
It is worth noting that even if you choose this approach, you can still have your core engines
decoupled from the game object. When writing the graphics engine, for example, instead of passing in
a game object as a parameter to a Draw function, we could pass in the members that the graphics engine
needs:
void Graphics::SetTexture(int textureID);
void Graphics::SetTextureCoords(const float* coordArray);
void Graphics::Draw(const M5Mtx44& worldMtx);
vs
void Graphics::Draw(const Object& obj);
Another argument for creating objects like this is that we know exactly what is in our game object.
Compared with other approaches, we never need to cast our object or search for properties within the
object. These operations make the code more complicated and have a slight performance cost. By
using a simple struct, we have direct access to the variables and the code is easier to understand.
The only time we might use this approach is if we know 100% that the number of object types won't
be large, for example, if you are making a puzzle game and the only game objects are sheep and
walls. Puzzle games are often very simple and use the same mechanics over and over. In this case,
this is a good approach because it is easy and doesn't require any time building a complicated system.
The idea that Players, Enemies, Missiles, and Medics should all derive from one base object is very
common to programmers new to object-oriented programming. It makes a lot of sense on paper that if
you have a Raider and a SuperRaider, one should inherit from the other. I believe this comes from
how inheritance is taught. When you are first learning about inheritance, you will almost always see a
picture similar to this:
Figure 3.1 - A typical inheritance diagram when learning to program
Many introductory programming courses are so focused on the mechanics of inheritance that they
forget to tell you how to use it properly. A picture like the one above makes it easy to understand that
ITWorker is an Employee, which is a Person. However, once you go beyond the mechanics, it is time
to learn how to use inheritance correctly. This is why books on design patterns exist.
Inheritance is a powerful tool that lets us extend classes by adding members and methods that are
specific to the derived classes. It allows us to start with general code and create more specialized
classes. This solves one of the original problems that we had with the extremely bloated object struct
in the first section. Inheritance lets us take an existing class, such as a Raider, and add more members
to create a SuperRaider:
//Inheritance Based Object:
class Object
{
public:
Object(void);
virtual ~Object(void);//virtual destructor is important
virtual void Update(float dt);
virtual void CollisionReaction(Object* pCollidedWith);
protected:
//We still need the basic data in all object
M5Vec2 m_pos;
M5Vec2 m_scale;
float m_rotation;
int m_textureID;
};
//Inheritance Based derived class

class Unit: public Object
{
public:
Unit(void);
virtual ~Unit(void);
protected:
M5Vec2 m_vel;//So that Units can move
float m_maxSpeed;
float m_health;
float m_damage;
};
class Enemy: public Unit

{
public:
Enemy(void);
virtual ~Enemy(void);
protected:
unsigned char m_color[4];
float m_textureCoords[4];//For animation
};
Figure 3.2 - An example of Space Shooter inheritance hierarchy
This hierarchy makes a lot of sense when first designing a space shooter. It allows us to separate the
details of a Raider class or a Bomber class away from the Player class. Adding a game object is easy
because we can extend a class to create what we need. Removing a game object is easy because all
the code is contained within each derived class. In fact, now that we have separate classes, each one
can be responsible for itself via class methods. This means we no longer need switch statements all
over our code.
Best of all, we can use the power of virtual functions to decouple our derived classes from the core
engines of our game. By using an array of base class pointers to the derived class instances, our core
engines such as graphics or physics are only coupled to the object interface instead of derived
classes, such as Planet or SpawnerStation.
Without inheritance hierarchy, the code would be as follows:

Object gameObjects[MAX_COUNT];

//...
for(int i = 0; i < objectsInUse; ++i)

{
switch(gameObjects[i].type)
{
case OT_PLAYER:
//Update based on input
break;
case OT_PLANET:
//Add intercept code here
break;
case OT_ENEMY_SPAWNER:
//Add case code here
break;
case OT_RAIDER:
//Add find target and chase code here
break;
case OT_BOMBER:
//Move slowly and do large damage code here
break;
default:
M5DEBUG_ASSERT(true, "Incorrect Object Type");
}
}
With inheritance and polymorphism, the code is as follows:

Object* gameObjects[MAX_COUNT];//array of pointers

//...
for(int i = 0; i < objectsInUse; ++i)

gameObjects[i]->Update(dt);
Organizing the code by what it does, not what it
is
What is the difference between the Raider and the Bomber, really? How are a Raider and a
SuperRaider different? Maybe they have a different speed, a different texture, and a different damage
value? Do these changes in data really require a new class? Those are really just different values, not
different behaviors. The problem is that we are creating extra classes because the concept of a Raider
and SuperRaider is different, but there aren't differences in behavior.
Our class hierarchy actually violates three principles I teach, two of which I learned from the Gang of
Four book:
"Keep your inheritance trees shallow"

"Favor object composition over class inheritance" - Gang of Four, p20
"Consider what should be variable in your design. This approach is the opposite of focusing on the
cause of redesign. Instead of considering what might force a change to a design, consider what
you want to be able to change without redesign. The focus here is on encapsulating the concept
that varies, a theme of many design patterns" - Gang of Four, p29
A different way to state the third principle is the following:
"Find what varies and encapsulate it"
These principles exist to eliminate, or completely avoid, the problems that can and will arise when
using inheritance.
The problem with our current design is that if we create a new class for every object type, we will
end up with a lot of little classes that are mostly the same. Raider, SuperRaider, Bomber, and
SuperBomber are mostly the same with just a few minor differences, some of which are only
differences in float and int values. While this approach may seem like an improvement over the easy
way, it becomes a problem because we will end up writing the same behavior code over and over
again in many classes. If we have a lot of enemies, we might end up writing the same basic
ChasePlayerAI code in every Update function. The only solution is moving the ChasePlayerAI up to a base
class.
Let's take another look at our Space Shooter hierarchy but this time, let's add in some different
behaviors to our classes:
Figure 3.3 - After adding behavior to our objects (refer graphic bundle)
We have decided that our base object class will at least be drawable to make things simple. If an
object such as a trigger region needs to be invisible, we can simply support disabling rendering by
putting a bool in the drawable behavior so it won't get drawn. However, with this game object
approach, I still have some duplicated code. Both the Raider class and the AttackStation class have some
AI that targets and shoots bullets at the Player. We have only duplicated our code once so maybe it
isn't a big deal.
Unfortunately, all game designs will change. What happens when our designer wants to add asteroids
to our game? Technically, they are structures so they need some of the data inherited from that class,
but they also move. Our designer also really liked the SpawnerStation class and wants to add that ability
to a new SpawnerPlanet class, and to a new BossSpawner class. Should we rewrite the code two more
times, or refactor the code into the base class? Our designer also wants to give the Station the ability
to slowly patrol an area. This means the Station class needs the Patrol AI ability as well. Let's take a
look at our hierarchy now:
Figure 3 4 - After refactoring duplicate code to our base class (refer graphic bundle)
As it turns out, this approach isn't as flexible as it originally seemed. In order for our design to be
really flexible, almost all of the behaviors need to be factored up into the base class. In the end, we
aren't much better off than when we wrote our game object the easy way. And it is still possible that
our designer will want to create the RepairHelper that chases the Player, meaning that everything will be
in the base class.
This might sound like a contrived example but remember that games take years to develop and are
likely to change. DMA Design's Grand Theft Auto was originally titled Race'n'Chase, but it was
changed because a bug caused the police to try and run the Player off the road instead of pull them
over. This ended up being way more fun. Another example is Blizzard's first-person shooter
Overwatch, which was originally in development for 7 years as a massively multiplayer online game.
The purpose of object-oriented programming is to recognize that designs will change

and to write code with that change in mind.
Another problem with our inheritance approach is that it isn't very easy to add or remove abilities at
runtime. Let's say our game has a special power-up item that will let the Player use a shield for 1
minute. The shield will absorb 50% of the damage done to the Player for 1 minute then remove itself.
We now have the problem of making sure that when a bullet collides with the shield, it will transfer
some of the damage to the Player. The shield isn't just responsible for itself; it is responsible for the
Player object too.
This same situation exists for all things that will affect another game object for some duration of time.
Imagine if we want our Raider to be able to do acid damage to the Player over 5 seconds instead. We
need a way to attach this acid damage to the Player, and to remember to remove it after 5 seconds. We
could add new variables such as bool hasAcid and float acidTime in the Player class that we can use to
know whether we should do acid damage in this frame. However, this still isn't a flexible solution,
because each new type of damage caused over time will need new variables like this.
In addition, there is no way to stack the acid damage effect if three enemies are attacking the Player
with acid damage. If we like this ability and want the Player to use it, we also need to give all game
objects these extra time-based damage variables and behavior code. What we would really like to do
is attach acid behavior (or any effect) onto a game object at runtime and have it automatically detach
itself when the effect is over. We are going to talk about how to do that later in this chapter, but first
we need to talk about one more problem related to inherence hierarchies in C++.
Avoiding the Diamond of Death
The final problem with our inheritance approach involves the situation where we take code reuse to
the extreme. In our hierarchy, we have SuperRaider, which is very fast, weak, and shoots little
bullets. We also have SuperBomber, which is slow, strong, and shoots big bombs. Someday, a clever
designer will want to create SuperBomberRaider that is very fast, strong, and shoots both little
bullets and big bombs. Here is our partial tree:
Figure 3.5 - Example of the Diamond of Death
This, of course, is the Diamond of Death (or Dreaded Diamond of Death), so named because the
inheritance tree forms a diamond. The problem is that our SuperBomberRaider inherits from both
the SuperBomber and SuperRaider. Those two classes each inherit from the Enemy, Unit, and object. That
means SuperBomberRaider will have two copies of m_pos, m_scale, m_rotation, and every other member of
object, Unit, and Enemy.
There will also be two copies of any functions that are contained in the Object, Unit, and Enemy. This
means we need to specify which version of the functions we wish to use. This might sound good,
since we get behavior from both classes, but remember that the individual base class function will
only modify their version of variables. After calling SuperRaider::Update and SuperBomber::Update, we now
need to figure out what version of m_pos (and m_scale and m_rotation) we want to use when we draw our
object.
C++ has ways of solving this problem, but most programmers agree that the solution makes things
more complicated to understand and more difficult to use. The rule of thumb is that we should just
avoid using multiple inheritance. We have seen some of the problems that it can cause and we haven't
even talked about bugs related to using new and delete in a situation like this.
The Strategy pattern and the Decorator
pattern
We saw that in trying to be more flexible with our game object, a lot of behavior was factored into the
base class. We also said that it would be nice to attach a behavior at runtime and have it detach itself
when we are done with it.
There are actually two design patterns that have the potential to help our design, the Strategy pattern
and the Decorator pattern. The Strategy pattern is all about encapsulating sets of behaviors instead of
inheriting. The Decorator pattern is all about dynamically adding responsibilities as needed.
The Strategy pattern explained
The Strategy pattern is about encapsulating a set of behaviors and having the client control the
behavior through an interface, instead of hardcoding the behavior into the client function itself. What
this means is that we want the game object to be completely independent of the behavior it uses.
Imagine that we want to give each enemy a different attack and flight AI. We could use the Strategy
pattern instead of creating an inheritance tree:
class Enemy: public Unit
{
public:
Enemy(void);
virtual ~Enemy(void);
protected:
unsigned char m_color[4];
FlightAI* m_flight;
AttackAI* m_attack;
};
In this case, our client is the Enemy class and the interfaces that the client controls are the AttackAI and
FlightAI. This is a much better solution than inheriting from the Enemy because we are only encapsulating
what varies: the behavior. This pattern allows us to create as many FlightAI derived classes as we
need and to reuse them to create different kinds of game object types, without needing to expand our
inheritance tree. Since we can mix different strategy combinations, we can get a large number of
different overall behaviors.
We are going to share the same strategies for both units and structures, so we should actually remove
our inheritance tree altogether and just use the Object as our client. This way, the Object class becomes
a collection of strategies, and our design is simpler. Plus, we are following some great programming
principles:
Programing to an interface means that our client depends on behavior in an abstract class instead
putting behavior in the client itself.
Our interfaces are opened for extension so we can easily add as many behaviors as we need.
The interface is simple so it won't need to be changed, which might break code.
Our inheritance trees are shallow so we don't need to worry about the Diamond of Death.
Figure 3.6 - Example of our Object using the Strategy pattern
The Strategy pattern allows our game object to be very flexible without the need for an inheritance
tree. With these six small classes shown in the preceding diagram, we can have a total of nine
different game object behaviors. If we add a new FlightAI, we have 12 possible game object
behaviors. Creating brand new strategies allows for an amazing amount of mixed behaviors.
However, if we only extend just the two strategies, we don't need to modify the Object at all. This
works for the Player as well, if we make an AttackAI and FlightAI that have access to input.
Staying with only two strategies is unlikely, which means that whenever we add a new strategy, we
will need to change the Object by adding a new member and modifying the Update function. This
means that while the pattern is flexible enough to let us change strategies at runtime, we can't add
behaviors dynamically. If we need to add acid damage as a debuff in our game, we would need a
Damage base class, and to give a Damage base class pointer to the object:
class Object
{
public:
//Same as before...
protected:
//Other Object Strategies
//...
Damage* m_damage.
};
This doesn't seem like a great solution because most damage will be instantaneous and, most of the
time, the player isn't even taking damage. That means this will be either null or an empty strategy
class, such as using a NoDamage derived class, that will be updated every frame but will do nothing.
This is also no way to stack corrosive effects or to have two types of damage affecting the Player,
such as corrosive damage and ice damage, which might cause the Player to move slower for 10
seconds. We really need a way to dynamically add and remove these abilities. Luckily, there is a
pattern for that.
The Decorator pattern explained
The purpose of the Decorator pattern is to dynamically add responsibilities to an object at runtime.
The goal is to be a flexible alternative to creating derived classes while still allowing for extended
behavior. What this means is that we can take our object and add decorations or, in our case,
behaviors at runtime.
This pattern requires that the Decorator and our object are derived from a common base class so they
share the same interface. Each Decorator will then layer itself on top of an object or another Decorator to
create more interesting object types and effects. When a function gets called on a Decorator, it will call
the corresponding function on the next layer down, eventually calling the function of the object. It is
similar in concept to the Russian Matryoshka dolls, the dolls that contain smaller and smaller
versions inside of themselves. The final, most nested object is always the object with the core
functionality:
Figure 3.7 - The layering effects of the Decorator pattern
Here is a simplified version in code:

class Component //Our base interface
{
public:
virtual ~Component(void) {}
virtual std::string Describe(void) const = 0;
};
class Object: public Component //Our core class to decorate

{
public:
Object(const std::string& name):m_name(name){}
virtual std::string Describe(void) const
{
return m_name;
}
private:
std::string m_name;
};
//Our base and derived Decorators

class Decorator: public Component
{
public:
Decorator(Component* comp):m_comp(comp){}
virtual ~Decorator(void) { delete m_comp; }
protected:
Component* m_comp;
};
class RocketBoosters: public Decorator

{
public:
RocketBoosters(Component* comp) : Decorator(comp) {}
{
return m_comp->Describe() + " with RocketBoosters";
}
};
class LaserCannons: public Decorator

{
public:
LaserCannons(Component* comp) : Decorator(comp) {}
{
return m_comp->Describe() + " with LaserCannons";
}
};
Figure 3.8 - The Decorator pattern using our Object
//Using this code:

int main(void)
{
Component* ship = new Object("Player");
std::cout << ship->Describe() << std::endl;
delete ship;
Component* rocketShip = new RocketBoosters(new

GameObject("Enemy"));
std::cout << rocketShip->Describe() << std::endl;
delete rocketShip;
Component* laserRocketShip = new LaserCannons(new

RocketBoosters(new GameObject("Boss")));
std::cout << laserRocketShip->Describe() << std::endl;
delete laserRocketShip;
}
The Decorator classes layer our concrete object class and add more information on top of the object.
However, right now, all we are doing is adding superficial decorations. Since the Decorator class
doesn't know whether it has a pointer to the object class or another Decorator, it can't modify the object.
A good analogy is that the Strategy pattern changes the guts of the object, while the Decorator pattern
changes the skin. This can be useful but doesn't help us with our buff/debuff problem. To solve this
problem, we would need to add a method to find the object down the chain, or give a pointer to the
object in the constructor of a Decorator.
Another problem is that this pattern was designed to add a Decorator dynamically, but doesn't allow us
to remove one. In the case of using a corrosive damage Decorator, we would only want it to exist for a
set time, and then automatically detach itself. This can't be done, since a Decorator doesn't have a
pointer to its parent.
The final problem for games is that our Decorators can't live in a vacuum. Sometimes, different
gameplay behaviors may need to interact with each other. For example, the corrosive damage Decorator
may affect the health of an object; however, it may first need to check whether the object has a shield
Decorator and remove health from the shield.
Unfortunately, neither the Decorator nor the Strategy pattern will work perfectly for us. What we
really need is a new pattern that is a combination of the Strategy and Decorator patterns that does the
following:
Encapsulates specific behavior into components so we avoid Object inheritance trees

Allows for a flexible number of components so we don't need to modify the Object each time we
create a new component type
Lets us add and remove components at runtime
Gives components direct access to the Object so it can be modified
Allows components to be searchable by other components so they can interact
The Component Object Model explained
The alternative can be found by many names, though none are definitive yet. In this book, we will call
it the Component Object Model, but others have called the Entity Component System or just
Component System. No matter what you call it, the concept is surprisingly simple to learn and easy
to implement.
The Component Object Model inverts the concept of the Decorator pattern, where each Decorator
added a new layer on top of the game object. Instead of layering our object, which we have already
seen problems with, we will put the decorations inside of it. Since we don't know how many we will
need, our object will hold a container of decorations, as opposed to a single pointer. In the simplest
form, our object is nothing more than a container for these components.
If you search for Component Object Model (or Component Based object Model) on the Internet, you
will get results that are similar to what we saw in the Strategy pattern. The object contains hardcoded
pointers to each possible strategy. While using this approach alone is much better than a monolithic
object or an inheritance-based object, we are stuck checking for null pointers or constantly modifying
what strategies exists in our object.
In this alternative method, every strategy type will derive from a common interface. This way, our
object can contain an array, or in our case an STL vector of base class Component pointers. This is like
the Decorator, except our object is a separate class; it doesn't derive from the Component interface. Instead,
a Component will have a pointer to its parent object class. This solves the problem in which a Decorator
didn't know whether it held a pointer to another Decorator, or to the actual object. Here we avoid that
problem by always giving our Component a pointer to the object it controls:
//Using only Strategy Pattern
class Object
{
public:
void Update(float dt);//Non virtual function to update
Strategies
//Other interface here
//...
private://Lots of different Strategies
GfxComp* m_gfx;
BehaviorComp* m_behavior;
ColliderComp* m_collider;
};
//Using Flexible Component Object Model

class Object
{
public:
void Update(float dt);//Non virtual function to update
Components
//Other interface here
//...
private:
std::vector<Component*> m_components.
};
//Our Base Component

class Component
{
public:
virtual void Update(float dt) = 0;
protected:
Object* m_obj;
};
Figure 3.9 - The Component Object Model
This approach allows us to be very flexible because our object is nothing more than components.
There is nothing in it that is specific to any type. There is no code that is strictly for the Player or
SuperRaider. We are free to add, change, or remove anything at runtime. This is important because in
the early stages of development, the game design and game objects will change a lot. If we were to
hardcode pointers to different base class Strategies, we would spend a lot of time changing those
pointer types in the game object.
Using the Component Object Model makes our code almost completely reusable as well. The game
object itself is just an empty container of Components, and they are often so simple that most of them,
such as a CircleCollider, can be used in any game. This means that a behavior component, originally
meant only for the Player or SpawnerStation, can be easily used for any game object.
Implementing the Component Object Model
Now that we have seen a basic version in code as well as a diagram, let's look at exactly how the
Mach5 Engine implements this system. As you will see, the M5object, as it is called, contains a
position, rotation, scale, and velocity. Of course, these elements could be contained in a transform
component; however, these are so common that most other components will need access to this
information. This is different to data such as texture coordinates or a circle collider's radius, which
might not need to be shared at all:
//Component based Game object used in the Mach 5 Engine
class M5Object
{
public:
M5Object(M5ArcheTypes type);
~M5Object(void);
//Public interface
void Update(float dt);
void AddComponent(M5Component* pComponent);
void RemoveComponent(M5Component* pComponent);
void RemoveAllComponents(void);
void RemoveAllComponents(M5ComponentTypes type);
int GetID(void) const;
M5ArcheTypes GetType(void) const;
M5Object* Clone(void) const;
template<typename T>
void GetComponent(M5ComponentTypes type, T*& pComp);
void GetAllComponents(M5ComponentTypes type,
std::vector<T*>& comps);
M5Vec2 pos;
M5Vec2 scale;
M5Vec2 vel;
float rotation;
float rotationVel;
bool isDead;
private:
//Shorter name for my vector
typedef std::vector<M5Component*> ComponentVec;
//Shorter name for my iterator
typedef ComponentVec::iterator VecItor;
ComponentVec m_components;
M5ArcheTypes m_type;
int m_id;
static int s_objectIDCounter;
};
The first thing you will notice is that there are two enumerations in this code, M5ArcheTypes and
M5ComponentTypes. These will become more useful later when we talk about creating Factories.
However, for now, it is enough to understand that these will allow us to search through a collection of
M5objects and get the components we need. For example, if we have a collection of M5objects but we
need to find the Player, the M5ArcheTypes enum will allow us to do that.
The next thing you will notice is the M5object is more than just a container of components. It has some
public and private data. The public data is unlikely to need validating or protecting. We could create
getters and setters but they would really just simply get and set the data, so it isn't 100% necessary.
Since they are public, we are locked into keeping them public forever. If you wish to make them
privateand create accessor methods, that is fine. There are some very important variables that we want
to be private. The ID and the type are set once and can't be changed, and the array of components is
accessed through functions to add, remove, and clear all components. Let's discuss the purpose of the
public variables first:
pos: The position of the M5Object. This is the rotational center, or pivot point, of the object.
scale: The height and width of the M5Object, before rotation.
vel: The velocity of the M5Object. This is used to do time-based movement instead of simply
setting the position to plus or minus some value.
rotation: The rotation in radians. Positive rotations are counterclockwise.
rotationalVel: The rotational velocity of the M5Object, used to do time-based rotations.
isDead: This allows the M5Object to mark itself for deletion. Other objects or components are free
to call one of the DestroyObject functions found in the M5ObjectManager; however, it isn't a good idea
for an object to delete itself in the middle of its own Update function.
We are keeping these as part of the M5object because they are so common that all or almost all
components will need access to them. We are marking these as public because there is no validation
or protecting that we need to do on the data.
The private section starts with two type defs. They let us create shorter names for templated types.
This is simply a style choice. Another style choice is to have an m_ in front of all of the private
member variable names. This or something similar is a common practice for class members. We
didn't do this with our public members because we are treating them more like properties. Now let's
look at the rest of the private data:
m_components: This is the array of M5Component pointers. Each component in the vector will get
updated in the Update function.
m_type: The type of object. It will get set in the constructor and never change. It allows the user to
use the M5ObjectManager to search or remove objects based on type.
m_id: This is a unique ID among M5Objects. It can be useful in cases such as a missile needing to
target a specific instance of an object. If the missile contains a pointer to the target object, it can't
know whether the object has been destroyed. If we instead know the ID, we can search to see
whether the target still exists.
s_objectIDCounter: This is the shared ID counter for all M5Objects. This guarantees that each object
will get a unique value because they are all using the same shared variable. Notice that this is
marked with an s_ to indicate that it is static.
That is all of the data in the object. Now, let's look at the functions.
M5objectis the constructor for the class. It sets starting values for the variables as well as setting the
type and giving a unique ID. Notice that we reserve an amount of starting space for the vector. A game
object can have as many components as it needs, but in an actual game, we don't expect them to have
more than a few on average. By pre-allocating, we may avoid any additional calls to new (we will be
a doing a lot anyway):
M5Object::M5Object(M5ArcheTypes type) :
pos(0, 0),
scale(1, 1),
vel(0, 0),
rotation(0),
rotationVel(0),
isDead(false),
m_components(),
m_type(type),
m_id(++s_objectIDCounter)
{
m_components.reserve(START_SIZE);
}
is the destructor for our game object. Here we want to make sure that we delete all of the
~M5object
components in our game object, so we make use of one of our public functions to help us:
M5Object::~M5Object(void)
{
RemoveAllComponents();
}
AddComponent adds the given component pointer to this object vector. You will notice that before the
component is added, you will need to first check to make sure the same pointer isn't already in the
list. While this isn't very likely to happen, it could be a hard bug to find later so it is worth the check.
It is also important when given a component to use the SetParent method of M5Component to make sure this
object will be controlled by the component:
void M5Object::AddComponent(M5Component* pToAdd)
{
//Make sure this component doesn't already exist
VecItor found = std::find(m_components.begin(),
m_components.end(), pComponent);
if (found != m_components.end())
return;
//Set this object as the parent

pComponent->SetParent(this);
m_components.push_back(pComponent);
}
Update is the most used function in the M5object. This will get called automatically by the M5ObjectManager
every frame. It is used to update every component as well as update position and rotation based on
their velocities. The other important role of the Update function is that it deletes any dead components.
Except for the RemoveAllComponents function, this is the only place where components are deleted:
void M5Object::Update(float dt)
{
int endIndex = m_components.size() - 1;
for (; endIndex >= 0; --endIndex)
{
if (m_components[endIndex]->isDead)
{
delete m_components[endIndex];
m_components[endIndex] = m_components[m_components.size()
- 1];
m_components.pop_back();
}
else
{
m_components[endIndex]->Update(dt);
}
}
//Update object data
pos.x += vel.x * dt;
pos.y += vel.y * dt;
rotation += rotationVel * dt;
}
RemoveComponentis used for cases such as when you have buffs or debuffs on an object and you want the
stage, or some other object, to delete it. For example, the Player may be using a shield but, after being
hit with ion damage, the physics collider finds the shield and immediately removes it. Instead of using
this method, it would also be fine to simply mark the component as dead and it will be cleaned up in
the next update loop.
This code follows a similar pattern to the AddComponent function. First, we test to make sure the
component exists. If it does exist, we swap places with the last item in the vector and pop back the
vector. After that, we use the SetParent method to remove this object as the parent pointer before
deleting it. This is a small precaution as, in case another pointer to this component exists, the program
will crash instead of causing an undefined error:
void M5Object::RemoveComponent(M5Component* pComponent)
{
//Make the sure the instance exists in this object
VecItor end = m_components.end();
VecItor itor = std::find(m_components.begin(), end, pToRemove);
if (itor != end)
return;
(*itor)->isDead = true;
}
is the helper function used in the destructor. It deletes all components in the object.
RemoveAllComponents
Except for the destructor, there probably isn't much use for it. However, it was made public for those
rare occasions where this is the behavior you need. This function simply loops through the vector and
deletes every component, then finally clears the vector:
void M5Object::RemoveAllComponents(void)
{
VecItor itor = m_components.begin();
VecItor end = m_components.end();
while (itor != end)
{
delete (*itor);
++itor;
}
m_components.clear();
}
The second version of RemoveAllComponents removes all components of a specific type. This is another
situation where external code, such as a stage, object, or even another component needs to remove a
group of the same component type. This could be used to remove all corrosive damage effects on the
Player, for example.
In this code, we are searching for the correct type, so we cannot use the std::vector::find method.
Instead, we use a for loop and check the type of each component. If we find the correct type, we
delete the current one, swap with the end and pop back. Since we are doing a swap, but continue
searching, we must make sure to check the current index again to see whether it matches as well:
void M5Object::RemoveAllComponents(M5ComponentTypes type)
{
for (size_t i = 0; i < m_components.size(); ++i)
{
if (m_components[i]->GetType() == type)
m_components[i]->isDead = true;
}
}
GetComponent and GetAllComponents are helper functions to find and cast specific component types in an
M5object, if they exist. As I said before, sometimes it is necessary that components interact. In that case,
we need a way to search for a specific component and to convert it to the correct type. These two
functions are almost the same. The first one finds the first instance of the correct component type and
assigns it to the pointer parameter. If one doesn't exist, we make sure to set the parameter to 0. The
second one finds all components of the correct type and saves them in the vector parameter. These are
template functions so the component can be cast to the correct type supplied by the user:
void M5Object::GetComponent(M5ComponentTypes type, T*& pComp)
{
size_t size = m_components.size();
for (size_t i = 0; i < size; ++i)
{
//if we found the correct type, set and return
{
pComp = static_cast<T*>(m_components[i]);
return;
}
}
pComp = 0;
}
void GetAllComponent(M5ComponentTypes type, std::vector<T*>& comps)
{
{
//if we found the correct type, add to vector
comps.push_back(static_cast<T*>(m_components[i]));
}
}
The GetID and GetType functions just return the private class data. The Clone method is more interesting
but we will go into more detail about it when we discuss the Prototype pattern.
Implementing components
Now that you have seen the M5object, let's take a look at how the Mach5 Engine creates and uses the
component hierarchy. Since this is an abstract class, there is no way to create an instance of an
M5Component. It is only an interface.
As you will see, the base component contains some of the same members as the M5object. Since we
will have a lot of components, it is important to give each one a type, so they can be searchable. It is
important to give each a unique ID. Since these components can be deleted at any time, it is important
to save an ID instead of a pointer, which can become invalid:
class M5Component
{
public:
M5Component(M5ComponentTypes type);
virtual ~M5Component(void);
virtual M5Component* Clone(void) = 0;
virtual void Update(float dt)= 0;
virtual void FromFile(M5IniFile&);
void SetParent(M5Object* pParent);
M5ComponentTypes GetType(void) const;
int GetID(void) const;
//public data
bool isDead;
protected:
M5Object* m_pObj;
private:
int m_id;
M5ComponentTypes m_type;
staticint s_compIDCounter;
};
The data section doesn't contain as much as the M5object, but now it is split into three sections which
are public, private, and protected:
isDead: This is the only public data and it serves a similar function to the member in the game
object. This allows the component to mark itself for deletion. It isn't a good idea for a
component to call RemoveComponent on itself during its own Update function.
m_pObj: This is a pointer to the M5Object that owns this component.
m_id: The unique ID of this component. This allows users to get access to this specific component
again, without the risk of saving a pointer which may become invalid.
m_type: The type of this component. This allows users to search for a specific component within a
game object.
s_compIDCounter: This is used to create a unique ID for each component.
The functions of the M5Component are not that interesting because they are mostly virtual. However, it is
worth going over their purpose.
M5Component is the non-default constructor for the component. This takes an argument of type
M5ComponentTypes so that the private data m_type is guaranteed to be set by a derived type:
M5Component::M5Component(M5ComponentTypes type):
isDead(false),
m_pObj(0),
m_type(type),
m_id(++s_componentID)
{
}
~M5Componentis the destructor for the class. Since this is meant to be a base class, it is important that a
virtual destructor exists so that the correct method will be called when using polymorphism:
M5Component::~M5Component(void)
{
//Empty Base Class virtual destructor
}
Update is where the component does the action. This method will be called every frame and its
intended purpose is to add a behavior and/or data to the M5object. It is marked as pure virtual (= 0) so
that the base class is forced to override it. That also means there is no body to the base class version.
FromFileis a virtual function that allows the component to read data from a preloaded INI file. It is not
marked as pure virtual, meaning that a component doesn't need to override this function. This might be
the case if the derived component has no data to be loaded from a file:
void M5Component::FromFile(M5IniFile&)
{
//Empty for the base class
}
The SetParent method is simply a setter for m_pObj. Recall the AddComponent function of M5object. When a
component is added to an object, the object uses this function so the component knows which object
to control.
The GetType and GetID functions are similar to the functions in M5object. They allow the component to be
searchable and saved without needing to use pointers that may become invalid. The M5Component also
has a pure virtual Clone method. There is no function body in the base class. We will discuss the Clone
method of both M5Component and M5object when we discuss the Prototype pattern.
To add a behavior to an object, we must derive from the M5Component base class,
overload the necessary methods, add a value to the M5ComponentTypes enumeration, then
finally register the class and the associated builder with the object manager. Of
course, these steps are prone to error and doing them repeatedly would be very
tedious.
For this reason, the Mach5 engine has included a batch file to do these steps
automatically. By adding components to the Source folder within the file hierarchy, the
batch file will find all files named *Component.h, where the asterisk is a wildcard
character that includes any valid C++ identifier.
For example, if a component named LazerComponent is located within a file named

LazerComponent.h, an enumeration value named CT_LazerComponent will be automatically
created along with the correct class builder, and both will be registered with
the M5ObjectManager.
Creating and removing objects and
components
In order to use the Component Object Model, first create a game object, then add some components,
then finally add it to the M5ObjectManager which calls an update on the game object every frame. Let's
look at some code for creating objects and components.
If we wanted to create a Player object to fly around on screen, but stay within the bounds of the screen,
we could do this inside the Init method of a stage:
M5Object* pObj = new M5Object(AT_Player);
GfxComponent* pGfxComp = new GfxComponent;
PlayerInputComponent* pInput = new PlayerInputComponent;
ClampComponent* pClamp = new ClampComponent;
pObj->AddComponent(pGfxComp);
pObj->AddComponent(pInput);
pObj->AddComponent(pClamp );
//Set position, rotation, scale here
//...
M5ObjectManager::AddObject(pObj);
This code works fine, but there are a few problems. First, we didn't specify what texture we want.
However, we could easily add a textureID or filename as a parameter to the GfxComponent constructor.
The larger problem is that this code was tedious to write and we don't want to write it again and
again. If we are creating a Player in another stage, it will likely contain the exact same code. So a
better approach is to factor this code into the M5ObjectManager:
M5Object* M5ObjectManager::CreateObject(M5ArcheTypes type)
{
switch(type)
{
case AT_Player:
AddObject(pObj);
//...
return pObj;
break;
case AT_Bullet:
//...More Code here
Now in our stage Init function, we can simply write the following:
M5Object* pObj = M5ObjectManager::CreateObject(AT_Splash);
//Set additional data here if needed
However, this is pretty hardcoded. This explicitly creates all of the components that a Player (and
every type) needs, which means that our M5ObjectManager now contains game-specific code. The benefit
of using the Component Object Model is that it is flexible, but we have lost some of that flexibility by
having a hardcoded switch statement. We really want our designers, not programmers, to choose what
goes into a Player, Raider, or SuperRaider. That means loading our object types from a file. In our
case, we will use INI files because they are simple to use and simple to understand. They consist of
global or labeled sections of key/value pairs. Here is an example Player archetype found in Player.ini:
posX = 0
posY = 0
velX = 0
velY = 0
scaleX = 10
scaleY = 10
rot = 0
rotVel = 0
components = GfxComponent PlayerInputComponent ClampComponent
[GfxComponent]
texture = playerShip.tga
[PlayerInputComponent]
forwardSpeed = 100
bulletSpeed = 7000
rotationSpeed = 10
Notice that the first (global) section of the INI file contains all data found in the M5object. Since we
know that those variables always exist in the object, they are placed at the top. This includes a list of
components that this archetype will use. Here we have GfxComponent, PlayerInputComponent, and
ClampComponent. The next sections are data associated with each component, for example, with the
GfxComponent we can specify our texture to load. The ClampComponent doesn't need any data loaded so we
didn't need to add a section for it.
Comparing the Component Object Model with the monolithic object or the inheritance tree, we can
see that the component-based approach is vastly more flexible and reusable. With this method, we can
write as many different components as we want and let the designer choose what behaviors each
object uses. The best part is that everything but the most game-specific components can be reused in
another game.
That means that the PlayerInputComponent probably can't be reused in another game, but the ClampComponent
and GfxComponent can be used whether we are making another Space Shooter, Platformer, or Racer.
One note about components used for graphics and physics, such as GfxComponent and
CircleColliderComponent: these are special in the sense that they need to interact with core engines in a
way that other components may not need to. For example, the Graphics engine might want to organize
these components based on whether they are world space objects or screen space objects (referred to
as HUD space, since these would be things such as buttons and health bars). The Physics engine might
want to use a special partition data structure to minimize the number of collision tests that need to be
performed. For this reason, these components are automatically registered to their respective core
engines when created through the object Manager and they automatically unregister when they are
deleted.
Performance concerns
There are a lot of benefits to using the Component Object Model. These days, many engines use this
approach because of the flexibility it provides. However, that flexibility comes at a cost to
performance. The biggest performance costs are calls to new/delete, cache coherency, and virtual
methods.
Our M5ObjectManager uses pointers to M5objects which uses an STL vector of pointers to components. This
means that as we create Bullets, Asteroids, Raiders, and Planets, we are constantly calling new and delete.
These are slow functions and have the chance to fragment our memory. In a later chapter, we will see
how object pools can help us solve both of these problems.
However, even with object pools, we still have problems with cache misses. The fact is that iterating
over an array of contiguous data is much faster than iterating over an array of pointers to data. When
using the Component object Model, the CPU will be spending a lot more time chasing pointers and
loading that data into the cache than if we just used arrays. Unfortunately, this is the price we pay for
flexibility. Depending on the game, this may or may not cause a problem.
Virtual methods are also a source of potential performance problems because the function to call must
always be looked up at runtime and they cannot be inlined by the compiler. Again, this is the price we
pay for flexibility. We have an approach that allows our designer to load a behavior from a file and
change that behavior at runtime. In my opinion, that outweighs the performance issues, at least at the
beginning of the development cycle.
You may have heard premature optimization is the root of all evil. It is more important to focus on
making a fun game and to solve the performance problems later. You always have the option of
hardcoding specific behaviors or data in the game object much later in the development cycle. If
possible, you might merge two or more components that always get used together once you are in the
polish stage. However, by limiting your flexibility early on, you may never discover a fun feature that
comes from mixing two components in a way that wasn't originally planned.
My advice is to focus first on algorithmic optimizations, then macro optimizations, then finally micro
optimizations. What I mean is that it is better to worry about the time complexity of your physics
engine and how many draw calls or collision tests you are performing, before worrying about what is
in the CPU cache or the performance cost of virtual functions. While they can be a problem, these
things fall under the category of micro optimizations.
However, before starting the long process of creating a game using an unfamiliar game engine, it can
be a good idea to do some simple prototype tests to make sure the engine can meet the needs of the
game. For example, a programmer could approximate the number of objects, and components, and test
for performance to see whether the engine will work.
Summary
In this chapter, we have explored many different ways to create a game object. We have seen the
problems with using monolithic objects or large inheritance trees. We now know that neither of those
approaches scale when creating a large game. They both suffer from the problem of giant bloated
classes and dependencies in our code.
We have also seen the flexibility that using the Component Object Model can bring to our games. It
lets programmers focus on writing new code, while allowing designers to use that code to create new
object types, even at runtime. Since we can now define objects completely in a file, we can create a
tool that will let our designer, or even players, make completely new objects, or possibly a new
game.
We also briefly touched on the performance issues related to using the Component Object Model.
While these can be a problem, it is much better to focus on algorithmic optimizations then very low-
level CPU instruction optimizations. We will revisit these problems in later chapters.
For now, let's move on to a design pattern that can help us implement one of the big core engines of
our game, as well as a small but important type of component. In the next chapter, we will discover
how the State pattern can help decouple our code and can give us a great way to implement artificial
intelligence in our game.
Artificial Intelligence Using the State Pattern
In the last chapter, we discussed the Component Object Model. Giving an entity a behavior is now as
simple as just creating a new component and having that control the game object.
Whenever someone starts to make a game, they begin by writing gameplay code. That is the fun stuff.
Everyone wants to see graphics and physics take effect on screen. Things such as a pause screen,
options menu, or even a second level are an afterthought. The same happens for organizing the
behaviors of a player. Programmers are excited to make a player jump and make a player dash, but
with each new ability a player has, there are combinations that you may want to disallow. For
example, the player might not be allowed to dash while jumping, or may only be able to dash every 3
seconds. The State pattern solves these problems.
By coding the Game State Manager first, the problem of switching to a menu or pausing is solved. By
coding finite State Machines as a component for the game object, the problem of complex behavior or
a player or enemy is solved. By adding multiple State Machines to the same game object, complex
behavior can be created with very simple code, as is seen in many games, and is a widely used
feature built into the Unreal Engine and given visual editors in Unity using Hutong Games LLC's
popular Playmaker extension.
Chapter overview
In this chapter, we will create a simple State Machine to control the player via input, as well as
create an enemy State Machine which will detect when the player gets close to it and will follow
them when in range. We will also look at the base StateMachineComponent class in the Mach5 Engine and
show that, by writing code for a few states, we can create a more complex behavior quite easily. We
will also show that adding more than one State Machine to an object can create multiple behaviors to
run at the same time, avoiding duplicated state code.
Your objectives

Doing things based on our states
Why if statements could get you fired
The State pattern in action--M5StateMachine
The State pattern in action--StageManager
Issues with FSMs
The State pattern is a way to allow a game object to change its behavior and functionality in response
to different stimuli within the game, especially when a variable or a condition within that object
changes, as those changes may trigger a change in state. The state of an object is managed by some
context (commonly referred to in the game industry as a machine), but the states tell the machine when
to change states and thereby functionality. The State pattern contains two parts: the state and the
context. The context object holds the current state and can be used by the state class to change what
state should be run, whereas the state object holds the actual functionality:
In the Mach5 Engine, there is a class that already uses the State pattern (M5StateMachine) but, before we
dive into a completed version, let's actually build one from scratch.
There are multiple ways to implement the State pattern or to get a State-like behavior. We'll go over
some of the commonly seen versions and the pros and cons of using them before moving to our final
version.
We often write code to react to things happening within the game environment based on the
expectations of us, as well as our players. For instance, if we are creating a 2D side-scrolling
platformer game, when the player presses one of the arrow keys, we're expecting the player's
character to move and, whenever we press the spacebar, we expect the sprite to jump into the air. Or
perhaps in a 3D game, when our player sees a panel with a large button, they expect to be able to
press it.
Tons of things in our ordinary lives act this way as well, reactive to certain stimuli. For instance,
when you use your television remote, you expect certain things to happen, or even when swiping or
tapping on your mobile phone. Based on the stimuli provided, the state of our object may change. We
call something that can be in one of multiple states at a time a State Machine.
Almost every program you write can be considered a State Machine of some sort. The second that
you add in an if statement to your project, you have developed code that can be in at least one of
those states. That being said, you don't want to have a bunch of switch and/or if statements inside of
your code as it can quickly get out of hand and make it difficult to understand exactly what it is that
your code is doing.
As programmers, we often want to take our problems and break them down over and over again until
they're in their simplest form, so let's see a possible way to do that. In game development, you'll hear
references to an FSM which stands for Finite State Machine. Finite means that there are only a
certain number of states and that they are all clearly defined for what they can do and how they will
change between states.
Let's say we're going to create a simple enemy. This enemy will not do anything by default, but if the
player is nearby, it will move toward them. However, if the player gets too far away from them, then
they will stop their pursuit. Finally, if the player shoots the enemy, it will die. So, keeping that in
mind, we can extract the states that we'll need. They are as follows:
Idle
Follow
Death
While we are creating our State Machine, we need some way to keep track of what state our objects
are going to be in. One may think a way to do this would be to have a bool (Boolean value of true or
false) for each possible state there is and then set them all to false, except for the state that we're in.
This is a very bad idea.
Another thought could be to just have an integer and then set a value for each one that there is. This is
also a bad idea, as using numbers in this way is basically the same thing as using magic numbers in
our code, since the numbers have no logic to them for people to read. Alternatively, you could have
#defines for each possible value, but that will allow people to put in whatever number they want
without any protections at all. Instead, whenever we see a series of things where only one of them is
true at a time, we can make use of the programming feature of enumerations, called enums for short.
The basic concept of using enumerations is that you get to create your own custom data types which
are restricted to only have a certain list of values. Unlike integers or #defines, these numbers are
expressed using constants and allow us to have all of the advantages of having a value, such as being
able to compare values. In our case, an enum for our states would look something like the following:
enum State
{
Idle,
Follow,
Death
};
Acting on states
Now that we have our states defined, let's now make it so that we can actually do something in our
code based on what state our object is in. For this first example, I'm going to update the
ChasePlayerComponent class that already exists in the EngineTest project.
From the Solution Explorer tab on the right-hand side, open up the SpaceShooter/Components/ChasePlayerComp
folder and access the ChasePlayerComponent.h file. From there, replace the class with the following
changes in bold:
enum State{
Idle,
Follow,
Death
};
//!< Simple AI to Chase the Player

class ChasePlayerComponent : public M5Component
{
public:
ChasePlayerComponent(void);
virtual void FromFile(M5IniFile& iniFile);
virtual ChasePlayerComponent* Clone(void);
private:
float m_speed;
float m_followDistance;
float m_loseDistance;
void FollowPlayer();
float GetDistanceFromPlayer();
State m_currentState;
};
The FollowPlayer and GetDistanceFromPlayer functions are going to be helper functions for our functionality.
We've added our state enum to store each of the possible states we can be in, and we added the
m_currentState variable to hold the current state we are in. To determine when we should switch states,
we have two other values, m_followDistance and m_loseDistance, which are the distance in pixels that our
player needs to be from the enemy to follow them, and then how far the player needs to get away to
escape, respectively.
Now that we have that finished, let's first go ahead and add in the helper functions at the bottom of the
ChasePlayerComponent.cpp file so that we can have the proper functionality, once we update our other
functions:
/*************************************************************************/
/*!
Makes it so the enemy will move in the direction of the player
*/
/*************************************************************************/
void ChasePlayerComponent::FollowPlayer()
{
std::vector<M5Object*> players;
M5ObjectManager::GetAllObjectsByType(AT_Player, players);
M5Vec2 dir;
M5Vec2::Sub(dir, players[0]->pos, m_pObj->pos);
m_pObj->rotation = std::atan2f(dir.y, dir.x);
dir.Normalize();
dir *= m_speed;
m_pObj->vel = dir;
}
/*************************************************************************/
/*!
Returns the distance of the object this is attached to the player
*/
/*************************************************************************/
float ChasePlayerComponent::GetDistanceFromPlayer()
{
std::vector<M5Object*> players;
M5ObjectManager::GetAllObjectsByType(AT_Player, players);
return M5Vec2::Distance(m_pObj->pos, players[0]->pos);

}
These functions use some basic linear algebra in order to move our object toward the player and to
get the distance between two positions.
Diving into the mathematics behind it is out of the scope of this book, but if you're
interested in learning more, I highly suggest you check out the following link. The
code is written for Cocos2D so it will not be exactly the same as what Mach5 would
use, but the concepts are explained very well: https://www.raywenderlich.com/35866/trigonometr
y-for-game-programming-part-1 .
Now that we have that functionality in, we need to update a couple of things. First of all, we will use
the constructor to set the initial value of our currentState variable:
/*************************************************************************/
/*!
Sets component type and starting values for player
*/
/*************************************************************************/
ChasePlayerComponent::ChasePlayerComponent(void):
M5Component(CT_ChasePlayerComponent),
m_speed(1)
{
m_currentState = Idle;
}
Next, we need to tell our object to read in the values of our object through its INI file:
void ChasePlayerComponent::FromFile(M5IniFile& iniFile)
{
iniFile.SetToSection("ChasePlayerComponent");
iniFile.GetValue("speed", m_speed);
iniFile.GetValue("followDistance", m_followDistance);
iniFile.GetValue("loseDistance", m_loseDistance);
FromFileis only called once on the first object that gets created in initialization. In order to make it
easy to tweak values without having to recompile the project, Mach 5 reads in information from a file
to set variables. We haven't modified the .ini file yet, but we will once we finish all of these
modifications:
M5Component* ChasePlayerComponent::Clone(void)
{
ChasePlayerComponent* pNew = new ChasePlayerComponent;
pNew->m_speed = m_speed;
pNew->m_followDistance = m_followDistance;
pNew->m_loseDistance = m_loseDistance;
return pNew;
}
We then need to go to Windows Explorer and move to the project's EngineTest/EngineTest/ArcheTypes

folder, and then access the Raider.ini file and add the new properties to the object:
posX = 0
posY = 0
velX = 0
velY = 0
scaleX = 10
scaleY = 10
rot = 0
rotVel = 0
components = GfxComponent ColliderComponent ChasePlayerComponent
[GfxComponent]
texture = enemyBlack3.tga
drawSpace = world
[ColliderComponent]
radius = 5
isResizeable = 0
[ChasePlayerComponent]
speed = 40
followDistance = 50
loseDistance = 75
If a text editor doesn't open for you, feel free to use Notepad. In this case, we are adding in two new
properties which represent the values we created earlier.
Then, we need to update our stage so it's a little easier for us to do some testing. Back in Windows
Explorer, open up the EngineTest/EngineTest/Stages folder and then open up the Level01.ini file and set it to
the following:
ArcheTypes = Player Raider
[Player]
count = 1
pos = 0 0
[Raider]
count = 1
pos = 100 10
With this, our level will just have our player in the center of the world and an enemy Raider
positioned at (100, 10). With all of that accomplished, save the files and dive back into our
ChasePlayerComponent.cpp file and replace the Update function with the following:
void ChasePlayerComponent::Update(float)
{
// Depending on what state we are in, do different things
switch (m_currentState)
{
case Idle:
// No longer move if we were
m_pObj->vel = M5Vec2(0, 0);
// If the player gets too close, the enemy notices them

if (GetDistanceFromPlayer() < m_followDistance)
{
// And will begin to give chase
m_currentState = Follow;
}
return;
case Follow:
// Follow the player
FollowPlayer();
// If the player manages to get away from the enemy

if (GetDistanceFromPlayer() > m_loseDistance)
{
// Stop in your tracks
m_currentState = Idle;
}
break;
case Death:
// Set object for deletion
m_pObj->isDead = true;
break;
}
Save everything and go ahead and run the project. If all goes well, you should see a scene like this:
Notice that our enemy is not moving at the beginning due to it being in the Idle state. However, if we
move closer to it, it would look something like this:
You'll see that it now follows us without stopping. If we manage to move far enough away from the
enemy though, they'll stop:
This clearly shows the basic principles of the State pattern in use, though there are a number of things
we can do to improve this, which we will talk about soon.
Issues with conditionals
The next thing we need to consider is how we should do something based on what state we are in.
When writing programs, conditional clauses such as the if and switch statements that we learned about
earlier may make your code more difficult to manage. Sometimes, when writing code for specific
functionality, writing if statements is completely understandable, especially if it makes sense when
you are writing it. For example, the following code makes perfect sense:
void MinValue(int a, int b)
{
if (a < b)
return a;
else
return b;
}
// Could also be written in the following way:

void MinValue(int a, int b)
{
return (a < b) ? a : b;
}
However, if you are writing something where you are checking what the type of an object is, or
whether a variable is of a certain type, that is a bit of an issue. For instance, look at the following
function:
void AttackPlayer(Weapon * weapon)
{
if (weapon.name == "Bow")
{
ShootArrow(weapon);
}
else if (weapon.name == "Sword")
{
MeleeAttack(weapon);
}
else
{
IdleAnimation(weapon);
}
}
As you can see, if we start going down this path, we will need to add many different checks
throughout our project, which will make our code hard to change if we ever decide to add more things
to support here. First of all, instead of a bunch of if/else statements, when we see something that's
comparing the same value and doing something based off of that value, we should be using a switch
statement, like we did earlier, with a few modifications:
{
// C++ doesn't support using string literals in switch
// statements so we have to use a different variable
// type, such as an integer
switch (weapon.type)
{
case 0:
ShootArrow(weapon);
break;
case 1:
MeleeAttack(weapon);
break;
default:
IdleAnimation(weapon);
But in this particular case, we are just calling a different function based on the value, with each of the
functions being some kind of attack. Instead, we should make use of polymorphism and have the code
automatically do the correct thing:
class Weapon
{
public:
virtual void Attack()
{
// Do nothing
};
};
class Bow : Weapon

{
public:
virtual void Attack()
{
// Attack with Bow
};
};

{
weapon->Attack();
}
Now whenever we call AttackPlayer, it will do the correct thing automatically.
Just remember that creating complex behavior leads to ugly code being written and increases the
likelihood of bugs. If you forget about a condition that needs to be there, your game hopefully would
break, letting you know there is a problem, but it could not do anything. Then, when you find your
game crashes down the road, your life becomes a lot more complex and your game could become
unplayable or just plainly not fun.
Robert Elder has a link of the subject which I think explains the kind of crazy things
that you can do with conditional statements, which would almost certainly get you
fired: http://blog.robertelder.org/switch-statements-statement-expressions/.
Don't lose sleep over having conditionals in your code, but make sure that you only include them
when you actually need them there. As you continue coding, you'll have a better idea as to when it's a
good idea or not, but it is something to keep in mind.
So currently, you'll notice that in the Idle state we are setting our velocity to 0,0 every single frame. In
this simple example, it's not a terribly big deal, but this overdoing of calculations is something that
we'd like to avoid in the future. We only really need to do it once, right when we enter the state. We
may also want to do certain actions when we leave the state, but we won't be able to do that in the
current form of our State Machine, so we are going to need to redo some stuff.
First, let's go back to the ChasePlayerComponent.h file and add the following bold function definitions:
class ChasePlayerComponent : public M5Component
{
public:
ChasePlayerComponent(void);
virtual M5Component* Clone(void);
virtual void EnterState(State state);
virtual void UpdateState(State state, float dt);
virtual void ExitState(State state);
virtual void SetNewState(State state, bool initialState = false);
private:
float m_speed;
float m_followDistance;
float m_loseDistance;
void FollowPlayer();
float GetDistanceFromPlayer();
State m_currentState;
};
So instead of having our Update function handle everything, we've now created three functions for each
of the different times that our state can be in: entering a new state, updating based on that state, and
then what to do when we leave the state. Aside from that, we also have a SetNewState function which
will take care of changing the state to something else. All of the functions take in a State enum to
choose how to execute, with the Update state also having the time that passed this frame, and the
SetNewState having an option for saying it's the first time you've set a state so you don't need to leave the
previous one. After that, we need to actually add in the functionality for these new functions:
void ChasePlayerComponent::EnterState(State state)
{
switch (state)
{
case Idle:
// No longer move if we were
if (m_pObj)
{
m_pObj->vel = M5Vec2(0, 0);
}
M5DEBUG_PRINT("\nIdle: Enter");
break;
case Follow:
M5DEBUG_PRINT("\nFollow: Enter");
break;
case Death:
M5DEBUG_PRINT("\nDeath: Enter");
break;
}
}
void ChasePlayerComponent::UpdateState(State state, float)

{
switch (state)
{
case Idle:
//M5DEBUG_PRINT("\nIdle: Update");
// If the player gets too close, the enemy notices them
if (GetDistanceFromPlayer() < m_followDistance)
{
// And will begin to give chase
SetNewState(Follow);
}
break;
case Follow:
//M5DEBUG_PRINT("\nFollow: Update");
// Follow the player

FollowPlayer();
// If the player manages to get away from the enemy

if (GetDistanceFromPlayer() > m_loseDistance)
{
// Stop in your tracks
SetNewState(Idle);
}
break;
}
}
void ChasePlayerComponent::ExitState(State state)

{
switch (state)
{
case Idle:
M5DEBUG_PRINT("\nIdle: Exit");
break;
case Follow:
M5DEBUG_PRINT("\nFollow: Exit");
break;
}
}
// initialState by default is false, so will only need to give

// second parameter the first time it is called
void ChasePlayerComponent::SetNewState(State state, bool initialState)
{
if (!initialState)
{
// Exit of our old state
ExitState(currentState);
}
// Then start up our new one

m_currentState = state;
EnterState(m_currentState);
}
And then, we need to update our Update function to just call our correct function:
void ChasePlayerComponent::Update(float dt)
{
UpdateState(m_currentState, dt);
}
We also need to change our constructor so that instead of setting the current state, we set it ourselves:
/*************************************************************************/
/*!
Sets component type and starting values for player
*/
/*************************************************************************/
ChasePlayerComponent::ChasePlayerComponent(void):
M5Component(CT_ChasePlayerComponent),
m_speed(1)
{
SetNewState(Idle, true);
}
First of all, note that I am calling the M5DEBUG_PRINT function. This is to make it easy to tell that we are
changing between different states. For the purposes of this demonstration, I commented out the Update
function's version, but it could be useful for you to check it out. Note in this version, we have a switch
statement for each of the functions and do something differently based on the state that is set in there.
In my version of the editor, by default the text will not be displayed on the screen. To fix this issue, go
to the SplashStage.cpp file and comment out the following bold code:
SplashStage::~SplashStage(void)
{
//We are done this with ArcheType so lets get rid of it.
M5ObjectManager::RemoveArcheType(AT_Splash);
//M5DEBUG_DESTROY_CONSOLE();
}
Now let's run our project!
You can tell from the editor when we are switching our states and that the code is being called
correctly!
This version works pretty well, but there are some issues with it; namely that it involves a lot of
rewriting, and we will need to copy/paste this functionality and make changes anytime we want to
make a new version. Next, we will take a look at the State Machine included in the Mach5 Engine and
the advantages that it has over what we've been talking about so far.
The State pattern in action - the
M5StateMachine class
The Mach5 Engine itself also has its own implementation of a State Machine, using inheritance to
allow users to not have to rewrite the base functionality over and over again and using function
pointers instead of having one function for each state. A function pointer is what it sounds like--a
pointer to the address in memory where the function is--and we can call it from that information.
To learn more about function pointers and how they are used, check out http://www.cprog
ramming.com/tutorial/function-pointers.html.
You can take a look at the base version of one here, starting with the Header file:
#ifndef M5STATEMACNINE_H
#define M5STATEMACNINE_H
#include "M5Component.h"
#include "M5Vec2.h"
//! Base State for M5StateMachines

class M5State
{
public:
//! Empty virtual destructor
virtual ~M5State(void) {}
//! Called when we first enter a state
virtual void Enter(float dt) = 0;
//! called once per frame
//! called before we exit a state
virtual void Exit(float dt) = 0;
};
//! Base class for Finite statemanchine component for AstroShot

class M5StateMachine : public M5Component
{
public:
M5StateMachine(M5ComponentTypes type);
virtual ~M5StateMachine(void);
void SetNextState(M5State* pNext);
private:
M5State* m_pCurr; //!< a pointer to our current state to be updated
};
#endif //M5STATEMACNINE_H
In the preceding code, note that we finally broke apart the StateMachine and the State object into their
own classes, with the state function having its own Enter, Update, and Exit functions. The State Machine
keeps track of the current state that we are in and updates appropriately using the Update and
SetNextState functions, and a SetStateState function is used to dictate what state we should start from.
The implementation for the class looks a little something like this:
#include "M5StateMachine.h"
M5StateMachine::M5StateMachine(M5ComponentTypes type):
M5Component(type),
m_pCurr(nullptr)
{
}
M5StateMachine::~M5StateMachine(void)
{
}
void M5StateMachine::Update(float dt)

{
m_pCurr->Update(dt);
}
void M5StateMachine::SetNextState(M5State* pNext)

{
if(m_pCurr)
m_pCurr->Exit();
m_pCurr = pNext;
m_pCurr->Enter();
}
This system provides a template that we can expand upon, in order to create more interesting
behavior that does something a bit more complex. Take, for example, the RandomGoComponent class, whose
header looks like this:
#ifndef RANDOM_LOCATION_COMPONENT_H
#define RANDOM_LOCATION_COMPONENT_H
#include "Core\M5Component.h"
#include "Core\M5StateMachine.h"
#include "Core\M5Vec2.h"
//Forward declation
class RandomGoComponent;
class RLCFindState : public M5State

{
public:
RLCFindState(RandomGoComponent* parent);
void Enter(float dt);
void Exit(float dt);
private:
RandomGoComponent* m_parent;
};
class RLCRotateState : public M5State
{
public:
RLCRotateState(RandomGoComponent* parent);
private:
float m_targetRot;
M5Vec2 m_dir;
};
class RLCGoState : public M5State
{
public:
RLCGoState(RandomGoComponent* parent);
private:
};
class RandomGoComponent : public M5StateMachine
{
public:
RandomGoComponent(void);
private:
friend RLCFindState;
friend RLCGoState;
friend RLCRotateState;
float m_speed;
float m_rotateSpeed;
M5Vec2 m_target;
RLCFindState m_findState;
RLCRotateState m_rotateState;
RLCGoState m_goState;
};
#endif // !RANDOM_LOCATION_COMPONENT_H
This class contains three states, Find, Rotate, and Go, which have been added as objects in the
RandomGoComponent. Each of the states has their own Enter, Update, and Exit functionality, in addition to the
constructor and a reference to their parent. The implementation for the classes looks something like
this:
#include "RandomGoStates.h"
#include "RandomGoComponent.h"
#include "Core\M5Random.h"
#include "Core\M5Object.h"
#include "Core\M5Intersect.h"
#include "Core\M5Gfx.h"
#include "Core\M5Math.h"
#include <cmath>
FindState::FindState(RandomGoComponent* parent): m_parent(parent)

{
}
void FindState::Enter()
{
M5Vec2 botLeft;
M5Vec2 topRight;
M5Gfx::GetWorldBotLeft(botLeft);
M5Gfx::GetWorldTopRight(topRight);
M5Vec2 target;
target.x = M5Random::GetFloat(botLeft.x, topRight.x);
target.y = M5Random::GetFloat(botLeft.y, topRight.y);
m_parent->SetTarget(target);
}
void FindState::Update(float)
{
m_parent->SetNextState(m_parent->GetState(RGS_ROTATE_STATE));
}
void FindState::Exit()
{
}
This class will just tell our main State Machine where its intended location is. This only needs to be
done once, so it is done in the Enter state. The Update state just states that after this is done, we want to
move to the Rotate state, and Exit does nothing. Technically, we could not create it, and that would be
fine as well since the base class doesn't do anything as well, but it is here if you wish to expand upon
it:
RotateState::RotateState(RandomGoComponent* parent): m_parent(parent)
{
}
void RotateState::Enter()
{
M5Vec2 target = m_parent->GetTarget();
M5Vec2::Sub(m_dir, target, m_parent->GetM5Object()->pos);
m_targetRot = std::atan2f(m_dir.y, m_dir.x);

m_targetRot = M5Math::Wrap(m_targetRot, 0.f, M5Math::TWO_PI);
m_parent->GetM5Object()->rotationVel = m_parent->GetRotationSpeed();
}
void RotateState::Update(float)
{
m_parent->GetM5Object()->rotation = M5Math::Wrap(m_parent->GetM5Object()->rotation, 0.f, M5Math::TWO_PI);
if (M5Math::IsInRange(m_parent->GetM5Object()->rotation, m_targetRot - .1f, m_targetRot + .1f))

m_parent->SetNextState(m_parent->GetState(RGS_GO_STATE));
}
void RotateState::Exit()
{
m_parent->GetM5Object()->rotationVel = 0;
m_dir.Normalize();
M5Vec2::Scale(m_dir, m_dir, m_parent->GetSpeed());
m_parent->GetM5Object()->vel = m_dir;
}
The Rotate state will just rotate the character till it is facing the location that it wants to go to. If it is
within the range of the rotation, it will then switch to the Go state. Before leaving though, it will set the
velocity of our parent to the appropriate direction in the Exit function:
GoState::GoState(RandomGoComponent* parent): m_parent(parent)
{
}
void GoState::Enter()
{
}
void GoState::Update(float)
{
M5Vec2 target = m_parent->GetTarget();
if (M5Intersect::PointCircle(target, m_parent->GetM5Object()->pos, m_parent->GetM5Object()->scale.x))
m_parent->SetNextState(m_parent->GetState(RGS_FIND_STATE));
}
void GoState::Exit()
{
m_parent->GetM5Object()->vel.Set(0, 0);
}
The Go state merely checks whether the enemy intersects with the target that we are set to go to. If it
does, we then set our state to move back to the Find state and start everything over again, and also stop
the player from moving in the Exit function:
RandomGoComponent::RandomGoComponent():
M5StateMachine(CT_RandomGoComponent),
m_speed(1),
m_rotateSpeed(1),
m_findState(this),
m_rotateState(this),
m_goState(this)
{
SetNextState(&m_findState);
}
void RandomGoComponent::FromFile(M5IniFile& iniFile)
{
iniFile.SetToSection("RandomGoComponent");
iniFile.GetValue("speed", m_speed);
iniFile.GetValue("rotationSpeed", m_speed);
}
RandomGoComponent* RandomGoComponent::Clone(void) const
{
RandomGoComponent* pNew = new RandomGoComponent;
pNew->m_speed = m_speed;
pNew->m_rotateSpeed = m_rotateSpeed;
return pNew;
}
M5State* RandomGoComponent::GetState(RandomGoStates state)

{
switch (state)
{
case RGS_FIND_STATE:
return &m_findState;
break;
case RGS_ROTATE_STATE:
return &m_rotateState;
break;
case RGS_GO_STATE:
return &m_goState;
break;
}
//In case somethings goes wrong

return &m_findState;
}
As you can see, this works in a very similar way to what we have done before--setting our first state,
getting the initial values from the INI file, and then setting things properly when cloned. Finally, we
also have a GetState function which will return the current state that the player has using a switch like
we talked about previously.
To see this in action, go ahead and go to the Raider.ini file and modify the code to fit the following:
posX = 0
posY = 0
velX = 0
velY = 0
scaleX = 10
scaleY = 10
rot = 0
rotVel = 0
components = GfxComponent ColliderComponent RandomGoComponent
[GfxComponent]
texture = enemyBlack3.tga
drawSpace = world
[ColliderComponent]
radius = 5
isResizeable = 0
[RandomGoComponent]
speed = 40
rotationSpeed = 40
If all went well, save the file and then run the project!
Now we will see the enemy continually move into new areas, rotating before going there!
The State pattern in action - StageManager
Another aspect of the Mach5 Engine that uses the State pattern is the M5StageManager class:
class M5StageManager
{
public:
friend class M5App;
//Registers a GameStage and a builder with the the StageManger

static void AddStage(M5StageTypes type, M5StageBuilder* builder);
//Removes a Stage Builder from the Manager
static void RemoveStage(M5StageTypes type);
//Clears all stages from the StageManager
static void ClearStages(void);
//Sets the given stage ID to the starting stage of the game
static void SetStartStage(M5StageTypes startStage);
//Test if the game is quitting
static bool IsQuitting(void);
//Test stage is restarting
static bool IsRestarting(void);
//Gets the pointer to the users game specific data
static M5GameData& GetGameData(void);
//Sets the next stage for the game
static void SetNextStage(M5StageTypes nextStage);
// Pauses the current stage, so it can be resumed but changes stages
static void PauseAndSetNextStage(M5StageTypes nextStage);
// Resumes the previous stage
static void Resume(void);
//Tells the game to quit
static void Quit(void);
//Tells the stage to restart
static void Restart(void);
private:
static void Init(const M5GameData& gameData, int framesPerSecond);
static void Update(void);
static void Shutdown(void);
static void InitStage(void);
static void ChangeStage(void);
};//end M5StageManager
Since there will only be one of these in the game, all of the functionality has been made static
similarly to a Singleton but, depending on the state that the project is in, it will do different things.
Take, for example, changing what stage we are in. I'm sure you'll find that it looks very similar to how
we changed states earlier:
void M5StageManager::ChangeStage(void)
{
/*Only unload if we are not restarting*/
if (s_isPausing)
{
M5ObjectManager::Pause();
M5Phy::Pause();
M5Gfx::Pause(s_drawPaused);
PauseInfo pi(s_pStage, s_currStage);
s_pauseStack.push(pi);
s_isPausing = false;
}
else if (s_isResuming)
{
/*Make sure to shutdown the stage*/
s_pStage->Shutdown();
delete s_pStage;
s_pStage = nullptr;
}
else if (!s_isRestarting) //Just changine the stage
{
delete s_pStage;
s_pStage = nullptr;
//If we are setting the next state, that means we are ignore all
//paused states, so lets clear the pause stack
while (!s_pauseStack.empty())
{
M5Gfx::Resume();
M5Phy::Resume();
M5ObjectManager::Resume();
PauseInfo pi = s_pauseStack.top();
pi.pStage->Shutdown();
delete pi.pStage;
s_pauseStack.pop();
}
}
else if (s_isRestarting)
{
}
s_currStage = s_nextStage;
}
I highly advise taking a closer look at the file and going through each function to see how they interact
with each other.
Issues with FSMs
We've seen some of the ways in which FSMs can be valuable things to add to your projects and how
they can make simple AI behaviors much easier, but there are some issues with them.
Traditional FSMs such as the ones we've displayed here can, over time, become unmanageable as you
continue to add many different states to them. The difficult part is keeping the number of states to a
minimum while also adding complexity by adding new contexts in which your characters can respond.
You'll also have a lot of similar code being written as you'll be rebuilding different behaviors that
have pieces of others, which can also be time-consuming. Another thing that's been going on recently
in the game industry is AI programmers moving on to more complex ways of handing AI, such as
behavior trees.
If you're interested in why some people believe that the age of Finite State Machines
is over, check out http://aigamedev.com/open/article/fsm-age-is-over/. A look at the issues with
FSMs, as well as some potential solutions to fix those issues, can be found here: http://
aigamedev.com/open/article/hfsm-gist/.
Summary
In this chapter, we learned about the State pattern, which is a way to allow a game object to change
its behavior and functionality in response to different stimuli within the game. We learned about the
State and the Context (Machine) and how they are used together. We then learned how we can use the
State pattern to gain some exposure toward AI programming, as well as how our project's Game State
Manager works and why it's important. Of course, FSMs are most popular in being used for AI, but
can also be used in UI as well as dealing with user input, making them another useful tool to have in
your arsenal.
Decoupling Code via the Factory Method
Pattern
Every project and every game design is going to change. One of the goals of object-oriented
programming is to program with that change in mind. This means writing flexible and reusable code
so that, when changes happen, the project doesn't fall apart. Unfortunately, the requirements are never
fully known and the vision of the designer is never 100% complete. The good news is that newly
added features may interact with old features in unexpectedly fun ways, leading to unplanned features
being created, which can make for a completely different, and more enjoyable, game.
In the worst-case scenario, the game design may not be fun at all, which can lead to drastic changes in
game object types, object behaviors, and even the design of the entire game. In this case, we want to
be able to rework our game to try new possibilities with the least amount of change to code. Changing
code takes time to write, test, and debug, and any time new code is added, it has the possibility of
causing bugs in older code.
Since we know that our design will change, we must plan for that change by following good design
principles and utilizing design patterns to solve common problems. This includes using flexible
designs such as the Component Object Model to avoid inheritance hierarchies. This includes using the
State pattern and finite state machines to avoid complex if statements and cascading if else chains that
lead to bugs any time a change occurs. This also includes things such as striving for low coupling and
avoiding hardcoding anything.
We all know that as programmers, we should avoid magic numbers. We don't want to see seemingly
random or odd numbers in our code. A hardcoded number isn't descriptive enough, which makes the
code hard to read, maintain, and debug. If we need to change a magic number, we have to spend time
searching our code for every time it appears. Unfortunately, these hardcoded values tend to change a
lot during development. In the real world, gravity may be 9.8 m/s2, but in the game world, we may
need to adjust it so the game is more fun, or possibly change it at runtime so the player can walk on
the ceiling.
When we think about hard coding, we usually think about magic numbers. However, a call to the new
operator using a concrete class type such as a new ChaseAIComponent may be readable, but it is just as
likely to change as the value of gravity or the size of an array.
Chapter overview
In this chapter, you will be focusing on a common interface to create new objects without the need to
call constructors directly. First, we will look at reasons why a switch statement can be harmful. Next
we will look at a design principle that leads us to our final solution, a factory. Then we will look at
few different ways to design our factories to make them flexible and reusable.
Your objective
Over the course of this chapter, we will be looking at a few more important concepts and principles
that can make our programs better. Here is an outline of what we will cover and your tasks for this
chapter:
Learning why using switch statements can be bad

Learning the Dependency Inversion Principle
Learning the Factory Method pattern
Building a Component, Stage, and Object Factory
Improve your Factories by using templates
The trouble with switch statements
When first learning to program, simply understanding the grammar of the language is very difficult.
Often, new programmers focus on the syntax of a function call or for loop and they don't even think
about making reusable, maintainable code. This is partly because they jump into the coding without
planning anything out. This is true for games as well. Often, new programmers want to get straight to
writing the game and they forget about things such as user interface and pause menus. Things such as
the window resolution, enemy placement in the stages, and even where the mouse should be when
clicking buttons will be hardcoded. Here is code from one of the authors' first games. This was a
section in the MainMenu function that would switch the game into the Load state when the button was
clicked. The variable p is the location of the mouse:
if ((p.x > .15 * GetSystemMetrics(SM_CXSCREEN)) &&
(p.x < .42 * GetSystemMetrics(SM_CXSCREEN)) &&
(p.y > .58 * GetSystemMetrics(SM_CYSCREEN)) &&
(p.y < .70 * GetSystemMetrics(SM_CYSCREEN)))
{
if (mousedown)
{
mGameState = TCodeRex::LOAD;
mGameLevel = L0;
}
}
Code similar to this was repeated (without a loop) four times, because there were four buttons in the
MainMenu. Of course, there were better (much better) ways to write this code. It is important to separate
game stages from the Graphics class for one, and buttons can manage themselves if they are written
correctly. However, the worst part of this old code base was how game stages were changed.
In that game, there was an Update function in the Engine class that contained a 507-line-long switch
statement. This included a switch within a switch for situations where there needed to be a menu stage,
but there were multiple menus to choose from. Here is some sample code from when a level needed
to be loaded:
case TCodeRex::LOAD:
{
StopGameMusic();
StopMenuMusic();
switch (mGameLevel)
{
case TCodeRex::L0:
{
grafx.LoadScreen(mLoading);
mLoading = true;
if (ObjectMgr->LoadLevel(".\\Levels\\Level_00.txt"))
{
mPlayer1->SetPhysics().setPosition(
(Physics::real)ObjectMgr->PlayerX,
(Physics::real)ObjectMgr->PlayerY, 0.0f);
grafx.GetCamera().Move(
ObjectMgr->PlayerX - 500,
ObjectMgr->PlayerY - 500);
ObjectMgr->AddToList(mPlayer1);
mPlayer1->SetHealth(100);
mBGTexture = grafx.GetTextureMgr().GetTile(49);
mplaying = true;
mGameState = TCodeRex::PLAY;
}
else
mGameState = TCodeRex::MENU;
break;
}//end case TCODEREX::LOAD
If you feel like that is hard to look at, you are right. For some reason, the stage-switching code was
responsible for setting the player's health and position, moving the camera, setting the background
texture, and going back to the menu if something failed to load.
The reason we are showing this is so you can see an example of really bad code. This Engine class
was tightly coupled with the graphics class, the physics class, the object manager class, and the player
class. If any of these classes changed or were removed, the code in the Engine class needed to change.
Another problem is that there are many dependencies here, which means a change to any of the
classes above will cause the code to be recompiled as well.
One of the authors was in a similar situation a few years later while working on an iPhone game as
contract programmer. The original design called for only two stages, the main menu and the gameplay.
Since the plan was so simple, code similar to what we saw above was used. This soon became a
problem, because all game designs change even if the project is only three months long.
After one month, the game was completed to specification, including a level editor so the designer
could create as many levels as were needed. The goal was to spend the rest of the time adding effects
and polish to the game. However, every week, the producer and designer would come and say they
needed a new menu or transition screen.
In the end, the game design called for an options menu, a level select menu, a credits screen, a
celebration screen every 10 levels, and a congratulations screen after completing the game. The
simple two-stage code that was originally written ended up being a hacked-together mess because of
the fast development timeline and the constant addition of new features. One horrible line of code
was similar to this:
if(levelCounter == 81)
ShowCongratsScreen();
This worked because there were 80 levels in the game and the levelCounter incremented after
completing a level. After seeing game code like this, we hope you understand why it is important to
plan for change to occur.
In Chapter 3, Creating Flexibility with the Component Object Model, we saw how to create
components for our game object so we can easily handle changes to our object design. This is an
important step in creating flexible code; however, when we create an object in our stage, we still
must use new to hardcode the concrete types that make up our object:
//...
M5ObjectManager::AddObject(pObj);
This means that our derived M5Stage class is tightly coupled to the components of this object type, and
really all object types that need to be created in this level. If the objects need different components, or
any of the components are changed or removed, then our Stage class must be changed.
One way to solve this (as we saw in Chapter 3, Creating Flexibility with the Component Object
Model) is to put this code into our M5ObjectManager class. This way, our stages don't need to be
constantly updated as our types are modified:
{
switch(type)
{
case AT_Player:
// Create player object
M5Object* pPlayer = new M5Object(AT_Player);
// Create the components we'd like to use

GfxComponent* pGfx = new GfxComponent;
// Attach them to our player

pObj->AddComponent(pGfx);
pObj->AddComponent(pClamp);
//Add this object to the M5ObjectManager

AddObject(pPlayer);
return pPlayer;
break;
case AT_Bullet:
//...More Code here
This solves the problem of needing to change our stages whenever an object is changed. However,
we will still have to change our switch statement and the object manager if our objects or components
change. Actually, except for this function, the Object Manager doesn't really care about any derived
component types. It only needs to be dependent on the M5Component abstract class. If we can fix this
function, we can completely decouple our derived component types from this class.
The solution to our problem is the same solution used to solve the stage management problem I faced
years ago, the Dependency Inversion Principle.
The Dependency Inversion Principle
The concept of avoiding concreate classes isn't new. Robert C. Martin defined this idea in The C++
Report in May 1996 in an article titled The Dependency Inversion Principle. It is the D in his SOLID
design principles. The principle has two parts:
High-level modules should not depend on low-level modules. Both should depend on
abstractions.
Abstractions should not depend on details. Details should depend on abstractions.
While this may seem like a mouthful, the concept is actually very easy. Imagine we have a StageManager
class that is responsible for initializing, updating, and shutting down all of the stages in our game. In
this case, our StageManager is our high-level modules, and the stages are the low-level modules. The
StageManager will control the creation and behavior of our low-level module, the stages. This principle
says that our StageManager code shouldn't depend on derived stage classes, but should instead depend on
an abstract stage class. To see why, let's look at an example of not following this principle.
Here our StageManager has a function called Update, which looks like this:
void StageManager::Update()
{
Level1 level1;
level1.StartGame();
while(nothingChanged)
level1.PlayGame()
level1.EndGame();
}
Of course, this is just the first level of our game, so we will need to include code to update the main
menu, and the second level of our game. Since these classes are unrelated, we will need a switch
statement to make a choice between stages:
{
switch(currentStage)
{
case LEVEL1:
{
SpaceShooterLevel1 level1;
level1.StartLevel();
while(currentStage == nextStage)
level1.PlayLevel()
level1.EndLevel();
break;
}
case LEVEL2:
{
SpaceShooterLevel2 level2;
level2.StartGame();
level2.PlayLevel()
level2.EndLevel();
break;
}
case MAIN_MENU:
{
SpaceShooterMainMenu mainMenu;
mainMenu.OpenMenu();
mainMenu.Show()
mainMenu.CloseMenu();
break;
}
}//end switch
}//end Update
You can see that as we continue to add more and more levels, this code will get large very fast and it
will quickly become impossible, or just really difficult to maintain. This is because we will need to
go into this function and remember to update the switch statement every time a new stage is added to
the game.
The first part of this section, tells us that that our StageManager class shouldn't depend on levels or
menus, but should instead depend on a common abstraction. That means that we should create an
abstract base class for all of our stages. The second part says that the abstraction shouldn't care
whether the stage is a level or a menu. In fact, it shouldn't even care that we are making a stage for a
space shooter game, a platformer, or a racing game. A major benefit of using this principle, besides
having more flexible code, is that our StageManager class will no longer depend on anything in this
specific game, so we can reuse it for our next project:
StageManager depends on specific classes
//Stage.h
class Stage
{
public:
virtual ~Stage(void) {}//Empty virtual destructor
virtual void Init(void) = 0;
virtual void Update(void) = 0;
virtual void Shutdown(void) = 0;
};
//StageManager.cpp
{
//We will talk about how to get the current stage soon
Stage* pStage = GetCurrentStage();
//Once we have the correct Stage we can run our code

pStage->Init();
pStage->Update();
pStage->Shutdown();
}
//Example of a derived class

//MainMenu.h
class MainMenu : public Stage
{
public:
virtual ~MainMenu(void);
virtual void Init(void);
virtual void Update(void);
virtual void Shutdown(void);
private:
//Specific MainMenu data and functions here...
};
StageManager only depends on abstraction
Now the Update function of StageManager is much simpler. Since we are only dependent on the abstract
class, our code no longer needs to change based on derived implementations. We have also simplified
the interface of all stages. The functions of each stage no longer change based on the details of the
class (menu, level, and so on); instead, they all share a common interface. As you can see, knowing
about the Dependency Inversion Principle will not only make you a big hit at parties, it will also
allow you to decouple your code base and reuse higher-level modules.
We still have the problem of selecting the correct derived class. We don't want to put a switch
statement in the Update function of StageManager. If we do that, we will have dependency problems like
before. Instead, we need a way to select the correct derived class while depending only on the base
class.
The Factory method pattern
The Factory method pattern is exactly the design pattern we need to solve our problem. The purpose
of this pattern is to have a way of creating the derived class that we want without needed to specify
the concreate class in our high-level module. This is done by defining an interface for creating objects,
but letting subclasses decide which class to instantiate.
In our case, we will create a StageFactory interface with a method called Build that will return a Stage*.
We can then create subclasses such as Level2Factory to instantiate our derived classes. Our StageManager
class now only needs to know about the Stage and StageFactory abstractions:
Creating a Stage Factory
//StageManager.cpp
{
Stage* pStage = m_stageFactory->Build();
pStage->Init();
pStage->Update();
pStage->Shutdown();
m_StageFactory->Destroy(pStage);//stage must be destroyed
}
Notice that we have moved the call to new from our StageManager::Update function into derived
StageFactory methods. We have successfully decoupled our StageManager from our derived Stage classes.
However, the call to Build represents the creation of only one derived Stage class. We still need a way
to choose which derived Stage we want to use and which derived StageFactory we need to instantiate.
We need a way to choose between different types of factories. Before we look at the solution that was
used in the Mach5 Engine, let's look at an alternative Factory method, the Static Factory.
The Static Factory
The simplest way to implement a factory method the way we want is with either a global function or a
static class function. We could define a function called MakeStage outside of StateMananger that is
responsible for instantiating the correct derived type based on a parameter. In this case, we will use
an enum called StageType to help us choose our correct type:
//MakeStage.cpp
Stage* MakeStage(StageType type)
{
switch(type)
{
case ST_Level1:
return new Level1;
case ST_LEVEL2:
return new Level2;
case ST_MainMenu:
return new MainMenu;
default:
//Throw exception or assert
}
}
If we use this style of factory, our StageManager::Update function will look like this:
{
Stage* pStage = MakeStage(currentStage);
pStage->Init();
pStage->Update();
pStage->Shutdown();
DestroyStage(pStage);//Clean up the stage when done
}
This version of the factory method works exactly as we want. We can now choose which derived
Stage class is instantiated. We still have a switch statement that we must maintain, but at least our
higher-level module is no longer dependent on derived classes. In the default case, where our switch
statement can't match the correct type, we are left with the choice of using an assert to crash the
program, throwing an exception and letting the client resolve the issue, or perhaps returning null,
which still gives responsibility to the client.
The Static Factory successfully decouples our StageManager class from specific derived Stage classes,
while allowing us to choose which stage will be instantiated at runtime. This is great, but as I said,
this isn't how the Mach5 Engine implements Stage or component factories. Instead, Mach5 uses a more
dynamic solution, so we will call it the Dynamic Factory.
The Dynamic Factory
While the Static Factory is simple enough for our purpose, the Mach5 Engine uses a different
approach. This approach combines the polymorphic solution of the classic Factory method with the
selection capability of the Static Factory. This new approach uses a searchable collection of derived
StageFactory classes.
Remember that problem with the classic Factory method is that the method represents only one class
to instantiate. This allows our code to be flexible because we don't depend on a derived Stage class or
a call to the new operator. However, we still needed a way to get specific derived StageFactory
instances.
In the Mach5 Engine, the names are changed a little. There is only one StageFactory class, which
contains a collection of M5StageBuilder pointers (these are classic Factories), which implement a Build
method. The design looks like this:
Design of Dynamic Factory
The first class we want to see is the base M5Stage class:

class M5Stage
{
public:
virtual ~M5Stage(void) {} //Empty virtual destructor
virtual void Load(void) = 0;
virtual void Init(void) = 0;
virtual void Shutdown(void) = 0;
virtual void Unload(void) = 0;
};
The base M5Stage class is a pretty simple abstract base class with a virtual destructor. The specific
virtual functions in M5Stage aren't important for the details of the factory. We are simply showing the
class here, because the M5StageManager and M5StageFactory will be using this abstraction.
C++ classes intended for polymorphic use, including abstract base classes, should
always implement a virtual destructor, otherwise the correct derived class destructor
cannot be called.
Creating our Stage Builders
Next let's look at our base builder class. Notice that this is the same type of interface that would be
used in the classic Factory method pattern. This abstraction declares a method that returns another
abstraction, in this case M5Stage.
Just like before, we need to have an empty virtual destructor so when we use this class in a
polymorphic way, the correct destructor will be called. Also like before, the other methods are
marked as purely virtual, which disallows direct instantiation of this class. That means we cannot
create an M5StageBuilder directly. We must derive from it, and implement the pure virtual methods:
class M5StageBuilder
{
public:
virtual ~M5StageBuilder() {} //Empty virtual destructor
virtual M5Stage* Build(void) = 0;
};
Even though the name is different, this is how the classic Factory method would be implemented. The
Mach5 Engine calls this a Builder instead of a Factory but it is only a change in the name and not the
functionality. The name of the Build method isn't important. Some programs will call the method Create
or Make. Mach5 calls it Build but any of these names are fine.
No matter what the name is, a user creating a game with the Mach5 Engine would want to derive their
own specific stage builders for the stages in their game. For this book, we have stages for a space
shooter game called AstroShot. In order to make builders for these stages, we need to derive from
M5StageBuilder and implement the Build method. For example, if we had M5Stage derived classes called
SplashStage and MainMenuStage, we would create builders like this:
//SplashStageBuilder.h
class SplashStageBuilder: public M5StageBuilder
{
public:
virtual M5Stage* Build(void);
};
//SplashStageBuilder.cpp
M5Stage* SplashStageBuilder::Build(void)
{
return new SplashStage;
}
//MainMenuStageBuilder.h
class MainMenuStageBuilder: public M5StageBuilder
{
public:
};
// MainMenuStageBuilder.cpp
M5Stage* MainMenuStageBuilder::Build(void)
{
return new MainMenuStage;
}
Note here that the use of the keyword virtual in the derived classes is completely optional. In the pre
C++ 11 days, programmers would mark the function virtual as a form of documentation to other
programmers. These days, you can add the override specifier to a virtual function so the compiler can
issue an error if the function isn't a true override.
To some, this might seem a little tedious. In fact, the biggest complaint I hear from people first
learning about object-oriented programming is that they feel like they waste a lot time creating many
files with small classes in them. To them, it feels like a lot of work for very little pay-off.
I agree that programming this way can require lots of files and lots of little classes. However, I
disagree with it being a waste of time. I think these arguments are a result of shortsighted thinking.
They are only thinking of the time it takes to write the original code, but they aren't thinking about the
time it will save when making changes to the design or the time it will take to test and debug.
It doesn't take much time to create new files. With an integrated development environment such as
Visual Studio, it takes less than 10 seconds to create a source and header file. Writing small classes
such as the builders above doesn't take much time either. In total, writing these two classes takes less
than five minutes. Of course, this is more than directly writing new into a high-level module, but
remember that the goal is to write code that can adapt to change in our game design.
These shortsighted arguments are similar to the complaints of new programmers learning to write
functions. We have already discussed the benefits of writing functions and those same benefits apply
here. We shouldn't think only of the time it takes to write the initial code. We need to factor in how
long it will take to test and debug the code, how likely it is to introduce new bugs into old code, how
long it will take to modify if our design changes a month or year from now, and if the design does
change, how likely is it that modifying our code will introduce bugs.
It is important to understand that by using design patterns we are writing more code upfront, so that
we can cut down on the time it takes to test, debug, integrate, maintain, and change our code in the
future. It is important to understand that writing the original code is easy and cheap, while changing it
later is much harder and more expensive. In this case, cheap and expensive could be referring to
hours worked or money spent paying programmers.
The template builder
Those worried about writing a lot of little classes are in luck. Most, if not all, of our builders will be
the same except for the specific derived class they instantiate. That means we can use the power of
C++ templates to create builders for us. Our templated builder will look like this:
//M5StageBuilder.h
class M5StageTBuilder : public M5StageBuilder
{
public:
};

M5Stage* M5StageTBuilder<T>::Build(void)
{
return new T();
}
This code works great for most of our stages. The only time it doesn't work is when we need to do
something more specific, such as call a non-default constructor, or a function specific to the derived
stage.
Notice that the implementation of the Build function is also included in the .h file. This is because
template functions are not the same as regular functions. They work as recipes so the compiler knows
how to generate the function for the specific type. Each time we need to use this function, the compiler
will need to know about the recipe. This allows the compiler to instantiate the function as opposed to
requiring the user to explicitly instantiate the Builder classes they need before using them. So when we
want to use this class, it will look something like this:
//SomeFile.cpp
#include "M5StageBuilder.h"
#include "MainMenuStage.h"
void SomeFunction(void)
{
//Creating the class needs the type
M5Builder* pBuilder = new M5StageTBuilder< SplashStage >();
//But using the Build function doesn't need the type

M5Stage* pStage = pBuilder->Build();
}
Creating the Dynamic Factory class
So far, we have only created our builders, which are equivalent to the classic Factory method pattern.
However, we haven't seen the Factory part of the Dynamic Factory. Let's look at how the Mach5
Engine implements the StageFactory class:
class M5StageFactory
{
public:
~M5StageFactory(void);
void AddBuilder(M5StageTypes name, M5StageBuilder* builder);
void RemoveBuilder(M5StageTypes type);
void ClearBuilders(void);
M5Stage* Build(M5StageTypes name);
private:
typedef std::unordered_map<M5StageTypes,
M5StageBuilder*> BuilderMap;
typedef BuilderMap::iterator MapItor;
BuilderMap m_builderMap;
};
As you can see, M5StageFactory isn't a very complicated class. Once you understand the design behind
the patterns, implementing them is usually not very difficult. As for this class, it is only five methods
and one member. The private section looks a little complicated because Mach5 prefers using a typedef
for templated containers. Since the container is used in all of the private functions, let's look at the
member before exploring the five methods.
Let's first look the typedefs:

typedef std::unordered_map<M5StageTypes, M5StageBuilder*> BuilderMap;
Since we want a container of M5StageBuilders, we have a few choices. We could use an STL vector or
list, but those containers are not ideal for searching because of the potential lack of efficiency if we
have many builders. However, this is exactly what the STL map and unordered_map are perfect for. They
allow us to save key/value pairs and later use the key to efficiently find the value, even if we had
thousands of builders. We will use the M5StageTypesenum as our key, and use a derived M5StageBuilder* as
our value.
An STL map is implemented as a tree, while unordered_map is implemented as a hash

table. In general, this means that the map will use less memory, but will be a little
slower to search. unordered_map will use more memory but search much faster. In our
games, we aren't likely to create thousands of stages, so the difference in speed isn't
going to matter that much, especially since we won't be searching very often either.
We choose the hash table because, on a PC, I am less concerned about memory and
more concerned about speed. If you are interested in learning more, check out http://ww
w.cplusplus.com/reference/ for lots of information about the standard library.
We should also prefer to have as readable code as possible. Using a typedef will help others
understand our code, because we only need to write the long std::unordered_map< M5StageTypes,
code one time. After that, we can use the shorted name, in this case BuilderMap. This also
M5StageBuilder*>
gives us the ability to easily change containers if we later decide to use a map instead:
The next typedef gives us a shorted name for our BuilderMap iterators.
This is unnecessary with the C++ 11 auto keyword, but doesn't make our code less
readable so we have chosen to use the typedef.
Finally, the actual member:

This will be our container that maps M5StageTypes to M5StageBuilder*. We should make it private because
we want all builders to be added and removed with the class methods so the data can be validated.
Now for the class methods. Let's start with the most important method of the factory:
M5Stage* M5StageFactory::Build(M5StageTypes type)
{
ArcheTypeItor itor = m_builderMap.find(type);
if (itor == m_builderMap.end())
return nullptr;
else
return itor->second->Build();
}
The Build method is where the magic happens, at least for the user. They pass in a stage type and we
build the correct stage for them. Of course, we use the find method first to make sure the type has been
added. If it can't be found, we use a debug assert to let the user know that this type wasn't added. In
general, the find method is safer to use than the operator[] that exists in the map and unordered map.
Using the operator[] will create and return a null value if the key doesn't exist. If that happened while
trying to build, we would get a null pointer exception, which would crash the program without giving
the user an explanation of why.
We have the choice of adding some default stage to the map, and building that if the correct type can't
be found. However, there is a chance that the programmer won't notice that a mistake has been made.
Instead, we have chosen to return a null pointer. The requires the user to check whether the builder is
valid before using it, but also means that the code will crash if they don't fix the problem:
bool M5StageFactory::AddBuilder(M5StageTypes name,
M5StageBuilder* pBuilder)
{
std::pair<MapItor, bool> itor = m_builderMap.insert(
std::make_pair(name, pBuilder));
return itor.second;
}
The AddBuilder method allows our user to associate an M5StageTypes value with a derived M5StageBuilder. In
this case, our code doesn't know or care whether pBuilder is pointing to a templated class or not. All
that matters is that it derives from M5StageBuilder.
Just as before, we should write our code to help the user find and fix bugs if they occur. We do that by
testing the return value of the insert method. The insert method returns a pair in which the second
element will tell us whether the insert was successful or not. Since a map and an unordered_map do not
allow duplicates, we can test to make sure the user isn't associating an M5StageTypes value with two
different builders. If the user tries to use an enum value twice, the second builder will not be inserted
and false will be returned.
The STL versions of map and unordered_map do not allow duplicate items. If you wish to
have duplicates, you can replace the container with multimap or unordered_multimap, which
do allow duplicates. It wouldn't be useful to use the multi versions in this class, but
they are good tools to know about.
void M5StageFactory::RemoveBuilder(M5StageTypes name)
{
BuilderMap::iterator itor = m_builderMap.find(name);
if (itor == m_builderMap.end())
return;
//First delete the builder

delete itor->second;
itor->second = 0;//See the note below
//then erase the element

m_builderMap.erase(itor);
}
By now, the pattern should feel routine. First we write code to make sure there are no errors, then we
write the actual function code. In this function, we first check to make sure the user is removing a
previously added builder. After we make sure the user didn't make a mistake, we then delete the
builder and erase the iterator from the container.
Since we are immediately erasing the iterator after deleting the builder, it is
unnecessary to set the pointer to 0. However, I always set the pointer to 0. This helps
find bugs. For example, if I forgot to erase the iterator and tried to use this builder
again, the program would crash, resulting from using the null pointer. If I didn't set
the pointer to 0 but still tried to use it, I would instead get undefined behavior.
void M5StageFactory::ClearBuilders(void)
{
MapItor itor = m_builderMap.begin();
MapItor end = m_builderMap.end();
//Make sure to delete all builder pointers first

while (itor != end)
{
delete itor->second;
itor->second = 0;
++itor;
}
//Then clear the hash table

m_builderMap.clear();
}
Just as with the RemoveAllComponents from the M5Object, the purpose of ClearBuilders is to help the destructor
of the class. Since this code needs to be written anyway (it would go in the destructor), we think it is
better to factor it into a separate function that the user can call if they need:
M5StageFactory::~M5StageFactory(void)
{
ClearBuilders();
}
Finally, we have our factory destructor. This just ensures that we don't have any memory leaks by
calling the ClearBuilders function.
Using the Dynamic Factory
Now that we have a completed Factory class, let's look at how we would use it. Since the goal of this
class was to decouple our M5StageManager from our specific derived M5Stage classes, it makes sense that it
will be used in the M5StageManager class:
{
public:
//Lots of other stuff here...
static void AddStage(M5StageTypes type, M5StageBuilder*
builder);
static void RemoveStage(M5StageTypes type);
static void ClearStages(void);
private:
//Lots of other stuff here
static M5StageFactory s_stageFactory;
};
Since the factory will be private in the M5StageManager, we will add interface functions so the user can
control the factory without knowing the implementation. This allows us to change the details, without
affecting the user.
Inside the M5StageManager::Update function, we will use the factory to get access to the current stage.
Notice that this class is completely decoupled from any specific M5Stage derived classes. This give the
user freedom to change the game design, including stage types, stage count, and stage names, without
needing to modify the M5StageManager class.
In fact, that is the purpose of creating the Mach5 Engine the way we are. It can be used and reused in
many game projects without changing the engine code. Here is a simplified version
(pausing/restarting code has been omitted) of the M5StageManager::Update showing code relevant to the
stages and factory:
void M5StageManager::Update(void)
{
float frameTime = 0.0f;
/*Get the Current stage*/
M5Stage* pCurrentStage = s_stageFactory.Build(s_currStage);
/*Call the initialize function*/

pStage->Init();
/*Keep going until the stage has changed or we are quitting. */

while ((s_currStage == s_nextStage) &&
!s_isQuitting &&
!s_isRestarting)
{
/*Our main game loop*/
s_timer.StartFrame();/*Save the start time of the frame*/
M5Input::Reset(frameTime);
M5App::ProcessMessages();
pStage->Update(frameTime);
M5ObjectManager::Update(frameTime);
M5Gfx::Update();
frameTime = s_timer.EndFrame();/*Get the total frame time*/
}
/*Make sure to Shut down the stage*/
pStage->Shutdown();
ChangeStage();
}
As you can see, the M5StageManager is completely decoupled from any derived M5Stage classes. This
allows the user to change, add, or remove any stages during development without needing to modify
the M5StageManager class. This also allows the M5StageManager and M5StageFactory classes to be reused in
another game, shortening the development time of that project.
Now that we have seen the dynamic factory and how to use it, an important question should come to
your mind: what are the benefits of the Dynamic Factory? Both the Static and Dynamic Factories
allow us to decouple our code. Since they both offer that benefit and the Static Factory is much easier
to implement, why should we bother spending time on the Dynamic Factory? Asking questions like
these is always a good idea. In this case, I think there are two benefits to using the Dynamic Factory
over the Static Factory.
The first benefit of the Dynamic Factory is that it is dynamic, meaning that we can load builders from
a file at runtime or remove a stage if we will never use it again (SplashStage). Being dynamic allows us
to swap out a builder at runtime. For example, depending on the difficulty the player selects, we can
swap out a difficulty component on enemies. The code to these difficulty component builders could be
put into the menu, and the rest of our game no longer needs to care about the difficulty, the levels just
create enemies the same way, no matter what.
The second and more important benefit of creating the Dynamic Factory comes in the next step. Since
we have successfully created a StageFactory, we should do the same thing for components and game
objects. In the next section, we will take a look at creating these factories.
Creating a component and Object Factory
Now that we have built a stage factory, building a component factory should be easy. Let's take a look
at what a component and object factory would look like:
//Component Factory
class M5ComponentFactory
{
public:
~M5ComponentFactory(void);
void AddBuilder(M5ComponentTypes type,
M5ComponentBuilder* builder);
void RemoveBuilder(M5ComponentTypes type);
M5Component* Build(M5ComponentTypes type);
private:
typedef std::unordered_map<M5ComponentTypes,
M5ComponentBuilder*> BuilderMap;
};
//Object Factory
class M5ObjectFactory
{
public:
~ M5ObjectFactory (void);
void AddBuilder(M5ArcheTypes type,
M5ObjectBuilder* builder);
void RemoveBuilder(M5ArcheTypes type);
M5Object* Build(M5ArcheTypes type);
private:
typedef std::unordered_map< M5ArcheTypes,
M5ObjectBuilder *> BuilderMap;
};
Looking at those classes, you will notice that they are almost identical to the M5StageFactory class. The
only things that are different are the types involved. Instead of M5StageTypes, we use M5ComponentTypes or
M5ArcheTypes. Instead of M5StageBuilder, we use M5ComponentBuilder or M5ObjectBuilder. Finally, instead of the
Build method returning an M5Stage*, we return either an M5Component* or M5Object*.
If we were to implement these classes, the code would be identical as well. You might think that the
M5ObjectFactory would be a little different since the M5Object isn't part of an inheritance hierarchy, but it
actually doesn't matter. Even though the derived class builders are all doing different work, they are
always just returning a single pointer type. The builders might be different, but the return type isn't.
The Templated Factory
Since we need to create different versions of the same algorithms using different type, we should
again make use of C++ templates. This will allow us to write the code one time, and reuse it for any
factory type we need.
First we need to factor out the types that are different. If you look at all three classes, you will see that
three types are different. The enumeration type, the builder type, and the return type for the Build
method is different in all three classes. If we make those template parameters, we can reuse the same
code instead of recreating the same class three times. Here is how we should refactor our code:
//M5Factory.h
template<typename EnumType,
typename BuilderType,
typename ReturnType>
class M5Factory
{
public:
~M5Factory(void);
void AddBuilder(EnumType type, BuilderType* pBuilder);
void RemoveBuilder(EnumType type);
ReturnType* Build(EnumType type);
private:
typedef std::unordered_map<EnumType, BuilderType*> BuilderMap;
typedef typename BuilderMap::iterator MapItor;
};
Notice that our class is now a template class with three template parameters, EnumType, BuilderType, and
ReturnType. Instead of any specific types such as M5StageTypes, we have used our template parameters.
One change that confuses many people is this line:
typedef typename BuilderMap::iterator MapItor;
In the original, non-templated M5StageFactory class, the compiler was able to look at the code
BuilderMap::iterator and know for certain that iterator was a type inside BuilderMap. Now that we have a
template class, the compiler can't be sure whether BuilderMap::iterator is a variable or a type, so we
need to help the compiler by using the typename keyword to say that this is a type.
Since our factory is now a templated class, we should again put all functions implementations into the
header file. In addition, each implementation must be marked as a template function. Here is an
example of one of the Build methods:
//M5Factory.h
template<typename EnumType,
typename BuilderType,
typename ReturnType>
ReturnType* M5Factory<EnumType,
BuilderType,
ReturnType>::Build(EnumType type)
{
MapItor itor = m_builderMap.find(type);
M5DEBUG_ASSERT(itor != m_builderMap.end(),
"Trying to use a Builder that doesn't exist");
return itor->second->Build();
}
Except for the change in the function signature, the Build function is exactly the same. This is true of
AddBuilder, RemoveBuilder, and all functions of the class. As I said, by making the Dynamic Factory a
template class, we can reuse the same code for our stage factory, our component factory, and our
object factory. Since that is the case, we won't spend time on making the template factory. However,
we still need to see how to use this new class. Let's look at our M5StageFactory class:
{
public:
//Same as before
private:
//static M5StageFactory s_stageFactory; //Our Old Code
static M5Factory<M5StageTypes,
M5StageBuilder,
M5Stage> s_stageFactory;//Our new code
};
That is the only change we need to make to our M5StageFactory. Everything else will work the exact
same way. The nice thing is that once we have our template factory completed, using a component
factory is easy. Here is how we can use our component factory and object factory inside our
M5ObjectManager class:
class M5ObjectManager
{
public:
//See M5ObjectManager.h for details
private:
static M5Factory<M5ComponentTypes,
M5ComponentBuilder,
M5Component> s_componentFactory;
static M5Factory<M5ArcheTypes,
M5ObjectBuilder,
M5Object> s_ObjectFactory;
};
Reusing the code is simple once we have created the template version. We should have created that
first, but most programmers have a hard time thinking about how to reuse a class until after the initial
code has been written. I think it is easier and more natural to create the stage factory first, then
refactor the code into a template class.
There is one more important thing to consider when using our factory: how to add our builders. For
now, let's only consider the M5Factory inside the M5StageManager since we have seen that code already.
Somewhere in our code, we must still instantiate our derived builders so we can add them to the
M5StageManager. For example, we would need a function like this:
#include "M5StageManager.h"
#include "M5StageTypes.h"
#include "GamePlayStage.h" //Example Stage
#include "SplashStage.h" //Example Stage
void RegisterStages(void)
{
M5StageManager::AddStage(ST_GamePlayStage,
new M5StageTBuilder< GamePlayStage >() );
M5StageManager::AddStage(ST_SplashStage,
new M5StageTBuilder< SplashStage >() );
}
As you can see, this function is dependent on all stages in our game and is likely to change as we
change our design. Unfortunately, this is as far as we can decouple our code. At some point, we need
to instantiate derived classes. Even though this is necessary, later we will look at how to minimize the
work involved in maintaining this code. In the case of the Mach5 Engine, this code is auto generated
with a Windows batch file before the code is compiled. By auto generating our file, we reduce the
chance of forgetting to add a stage, and also minimize the work when our code changes.
Architecting versus over-architecting
Over-architecting is the concept of spending time planning as well as writing code that includes
completely unneeded and ultimately unused features. Since every project has a deadline, over-
architecting means wasting time that could be better spent writing code that will be used.
In our effort to learn design patterns, we want to know not only how to use them, but also about when
not to use them. When you are working on a project, you must always find the balance between
writing flexible code and getting the project finished on time. It always takes more time to write
flexible, reusable code, so you have to consider whether it is worth the extra time to write that code.
It would be great to spend time creating the ultimate graphics engine, or creating a content creation
tool that can rival Unreal or Unity. However, if you strive to write perfect, flexible, 100% reusable
code, you may never complete your game. You may end up writing a great particle system and have
your designers only use 10% of the capabilities. This is why many companies choose to use a
premade engine in the first place. Those companies don't want to spend time or money on creating a
tool. They want to spend time making a game that is fun.
The opposite of this situation is just as bad. We don't want to write code that breaks whenever a
change is introduced, or is impossible to use again. We can all imagine how ugly the code would be if
an entire game was written in the standard main function. We might laugh at the thought of someone
doing that, while at the same time hardcoding behavior with large if/else chains instead of using
Finite State Machines.
Finding the balance between these two extremes is difficult. I already mentioned that besides writing
the initial code, there are additional factors to consider. These include the time it takes to test and
debug the code as well as time to modify code if and when change occurs.
Determining whether writing flexible code is worth the time also includes determining how likely that
code is to change. This is why we are using Singleton classes for our core engines. These are unlikely
to change during the project. Of course, if we need to support multiple graphics APIs, multiple
platforms, or even a multithreaded environment, we might make a different decision. This is also why
using the Component Object Model and Finite State Machines are very useful, since our game objects
and their behavior are likely to change constantly.
In this case, we need to choose between using the Static Factory or the Dynamic Factory. The Static
Factory is very simple to write and use. Since it is so simple, testing and debugging should be easy. It
is likely to change, but those changes should be easy as well. However, when using the Static Factory,
we must write, test, debug, and maintain code for at least three separate types in our game: the stages,
components, and objects. These will change often during the development cycle. Each time a change
occurs, you would need to go back and modify these functions.
The templated Dynamic Factory is a little more difficult to implement, especially if you aren't very
familiar with using templates. However, the major benefit of using the templated Dynamic Factory is
that we only need to write the code once, and we can use it for stages, components, and objects. In
addition, we have the ability to add, remove, or change items in our factories at runtime. As I
mentioned, this could mean changing archetype builders based on difficulty to create harder versions
of the same archetype without needing new enumeration values. Finally, we have the option of using
this code again in another project, which is unlikely if we stick with the Static Factory.
In the end, the M5Factory.h file that contains the templated Dynamic factory is only around 125 lines of
code, with maybe 30% of that being comments and white space. It may be a little more difficult, but I
don't think it is so difficult that someone should prefer the Static Factory.
Summary
In this chapter, we focused a lot on decoupling code. Since our game design is very likely to change,
we want to make sure that our high-level modules don't depend on derived stages or components.
That is why we should follow the Dependency Inversion Principle, which says the following:
High-level modules should not depend on low-level modules. Both should depend on
abstractions.
Abstractions should not depend on details. Details should depend on abstractions.
In simpler terms, this means that all our code should be built around the interfaces. We used the
example of our M5StageManager not being dependent on derived M5Stage classes. Since we want to avoid
class dependencies like this, we learned that we should also avoid hard coding, including using calls
to the new operator. In order to avoid direct calls to the new operator, we learned about three ways to
make Factories.
The first method was the classic Gang of Four Factory method, which says we should create a
hierarchy of classes, each capable of instantiating a single class. This method helped us get to our
final solution, but wasn't quite good enough, because we want to be able to select which derived
class to instantiate via a string or an enum.
The second method we learned was the Static Factory Method, which uses a simple global or static
function and a switch statement to allow us to select our desired derived class. This works great for
our needs, but we decided to go one step further and create an even more flexible, reusable factory.
Finally, we learned about the Dynamic Factory, specifically the templated version of the Dynamic
Factory, which combines both the classic Factory method and the Static Factory. The best part is that
since we are using the power of C++ templates, we can reuse the code for stages and components, as
well as objects.
Throughout this chapter, but especially in the last section, we discussed the balance between spending
time writing flexible code versus over-architecting. Of course, one strong reason to study design
patterns is to learn how to write great reusable code, but we always want to make sure it fits the
project and fits within the project deadline.
The goal of this book is to help you understand when and where we should apply these patterns
throughout a game. We already know how to create flexible game objects using components and Finite
State Machines. Now that we know about the Dynamic factory, we have decoupled our stage and
component creation from our core engine, making everything more reusable.
However, the ultimate goal is to make things so flexible that they can be changed from a text file or
tool without the need to recompile anything. That is what we will learn how to do next.
Creating Objects with the Prototype Pattern
We saw in the last chapter how using a Dynamic Factory can help us decouple our high-level
modules, such as the M5StageManager or M5ObjectManager from the implementation details of our derived
M5Stage or M5Component classes. We did this by pushing those dependencies into derived builder classes
that would be used by a Dynamic Factory. This allowed us the freedom to change our derived stage
and component classes without needing to modify our higher level modules. C++ template classes
made using the Dynamic Factory very easy, since we were not required to create a derived class
builder for every stage and component.
However, we are required to create a builder for each M5Object type we want, since they will contain a
set of components that are unique to each object. Unfortunately, these builders may require frequent
changes as we playtest, balance, and modify our game design. Each time these builders change, the
game will be need to be recompiled.
The goal would be to have our game object types be completely defined in a file. This would give the
game designer the ability to test, tweak, and balance the game without needing to touch C++ code or
ever recompile. Of course, all of this could be done in a level editor tool, which could also be given
to players, allowing them to create additional game content.
In this chapter, we will explore the Prototype pattern and how it can help us define objects
completely within a text file. We will do this by first looking at a simple example of the pattern, then
looking at how the Mach5 Engine, and the M5ObjectManager specifically, makes use of this pattern to read
game object definitions from a file. Along the way, we will look at some language features of C++
that will help us write better, safer code.
Your objectives
Here is an outline of the topics we will cover and your tasks for this chapter:
Learn the trouble with using a factory for game objects
Implement the Prototype pattern
Learn how the Mach5 engine uses the Prototype pattern
Implement components within the Mach5 Engine
Learn how to define objects completely in a file
The trouble with using a factory for game
objects
In Chapter 5, Decoupling Code via the Factory Method Pattern we learned how to use a Dynamic
Factory to decouple our stages, components, and objects from higher level modules. We did this by
placing the dependencies of each derived class into a separate builder class instead of a high-level
module. Let's look at an example of creating a derived type stage builder:
//SplashStageBuilder.h--------------------
class SplashStageBuilder: public M5StageBuilder

{
public:
};
//SplashStageBuilder.cpp--------------------
#include "SplashStageBuilder.h"
#include "SplashStage.h"
M5Stage* SplashStageBuilder::Build(void)
{
return new SplashStage();
}
The reason we did this is so that changes to the SplashStage class only affect this file as opposed to the
M5StageManager, for example. This means any changes to derived stage or stage builder classes cannot
break other code, because the other code will only be using an M5Stage pointer. Changes to this class
could still break other code, particularly if this stage was removed from the game altogether.
However, by minimizing dependencies, we reduce the chances that other code will need to be
changed in the future.
In the Mach5 Engine, the M5Stage derived classes only need a default constructor. This is by design to
keep the builder classes as simple as possible. Each stage class will read its own required data from
a file. The logic for which file to read is written into the constructor. The same simple default
constructor design is used in the M5Component derived classes as well. What this means is that instead of
needing to create a builder class for each stage or component, we can instead use the power of C++
templates:
//M5StageBuilder.h
class M5StageBuilder
{
public:
virtual ~M5StageBuilder(void) {} //empty virtual destructor
virtual M5Stage* Build(void) = 0;
};

class M5StageTBuilder: public M5StageBuilder
{
public:
};
M5Stage* M5StageTBuilder<T>::Build(void)
{
return new T();
}
By using C++ templates, we can reduce the number of small classes that need to be manually created
while still gaining the decoupling benefits of the factory. Of course, it is always possible that a few
stages or components will need more complex constructors or builder classes. In that case, we can
easily create the required classes when the need arises. Unfortunately, we don't get the option of using
templates with our object types.
Using builders with objects
Our game objects are mostly just collections of components. Each object type will have different
components and component data based on the decisions of the designer. As the development
progresses, those collections are likely to change. Remember, even though each individual component
has a builder in the factory, the object will need to somehow instantiate the separate components.
Let's look at shortened examples of using builders for the Player and Raider objects:
//PlayerBuilder.h--------------------
#include "M5ObjectBuilder.h"
class PlayerBuilder: public M5ObjectBuilder

{
public:
virtual M5Object* Build(void);
};
//PlayerBuilder.cpp--------------------
#include "PlayerBuilder.h"
#include "M5Object.h"
#include "M5ObjectManager.h"
M5Object* PlayerBuilder::Build(void)
{
M5Object* pObj = new M5Object;
//Build and set Gfx component for player
GfxComponent* pGfx =
M5ObjectManager::CreateComponent(CT_GfxComponent);
pGfx->SetTexture("playerShip.tga");
pGfx->SetDrawSpace(DS_WORLD);
//Build and set input component for player

PlayerInputComponent* pPI =
M5ObjectManager::CreateComponent(CT_PlayerInputComponent);
pPI->SetSpeed(100);
pPI->SetRotationSpeed(10);
pObj->AddComponent(pPI);
//...add more components here
return pObj;
}
//RaiderBuilder.h--------------------
#include "M5ObjectBuilder.h"
class RaiderBuilder: public M5ObjectBuilder

{
public:
virtual M5Object* Build(void);
};
// RaiderBuilder.cpp--------------------
#include "RaiderBuilder.h"
#include "M5Object.h"
#include "M5ObjectManager.h"
M5Object* RaiderBuilder::Build(void)
{
M5Object* pObj = new M5Object;
//Build and set Gfx component for Raider
GfxComponent* pGfx =
M5ObjectManager::CreateComponent(CT_GfxComponent);
pGfx->SetTexture("enemyBlack3.tga");
pGfx->SetDrawSpace(DS_WORLD);
//Build and set behavior for Raider
ChasePlayerComponent* pCP =
M5ObjectManager::CreateComponent(CT_ChasePlayerComponent);
pPI->SetSpeed(40);
pObj->AddComponent(pCP);
return pObj;
}
In these two simple examples, we can see why each object builder needs to be different. Each
specific object type will have a unique set of components. These examples only use two components
each, but we haven't considered our physics colliders, any weapon components, or additional
behaviors. Even in these short examples, both object types use a GfxComponent but the differences in
data, such as texture, mean that we need different code. Since there is only one object class, as
opposed to a hierarchy of derived classes, there is no way to let the object manage the creation of the
necessary components.
To deal with this, we will either need a builder class for each object type or one object builder with a
switch statement and a case for each object type. The problem with both solutions is that the list of
components and the data for each component is likely to change often. Now, instead of being worried
about high-level modules and dependencies, we have two new problems to be worried about.
The first problem is constant, and possibly long, compile times. We already know that as
development continues, the game design will change. This could mean not only changing components,
but also changing values inside those components. At some point, especially toward the end of
development, the game may be complete, but not quite balanced. During this phase, game objects and
the code responsible for them will be constantly tweaked. Changes in health, damage, speed, and
other attributes could be changed frequently causing the code to be recompiled.
There are a lot of factors that can affect how long it takes to compile a project. In the best-case
scenario, we would only change one file, and recompiling could be very fast. However, even a short
build time such as 10 seconds can get annoying if the only thing that changed is a speed value from
10 to 11. Something as simple as balancing a single unit can take all day, and then still need
additional tweaks as other units get balanced. Our goal here is to make modifying the object and
seeing the results of the modification as fast as possible.
The second problem that comes up is related to who is responsible for making these balance changes.
One scenario is that the programmer is responsible because all balance changes are code changes.
This is a bad situation because, as we already mentioned, balancing can take a long time, and now
both the designer and the programmer are involved. The designer may not be able to properly explain
the desired result to the programmer, so they must sit together tweaking and compiling, repeatedly. It
would be better if the designer was free to balance, and the programmer was free to fix bugs or
optimize code where needed.
The other scenario is that the designer is responsible for making these balance changes to the code.
The problem here is that the designer may not be as familiar with the engine, or with the programming
language. It is very possible that the designer doesn't have programming experience at all, so they
would not be familiar with IDEs or version control systems. The chance of introducing bugs or
breaking the project could be very high.
Solution - reading from files
Our solution to both problems is to move all hard-coded values into text or binary files that will be
read by the engine at runtime. Since we are using a component-based system in our game, this means
being able to define which components belong to an object in the file as well. The benefit of this
approach is that these files could be created with a level editor or some other content creation tool by
a designer. For now, we won't focus on how these files get created, but instead on how they are read
into our engine and how we can use them in place of an object builder class.
Now that we have decided to define our objects in files, we need to think about when they will be
read by our engine. There are a few different solutions to our problem. The first solution is to simply
have our builder read a file every time we need a new Raider, Bullet, or Asteroid. This is a common
first thought, but is a very bad idea. Reading files from the hard disk is very slow compared to
accessing data already in RAM. Depending on the factors, it can be anywhere from 10,000 to 50,000
times slower or more. This is like a normal 5 minute drive to the store taking 173 days, 14 hours, and
40 minutes. If you started your drive on January 1st, you would arrive at the store on June 21st.
This doesn't mean that we should never read files from the hard disk. It just means that we need to be
more strategic about it. Using this solution would mean we could be reading the exact same file many
times during the same frame. This would be like taking that half-year trip to the store to buy eggs,
going home, then immediately going back to the store to buy milk. Instead of reading the same file
many times we should read it once and store the data into RAM. Since it is so slow, we should avoid
reading and writing files during gameplay and instead load as much from our files during load time as
we can.
In large scale games, such as Rockstar's Grand Theft Auto 3, there is too much data to keep in RAM
all at once. A game like this will be constantly reading data from files that aren't currently in memory,
and releasing resources that are no longer in use. This is done by reading files on separate threads so
the main gameplay isn't paused or slowed down. There is a lot of work involved in this file streaming
process, including making sure that key gameplay elements are in memory when the player needs them
so gameplay isn't affected. However, even games that stream data like this don't waste time reading
files repeatedly once they are in memory.
We won't be worrying about threads for now. So, a more common solution for us would be to read the
file once during a load screen to create the object types we need, then simply to copy these objects
whenever we need a new instance of them. This will allow us to focus on the pattern, as well as
solve key issues related to this problem, without worrying about the difficulties involved in
multithreaded architecture.
Even without using threads, we still have a problem. How do we copy the object that contains
components without knowing the types? Typically, when creating new objects, we need the new
keyword, which means we also need a constructor to call. Of course, we have an enum for the type,
which means we could use a switch statement and call the correct copy constructor. However, we
already know that switch statements can be difficult to maintain and should be avoided.
The component factory will create a new component of the correct type, but we would still need to
write a copy function that can copy the data once it has been created. It would be nice if we had a
function that would construct and copy the data, based on the type of object it is. We need a
constructor that works like a virtual function.
The Prototype pattern explained
The Prototype pattern gives us a way to copy a class without knowing the actual type of that class.
This is often referred to as a virtual constructor because we can use it to create and copy a derived
class while only using a base class pointer or reference. This pattern is most powerful when used
with object hierarchies, but it doesn't only need to be used with virtual functions. Another purpose of
the Prototype pattern is simply to create a prototypical (or archetypical) instance of an object and to
use that to copy from.
Imagine that we are creating a level editor. In the middle of the tool we would have the map of the
level. Here we can place tiles, power-ups, enemies, and players. Along the side of the map, we
would have all the objects and tiles used in our game that can be placed on the map. This can be seen
in the following screenshot:
Figure 6 1 - Example of a Simple Level Editor
Since we are striving for clean code, we have a clear separation between the part of the code that
handles drawing, the part of the code that handles clicking and manipulating the objects on the side,
and the part of the code that defines our specific object type. However, when we click and drag an
object from the side, we would make a new instance of the object that was clicked and would draw it
at the mouse location, and eventually on the map where the user places it. Based on the Dependency
Inversion Principle, we know we don't want our high-level modules depending on our low-level
objects. Instead, they should both depend on abstractions.
Prototype versus Archetype

Before we learn more about this pattern we should talk a little about word choice. In
the Mach5 Engine, the files we will be reading and creating, as well as the
enumerations we will be using, are called archetypes instead of prototypes. An
Archetype is a perfect, unchanging, ideal example of a thing. A prototype is typically
an early, usually unrefined version of something that later versions can depart from.
While both words could be correct, the authors use the word Archetype to refer to the
object definitions in the files, even though the pattern is called the Prototype pattern.
The virtual constructor
The Prototype pattern is simple. It involves giving an object for which you might want to copy a Clone
method, and letting the object know about the details of how to perform the clone. The actual name of
the method isn't important. Clone is just one common example. This is most powerful when used with
object hierarchies in which you want to copy an object, but don't know the type of derived object you
might be holding. Simply add a Clone method to the interface of a hierarchy and have the derived
classes each implement the method. Let's look at a simple example to start off:
In the following code, we have our interface object, Shape. We will just be talking about simple shapes
in this example. Since this class will be used as an interface, we must mark the destructor as virtual
so the correct destructor will get called when derived classes are deleted. Next, we have two pure
virtual methods. The Draw method could be any action that we need virtual behavior with. In this
simple case, we will just be using printed statements instead of drawing shapes on the screen. The
Clone method will be our virtual constructor. This method will know how to copy itself and return a
new instance:
class Shape
{
public:
virtual ~Shape(void) {}//empty base class constructor
virtual void Draw(void) const = 0;
virtual Shape* Clone(void) const = 0;
};
Now, let's look at the derived class examples:

class Circle : public Shape
{
public:
virtual void Draw(void) const
{
std::cout << "I'm a Circle" << std::endl;
}
virtual Shape* Clone(void) const
{
return new Circle(*this);
}
};
class Square : public Shape

{
public:
{
std::cout << "I'm a Square" << std::endl;
}
{
return new Square(*this);
}
};
class Triangle : public Shape

{
public:
{
std::cout << "I'm a Triangle" << std::endl;
}
{
return new Triangle(*this);
}
};
Of course, our derived classes know how to draw themselves. To keep things simple, the Draw methods
are just printing to the console. The important part here is that each Draw method has a different
behavior; in this case, a hardcoded string to print. The Clone methods are where the real magic
happens--each one returns a new instance of itself. Specifically, they are calling their own copy
constructors. This will allow the client to hold a pointer to any Shape, and get a copy of the correct
derived type without knowing or caring which constructor to call. Let's look at an example of this
using code:
int main(void)
{
//seed the RNG
std::srand(static_cast<unsigned>(time(0)));
//Create my shapes
const int MAX = 3;
Shape* shapes[MAX] = { new Circle(),
new Square(),
new Triangle() };
for (int i = 0; i < MAX * 2; ++i)

{
Shape* pCopy = shapes[std::rand() % MAX]->Clone();
copy->Draw();
delete pCopy;
}
//make sure to delete my original shapes

for (int i = 0; i < MAX; ++i)
delete shapes[i];
return 0;
}
The first few lines are simply initializing the random number generator and the array of shapes. In the
array, you can see that we create a new instance of a Circle, Square, and Triangle. These will be our
prototype Shapes to clone.
The next section is a loop to show the Clone method at work. We use a random index into the array and
clone an object. Since it is random, we can't know which Shape will be cloned. This is simulating a
random click by the user. Once we clone the Shape, we are free to call Draw or any other interface
method we need. At the end of the loop, we delete the cloned Shape but, of course, that doesn't delete
our prototype Shape from the array because it was a copy of the object. After the loop, we go through
our prototype array and delete each of those shapes as well.
The following is the output for the preceding code:

I'm a Triangle
I'm a Square
I'm a Square
I'm a Circle
I'm a Circle
I'm a Square
The problem with constructors
Now that we have seen a little about virtual constructors and the Prototype pattern, let's look at
exactly what problem we are trying to solve.
To understand the problem of construction, we first need to understand the difference between a class
and an object of that class. Classes are what the programmer creates. They are the code template or
recipe that the program uses to create objects. In C++, we cannot create classes at runtime. We don't
have a way to introduce new code while the program is running.
This is because C++ is a statically typed language. This means that the language tries to prevent
operations if they can't be performed on that type at compile time. For example, we can't divide a
float by a void* because the C++ compiler checks a compile time if the operation makes sense, and
will issue an error if it doesn't.
This static typing in C++ is why we are forced to declare a type for every variable. It is also why we
are required to specify the constructor in a case like this:
Base* p = new Derived;
In this case, the compiler must know the class that we are trying to create. Unfortunately, in C++,
classes do not have first class status. That means we can't pass a class as an argument to a function or
use a class as a return value. We can't copy a class, save it in a variable, or create one at runtime.
Some languages do have these features. Here is an example of what you could do if classes had first
class status in C++:
//This isn't real C++ code
Shape* CreateShape(class theShape)
{
Shape* pShape = new theShape;
return pShape;
}
While this might be useful, we would be trading some type safety for flexibility. The static type
checking that the C++ compiler performs has the chance to catch problems before they become bugs
at runtime. This is a good thing. We should enjoy the type checking C++ provides. We should also
recognize when it has an impact on our code flexibility.
Even though C++ is a statically typed language, we have ways of getting around this problem. One
such way was with the factories that we created in the last chapter. We were forced to write the
factories ourselves, but we still can choose which class gets created at runtime, while getting the
benefits of static typing for all other classes and types. Factories are just one way we can avoid the
rigidity of classes not having first class status. The virtual constructor is another way.
The benefits of a virtual constructor
Using a virtual constructor can be very powerful. The biggest benefit is that we can now treat classes
as if they had first class status. We can make copies of an object without knowing the exact type of the
object. We can use the Prototype pattern in our level editor example, but we could also use it in our
game. Anytime we need to make a copy without truly knowing the type, we can use this pattern. As
we said before, C++ is a statically typed language, meaning the compiler will make sure we are using
the correct types at compile time. This static type checking helps us write safer code.
By using Dynamic Factories and virtual constructors, we are bypassing this type checking a little. The
compiler is still doing the type checking on the pointers we are working with, but we are choosing
our derived class at runtime. This has the chance to lead to bugs that are difficult to find if we
somehow mix up the types we are cloning. That doesn't mean we shouldn't use the patterns; it is just
good to understand that we are giving up a little safety for flexibility.
We don't need to know the type
As we said, the biggest benefit of using the Prototype pattern is that we can make copies without
knowing the type. This means we can make a copy of a function parameter or function return type
without caring about the derived class involved. This also means that we can share a pointer with
another class or method and we don't need to care if the type is modified.
In the following example, we have a SpawnShape class with the ability to spawn a specific type of shape.
By using the Prototype pattern, the class doesn't need to care what type it is spawning. The
constructor takes a pointer to some shape and then it just needs to call the Clone method. If the base
class pointer is pointing to a Circle, then a circle will be created. However, if we have a pointer to a
Triangle, then a triangle will be created. Here is an example that shows the Prototype pattern in action:
class ShapeSpawner
{
public:
ShapeSpawner (Shape* pShape, float maxTime):
m_pToSpawn(pShape),
m_spawnTime(0.f),
m_maxSpawnTime(maxTime)
{
}
void Update(float dt)
{
m_spawnTime += dt;
if(m_spawnTime > m_maxSpawnTime)
{
//The class doesn't care what type it is cloning
Shape* pClone = m_pToSpawn->Clone();
//...Register the clone somehow
//Reset timer
m_spawnTime = 0;
}
}
private:
Shape* m_pToSpawn;
float m_spawnTime;
float m_maxSpawnTime;
};
Our SpawnShape class doesn't care if it is spawning a Circle, Square, or Triangle, or any new shape we
might create later. It can make copies without knowing the real type of the shape. If we add a public
SetShape method, we could even change the type that spawns at runtime. Compare this with a more
rigid example that can only spawn Circles:
class CircleSpawner
{
public:
CircleSpawner (Circle* pCircle, float maxTime):
m_pToSpawn(pCircle),
m_spawnTime(0.f),
m_maxSpawnTime(maxTime)
{
}
void Update(float dt)
{
m_spawnTime += dt;
if(m_spawnTime > m_maxSpawnTime)
{
//Use copy constructor
Circle* pClone = new Circle(*m_pToSpawn);
//...Register the clone somehow
//Reset timer
m_spawnTime = 0;
}
}
private:
Circle* m_pToSpawn;
float m_spawnTime;
float m_maxSpawnTime;
};
In the second example (not using the Prototype pattern) we are forced to use the copy constructor of
the derived class, in this case the Circle. If we want to spawn a Square or a Triangle, we would need to
create a SquareSpawner or TriangleSpawner. That is a lot of repeated code. It could get worse as we add
even more shapes. By using the Prototype pattern, we can reduce the number of classes we need.
No need to subclass
The reduction in classes is another big benefit of using a virtual constructor. In the case above, we
only needed a single SpawnShape instead of duplicating our spawn class or making derived versions.
Consider our factory builder classes that we saw before. We were forced to create an abstract base
class and create derived classes for every new M5Component and M5Stage that we used. C++ templates
helped us autogenerate that code, but the code still exists.
By using a virtual constructor, we don't need derived builder classes for every M5Stage, M5Component,
Shape, or other inheritance hierarchy. We can let the objects copy themselves. Does that mean we
should remove our factories and always use the Prototype pattern? It depends.
Remember that with the Prototype pattern, we must first instantiate an object before it can be cloned.
This is fine with shapes or components because these types are very small. However, the M5Stage
derived classes may be very big and they also may cause side effects. This means that the constructor
of a stage may add objects to the M5ObjectManager or may load textures or another large resource.
Since using C++ templates made creating our builders so easy, we can keep using them for stages and
components. However, we want to avoid making builders for the M5Object because those builders are
very likely to change during development. By using virtual constructors and creating prototypes (or
archetypes) that can clone themselves, the rest of our code will not be affected by changes to our
types.
It's easy to make exact copies
Here, the concept of a prototype doesn't need to be used with virtual functions. We may have a group
of objects, for example, easy, hard, and insane enemies that we want to copy. These objects may be
the exact same class type, but they could have very different values for health, speed, and damage. In
this case, we have our prototypical example that we want to copy from. Of course, in this case we
could also just use a copy constructor, but it is possible, as in the case of the Mach5 Engine, that the
copy constructor doesn't exist.
Either way, since we don't need to create derived builder classes, we can add prototypes at runtime.
Take the above example of the easy, hard, and insane enemy types. We could have just a single file
that defines the health of an easy enemy with a value of 50, for example. Then at runtime we could
create a hard enemy prototype with a health of 100 and an insane enemy with a health of 200. We could
always just double the values of health and damage for each hard and insane enemy, or the file could
contain the scale factor for the hard and insane versions.
Another example where we might want to modify the data would be if we had an enemy base that
spawned enemies after a set time (like our ShapeSpawner from above). In this example, the base might
increase the health and damage of the object over time. So, the base might initially create enemies
with health of 50, but after each spawn, the health is increased by 5. So, the second enemy has a health
of 55. The third enemy has a health of 60. Since each base has a specific instance, each base would be
spawning enemies with different health values.
The object could also be modified via an in-game level editor. Imagine the benefit of play testing a
level only to realize that enemies are killed too easily. With the Prototype pattern and an in-game
level editor, we could pause the game, edit attributes of an object type, and continue playing. This
method wouldn't require programmers, or recompile times. It doesn't even require a game restart. Of
course, this same effect could also be done with a separate level editor or by just modifying the
Archetype files and reloading the files at runtime. In these situations, we can see that creating copies
of a specific instance is very easy and very useful.
Examples of the clone method in Mach5
So far, we have seen simple examples of how the Prototype pattern is implemented. If you thought that
the examples were easy, you are in luck--it doesn't get much harder than that. We also discussed a few
ways in which an object instance and a virtual constructor can be useful in a game. Now let's look at
how the Mach 5 Engine uses the Prototype pattern in the M5Component and M5Object classes. Since the
M5Object class uses the M5Component Clone method, let's look at the components first.
In Chapter 3, Improving on the Decorator Pattern with the Component Object Model, we examined
almost all the methods and member variables in the M5Component class. However, the method we didn't
talk about was the Clone method:
//! M5Component.h
class M5Component
{
public:
//! virtual constructor for M5Component, must override
virtual M5Component* Clone(void) const = 0;
//The rest of the class is the same as before

};
As you can see, the M5Component class implements a pure virtual Clone method just like we saw above in
the Shape class. Since the M5Component class is used only as an abstract base class, we don't want to
provide any default behaviors for cloning. Cloning only makes sense for derived classes. This part of
the component was shown again so we can understand what the interface for overloading this method
should be.
The Gfx and collider components
Now that we have seen the interface, let's look at two very important components. These are
important because they allow game objects to interact with two of the other core pieces of our engine,
graphics and physics.
The first one we will look at is the GfxComponent class. This class allows the game object to have a
visual representation in the game. It contains the two minimum pieces of information to draw an
object in the game:
//GfxComponent.h
enum DrawSpace
{
DS_WORLD,
DS_HUD
};
class GfxComponent : public M5Component

{
public:
GfxComponent(void);
~GfxComponent(void);
void Draw(void) const;
virtual GfxComponent* Clone(void) const;
void SetTextureID(int id);
void SetDrawSpace(DrawSpace drawSpace);
private:
int m_textureID; //!< Texture id loaded from graphics.
DrawSpace m_drawSpace; //!The space to draw in
};
The two pieces of information we need to draw an object are which texture to draw and which space
to draw in. Of course, we need a texture to draw, but the draw space may be a bit more confusing. It
is just an enum letting us know which type of graphics projection should be used with the object. For
now, it is enough to know that the HUD draw space is always on top and isn't affected by camera
movement or camera zoom. Of course, there could be more data, such as texture color and texture
coordinates. These could be added in a derived class if we wanted. Here we are just showing the
basics.
There are a few functions used to set these values, as well as a FromFile function that we talked about a
little before. The Update function doesn't do anything for this component because there is nothing to
update. The Draw function will be called by the graphics engine, making each M5Component responsible for
drawing itself. However, the most important function for this chapter is the Clone:
GfxComponent* GfxComponent::Clone(void) const
{
//Allocates new object and copies data
GfxComponent* pNew = new GfxComponent;
pNew->m_pObj = m_pObj;
pNew->m_textureID = m_textureID;
pNew->m_drawSpace = m_drawSpace;
if (m_drawSpace == DrawSpace::DS_WORLD)
M5Gfx::RegisterWorldComponent(pNew);
else
M5Gfx::RegisterHudComponent(pNew);
return pNew;
}
In this function, we simply create a new GfxComponent and copy the relevant data from this object to the
newly created one. What you don't see is that in the GfxComponent constructor, the component type is set
by calling the M5Component component constructor which, of course, also gives this component a unique
ID. The last thing we do is register this component with the graphics engine depending on the draw
space. This class automatically unregisters itself when it is destroyed:
GfxComponent::GfxComponent(void):
M5Component(CT_GfxComponent),
m_textureID(0),
m_drawSpace(DrawSpace::DS_WORLD)
{
}
GfxComponent::~GfxComponent(void)
{
M5Gfx::UnregisterComponent(this);
}
Now that we have seen the GfxComponent, let's look at the most basic of all physics colliders. The
Mach5 Engine ColliderComponent is as simple as possible for a 2D game. For now, it is only concerned
with circle versus circle collisions. It could easily be extended to test for rectangle collision as well:
//ColliderComponent.h
class ColliderComponent : public M5Component
{
public:
ColliderComponent(void);
~ColliderComponent(void);
virtual ColliderComponent* Clone(void) const;
void TestCollision(const ColliderComponent* pOther);
private:
float m_radius;
};
This class is a lot like the previous one because it is connected to one of the core pieces of the game
engine. Just like all components, FromFile must be overloaded to read component data from an .ini file.
Update must also be overloaded but, just like with the GfxComponent, this doesn't do anything in this simple
version. If the class used oriented bounding boxes, the Update function could be used to update the
corner points of the oriented box. The TestCollision function is also important. It is called by the
Physics Engine to test if this object is colliding with another object. If it is, the two objects are added
to a list of colliding pairs that can be resolved later. Again, the most important function for this
chapter is the Clone:
ColliderComponent* ColliderComponent::Clone(void) const
{
ColliderComponent* pNew = new ColliderComponent;
pNew->m_radius = m_radius;
M5Phy::RegisterCollider(pNew);
return pNew;
}
Just like the GfxComponent, this component first creates a new version of itself, then copies the important
information into the new component. Before returning the new component, it first registers itself with
the physics engine. Since it is registered, it must be unregistered when it is destroyed, so we do that in
the destructor:
ColliderComponent::ColliderComponent(void) :
M5Component(CT_ColliderComponent), m_radius(0)
{
}
ColliderComponent::~ColliderComponent(void)
{
M5Phy::UnregisterCollider(this);
}
There are a few things to point out with both classes. First, notice that we don't clone the m_type, the
m_id, or the isDead variables. This isn't necessary. The type is set by the constructor in the M5Component
base class when we call the constructor. The id is also set in the base class, but it is important to point
out that the purpose of the m_id is to be unique. It wouldn't serve the correct purpose if we also copied
the id. Instead, we are copying the rest of the important data but we recognize that this is a separate
component, not just an exact copy. For the same reason, we also don't copy the isDead variable. We are
creating a new component that is like the old one, but still a unique component. If we copied the
isDead, this component would be deleted in this frame or the next.
Next, both classes register themselves with engines in the clone method instead of the constructor.
This is because of how they are intended to be used. Our object manager will hold a collection of
these pre-created prototype components at the start of the game so they are ready to be cloned. We
don't want these initial components to pollute the graphics or physics engine.
However, we are assuming that an object is being cloned and it also needs to live in the game world,
so we register at that time. This seems like the most standard reason for cloning. It is better for the
user to only worry about cloning, than to worry about cloning, registering, and then unregistering. If
the user wishes to do something non-standard, they are free to unregister after cloning.
Cloning an object
We have seen a few examples of how M5Component derived classes use the Prototype pattern. We will
look at a few more important ones a little later, but for now let's look at how the M5Object class uses
these Clone methods, and how the M5Object itself is cloned. Recall that the M5Object also had a Clone
method. Even though this class isn't part of a hierarchy, it can still use the concept of the Prototype
pattern to create cloneable instances. Here is the Clone method of the M5Object:
//M5Object.h
class M5Object//Everything is the same as before
{
public:
M5Object* Clone(void) const;
};
//M5Object.cpp
M5Object* M5Object::Clone(void) const
{
//create new object
M5Object* pClone = new M5Object(m_type);
//copy the internal data
pClone->pos = pos;
pClone->vel = vel;
pClone->scale = scale;
pClone->rotation = rotation;
pClone->rotationVel = rotationVel;
//clone all components

{
M5Component* pComp = m_components[i]->Clone();
pClone->AddComponent(pComp);
}
return pClone;
}
The important thing when we clone is that we copy all the relevant data from the old object. This
includes not only things like the position and velocity but also all the components. So, we start off the
function by both creating a new instance of the object we will be creating and passing the correct type
to the constructor. This will set the m_type and m_id variables. Remember, even though we are cloning,
we want to make sure that each object has a unique ID. Next, we copy the data. Just as with
components, we don't need to copy the isDead value.
Finally, we have a loop to clone all components in the current object. This shows the power of the
Prototype pattern. We don't need to know the types of each component--all we need to do is loop, call
Clone to make our copy, then add that copy to the newly created object. Remember that the AddComponent
method will change the m_pObj in each component. This will ensure that all components point to their
correct owners.
In the end, the Prototype pattern is easy. Each component's Clone method is simple, and using them to
clone an object is simple. Even using these cloned objects in the M5ObjectManager is easy. We will look
at that in the next few pages, but first let's talk about a few details that some readers may have
noticed. The first is that we didn't use a copy constructor in any of the Mach5 Clone methods, even
though we did in the Shape example. The next is that the return types of the GfxComponent and
CollideComponent are different from the return type in the M5Component interface.
Choosing a copy constructor
As we said, and as you saw in the code examples above, we didn't use copy constructors in any of the
component Clone methods. We also didn't use them in the Clone method of M5Object. By default, classes
have copy constructors and assignment operators generated by the compiler. In the Shape example
above, we used the compiler generated copy constructor in our Clone methods.
However, in the Mach5 Engine there is an important choice to consider. What should the copy
constructor do about the m_id variable value? Remember that this ID is supposed to be unique for
every object and every component. However, if we use the compiler generated copy constructor, each
variable, including m_id, will be copied by value. This would mean that any time we use the copy
constructor, we would have two objects with the exact same ID.
Sometimes this is what we want, such as if we wanted to have a vector of objects instead of pointers
to objects, for example. When using standard vectors (and other containers), the copy constructor gets
called when adding to the container. If we were to add an object, we probably would want it to copy
the ID. It is also possible that we would want to shift the position of the object around the container.
Most likely, we would want it to keep the same ID.
However, this is not the behavior we want in our Clone methods. We want each clone to be a separate
entity with a different unique ID. Of course, we could write our own copy constructor and give a
different ID to each newly created object or component, just as we do in the default constructors.
Unfortunately, if we did that using standard containers we would be generating new IDs for every
time they internally called the copy constructor. In this case, the IDs wouldn't match up to the correct
objects or components.
In the Mach5 Engine we use containers of pointers instead of containers of objects or components, so
the authors decided to remove the copy constructors (and assignment operators) all together. This will
remove all confusion about the process. If you want a copy, you call the Clone method, because you
can't call the copy constructor. It is fine to make a different decision. In a different engine that uses
containers of objects instead of pointers, a different decision would probably be made.
In the Mach5 Engine, we remove these methods from the object by making them private so that they
can't be called. In C++ 11, you can mark them as deleted so the compiler won't generate them for you.
The assignment operators will already be removed because these classes contain const data that can't
be reassigned:
//In M5Object.h
M5Object(const M5Object& rhs) = delete;
//In M5Component.h
M5Component(const M5Component& rhs) = delete;
Covariant return types
Astute readers will also have noticed that the return types of each Clone method in the Mach5 Engine
are actually different. The base M5Component class returns an M5Component*, however the derived classes
return pointers to their own class types. This is a feature of C++ (as well as some other languages)
known as covariant return types. Let's look at this feature using the Shape class from above:
class Shape
{
public:
virtual ~Shape(void) {}//empty base class constructor
virtual void Draw(void) const = 0;
virtual Shape* Clone(void) const = 0;
};

{
public:
virtual void Draw(void) const;
virtual Shape* Clone(void) const;
};
int main(void)
{
Circle* pCircle = new Circle();
//The line won't compile

Circle* pClone = pCircle->Clone();
delete pClone;
delete pCircle;
return 0;
}
If the Clone method of the Circle class returns a Shape*, the compiler won't allow us to directly assign the
result to a Circle*. We would need to do either a static_cast or dynamic_cast, meaning we would have to
write code like this:
Shape* pClone = pCircle->Clone();
Circle* pCircle2 = dynamic_cast<Circle*>(pClone);

if (pCircle2)
{
//do something specific to circle
}
In both cases, the Clone method will return a circle. However, there is no way for the compiler to
know this, so we are forced to do the cast. The standard rule for using virtual functions is that the
function signature must be the same, including the return type. With covariant return types, the
compiler will allow us to replace the base class of a return type with a more specific type in the
inheritance hierarchy:
{
public:
virtual void Draw(void) const;
//Example of using a covariant return type
virtual Circle* Clone(void) const;
};
int main(void)
{
//No need to cast
Circle* pClone = pCircle->Clone();
//... Do something Circle specific with pClone
delete pClone;
delete pCircle;
return 0;
}
By using covariant return types, we can eliminate the need for unnecessary casting in cases where we
clone an object that needs to access properties of its true type. It is worth noting that this feature only
works with pointers or references. That means if the Clone method of Shape was returning a shape,
instead of Shape*, we wouldn't have the option of doing this.
Loading archetypes from a file
Now that we have seen the Prototype pattern in detail and discussed how it is used with the
components of the Mach5 Engine, let's look at how we can use it to load object data from a file. To do
that, we will need to first look at the object files, then look at specific methods in the M5ObjectManager
used to load and create these objects.
Archetype files
The first thing we need to do is look at how we define our object archetype within a file. The Mach5
Engine uses .ini files for archetypes, levels, and anything related to initialization of the engine. A
more standard file format would be XML or JSON if you wanted to keep them as human readable and
modifiable. If you didn't want them to be modified by users, the files could always be saved as
binary.
We have chosen .ini files because they are easy to read by both humans and a computer program.
They only have a few simple rules, so they are easy to explain in just a few sentences. They only
contain named sections which are defined by square brackets [ ], and key value pairs in the form of
key = value. The only exception is the global section, which doesn't have a name and therefore no
square brackets. Let's look at a basic example of an Archetype file. This is an example of Player.ini:
posX = 0
posY = 0
velX = 0
velY = 0
scaleX = 10
scaleY = 10
rot = 0
rotVel = 0
components = GfxComponent PlayerInputComponent ColliderComponent
[GfxComponent]
texture = playerShip.tga
drawSpace = world
[PlayerInputComponent]
forwardSpeed = 100
speedDamp = .99
bulletSpeed = 7000
rotationSpeed = 10
[ColliderComponent]
radius = 5
As you can see, the global section of the Player.ini file contains values for everything variable that is
defined in M5Object. Except for the components key, everything is read in the FromFile method of the
M5Object. In this case, most of our starting values are zero. This is because things like the starting
position for the player object will depend on the level, so this data will be modified after creation.
The more important part is the components. The components key contains a list of components that the
object will use. These strings will be used by the M5ObjectManager to create a component and then read
the specific component data defined in each section. This allows us to reuse components, such as the
ColliderComponent, because each object that uses them can have different component data. In this case,
the player object will have a radius of 5, but a bullet might have a radius of 1.
The object manager
The M5ObjectManager is a singleton class that is responsible for, among other things, loading archetypes
and creating objects. There are a lot of members and methods in this class so looking at everything
would take too long. In this section, we will only go over the methods specifically related to loading
and creating an object from an Archetype file. Remember that since the class is a singleton, we have
global access. For that reason, every member and method is static:
class M5ObjectManager
{
public:
static M5Object* CreateObject(M5ArcheTypes type);
static void AddArcheType(M5ArcheTypes type,
const char* fileName);
static void RemoveArcheType(M5ArcheTypes type);
//Plus other methods
private:
typedef M5Factory<M5ComponentTypes,
M5ComponentBuilder,
M5Component> ComponentFactory;
typedef std::unordered_map<M5ArcheTypes,
M5Object*> ArcheTypeMap
static ComponentFactory s_componentFactory;
static ArcheTypesMap s_archetypes;
//Plus other members

};
Here we have the most important members and methods to show how objects are loaded from a file.
What we haven't shown here are methods related to destroying or searching for specific objects. If
you are interested in those functions, feel free to review the full source code that comes with this
book.
In the public section, the AddArcheType method will be used to read an archetype file, create the object,
and store it for later. The RemoveArcheType method is used to delete the object when it is no longer
needed. Finally, the CreateObject method will be used to clone one of the previously loaded archetypes.
In the private section, we have a few types defined for creating easier names. You can see we are
using the templated Dynamic Factory that we created in Chapter 5, Decoupling Code via the Factory
Method Pattern. We also have a map of the loaded Archetype objects.
Let's take a closer look at these methods:

void M5ObjectManager::AddArcheType(M5ArcheTypes type,
const char* fileName)
{
MapItor found = s_archetypes.find(type);
M5DEBUG_ASSERT(found == s_archeypes.end(),
"Trying to add a prototype that already exists");
M5IniFile file;
file.ReadFile(fileName);
M5Object* pObj = new M5Object(type);
pObj->FromFile(file);
std::string components;//A string of all my components

file.GetValue("components", components);
//parse the component string and create each component

std::stringstream ss(components);
std::string name;
//Loop through the stream and get each component name

while (ss >> name)
{
M5Component* pComp = s_componentFactory.Build(
StringToComponent(name));
pComp->FromFile(file);
pObj->AddComponent(pComp);
}
//Add the prototype to the prototype map
s_archeypes.insert(std::make_pair(type, pObj));
}
This might seem like a difficult function, but this is where the magic happens. Let's start at the
beginning. This function takes two parameters, an enumeration ID specifying the type to create, and a
file name to associate with that enum ID. Next, we need to check if this M5Archetypes ID has been loaded
before. If it has, there must be an error. After checking for the enum error, we read the .ini file. If the
file does not exist, the ReadFile method will assert.
If there haven't been any errors, we create a brand new M5Object, and pass the M5ArcheTypes ID to the
constructor. This simply sets the type of the object, but doesn't do anything else. To set the data for the
object, we call the FromFile method to read the global section from the .ini file. This will set the
position, scale, rotation, and everything else in the object except the actual component, which needs
to be handled differently.
The trouble with the components is that the file contains the component names as strings but, for the
sake of performance during the game, we want to avoid doing string comparisons. This means we
need to somehow convert these strings to an enum value. This is the purpose of the StringToComponent
function. This function is an if/else chain that will return the correct enum based on the parameter.
Functions like this can be a problem to maintain. We will discuss in a later chapter about how to use
Windows batch files to automate this process.
After we read the object data from the file, we read the component list from the file. This is a list of
component names separated by a space. There are lots of ways we could extract each individual
component name, but one of the easiest ways is to use an STL stringstream object. This allows us to
extract separate strings from the stream, just like std::cin.
After creating our stringstream object, we loop through the stream and extract the name. We then use
s_componentFactory to build the correct component, after it has been converted to a M5ComponentTypes enum.
After the correct component is built, we pass the .ini file to the component's FromFile method to let the
derived component read its own data. Then we make sure to add the component to the object. Finally,
after all components have been read, we add the type and object pointer to our s_archetypes map.
This may seem like a complicated way of loading objects. However, this function doesn't need to
know about any derived component types, or which components go with a specific object type. If our
archetype .ini files change, we don't need to recompile this code. We are free to add, remove, or
change objects in our game and our high-level module, the M5ObjectManager, doesn't need to change:
void M5ObjectManager::RemoveArcheType(M5ArcheTypes type)
{
M5DEBUG_ASSERT(found != s_archetypes.end(),
"Trying to Remove a prototype that doesn't exist");
delete found->second;
found->second = 0;
s_archetypes.erase(found);
}
The RemoveArcheType method is much simpler than the AddArcheType. All we need to do here is make sure
the type to delete exists in the map, which we do by first finding and using a debug assert if it isn't
there. Then we delete the prototype object and erase the iterator within the map.
The RemoveArcheType method doesn't need to be called since all archetype objects will be deleted when
the game exits. However, this could be used if the user wanted to minimize which archetypes existed
throughout the game. By default, the Mach5 Engine automatically loads all archetype .ini files before
the game begins:
{
M5DEBUG_ASSERT(found != s_archetypes.end(),
"Trying to create and Archetype that doesn't exist");
M5Object* pClone = found->second->Clone();

s_objects.push_back(pClone);//A std::vector<M5Object*>
return pClone;
}
Finally, we have the method that allows the user to create Archetype objects. Here, the user supplies
the M5ArcheTypes type that they want to create. First, the method does the standard error checking that we
are familiar with. Then, after finding the correct iterator, we make use of the Prototype pattern's Clone
method to copy all data and components from the Archetype object. After creating the object, we
automatically add it to the list of active game objects and return the pointer to the user so they can
modify things such as position and velocity if needed.
Summary
In this chapter, we focused a lot on creating flexible code. Since we are using the Component Object
Model with our game objects, we want to make sure that, as our objects change, they handle that
change well. This means we don't want to modify lots of other files as we playtest and balance our
objects.
We said at the beginning of this chapter that the goal for our game objects is to completely define them
in a file. Since we are using components in our objects, we want to define the components that are
used by the objects within the file as well. By defining objects in a file, our programmers are free to
work on other code and the designers can work on balance and play testing without fear of breaking
the game or introducing bugs.
After looking at a simple example of the Prototype pattern, we looked at how it is used in the Mach5
Engine. We saw both the M5Component class and the M5Object use a Clone method to make copying objects
easy. These of course, were used by the M5ObjectManager to allow the user to create objects based on the
M5ArcheTypes enum.
Now that creating objects can be done through a file, we should focus on a problem that is more
difficult to see. Since we are using lots of object pointers that will have lots of component pointers,
we should talk about a few problems related to memory. This is what we will cover in the next
chapter.
Improving Performance with Object Pools
In programming languages, one of the most time-consuming things for the computer to do is deal with
memory allocation. It's fairly inefficient and, depending on the resources being used, could slow
down your game drastically.
A common element found in shooter games, or any game with explosions or bullets, is to create and
destroy many objects in quick succession. Take, for example, the Touhou Project series of games,
where there are many bullets being fired by both the player and enemies. When done in the simplest
manner, calling new when you want to create a bullet and delete when you want to remove it will cause
our game to lag or freeze over time.
To prevent this from happening, we can make use of the Object Pool pattern.
Chapter overview
In this chapter, we will create an object pool that will allow players to spawn a large number of
bullets on the screen for the game.
Your objectives
Why we should care about memory

Using memory pools--overloading new and delete
Design decisions
Why you should care about memory
As a programmer, you're probably already used to using new and delete (or malloc and free if you're
writing C), and you may be wondering why you would want to handle memory by yourself when it's
already built into the language and is easy to use. Well, the first thing is that like most aspects of using
a high-level programming language, you do not know what is going on behind the scenes. If you write
your own logic to handle memory, you can create your own statistics and additional debugging
support, such as automatically initializing data. You can also check for things such as memory leaks.
However, for game developers the most important aspect to look into is that of performance.
Allocating memory for a single object or thousands of them at once is approximately the same time as
the computer needs to look through your computer's memory for an opening that isn't being used, and
then give you the address to the beginning of that contiguous piece of memory. If you keep requesting
small pieces of memory over and over again this can lead to memory fragmentation, which is to say
that there isn't enough free continuous space when you want to get a larger object.
We may start off with some memory like this, with the gray sections being free memory and black
being memory set aside because we called new for that amount of data. Each time we call for new, the
computer needs to look for the first address that is open which has enough space to fit the object type
we provide:
Later on, we remove some of the memory and that opens up some space, but the computer will need to
look at each address and spend more time searching:
Finally, we get to a spot where we have very little open data and it requires a lot of work to find a
place to insert new data, due to the memory becoming fragmented.
This is especially important if you are developing titles for a console or for a mobile device, as the
size of memory you have to work with is much smaller than what you're used to working with on a
PC. If you used computers five or more years ago you may remember the idea of defragging your
computer, in which your computer would shift pieces of memory over in order to create larger blocks
that could be used later. But this was a very time-consuming process.
Mach5 doesn't easily give us the ability to support having game objects being created in this way but,
if you are interested in doing this, we do have a way that can use the concepts of an object pool in
order to not waste resources; we will discuss that later on in the chapter.
An excellent article on writing memory managers for game programming can be

found at http://www.gamasutra.com/view/feature/2971/play_by_play_effective_memory_.php.
Previously, we talked about the Singleton design pattern and how it's used to create a single instance
of something inside of our project, often something static. We know there is only one and it's only
created once, and that we can share it with the rest of our project without issues. However, the
Singleton pattern only works when the instance is initialized.
The object pool is similar but, instead of one object, we want to have a group (or pool) of objects (or
instances) that we can refer to within the rest of the project. Whenever the project wants to access
these objects, we have another object called an object pool, which acts as a liaison between the
project and the objects themselves.
Also called a resource pool or an N-ton elsewhere in computer science (but most frequently in game
development referred to as an object pool) you can think of the object pool as having a similar role to
a manager. When our program wants an object to work with, the manager knows which objects are
currently being used and will give you one that isn't, or expand to create a new one. This promotes
reusing previously created objects instead of creating and deleting them on the fly. This provides a
number of advantages when the cost of initializing a class instance is an expensive operation, or when
the rate of instantiation is high and the amount of time the objects are being used is low.
Let's consider our space shooter game as an example. Any time we press the spacebar, we have been
creating a new object of the laser we want to shoot out. Also, any time we shoot something, we have
to destroy them. This will slow down the performance of our game. This is not a huge issue now with
such a simple game, but in AAA games we use this idea a lot, for example, in any of the games in the
Naughty Dog's Uncharted series or most FPS titles. The enemies in these games are very complex,
and having them in the game is very expensive. Instead of having a bunch of enemy objects in the
game what will often be done instead is, after using enemies and having them die, they just turn
invisible and when you need a new enemy, the dead object gets moved to a new position and gets
turned on again.
The very basic elements of an object pool will look something like this:
In our object pool's case, we have some type of variable that we want to hold the copies of. In this
case, I've named it GameObject, but you'll also hear it referred to as a Reusable or Resource class instead.
We use the AcquireObject function to get an object from our object pool, and we use the ReleaseObject
function when we are finished working with it. The GetInstance function works in a similar manner to
how it did with the Singleton class we talked about earlier, by giving us access to the ObjectPool
referred to by it.
In the Mach5 Engine, there isn't an included object pool by default, so we will actually need to extend
the engine to support it. That means we'll need to actually build one from scratch.
There are multiple ways to implement the Object Pool pattern or to get a similar behavior. We'll go
over some of the commonly seen versions and the cons before moving to our final version.
Implementing a basic object pool
Let's first start off by creating an object pool for a simple class that we can create multiples of:
class GameObject
{
private:
// Character's health
int currentHealth;
int maxHealth;
// Character's name
std::string name;
public:
GameObject();
void Initialize(std::string _name = "Unnamed",
int _maxHealth = -1);
std::string GetInfo();
};
So, this sample GameObject class contains a name of the object to identify it by and some example
properties to make the class seem more game-object-like. Obviously, you can easily add more
properties and the same principles apply. In this case, we have a function called Initialize, which
provides both a set and reset of values for the class. Finally, I added in a GetInfo function to print out
information about the class so we can verify that things are working correctly.
The implementation for the class will look something like this:
/*************************************************************************/
/*!
Constructor that initializes the class' data
*/
/*************************************************************************/
GameObject::GameObject()
{
Initialize();
}
/*************************************************************************/
/*!
Initializes or resets the values of the class
*/
/*************************************************************************/
void GameObject::Initialize(std::string _name, int _maxHealth)
{
name = _name;
maxHealth = _maxHealth;
currentHealth = maxHealth;
}
/*************************************************************************/
/*!
Prints out information about the class
*/
/*************************************************************************/
std::string GameObject::GetInfo()
{
return name + ": " + std::to_string(currentHealth) + "/" +
std::to_string(maxHealth);
}
Now that we have our game objects created, we need the pool to be created:
class GameObject;
class ObjectPool
{
private:
std::list<GameObject*> pool;
static ObjectPool* instance;
// Private constructor so users are unable to create without

// GetInstance
ObjectPool() {}
public:
static ObjectPool* GetInstance();

GameObject* AcquireObject();
void ReleaseObject(GameObject* object);
void ClearPool();
};
To start off with there are two variables: pool, which will contain all of the available objects within
our object pool, and instance, which is a way for us to access it. Note that our object pool uses the
Singleton design pattern, in that there can only be one per type of object you'd like to have copies of.
In this instance, we have the same issues that we talked about earlier, namely that you have to actually
delete the pool and remove all of the elements that were created, which is why we added in a ClearPool
function which does exactly that. The implementation for the class will look something like this:
ObjectPool* ObjectPool::GetInstance()
{
if (instance == nullptr)
{
instance = new ObjectPool();
}
return instance;
}
In this preceding function, we first check if instance is set. If it's not, we dynamically allocate memory
for it and set the instance variable to it. Either way, we'll have an instance afterwards, and that is what
we return:
/*************************************************************************/
/*!
Returns the first available object if it exists. If not, it will create a new
one for us
*/
/*************************************************************************/
GameObject* ObjectPool::AcquireObject()
{
// Check if we have any objects available
if (!pool.empty())
{
// Get reference to an avaliable object
GameObject* object = pool.back();
// Since we are going to use it, it's no longer available,

// so we need to remove the last element from our list
pool.pop_back();
// Finally, return the reference

return object;
}
else
{
// If none are available, create a new one
return new GameObject()
}
}
/*************************************************************************/
/*!
Marks an object as being available again
\param
The object to be made available again
*/
/*************************************************************************/
void ObjectPool::ReleaseObject(GameObject* object)
{
// Reset the object
object->Initialize();
// Add it to our avaliable list

pool.push_back(object);
}
/*************************************************************************/
/*!
Takes care of removing all of the objects from the pool whenever we're finished
working with it.
*/
/*************************************************************************/
void ObjectPool::ClearPool()
{
while (!pool.empty())
{
GameObject * object = pool.back();
pool.pop_back();
delete object;
}
}
The ClearPool function will continuously remove objects from the pool until it is empty. We first get a
reference to the object, by retrieving the last element using the back function.
We then remove the element from the pool before deleting the object itself safely:
ObjectPool* ObjectPool::instance = 0;
Finally, C++ requires that we initialize the instance variable, so we add that last.
Once we have this foundational code in, we can start to use the classes. An example usage could be
the following:
ObjectPool* pool = ObjectPool::GetInstance();
GameObject * slime = pool->AcquireObject();
std::cout << "Initial: " << slime->GetInfo() << std::endl;
slime->Initialize("Slime", 10);
std::cout << "After Assignment: " << slime->GetInfo() <<

std::endl;
pool->ReleaseObject(slime);
slime = pool->AcquireObject();
std::cout << "Reused: " << slime->GetInfo() << std::endl;

pool->ClearPool();
delete pool;
If we save this script and run it in a blank project, you'll see the following:
In this case, we first get our ObjectPool that makes use of the GetInstance function, and then obtain an
object from the object pool using the AcquireObject function (which calls new to create the object). From
there we print out its values and, due to the constructor, it is set to our predefined default values. We
then assign the values and use it. Afterward, we release it from the list in which we placed it on the
pool, to be reused when we are ready. We then get the object again and show that it's already reset to
be reused in exactly the same way as before!
Operator overloading in C++
We now have a good foundation to build on, but we can actually make our object pool much nicer to
use. One of the cooler features in C++ is the fact that you can override the default behaviors of
operators, typically referred to as operator overloading. This is done with functions being created
with specific names that contain the operator keyword, followed by what operator you want to define.
Just like regular functions, they have return types as well as parameters that get passed to them.
For more information on operator overloading and how it works in C++, check out ht
tp://www.cprogramming.com/tutorial/operator_overloading.html.
In addition to common operators, such as +, -, and /, we also have the ability to overload the new and
delete operators as well, allowing us to use our own custom object pool instead!
To do this, we will need to add the following to the end of the GameObject class, and add the following
bold lines to the class definition:
class GameObject
{
private:
// Character's health
int currentHealth;
int maxHealth;
// Character's name
std::string name;
public:
GameObject();
void Initialize(std::string _name = "Unnamed",
int _maxHealth = -1);
std::string GetInfo();
void* operator new(size_t);

void operator delete(void* obj);
};
Here, we added two new functions to the GameObject class--one for us to create our own version of new
and another for our version of delete. Then, we need to add the implementations:
void* GameObject::operator new(size_t)
{
return ObjectPool::GetInstance()->AcquireObject();
}
void GameObject::operator delete(void* obj)

{
ObjectPool::GetInstance()->ReleaseObject(static_cast<GameObject*>(obj));
}
In our case, we are just using the ObjectPool class' functions to acquire and release our objects when
needed, instead of just allocating memory all of the time. Then, we can modify the original
implementation code as follows:
ObjectPool* pool = ObjectPool::GetInstance();
GameObject * slime = new GameObject();
std::cout << "Initial: " << slime->GetInfo() << std::endl;
slime->Initialize("Slime", 10);
std::cout << "After Assignment: "

<< slime->GetInfo() << std::endl;
delete slime;
slime = new GameObject();
std::cout << "Reused: " << slime->GetInfo() << std::endl;
pool->ClearPool();
delete pool;
return 0;
Now, don't run the code just yet. If you remember, we call the new and delete operators inside of our
ObjectPool class, so running the code now will cause a stack overflow error, because when AquireObject
calls new, it will call the GameObject class' version of new, which in turn calls the AquireObject function, and
so on and so forth. To fix this, we'll need to use the C version of allocating memory, the malloc and free
functions, in order to get memory from the system:
/*************************************************************************/
/*!
one for us
*/
/*************************************************************************/
GameObject* ObjectPool::AcquireObject()
{
// Check if we have any objects available
if (!pool.empty())
{
// Get reference to an avaliable object
GameObject* object = pool.back();
// Since we are going to use it, it's no longer available, so

// we need to remove the last element from our list
pool.pop_back();

return object;
}
else
{
// If none are avaliable, create a new one
return static_cast<GameObject*>(malloc(sizeof(GameObject)));
}
}
/*************************************************************************/
/*!
Takes care of removing all of the objects from the pool whenever we're finished
working with it.
*/
/*************************************************************************/
void ObjectPool::ClearPool()
{
while (!pool.empty())
{
GameObject * object = pool.back();
pool.pop_back();
free(object);
}
}
Now we should be able to run and see if everything is working the way we intended! This version
works fairly well as long as you want your users to still call new and delete; however, it gives you a
performance boost over time.
Building the object pool for Mach5
Now that we've seen an object pool in action, let's next learn how we can integrate the Object Pool
pattern into the Mach5 game engine. Since we are creating a shooter game, one of the things that we
spawn a lot during gameplay are the laser bullets from our ship, which makes it perfect for using
object pool functionality. And unlike the previous examples, we'll see a version of the object pool
that will not need to use pointers to access the pool, and we'll not have to worry about the pool being
created. To do this, we'll need to make some adjustments to the starter project. First, we are going to
need to change how our bullets are destroyed.
If you go into the Bullet.ini file located at Mach5-master\EngineTest\EngineTest\ArcheTypes, you'll see the
following:
posX = 0
posY = 0
velX = 0
velY = 0
scaleX = 2.5
scaleY = 2.5
rot = 0
rotVel = 0
components = GfxComponent ColliderComponent OutsideViewKillComponent
[GfxComponent]
texture = bullet.tga
drawSpace = world
texScaleX = 1
texScaleY = 1
texTransX = 0
texTransY = 0
[ColliderComponent]
radius = 1.25
Go in and remove OutsideViewKillComponent and replace it with BulletComponent. We are replacing

OutsideViewKillComponent because when it leaves the screen, it will set the object's isDead property to true,
which will call delete on it and remove it from the world. We are actually going to take care of this
ourselves, so let's replace this with our own behavior, which we will write inside of the
BulletComponent script that we will write later on in this chapter.
Next, we will want to create a new place for our ObjectPool so, with that in mind, go to the Solution
Explorer tab and then right-click on the Core/Singletons folder and select New Filter. Once you
create one, name it ObjectPool. From there, right-click on the newly created folder and select New
Item.... Then, from the menu select the Header File (.h) option and give it a name of M5ObjectPool.h.
In the .h file, we'll put in the following code:

/*************************************************************************/
/*!
\file M5ObjectPool.h
\author John P. Doran
\par email: john\@johnpdoran.com
\par Mach5 Game Engine
\date 2016/11/19
Globally accessible static class for object caching to avoid creating new objects
if we already have one not being used.
*/
/*************************************************************************/
#ifndef M5OBJECT_POOL_H
#define M5OBJECT_POOL_H
#include <vector>
#include <queue>
#include "EngineTest\Source\Core\M5Object.h"
template <M5ArcheTypes T>

class M5ObjectPool
{
public:
// Gives to us the first available object, creating a new one if none is available
static M5Object * AcquireObject();
// Returns the object to the pool making it available for reuse

static void ReleaseObject(M5Object* object);
// Removes all of the objects in the pool and removes references

// as needed
static void ClearPool();
private:
// All of the objects in the object pool
static std::vector<M5Object*> pool;
// All of the objects that are currently available

static std::deque<M5Object*> available;
};
#endif //M5OBJECT_POOL_H
You'll notice that the class is very similar to what we've done in the past but, instead of using the
GameObject class, we are going to use the Mach5 engine's M5Object class. We've also templatized the
class to make it so that this will work with any kind of object archetype that exists (including our
bullet, which is represented by AT_Bullet). I've also added a new variable called available, which is a
deque (pronounced deck), which stands for a double-ended queue. This variable will contain all of
the objects that both exist and are unused, so we can easily tell if we have any objects that we can use
or if we need to create a new one.
If you want to learn more about the deque class, check out http://www.cplusplus.com/referenc
e/deque/deque/.
We'll also want to create an M5ObjectPool.cpp file as well. In the .cpp, we'll write the following code:
/*************************************************************************/
/*!
\file M5ObjectPool.cpp
\author John P. Doran
\par email: john\@johnpdoran.com
\date 2016/11/19
Globally accessible static class for object caching to avoid creating new objects
if we already have one not being used.
*/
/*************************************************************************/
#include "M5ObjectPool.h"
#include "Source\Core\M5ObjectManager.h"
template class M5ObjectPool<AT_Bullet>;// explicit instantiation
/*************************************************************************/
/*!
one for us
*/
/*************************************************************************/

M5Object * M5ObjectPool<T>::AcquireObject()
{
// Check if we have any available
if (!available.empty())
{
// Get reference to an available object
M5Object * object = available.back();
// Since we are going to use it, it's no longer available,

// so we need to remove the last element from our list
available.pop_back();

return object;
}
else
{
M5Object * object = M5ObjectManager::CreateObject(T);
pool.push_back(object);
return object;
}
}
In this instance, we are first going to check if we have any objects inside of the available list. If none
exist we will spawn a new object, making use of the M5ObjectManager class' CreateObject function. We then
add it to the pool as it is an object in our object pool, but we do not make it available as it's going to
be used upon being acquired:
/*************************************************************************/
/*!
Marks an object as being available again
\param
The object to be made available again
*/
/*************************************************************************/
void M5ObjectPool<T>::ReleaseObject(M5Object * object)
{
// If it's valid, move this object into our available list
if ((object->GetType() == T) &&
(std::find(pool.begin(), pool.end(), object) != pool.end()))
{
//Make sure we haven't already been added already
if(std::find(available.begin(), available.end(), object) == available.end())
{
available.push_back(object);
}
}
In this case, the ReleaseObject function marks an object as being available for reuse. But, we want to do
some error checking to make sure that the function is being used properly and isn't being provided
with an invalid object.
First, the code makes sure that the object is the same type as the object pool's and that it is actually
located inside the pool somewhere. This ensures that we will only be adding objects that are valid
into our available deque. If we know the object is valid, we then look through the objects we already
have in the deque and make sure that the object hasn't already been added before. If it hasn't, we then
add it into the available deque:
/*************************************************************************/
/*!
Takes care of removing all of the objects from the pool whenever we're finished working with it.
*/
/*************************************************************************/
template<M5ArcheTypes T>
void M5ObjectPool<T>::ClearPool()
{
// Go through each of our objects and destroy them
for (int i = pool.size() - 1; i >= 0; --i)
{
M5ObjectManager::DestroyObject(pool[i]);
pool.pop_back();
}
// Now clear out the available queue

available.clear();
}
In the ClearPool function, we just go through every object in the pool and destroy that game object.
Then, we clear out the available list:
std::vector<M5Object*> M5ObjectPool<T>::pool;
std::deque<M5Object*> M5ObjectPool<T>::available;
Finally, we need to declare the pool and available objects so they can be created in the future.
Now that we have this base functionality, we need to return these objects back to our available pool.
To do this, we'll need to add the BulletComponent component we mentioned previously. Since this
component is exclusive to our game, let's move over to the SpaceShooter/Components filter and create a
new filter called BulletComp. From there, create two new files, BulletComponent.h and BulletComponent.cpp,
making sure the location is set to the Mach5-master\EngineTest\EngineTest\Source\ folder.
In the .h file, put in the following:

#ifndef BULLET_COMPONENT_H
#define BULLET_COMPONENT_H
//!< Removes The parent Game Object if it is outside the view port
class BulletComponent : public M5Component
{
public:
BulletComponent();
};
#endif // !BULLET_COMPONENT_H
Next, inside of the .cpp file, use the following:

#include "BulletComponent.h"
#include "EngineTest\M5ObjectPool.h"
BulletComponent::BulletComponent():
M5Component(CT_BulletComponent)
{
}
void BulletComponent::Update(float /*dt*/)

{
M5Vec2 pos = m_pObj->pos;
M5Vec2 scale = m_pObj->scale;
scale *= .5f;
M5Vec2 botLeft;
M5Vec2 topRight;
if (pos.x + scale.x > topRight.x || pos.x -

scale.x < botLeft.x ||
pos.y + scale.y > topRight.y || pos.y - scale.y < botLeft.y)
{
M5ObjectPool<AT_Bullet>::ReleaseObject(m_pObj);
}
M5Component * BulletComponent::Clone(void)
{
BulletComponent * pNew = new BulletComponent;
return pNew;
}
Save your files. This will make it so that if the object has a bullet component, it'll be returned to the
list; but we have to first make our objects. Go into the PlayerInputComponent.cpp file and update the
section of code for creating bullets in the Update function, as follows:
//then check for bullets
if (M5Input::IsTriggered(M5_SPACE) || M5Input::IsTriggered(M5_GAMEPAD_A))
{
M5Object* bullet1 = M5ObjectPool<AT_Bullet>::AcquireObject();

M5Object* bullet2 = M5ObjectPool<AT_Bullet>::AcquireObject();
bullet2->rotation = bullet1->rotation = m_pObj->rotation;
M5Vec2 bulletDir(std::cos(bullet1->rotation), std::sin(bullet1->rotation));

M5Vec2 perp(bulletDir.y, -bulletDir.x);
bullet1->pos = m_pObj->pos + perp * .5f * m_pObj->scale.y;
bullet2->pos = m_pObj->pos - perp * .5f * m_pObj->scale.y;
M5Vec2::Scale(bulletDir, bulletDir, m_bulletSpeed * dt);
bullet1->vel = m_pObj->vel + bulletDir;

bullet2->vel = m_pObj->vel + bulletDir;
Notice that we've replaced the creation of bullet1 and bullet2 to use our ObjectPool class' AcquireObject
function, instead of our ObjectManager class' version.
Now it will be difficult for us to see if we are using objects that have just been created or if they are
things we are reusing. Let's go back into BulletComponent and modify a property before we give it back
to the object pool:
void BulletComponent::Update(float /*dt*/)
{
M5Vec2 scale = m_pObj->scale;
scale *= .5f;
M5Vec2 botLeft;
M5Vec2 topRight;
if (pos.x + scale.x > topRight.x || pos.x - scale.x < botLeft.x ||

pos.y + scale.y > topRight.y || pos.y - scale.y < botLeft.y)
{
m_pObj->scale = M5Vec2(1.5f, 1.5f);

M5ObjectPool<AT_Bullet>::ReleaseObject(m_pObj);
}
Now we can go ahead and save our scripts and run our game!
You'll notice that, at the beginning of play, the objects have a scale of 2.5, 2.5. However, once some
objects go off the screen, you'll see something similar to the following screenshot:
When we shoot the new bullets, they have been scaled down! With this, we know that our pool is
working correctly, and that we are reusing the objects we've made before!
Issues with object pools
Now, as great as object pools are, we should take some time to talk about times when you would not
want to use object pools, and the alternatives out there.
First of all, you need to remember that when you are using a memory manager, you are telling the
computer that you are smarter than them and that you know how the data should be handled. This is
more power than other languages tend to give you, and using Uncle Ben's famous line, "with great
power comes great responsibility" as we mentioned previously in this book in Chapter 2, One Instance
to Rule Them All - Singletons. When using an object pool, you typically want to use it when objects
only have a limited lifetime and a lot of them will be created, but not all at the same time. If at one
point in time you'll have 10,000 on the screen, but the rest of the game you'll have 30 max, that 9,970
other objects' worth of memory will just be standing there waiting for you in the unlikely event that
you want to use it again.
An alternative method of handling a lot of objects at once is through a circular linked list, in which
the last element connects to the first. This will guarantee that you'll never create more things than you
have allocated memory for. If you happen to go all the way around you'll just be replacing the oldest
one and, if you have so many things on the screen at once, users will not notice the oldest one being
removed. This can be useful for things such as particle systems which we will be talking about in Chap
ter 10, Sharing Objects with the Flyweight Pattern. If you're spawning many particles, people
probably will not notice the game replacing the oldest particles with new ones.
For more information on circular linked lists, check out https://www.tutorialspoint.com/data_s

tructures_algorithms/circular_linked_list_algorithm.htm.
We were also using a type of object pool that allocated one element at a time. Alternatively, you
could allocate memory for a large number of them at a time to ensure that you'll always have that
memory reserved. While it's not needed in this case, it's definitely something to use for large classes.
While the code samples listed are in C#, Michal Warkocz lists some very good
examples of why an object pool may be a bad choice to use here: https://blog.goyello.com/2
015/03/24/how-to-fool-garbage-collector/.
Summary
In this chapter, we have used object pools to reduce system resources and user frustration by storing
and reusing objects instead of creating and removing them. After spending this time polishing your
work, you'll probably want to spend time modifying the UI of your game, which is what we will be
talking about in the next chapter!
Controlling the UI via the Command Pattern
In the last chapter, we dived deeply into the bits and bytes of computer memory so we could make our
components more efficient and easier to debug. Understanding these details can be the difference
between a game running at 60 frames per second or 30 frames per second. Knowing how to control
memory usage is an important aspect of becoming a great programmer. It is also one of the hardest
things about programming. In this chapter, we will take a break from low-level programming and look
at something high level.
The user interface, or UI, is just as important as memory management or stage switching. You could
even argue that it is more important because the player doesn't care about the low-level details. They
just want a game that is fun. However, it doesn't matter how fun the gameplay is; if the UI is difficult
to navigate or control, the fun level drops fast.
Can you remember a time when you played a game with terrible controls? Did you keep playing the
game? It is interesting because, for something so important, it often has the chance of being left until
the end of a project. Even in this book, we had to wait until the eighth chapter to cover it. However,
great games make the design of the UI and the user experience a top priority.
There are a lot of great books on how to design user interfaces and craft the user experience. This
isn't one of those books. Instead, we will look at how the code behind the UI can be implemented in a
flexible way that works with the rest of our engine. The first step to making a great UI is designing the
code that will make buttons and other input easy to create and change.
We will start by looking at a very simple but powerful pattern that allows us to decouple our function
calls from the objects that want to call them. While we are discussing the pattern, we will look at
some of the syntactically ugly and confusing ways C++ allows us to treat functions as if they were
objects. We will also see how the Mach5 Engine uses this pattern to create clickable UI buttons.
Chapter overview
This chapter is all about separating the user interface and input from the actions they perform. We will
learn about the Command pattern and how it can help us decouple our code. We will do this by first
understanding the problem then looking at how this could be solved in a C style fashion. Then after
looking at the Command pattern in depth, we will see how it is implemented in the Mach5 Engine.
Your objectives
The following lists the things to be accomplished in this chapter:
Learn the naive approach to handling input and why it should be avoided
Implement the Command pattern using function pointers and the class method pointer
Learn how the Mach5 Engine uses the Command pattern
Implement UI buttons within the Mach5 Engine
How can we control actions through buttons?
In Chapter 3, Improving on the Decorator Pattern with the Component Object Model, we
implemented game objects. Now that we have them, it seems trivial to create buttons on the screen. In
fact, in genres such as real-time strategy, there is no difference between clickable buttons and game
objects. The player can click on any unit or building and give them orders.
At first thought, our buttons could just be game objects. They both have a position, scale, and texture,
and that texture will be drawn to the screen. Depending on the game, you might draw your buttons
using orthographic projection while the objects will be drawn using perspective projection.
However, the differences go deeper than that.
At its core, a button has an action that needs to be performed when it is clicked or selected. This
behavior is usually simple; it doesn't require creating an entire state machine class. It does however,
require a little thought so we don't end up hardcoding button functionality all over our high-level
modules or repeating similar code in many different places.
In Chapter 5, Decoupling Code via the Factory Method Pattern, we saw an extremely naive way to
handle a button click on a menu screen. Recall that this code was written by one of the authors early
in their programming career:
if ((p.x > .15 * GetSystemMetrics(SM_CXSCREEN)) &&
(p.x < .42 * GetSystemMetrics(SM_CXSCREEN)) &&
(p.y > .58 * GetSystemMetrics(SM_CYSCREEN)) &&
(p.y < .70 * GetSystemMetrics(SM_CYSCREEN)))
{
if (mousedown)
{
mGameState = TCodeRex::LOAD;
mGameLevel = L0;
}
}
There are a lot of problems with this code:
First, the rectangular click region is hardcoded to the aspect ratio in full screen mode. If we
were to switch from widescreen 16:9 aspect ratio to standard 4:3 aspect ratio or even if we
changed from full screen to windowed mode, this code wouldn't work correctly.
Second, the click region is based on the screen and not the button itself. If the button position or
size were to change, this code wouldn't work correctly.
Third, this menu screen is coupled to the Windows GetSystemMetrics function instead of an
encapsulated platform code class like the M5App class. This means if we want to run on a different
operating system or platform, this menu and possibly all menus need to be modified.
Finally, the state (stage in Mach5) switching action is hardcoded to the menu. If we decide to
perform a different action, we need to modify the menu. If this action can be performed by both a
button click and keyboard input, we need to update and maintain both sections of code.
As you can see, this isn't an ideal way to handle buttons in a game. This is basically the worst way
you can implement buttons. This code is very likely to break if anything changes. It would be nice if
the author could say this code was only written as a demonstration of what not to do. Unfortunately, a
book like the one you are reading didn't exist at the time, so he had to learn the hard way.
Callback functions
A better way to deal with these button actions is with callback functions. Callback functions in C/C++
are implemented using pointers to functions. They allow you to pass functions around as if they were
variables. This means functions can be passed to other functions, returned from functions, or even
stored in a variable and called later. This allows us to decouple a specific function from the module
that will call it. It is a C style way to change which function will be called at runtime.
Just as pointers to int can only point at int, and pointers to float can only point at float, a pointer to a
function can only point at a function with the same signature. An example would be the function:
int Square(int x)
This function takes a single int as a parameter and returns an int. This return value and parameter list
are the function's signature. So, a pointer to this function would be:
int (*)(int);
We haven't given the function pointer a name, so it should look like this:
int (*pFunc)(int);
Note that the parentheses around the variable name pFunc are required, otherwise the
compiler will think this is a prototype of a function that returns a pointer to an int.
We can now create a pointer to a specific function and call that function through the variable:
int (*pFunc)(int);
pFunc = Square;
std::cout << "2 Squared is "<< pFunc(2) << std::endl;
The output for the preceding code is as follows:

Figure 8 1 - Function pointer output
Notice that we didn't need to take the address of the Square function (although that syntax is allowed);
this is because in C and C++ the name of the function is already a pointer to that function. That is why
we can call pFunc without needing to dereference it. Unfortunately, everything about function pointers
is weird until you get used to them. You must work at remembering the syntax since it doesn't work the
same as pointers to variables.
By looking at a larger example, we can get familiar with this syntax. Let's write a program with three
different ways to fill an array with values and a way to print the array:
//Fills array with random values from 0 to maxVal - 1
void RandomFill(int* array, int size, int maxVal)
{
for (int i = 0; i < size; ++i)
array[i] = std::rand() % maxVal;
}
//Fills array with value

void ValueFill(int* array, int size, int value)
{
array[i] = value;
}
//Fills array with ordered values from 0 - maxVal - 1 repeatedly

void ModFill(int* array, int size, int maxVal)
{
array[i] = i % maxVal;
}
//Helper to print array

void PrintArray(const int* array, int size)
{
std::cout << array[i] << " ";
std::cout << std::endl;
}
Our goal with this program is to write a function that can fill an array with any fill function, including
one that hasn't been written yet. Since we have a common function signature, we can create a function
called FillAndPrint that will take a pointer to any function with a matching signature as a parameter.
This will allow FillAndPrint to be decoupled from a specific fill function and allow it to be used with
functions that do not exist yet. The prototype for FillAndPrint will look like this:
void FillAndPrint(void (*fillFunc)(int*, int, int), int* array, int size, int param);
This is incredibly ugly and difficult to read. So, let's use a typedef to clean up the code a little.
Remember that a typedef allows us to give a different, hopefully more readable, name to our type:
//Defines a function pointer type named FillFUnc
typedef void(*FillFunc)(int*, int, int);
void FillAndPrint(FillFunc pFunc, int* array, int size, int param)

{
pFunc(array, size, param);
PrintArray(array, size);
}
In main, the user of this code can pick which fill function they want to use or even write a completely
new one (if the signature is the same), without changing FillAndPrint:
int main(void)
{
const int SIZE = 20;
int array[SIZE];
//See the Random number generator
FillAndPrint(ValueFill, array, 20, 3);
FillAndPrint(RandomFill, array, 10, 5);
return 0;
}
Here is what this code would output to the command line:

Figure 8 2 - Using FillAndPrint in different ways
We could even allow the user to pick the fill at runtime if we included a helper function to select and
return the correct fill function:
FillFunc PickFill(int index)
{
switch (index)
{
case 0:
return RandomFill;
case 1:
return ValueFill;
default:
//We could report an error if the value is outside of the
//range, but instead we just use a default
return ModFill;
}
}
//Our Second main example

int main(void)
{
const int SIZE = 20;
int array[SIZE];
int fillChoice;
int param;
//This doesn't properly explain to the user,

//but it is just an example
std::cout << "Enter a Fill Mode and parameter to use"
<< std::endl;
std::cin >> fillChoice;
std::cin >> param;
//See the Random number generator
FillAndPrint(PickFill(fillChoice), array, 20, param);
return 0;
}
This is a very simple example, but you can already see how using function pointers allows us to write
flexible code. FillAndPrint is completely decoupled from any specific function call. Unfortunately, you
can also see two flaws with this system. The functions must have the exact same signature, and the
parameters of the function must be passed to the user of the function pointer.
These two problems make function pointers interesting and powerful, but not the best solution for in-
game buttons that need to support a wide variety of actions with varying parameter lists. Additionally,
we might want to support actions that use C++ member functions. So far, all the examples that we
have seen were C style global functions. We will solve these problems in a moment, but first we
should look at how we will trigger our button click.
Repeated code in the component
We have the problem of wanting to decouple a specific function call from the place that calls it. It
would be nice to be able to create a button component that could save a function pointer or something
like it, and call it when the component is clicked.
One solution could be to create a new component for every action we want to execute. For example,
we might want to create a component that will change the stage to the main menu. We could create a
component class that knows how to perform that exact action:
//MainMenuComponent.h
class MainMenuComponent : public M5Component
{
public:
MainMenuComponent(void);
~MainMenuComponent(void);
virtual MainMenuComponent* Clone(void) const;
private:
};
//MainMenuComponent.cpp
void MainMenuComponent::Update(float /*dt*/)
{
M5Vec2 mouseClick;
M5Input::GetMouse(mouseClick);
if(M5Input::IsTriggered(M5_MOUSE_LEFT) &&
M5Intersect::PointRect(clickPoint,
m_pObj->pos, m_pObj->scale.x, m_pObj->scale.y))
{
M5StageManager::SetNextStage(ST_MainMenu);
}
}
The preceding case is a very simple example because it is only calling a static function with the
parameter hardcoded, but the function pointer as well as the function parameters could easily be
passed in to the constructor of this component. In fact, we could pass any object to the constructor and
hardcode a specific method call in the update function. For example, we could pass an M5Object to a
component such as the one above. The button click might change the texture of the object. For
example:
// SwapTextureComponent.cpp
void SwapTextureComponent::Update(float /*dt*/)
{
M5Vec2 mouseClick;
M5Input::GetMouse(mouseClick);
if(M5Input::IsTriggered(M5_MOUSE_LEFT) &&
M5Intersect::PointRect(clickPoint,
m_pObj->pos, m_pObj->scale.x, m_pObj->scale.y))
{
//Get the Graphics Component
M5GfxComponent* pGfx = 0;
m_savedObj->GetComponent(CT_GfxComponent, pGfx);
//Do something to swap the texture...

}
}
Unfortunately, there is a big problem with code like this; the action is completely coupled to the
button click. This is bad for two reasons. First, we can't use this action for a keyboard or controller
press unless we add additional keys to our UI button click component. Second, what happens when
we have a list of actions that we want to perform? For example, synchronizing the movement of
multiple UI objects, or scripting an in-game cut scene. Since the actions require the mouse to be
pressed on the object, our action is very limited.
The other reason this approach is bad is because we must repeat the exact same mouse click test code
in every button component that we create. What we would like to do is decouple the action from the
button click component. We would need to create a separate UI button component and an action class.
By doing that, we would factor out the part of the code that repeats, and we would gain the ability to
use the actions on their own.
The Command pattern explained
The Command pattern is exactly the pattern that solves our problem. The purpose of the Command
pattern is to decouple the requester of an action from the object that performs the action. That is
exactly the problem we have. Our requester is the button, and it needs to be decoupled from whatever
specific function call will be made. The Command pattern takes our concept of a function pointer and
wraps it into a class with a simple interface for performing the function call. However, this pattern
allows us more flexibility. We will easily be able to encapsulate function pointers with multiple
parameters, as well as with C++ object and member functions. Let's start off easy with just two
simple functions that have the same parameter count and return type:
int Square(int x)
{
return x * x;
}
int Cube(int x)
{
return x*x*x;
}
The Command pattern encapsulates a request into an object, and it gives a common interface to
perform that request. In our example, we will call our interface method Execute(), but it could be
called anything. Let's look at the Command abstract class:
//Base Command Class
class Command
{
public:
virtual ~Command(void) {}
virtual void Execute(void) = 0;
};
As you can see, the Command pattern interface is very simple--it is just a single method. As usual, we
mark the method as pure virtual so the base class can't be instantiated. Additionally, we create an
empty virtual destructor so the correct derived class destructor will be called when needed. As I
said, the name of the method isn't important. I have seen examples such as Do, DoAction, Perform, and so
on. Here we call it Execute because that was the name in the original book written by the Gang of Four.
Right from the start, we gain a benefit over function pointers by using this pattern. For every derived
class we are writing the Execute method, which means we can directly hardcode any function and any
parameters in that Execute function. Recall that when using function pointers, we needed to pass in
parameters at the time of the call:
//Derived command classes
class Square5Command: public Command
{
public:
virtual void Execute(void)
{
std::cout << "5 squared is " << Square(5) << std::endl;
}
};
In this example, we are just hardcoding the function call and the function parameter in place. This may
not seem very useful now for such a simple function, but it could be used in-game. As we will see
later, the Mach5 Engine has a command to quit the game. The command directly calls
StageManager::Quit().
In most cases, we probably don't want to hardcode the function and parameters. This is where the
power of this pattern shows. In this next example, we can use the fact that both functions have the
same signature. That means we can create a function pointer, and pass the function, and parameters to
the command. The benefit here is that because the command is an object, it has a constructor. So, we
can construct an object with an action and the parameters that will be used by that action:
//The function signature of both Square and Cube
typedef int (*OneArgFunc)(int);
//Command that can use any function of type OneArgFunc

class OneArgCommand: public Command
{
public:
OneArgCommand(OneArgFunc action, int* pValue):
m_action(action), m_pValue(pValue)
{
}
{
*m_pValue = m_action(*m_pValue);
}
private:
OneArgFunc m_action;
int* m_pValue;
};
There are a few interesting things going on here. The first is that this command can call any function
that returns an int and takes one int as a parameter. That means it can work for Square and Cube, but
also any other functions that we come up with later. The next interesting thing is that we can set the
action and parameter in the constructor; this allows us to save parameters within the class and use
them later. We could not do this by using function pointers alone. Finally, you may have noticed that
we are passing in a pointer to an int, instead of just an int. This demonstrates how we can save the
return value of a function call, and also allows us to think about these commands in a more flexible
way.
Commands are not just for quitting the game or changing the stage. We could have a command that
changes the position of a game object when executed, or perhaps swaps the position of the player and
an enemy based on some user input or a button click. By using commands, we can control everything
about the game via the UI. That sounds a lot like a level editor.
Now that we have seen two types of commands, let's look at how the client would use them. We will
start out with a simple main function. We will be constructing the command in the same function that
calls it, but these could be set via a function call instead. The important thing is that at the point where
the client calls Execute, they don't need to know which function is being called, or what parameters
(if any) are needed:
int main(void)
{
const int SIZE = 3;
int value = 2;
//This commands could be loaded via another function
Command* commands[SIZE] = {
new Square5Command,
new OneArgCommand(Square, &value),
new OneArgCommand(Cube, &value),
};
//The Client Code

commands[0]->Execute();//Square5
std::cout << "value is " << value << std::endl;
commands[1]->Execute();//OneArg Square
commands[2]->Execute();//OneArg Cube
for (int i = 0; i < SIZE; ++i)

delete commands[i];
return 0;
}

5 squared is 25
value is 2
value is 4
value is 64
As we can see, the client could call different functions using the same interface, and does not need to
care about function parameters. For such a simple pattern, the Command pattern is amazing. And it
gets even better.
Two parameters and beyond
We saw that one limitation of using function pointers was that the signatures must be the same. They
must have the same return type, as well as the same parameter types and count. We can already see
that this isn't true with the Command pattern. The client doesn't need to know or care about the
specific signature at call time since every command shares the common Execute interface. As an
example, let's look at a function with more than one parameter and create a command for that type.
Here is the function:
int Add(int x, int y)
{
return x + y;
}
As we mentioned before, the complexity of the function isn't important. For now, let's focus on
functions that take more than one parameter, as in the case of this Add function. To make our code
easier to read, let's create a typedef for this signature too:
typedef int (*TwoArgsFunc)(int, int);
Finally, let's create a Command for all functions that match this signature:
class TwoArgCommand: public Command
{
public:
TwoArgCommand(TwoArgsFunc action, int x, int y) :
m_action(action), m_first(x), m_second(y)
{
}

{
std::cout << "The Result is "
<< m_action(m_first, m_second)
<< std::endl;
}
private:
TwoArgsFunc m_action;
int m_first;
int m_second;
};
The main function is now updated to the following. Here we are only showing the parts of the code that
changed:
new Square5Command,
new TwoArgCommand(Add, 5, 6)
};
//The Client Code

commands[3]->Execute();//TwoArg

5 squared is 25
value is 2
value is 4
value is 64
The Result is 11
As you can see, we can easily create a new command for every function pointer signature we need.
When the client calls the method, they don't need to know how many parameters are used.
Unfortunately, even though our commands can take multiple arguments, those arguments are stuck
using only the int. If we wanted them to use the float, we would need to make new commands or use
the create a template command.
In a real-world scenario, you could get away with creating the commands as you need them, and only
creating them for the types you need. Another option, and one that is more common, is to have
commands call C++ class methods, since the method has the option to use class variables instead of
passed in parameters.
Pointers to member functions
So far, we have seen how we can use the Command pattern with function pointers and allow the client
to call our functions without caring about the parameter types or counts. This is incredibly useful. But
what about using commands with C++ objects? While we can get commands to work with objects, we
need to think about the problem a little first.
The most basic way to call member functions is to simply hardcode them in the Execute method. For
example, we could pass in an object to a command constructor and always call a very specific
function. In the example, m_gameObject is a pointer to an object that was passed to the constructor.
However, Draw is the hardcoded method that we always call. This is the same as hardcoding the
function in Square5Command:
//Example of hard-coding a class method
{
m_gameObject->Draw();
}
Since m_gameObject is a variable, the object that will call Draw can change, but we are still always calling
Draw. In this case, we don't have the option to call something else. This is still useful, but we would
like the ability to call any method on a class type. So, how do we get this ability? We need to learn
about pointers to member functions.
Using pointers to member functions isn't that different from pointers to non-member functions.
However, the syntax is a little stranger than you might expect. Recall that when calling a non-static
class method, the first parameter is always implicit to the pointer:
class SomeClass
{
public:
//Example of what the compiler adds to every
//Non-static class method. THIS IS NOT REAL CODE
void SomeFunc(SomeClass* const this);
private:
int m_x;
};
The this pointer is what allows the class method to know which instance of the class it needs to
modify. The compiler automatically passes it in as the first parameter to all non-static member
functions, and the address of the this pointer is used as an offset for all member variables:
SomeClass someClass;
//when we type this
someClass.SomeFunc();
//The compiler does something like this

SomeClass::SomeFunc(&someClass);
Even though it is implicitly passed in and is not part of the parameter list, we still have access to the
this pointer in our code:
void SomeClass::SomeFunc(/* SomeClass* const this */)
{
//We can still use the this pointer even though it isn't
//in the parameter list
this->m_x += 2;
//But we don't have to use it.

m_x += 2;
}
It is important to understand this because normal functions and member functions are not the same.
Class members are part of the class scope and they have an implicit parameter. So, we can't save
pointers to them like normal functions. The signature of a class method includes the class type,
meaning we must use the scope resolution operator:
//This doesn't work because they are not the same type
void (*BadFunc)(void) = &SomeClass::SomeFunc;
//We must include the class type

void (SomeClass::*GoodFunc)(void) = &SomeClass::SomeFunc;
Just having the correct pointer type is not enough. The class member access operators, known as the
dot operator (.) and arrow operator ( -> ), are not designed to work with arbitrary function pointers.
They are designed to work with known data types or known function names as declared in the class.
Since our function pointer isn't known until runtime, these operators won't work. We need different
operators that will know how to work with member function pointers. These operators are the pointer
to member operators, ( .* ) and ( ->*).
Unfortunately, these operators have lower precedence than the function call operator. So, we need to
add an extra set of parentheses around our object and our member function pointer:
void (SomeClass::*GoodFunc)(void) = &SomeClass::SomeFunc;
//this doesn't work. GoodFunc isn't part of the class

someClass.GoodFunc();
//Extra parenthesis is required for .* and ->*
(someClass.*GoodFunc)();
There is a lot more to pointers to members. This section here was just a short
introduction. If you want more information, please go to https://isocpp.org/wiki/faq/pointers-t
o-members.
Pointer to member command
Now that we know how to use pointers to member functions, we can create commands that can take
an object and a specific member function to call. Just like before, we will use a simple example. The
example class isn't designed to do anything interesting, it is just used to demonstrate the concepts:
class SomeObject
{
public:
SomeObject(int x):m_x(x){}
void Display(void)
{
std::cout << "x is " << m_x << std::endl;
}
void Change(void)
{
m_x += m_x;
}
private:
int m_x;
};
Here is a simple class called SomeObject. It has a constructor that takes an int parameter and uses it to
set the private member variable m_x. It also has two functions: one that will print the value to the
screen and one that changes the value. For now, we are keeping things simple by giving both member
functions the same signature and not taking any arguments. This allows us to create a typedef for this
type of method. Remember that the class type is part of the function signature:
typedef void (SomeObject::*SomeObjectMember)(void);
This creates a type called SomeObjectMember that can easily be used as a function parameter, function
return type, or even saved as a member to another class (of course, that is exactly what we will do
next). Even if you feel very comfortable with the syntax of function pointers and pointer to member
functions, it is still good practice to make these typedefs. They make the code more readable for
everyone, as you will see in the next code example:
class SomeObjectCommand: public Command
{
public:
SomeObjectCommand(SomeObject* pObj, SomeObjectMember member) :
m_pObj(pObj), m_member(member)
{
}
{
(m_pObj->*m_member)();
}
private:
SomeObject* m_pObj;
SomeObjectMember m_member;
};
Since the syntax of calling a member function pointer can be tricky to get right, it can be useful to use
a #define macro. While most of the time, macros should be avoided, this is one of the few times they
can help by making your code more readable:
#define CALL_MEMBER_FUNC(pObj, member) ((pObj)->*(member))
This changes our Execute function to this:

{
CALL_MEMBER_FUNC(m_pObj, m_member)();
}
All we have done is hide the ugliness away in a macro, but at least people will have a better
understanding of what it is doing. It is important to note that this macro only works with object
pointers because it uses the arrow star operator ( ->* ).
Now, in main we can create commands to object members:

int main(void)
{
const int SIZE = 6;
int value = 2;
SomeObject object(10);
new Square5Command,
new TwoArgCommand(Add, 5, 6),
new SomeObjectCommand(&object, &SomeObject::Display),
new SomeObjectCommand(&object, &SomeObject::Change)
};
//The Client Code

commands[3]->Execute();//TwoArg
//Member function pointers

commands[4]->Execute();//Display
commands[5]->Execute();//Change
commands[4]->Execute();//Display
for (int i = 0; i < SIZE; ++i)

delete commands[i];
return 0;
}
The following is the class diagram of command hierarchy:

Figure 8.3 - The command hierarchy
Even though this is just a simple demo, we can see the client code is the same whether they are
calling a function pointer or a pointer to a member function, and regardless of parameter count.
Unfortunately, we still need to create a typedef for every function and class type we need. However,
C++ templates can help us here too. We can create a template command class that can call class
methods of a specific signature (in our case, void (Class::*)(void)) that will work for all classes:
template<typename Type, typename Method>
class TMethodCommand: public Command
{
public:
TMethodCommand(Type* pObj, Method method) :
m_pObj(pObj), m_method(method)
{
}

{
(m_pObj->*m_method)();
}
private:
Type* m_pObj;
Method m_method;
};
As you can see in the Execute method, this is limited to only calling methods without arguments, but it
could easily be modified to suit your game's needs.
The benefits of the command pattern
If looking at all that crazy code makes your eyes glaze over, you are not alone. The complex syntax of
function pointers and pointer to member functions calls are some of the most difficult parts of C++.
For that reason, many people avoid them. However, they also miss out on the power offered by such
features.
On the other hand, just because something is powerful, it doesn't mean it is always the right tool for
the job. Simple is often better and, because of the many levels of indirection, code like we just saw
has the chance to cause a lot of bugs. It will be up to you to decide if using these tools is right for your
project. That being said, let's discuss some of the benefits of using the Command pattern so you can
better decide when and where to use it.
Treating a function call like an object
The biggest benefit of using the Command pattern is that we are encapsulating the function or method
call and the parameters. This means that everything needed for the call can be passed to another
function, returned from a function, or stored as a variable for later use. This is an extra level of
indirection over only using function or method pointers, but it means the client doesn't need to worry
about the details. They only need to decide when to execute the command.
This might not seem very useful since we need to know all the function arguments before we pass it to
the client. However, this situation can happen more often than you might think. The fact that the client
doesn't need to know the details of the function call means that systems such as the UI can be
incredibly flexible, and possibly even read from a file.
In the above example, it is obvious that at the time of the call, the client doesn't know which command
exists at a given array index. This is by design. What might not be so obvious, is that the array could
have been populated using the return value from a function instead of hardcoded calls to a new
operator (which we learned in Chapter 5, Decoupling Code via the Factory Method Pattern, leads to
inflexible code). This flexibility means that the function to be executed can be changed at runtime.
A perfect example of this is a context sensitive action button in a game. Since there is a limited
number of buttons on a gamepad, it is often useful to have the action of button change depending on
what the player is doing. This could mean one button is responsible for talking to an NPC, picking up
an item, opening a door, or triggering a quick time event depending on the player's location and what
they are doing.
Without the Command pattern, the logic involved in organizing, maintaining, and executing all the
possible actions in a game would be incredibly complex. With the Command pattern, it is giving
every actionable item a command, and making it available when the player is near.
Physically decoupling the client and the
function call
One aspect of good design is low coupling. We have talked about this a lot before, and it applies here
as well. First, since the client is only dependent on the base Command class, it is easier to test. This is
because both the client and the specific function calls or actions can be tested independently to ensure
that they work. Furthermore, since these unit tests are testing smaller amounts of code, we can be
more confident that all possible cases are tested. This also means that the client or the commands
have a better chance to be reused because of the low coupling within this project.
Second, the client is less likely to break when changes to the code base occur. Since the client doesn't
know which functions or methods are called, any changes to parameter counts or method names are
local only to the commands that implement the changed methods. If more commands need to be added,
those commands will automatically work with the existing client because they will use the Command
class interface.
Finally, compile times can be reduced because the client needs to include fewer header files.
Including fewer header files can lower the compile time since every time the header changes, every
source file that includes it must be recompiled. Even the smallest change to a comment in a header
file means that all the function calls from that header need to be rechecked for correct syntax at
compile time and relinked at link time. Since our client doesn't know the details of the functions calls,
there are no header files to include.
Temporal decoupling
This type of decoupling isn't talked about much because it only applies to a few situations and, most
of the time, this isn't what we want. Usually, when we call a function we want it to execute
immediately. We have a specific algorithm in our code and the timing and order of that code is very
important. This isn't true of all code. One situation is multithreaded code, in which multiple paths of
code are executing simultaneously. Other situations are UI or context sensitive buttons, where the
action to be executed is set up in advance instead of hardcoded in place. Let's look at some code as
an example:
//Examples of setting up function calls
//Immediate execution
Add(5, 6);
//Delayed execution
Command* p1 = new TwoArgCommand(Add, 5, 6);
//Immediate execution
someObject.Display();
//Delayed execution
Command* p2 = new SomeObjectCommand(&object,&SomeObject::Display);
In all four of the above situations, the functions and parameters are given. However, the command
versions can be passed to other methods, called and/or recalled based on the need of the client.
Undo and redo
Another major benefit of having the call details packaged together in a class is the ability to undo an
operation. Every modern desktop application, as well as the best web applications being made these
days, features the ability to undo the last action or actions. This should be a standard that you strive to
follow when implementing a level editor for your game.
Implementing a single level of undo in an application can seem like a large task. The naive approach
might be to save the entire state of the application, possibly to a file, and reload that state when we
need to undo. Depending on the application, there might be a lot of data to save. This method doesn't
scale well in applications that can have dozens or hundreds of levels of undo. As the user does more
actions, you would need to make sure to delete the oldest state before saving the current one.
This simple approach is even more difficult when you also need to implement redo. Obviously, the
text editors and tools that we use every day don't store hundreds of undo and redo files on the hard
drive. There must be a better way.
Instead of saving the entire state of the program, you only need to save information about the action
that happened, and what data was changed. Saving a function and the parameters to the function
sounds a lot like the Command pattern. Let's look at a simple example of moving a game object from
one place to another in a level editor. We could create a command like this:
class MoveCommand: public Command
{
public:
MoveCommand (Object* pObj, const Vec2D& moveTo) :
m_pObj(pObj), m_method(method), m_oldPos(pObj->pos)
{
}
{
m_pObj->pos = m_moveTo;
}
//Add this method to the Command Interface

virtual void Undo(void)
{
m_pObj->pos = m_oldPos;
}
private:
Object* m_pObj;
Vec2D m_moveTo;
Vec2D m_oldPos;//Save the old position so we can redo
};
By adding the Undo method to the command interface and making sure to save the old data that will be
modified in the Execute method, performing undo and redo becomes incredibly simple. First, we need
to implement a command for every action that can be performed in our editor. Then, when the user
interacts with the editor, instead of directly calling a function, they always call a command and add it
to the end of our array of commands. Undoing and redoing is just a matter of calling the Execute or Undo
method of the current array index.
It might seem like a lot of work to create all those commands, and it is. However, that work is
replacing the work of hardcoding function calls when a user presses keys or clicks the mouse. In the
end, you will build a better system that people will want to use.
Easy UI with commands in Mach5
Now that we have seen what the Command pattern is, let's look at how it is used in the Mach5 Engine.
You will be surprised that there isn't much code here. That is because using the Command pattern is
easy once you understand the code behind it. In this section, we will look at both the component
responsible for the mouse click and the commands that are used within the engine.
Let's have a look at the M5Command class:

class M5Command
{
public:
virtual ~M5Command(void) {}//Empty Virtual Destructor
virtual void Execute(void) = 0;
virtual M5Command* Clone(void) const = 0;
};
Here is the M5Command class used in the Mach5 Engine. As you can see, it looks almost identical to the
Command class we used in the example. The only difference is that since we plan on using this within a
component, it needs to have a virtual constructor. That way we can make a copy of it without knowing
the true type.
The code for the UIButtonComponent class is as follows:

class UIButtonComponent: public M5Component
{
public:
UIButtonComponent(void);
~UIButtonComponent(void);
virtual UIButtonComponent* Clone(void) const;
void SetOnClick(M5Command* pCommand);
private:
M5Command* m_pOnClick;
};
As you can see, our UI button is a component. This means that any game object has the potential to be
clicked. However, this class is specifically designed to work with objects that are in screen space,
which is how the operating system gives us the mouse coordinates. The rest of the code here looks
like you might expect. As part of the UIButtonComponent class, we have a private M5Command. Although this
class is simple, it will be worth it for us to go through and see what each method does:
UI Button Component::UI Button Component(void) :
M5Component(CT_UIButtonComponent), m_pOnClick(nullptr)
{
}
The constructor is simple (as are most component constructors) since they are designed to be created
via a factory. We set the component type and make sure to set the command pointer to null so we set
ourselves up for safer code later:
UIButtonComponent::~UIButtonComponent(void)
{
delete m_pOnClick;
m_pOnClick = 0;
}
The destructor is where that null pointer comes in handy. It is perfectly legal to delete a null pointer,
so we know that this code will work, even if this component never receives a command:
void UIButtonComponent::Update(float)
{
if (M5Input::IsTriggered(M5_MOUSE_LEFT))
{
M5Vec2 clickPoint;
M5Input::GetMouse(clickPoint);
if (M5Intersect::PointRect(clickPoint, m_pObj->pos,
m_pObj->scale.x, m_pObj->scale.y))
{
M5DEBUG_ASSERT(m_pOnClick != 0,
"The UIButton command is null"):
m_pOnClick->Execute();
}
}
}
The Update function is where we perform the test to see if the mouse click intersects the rectangle
created by the object. As we mentioned before, this class could work with all objects, but to simplify
the code we decided we would only use this class for screen space items. The code that is important
in this decision is the GetMouse function. This function always returns coordinates in screen space. It
would be possible to check if the object was in screen space or world space and convert the
coordinates using the M5Gfx method ConvertScreenToWorld.
That null pointer comes in handy here as well. Since we know that the command pointer is valid or
null, we can do a debug assert to test our code before we execute it:
UIButtonComponent* UIButtonComponent::Clone(void) const
{
UIButtonComponent* pClone = new UIButtonComponent();
pClone->m_pObj = m_pObj;
if(pClone->m_pOnClick != nullptr)
pClone->m_pOnClick = m_pOnClick->Clone();
return pClone;
}
The Clone method looks like you might expect after reading Chapter 6, Creating Objects with the
Prototype Pattern. This is one situation where we always need to test for null before using the
command. We can't clone a null command and it is completely valid to clone this component, whether
the command has been set or not:
void UIButtonComponent::SetOnClick(M5Command* pCommand)
{
//Make sure to delete the old one
delete m_pOnClick;
m_pOnClick = pCommand;
}
The SetOnClick method allows us to set and reset the command that is associated with this component.
Again, we don't need to test our command before deleting. We also don't need to test if the method
parameter is non-null, because a null value is perfectly acceptable.
Even though we haven't done it for this class, this class could easily be expanded to include an
OnMouseOver event that gets triggered when the mouse is inside the object rectangle but the mouse isn't
clicked. A feature like this could have lots of uses for both UI and world objects. Implementing it
would be as easy as swapping the two conditional statements in the Update function:
void UIButtonComponent::Update(float)
{
M5Vec2 clickPoint;
M5Input::GetMouse(clickPoint);
if (M5Intersect::PointRect(clickPoint, m_pObj->pos,
m_pObj->scale.x, m_pObj->scale.y))
{
if (M5Input::IsTriggered(M5_MOUSE_LEFT))
{
//Do onClick Command
}
else
{
//Do onMouseOver Command
}
}
}
Using commands
Now that we have seen the base M5Command class and UIButtonComponent class, let's look at one of the
derived commands to see how it is used in the game. The command that we will look at is a common
one needed in games. This is the action that will allow us to change stages from one to the next:
class ChangeStageCommand: public M5Command
{
public:
ChangeStageCommand(M5StageTypes nextStage);
ChangeStageCommand(void);
virtual void Execute(void);
void SetNextStage(M5StageTypes nextStage);
virtual ChangeStageCommand* Clone(void) const;
private:
M5StageTypes m_stage;
};
When used with a UIButtonComponent, this will allow the user to click a button and change to a new stage.
As you can see, there are two ways to change the stage in the constructor and in the SetNextStage
method. This allows the user the ability to create a command and decide later what stage it will
switch to. The Execute method is as simple as can be since the StageManager is a Singleton:
void ChangeStageCommand::Execute(void)
{
M5StageManager::SetNextStage(m_stage);
}
The following is the output:
Figure 8 4 - An example of UIButtons in the Mach5 Engine
To be truly flexible, we would want all UIButtons loaded from a file. As with game objects, it would
be best if menus and levels were not coupled to specific commands. At the very least, we would
prefer to avoid hardcoding positions and sizes for each button. This proved to be easy with game
objects. The Player or Raider game objects are so specific that when reading a level file, we only
need to overwrite the position of each object. The size, texture name, and other attributes can be read
from the more specific archetype file.
Buttons are more difficult since each one may use a different texture name, have a different size, and
use a different command. We can't set this data in a button archetype file because all buttons will be
different. Furthermore, game commands that need to control a specific game object are difficult to
load from file since we have no information about the object except the type. This means that while
we could create and load a command that controls the player, which we only have one of, we can't
create and load a command that controls an arbitrary Raider, since we could have many per stage.
Having a high-quality level editor would solve both issues because the tool can manage data better.
This could even include assigning object IDs that could be used by commands in the game. For this
book, defining archetypes for every button worked well. While this may seem like a lot of work, the
data in each archetype file would have otherwise been hardcoded into a .cpp file.
Summary
In this chapter, we focused on creating flexible, reusable buttons. Even though the UI may not be as
fun to code or talk about as gameplay mechanics, to the player, it is just as important. That is why
creating a good system to add and manage the UI in an intelligent way is so vital to making a great
game.
We took an in-depth look at C++ function pointers and pointers to members. This is well known for
being confusing and difficult. However, by mastering the techniques, we could create flexible
commands that can call any C style function or C++ object method.
While this technique isn't always needed, in the case of UI, it allowed us to create an incredibly
flexible system. Our UI objects and most commands can be set up and read from a file. If you were to
create a level editor, you could easily use this system to create and read all UI buttons and commands
from a file.
Now that we have a flexible system for creating the UI, let's move on to another problem everyone
has when making games. In the next chapter, we will talk about a pattern that will allow us to better
separate our engine code from our gameplay code.
Decoupling Gameplay via the Observer
Pattern
Wow! The last two chapters were full of pointer craziness. Those two chapters combined, cover what
are probably considered the hardest parts of C++. While all design patterns deal with pointers and
using virtual functions at some level, it doesn't get much more difficult than what was covered in Chapt
er 7, Improving Performance with Object Pools and Chapter 8, Controlling UI via the Command
Pattern.
In Chapter 7, Improving Performance with Object Pools, we went very low-level into the guts of C++
memory. This involved casting and strange pointer manipulation that most people tend to avoid. In Cha
pter 8, Controlling UI via the Command Pattern, we dealt with controlling the UI, which is more
high-level. However, we learned how to control C++ objects and their methods in a way that allows
us to create flexible code, but can also be very confusing.
If you feel comfortable with those chapters, then you are doing great. If you felt that those topics were
a little difficult, you are not alone. Either way, you should be excited to know that this chapter covers
a pattern that is easy to understand, easy to implement, and easily allows us to decouple our core
systems from the gameplay specific code that is likely to change often.
In this chapter, we will cover a pattern called Observer, which is used to connect objects together in
a decoupled way. As you will see, it is very simple to implement and can be applied in many places
throughout our code base.
Chapter overview
This chapter is all about learning how to decouple code using the Observer pattern. This chapter is a
little different than the others because it isn't about using the Observer pattern to solve one big
pattern; it is about learning to apply it to the many little situations that lead to sloppy code throughout
a game's development. The Observer pattern will show you that there is a better way.
First, we will explore the inevitable situation of gameplay code leaking into our engine code. Then,
we will learn about the Observer pattern and how it can improve these situations. Along the way, we
will look at some example code from the Mach5 Engine. However, since this pattern has so many
uses, we will focus on how it can be incorporated into a game, as opposed to showing one large
specific example.
Your objectives
Learn two ways that gameplay code can cause trouble for your engine code
Implement the simple Observer pattern example
Learn the pros and cons of the Observer pattern
How gameplay creeps into every system
The first time someone makes a game, there is very likely no distinction between the game and the
engine. This is usually because there is no engine. A common first game would include Ticktacktoe or
Hangman. Games like these are simple enough that they can be completely written in main, or
possibly using a few functions. They are also simple enough that they don't require complex systems
like graphics or physics. There is no need for reusable engine code.
As you are learning to program more, you may decide to try making a 2D game using a graphics API,
like DirectX or OpenGL. Code like this can be very difficult the first time it is used, so writing
cleanly separated code isn't a top priority. Just as before, the game is made with just a few functions
or classes. Drawing code is often mixed with collision code in one single file.
At some point, we all get to a place where the code is too complex and fragile. Hardcoding too many
game object types or too many levels make us wish for a better way. Of course, that is the reason for
this book. We are trying to find a better way to create games. Throughout this book there has been one
major theme: things always change!
To cope with this change, we have tried to create a clear distinction between the parts of our game
that will change and the parts of the game that are unlikely to change. In clear terms, we are trying to
separate our engine code from our gameplay code. This clear separation of parts has lead us through
eight chapters of patterns that solve very large and specific problems in games. Every game must deal
with creating flexible game objects with complex behaviors. So, we learned about the Component
Object Model and Finite Stage Machines. Every game must deal with creating an easy to modify UI.
So, we learned about using the Command pattern to read actions from a file. These are common
problems with common solutions.
However, as you write more code and you start to add more features to your game, you will always
find that the clear separation between engine and gameplay starts to blur. One place that this becomes
obvious is physics. The physics engine is responsible for moving objects as well as testing for and
resolving collisions.
While this engine should be purely mathematical, the fact is that a game is made up of more than just
physics objects. It is made of bullets, raiders, players, and more. When a bullet collides with a
player, we must execute some very specific gameplay code, such as deleting the bullet, creating a
small particle effect at the collision point, and subtracting player health. The question is, where
should this code be executed? If the code is placed inside the physics engine, it is highly coupled with
every game object type. If it is executed outside of the engine, we need to get the collision information
to the correct location in a clean way.
The same problem of creeping gameplay code occurs with achievements. Achievements are always
game specific but they end up getting mixed all throughout a code base. They can range from tracking
behavior such as how many bullets the player fired, to tracking total time played or how long the
game has been paused. However, they could always be related to engine specific behavior such as
how many times the resolution has been changed, how many network connections have been made,
how many times a UFO game object was created or destroyed, or how many collision events of a
certain kind have occurred. This blurred line between engine and gameplay code, as well as general
increased dependencies, makes code reuse very difficult.
Hardcoding requirements
We know that introducing gameplay code to our engines increases dependencies and limits code
reuse. We also know that for a given action, the requirements are likely to change as gameplay
features are added. Imagine the situation of adding controller support for split screen multiplayer in
our space shooter. As more controllers are plugged in, additional players are created, the game
difficulty is increased, and we split the screen to follow the new player. In this case, the original
controller detection occurs in the Input Manager, but we need to notify a few other engines that
something has changed. An example of this code might look like this:
//Not real code, just an example!
void InputManager:Update(void)
{
int controllerCount = GetControllerCount();
if(controllerCount > m_currentControllerCount)
{
m_currentControllerCount = controllerCount;
Object* pObj = ObjectManager::CreatePlayer(controllerCount);
GameLogic::SetDifficulty(controllerCount);
//player position is the camera location for the screen
Graphics::SetScreenCount(controllerCount, pObj->pos);
}
}
We might be reasonably certain that this code won't change. If we are, then hardcoding the
requirements is fine. However, if we are not certain, it is good to assume that requirements always
change. We might need to support online multiplayer and send a message to the Network Manager. We
might allow the player to choose which type of space ship they want from a list of possible player
ships, so we need to call a different Object Manager function or pause and switch to a new Ship
Selection Stage via the Stage Manager.
In this case, we have a set of objects that need to be notified when an event occurs. We want the
notification to happen automatically, but we don't want to change the Input Manager every time a new
object needs to be notified. More generally, we have a broadcasting object that we don't want to
change every time there is a new object that needs to listen. This would be like a Wi-Fi router
needing to be updated every time a new device is in range.
The situation above describes interactions between different core engines. However, these
interactions only happen this way because of the specific requirements of the game. Even though there
is no gameplay code, the functionality of the specific game has crept into the input engine, and would
need to be changed if we were making a different game. Of course, we could try to factor out some of
this code into a game logic engine or just put similar code into a stage. Is there another way? We will
consider that, but first, we will explore the problem from the other side.
Polling
Despite your best efforts, gameplay often turns into a mess. There are so many interacting parts that it
is impossible to completely reduce coupling of gameplay code. While it makes sense that a graphics
engine and a physics engine can be completely independent and decoupled, we shouldn't even try to
do this with gameplay. Gameplay code is the game. The best we can do is to attempt to handle
modifications to the game design gracefully.
The most obvious case of this is the state machines that we saw from Chapter 4, Artificial Intelligence
using the State Pattern. The states often need access to the Object Manager and so scan the entire list
of objects looking for an object. They may also need access to physics to see if they are colliding
with this frame or the next. The gameplay code is the code that glues everything else together, so it
can't really be completely decoupled.
Another example of this could be drawing HUD objects in screen space. If the window resolution
changes, the objects need to be reoriented. A button in the center of the screen at 800 x 600 must still
be in the center of the screen at 1280 x 1024. That means the position can't be hardcoded and must
automatically adjust when the resolution changes. There are two ways of doing this. The first is the
same as the example above; we could have the change resolution button call methods of other systems
that care. The second would be for objects that care about resolution changes to ask the application
for the resolution:
void RepositionComponent::Update(float /*dt*/)
{
M5Vec2 windowSize = M5App::GetResolution();
m_pObj->pos.x = windowSize.x * m_xScale;
m_pObj->pos.y = windowSize.y * m_yScale;
}
Here is an example RepositionComponent. Every frame, it asks the M5App for the window resolution and sets
the object to the specified scale of the window. An object that needed to be in the center of the screen
would have x and y scale values of .5. If the resolution is 800 x 600, the position of the object will be
x = 400, y = 300. If the resolution changes to 1280 x 1024, the position of the object will be x = 640, y =
512.
This works exactly as expected, but it does a lot of unnecessary work. The problem isn't that this code
will make your game slow; you could have thousands of objects doing this before seeing slowdown.
Still, this component is asking for the resolution 60 times a second, when a resolution change might
only happen once per play session. More likely, the player will choose a resolution once, and the
game settings will save it to a file and automatically load it every session after that.
The problem with this polling method is that the RepositionComponent has no way of knowing when the
resolution changes. It would be nice if this code ran only when the data changed. Instead, it constantly
asks for the exact same data, and calculates the exact same position every frame.
The example above is small. Problems like this may seem trivial. Individually, they won't even be a
problem but, when added together, they can affect your game. This is true of all the examples so far in
this chapter. They seem easy to solve one by one, but they can add up to lots of wasted CPU cycles as
well as wasted developer time. It would be nice to have a pattern that is easy to implement that can
solve little problems like these that pop up all over a game. Luckily, the Observer pattern will do
exactly that.
The Observer pattern explained
The intent of the Observer pattern is to define a one-to-many relationship between objects. When the
state of one object changes, all its dependents are notified. The typical names for the objects in this
pattern are the Subject (the one), and the Observers (the many). The Subject will contain data that
the Observers need to know about. Instead of the usual situation of classes requesting data from
another (polling), our Subject will notify a list of Observers when the data has changed.
The terms Subject and Observers may seem a little confusing at first. However, this concept is very
easy, and one that most of us are familiar with. When trying to understand the Observer pattern, think
of a blog and subscribers. In this case, the blog is the Subject and the subscribers are the Observers.
A blog may be updated once a day, once a week, once a month, or even less. Readers of the blog have
the option to check for updates as much as they want, however this can waste a lot of time if the
readers check more frequently than the blog is updated. Instead, fans will often choose to subscribe to
an e-mail list so they can be notified of updates as they come. The blog keeps a list of subscribers and
sends an e-mail to everyone on the list when an update is posted.
The Subject and Observer
Let's look at a code example to understand this pattern better. There are a few different ways to
implement this pattern, so we will discuss implementation strategies along the way:
class Observer
{
public:
virtual ~Observer(void) {}
virtual void Update(float currentHealth, float maxHealth) = 0;
};
We start off with our Observer interface. As always, we make our destructor virtual. The only method
we need is an Update method however, as always, the name isn't important. This is the method that the
Subject will use to notify the Observers that something has changed. One thing you might notice is that
the update is very specific. In this case, it has two floats as arguments. This is a dependency that
might change and cause our code to break. We will address improvements a little later.
You might also notice that there is no member data. It could be possible to give the base class a
pointer to the Subject, and have this class be responsible for registering and unregistering
(subscribing and unsubscribing) with the Subject. We decided to move that behavior into the derived
classes so we could keep the base class as simple as possible:
class Subject
{
public:
virtual ~Subject(void) {}
virtual void RegisterObserver(Observer* pToAdd) = 0;
virtual void UnregisterObserver(Observer* pToRemove) = 0;
virtual void Notify(void) = 0;
};
Our Subject is almost as simple as the Observer. The key methods we need are ways for the
Observers to subscribe and unsubscribe to the Subject. Here, we called those methods RegisterObserver
and UnregisterObserver. We have also added a Notify method which will be used to call Update on all
registered Observers. There is no reason that this method needs to be public or even exist at all. As
long as the derived class calls Update on the registered Observers, we are using the pattern correctly.
Again, you will notice that there are no data members in this class. We could easily add a vector of
Observer pointers here in the base class. In fact, we could easily implement these methods because
they will almost always be the same. However, we have chosen to keep this class simple and let the
derived classes choose how to implement these methods.
The Player
To demonstrate how the Observer pattern can be used, we will look at a common situation in games.
We will have a Player with some health that will need to be shared. The health of the Player can often
be used for many things in a game. The value of health might be displayed as part of the HUD. It can
also be displayed as a colored health bar in either the HUD or directly at the top or at the bottom of
the Player. Additionally, the game may switch to a game over screen when the Player health is at or
below zero.
These display elements, as well as the stage switching mechanisms, are dependent directly on the
Player's health. Since it is very unlikely that these variables are all in the same scope, it would take
some work if we tried to implement this via polling. In that case, each object would need to find the
Player and ask for the health value. Since the health of the Player is unlikely to change every frame,
most of this work is being wasted. Instead, we will make the Player derive from the Subject so it can
notify all Observers when the health has changed:
class Player: public Subject
{
public:
Player(float maxHealth);
void AdjustHealth(float health);
virtual void RegisterObserver(Observer* pToAdd);
virtual void UnregisterObserver(Observer* pToRemove);
virtual void Notify(void);
private:
typedef std::vector<Observer*> ObserverVec;
float m_maxHealth;
float m_health;
ObserverVec m_observers;
};
This Player class is very simple. Since this is only an example, we will only be focusing on the health.
In the constructor, we can set the max health. The AdjustHealth method will be used to make the health
change. Of course, we also implement each of the base class virtual methods. In the private section,
we use an STL vector to keep track of our Observers. We also store our constructor value, as well as
a variable for our current health:
Player::Player(float maxHealh):
m_maxHealth(maxHealth),
m_health(maxHealth)
{
}
The Player constructor sets data passed in by the user. Since the base Subject class has no data, there is
nothing special to do here:
void Player::RegisterObserver(Observer* pToAdd)
{
ObserverVec::iterator itor;
itor = std::find(m_observers.begin(),
m_observers.end(),
pToAdd);
assert(itor == m_observers.end());
m_observers.push_back(pToAdd);
}
The RegisterObserver method takes a pointer to an Observer and adds it to the vector of the Observers.
Depending on the behavior of the Observer, being added to the list twice could cause a lot of
problems and could be a difficult bug to track down. In this example, we have chosen to assert if the
same Observer is added twice. After that, we add it to our vector:
void Player::UnregisterObserver(Observer* pToRemove)
{
ObserverVec::iterator itor;
itor = std::find(m_observers.begin(),
m_observers.end(),
pToRemove);
if (itor != m_observers.end())
{
std::swap(*itor, *(--m_observers.end()));
m_observers.pop_back();
}
}
Our UnregisterObserver class is a little more forgiving. If we don't find the Observer in the vector, we
ignore it, instead of throwing an assert. This will make a little more sense later. You will see that our
Observers will automatically remove or unregister in their own destructors. However, unregistering
twice is unlikely to cause a problem. The line std::swap(*itor, *(--m_observers.end())) might look a little
scary. Remember that the end method returns an iterator to one past the end of the vector. So, before
we dereference, we decrement our iterator so it is pointing at the last element in the vector. Then we
swap and pop, removing the correct element:
void Player::Notify(void)
{
size_t size = m_observers.size();
m_observers[i]->Update(m_health, m_maxHealth);
}
As we said before, the Notify method doesn't need to exist. It would be fine if the class logic notifies
the Observers internally, perhaps in Setter methods or when the data changes as in our AdjustHealth
method. However, if there was more than one piece of data that the Observers cared about, the user
could make many changes and send all the data to the Observers just once. Or, perhaps initializing the
Observer data before the game has started.
This method is simple. It loops through the vector of the Observers and calls the Update method,
sending the health data to those that care:
void Player::AdjustHealth(float adjustHealth)
{
m_health += adjustHealth;
Notify();
}
This method simulates the Player gaining or losing health. As you can see, after the health is modified,
the class calls its own Notify method, letting all Observers know about the change.
The Observers
For this example, we have the three Observers that we mentioned before. Two are related to
displaying the health of the Player in different ways; the other is used for quitting when the health of
the Player is zero or less:
//Used to quit the game when the "game", when the player's health
//is less than or equal to 0
class StageLogic : public Observer
{
public:
StageLogic (Subject* pSubject);
bool IsQuitting(void) const;
~StageLogic(void);
virtual void Update(float currentHealth, float maxHealth);
private:
bool m_isQuitting;
Subject* m_pSubject;
};
//Used to Color the player health bar based on the how full it is
class PlayerHealthBar : public Observer
{
public:
PlayerHealthBar(Subject* pSubject);
~PlayerHealthBar(void);
void Display(void) const;
private:
float m_percent;
std::string m_color;
};
//Used to Display the health of the player as a value

class PlayerDisplay : public Observer
{
public:
PlayerDisplay(Subject* pSubject);
~PlayerDisplay(void);
void Display(void) const;
private:
float m_health;
};
As you can see, each derived Observer class overloads the Update method from base. You will also
notice that each constructor takes a pointer to a Subject as the only parameter, and saves that pointer
into a member variable. This isn't necessary, but it makes registering and unregistering more
convenient because the objects take care of themselves. In this example, all three of the Observers'
constructors and destructors do the exact same thing. Here is one:
PlayerDisplay::PlayerDisplay(Subject* pSubject):
m_health(0.0f),
m_pSubject (pSubject)
{
m_pSubject ->RegisterObserver(this);
}
PlayerDisplay::~PlayerDisplay(void)
{
m_pSubject ->UnregisterObserver(this);
}
The choice of keeping a pointer to the Subject is up to you. It has some problems, which we will look
at a little later; however, it allows the Observer to unregister in the destructor. This means that the
user doesn't need to do it, which makes using the Observers classes very easy. If we don't keep this
pointer, unregistering must be done manually and could be difficult depending on how you access the
Subject and Observers.
The rest of the Observer methods are simple and don't interact with the Subject at all. Instead, the Update
methods do some logic based on the values of currentHealth and maxHealth. For the two display elements,
this means calculating some values; for the StageLogic class, this means setting m_isQuitting to true if the
value of current health is zero or less. Let's look at an example Update from one of the Observers:
void PlayerHealthBar::Update(float currentHealth, float maxHealth)
{
m_percent = (currentHealth / maxHealth) * 100.f;
if (m_percent >= 75.0f)

m_color = "Green";
else if (m_percent < 75.0f && m_percent > 35.0f)
m_color = "Yellow";
else
m_color = "Red";
Figure 9 1 Interaction of Subjects and Observers
As you can see, the Update methods aren't very complicated. Nothing about the above method is using
the Subject. The data could have come from anywhere. The part that is most interesting is how simple
these objects are to use now. All three Observers are using the Player's health, but they don't need to
call any Player methods. Even though these four objects interact, using them is incredibly simple. Let's
look at how we can use these objects together:
int main(void)
{
//Our value to decrement by
const float DECREMENT = -1.0f;
const float STARTING_HEALTH = 5.0f;
//creating our objects

Player player(STARTING_HEALTH);
PlayerDisplay display(&player);
PlayerHealthBar bar(&player);
StageLogic stageLogic(&player);
//Set the initial values to print

player.Notify();
//loop until player is dead

while (!stageLogic.IsQuitting())
{
display.Display();
bar.Display();
player.AdjustHealth(DECREMENT);
}
return 0;
}
The main function starts out with a few const values to improve readability. After that, we create our
objects. We first create our Player, which is our Subject. Then we create our Observers. Each
Observer gets a pointer to the Subject. Remember, they are only dependent on the Subject interface,
not to the derived Player class. Once all the Observers are created, the Player does an initial Notify call
so the Observers start out with the correct data. Finally, we use our objects. The simplicity of this
while loop is amazing. Since the code linking them together is all internal, using the objects together
becomes very easy. Compare the example above with a version of the code that doesn't use the
Observer pattern:
//Alternate code without Observer pattern
while (!stageLogic.IsQuitting())
{
display.SetHeath(player.getHealth());
display.Display();
bar.setHealth(player.getHealth(), player.getMaxHealth());
bar.Display();
player.AdjustHealth(DECREMENT);
stageLogic.SetQuit(player.GetHealth() <= 0);

}
Using the Observer pattern allows us to write code that is more elegant and simple to use.
Unfortunately, many programmers write code that is closer to the second version. They don't realize
that with just a little thought as to how the objects will interact, the code is easier to use, easier to
read, and is more efficient because it only gets data from the Player when the data changes.
In this simple example, it may not seem like the code is much different, but remember that this is just
demonstrating the pattern as simply as possible. The second version looks reasonable because all the
objects are in the same scope. Except for the constructor, in a real project, the Observer code stays
the same. However, the second version can become a mess of Singleton method calls and object look
ups.
Push versus Pull
One big problem with the Observer pattern that we have looked at so far is that the Update method in
the base class is limited. In our case, it can only be used with Observers that expect two floats. If we
want a different style of Observer, we need to create a new Observer class and a new Subject to work
with it.
This Push version of the pattern is great because the classes are completely decoupled. The derived
classes don't need to know about each other at all. The price to pay for this decoupling is that we
need to write lots of Subject and Observer base classes for each method signature we want to use. An
alternative version of this pattern lets the Observers pull the data they want from the Subject. In the
Pull version, the Subject sends itself as a parameter in the Update method, and the Observer uses Getter
methods from the Subject to only pull the data it wants.
That is exactly what is happening in our next example. The PlayerHealthBar class now takes a pointer to
a Subject. In this case, we are expecting the Subject to be of type Player. The Update method can then
use any Player data it needs to complete its task:
//Now the update method takes a pointer to the subject
class Observer
{
public:
virtual ~Observer(void) {}
//Pull version of the Observer Pattern
virtual void Update(Subject* pSubject) = 0;
};
//Example of an Observer Update method that pulls data

void PlayerHealthBar::Update(Subject* pSubject)
{
//Make sure we have a Player
Player* pPlayer = dynamic_cast<Player*>(pSubject);
if(pPlayer == 0)
return;
m_percent = (pPlayer->GetHealth() / pPlayer->GetMaxHealth());

m_percent *= 100.f;
if (m_percent >= 75.0f)

m_color = "Green";
else if (m_percent < 75.0f && m_percent > 35.0f)
m_color = "Yellow";
else
m_color = "Red";
}
In the Pull version of the pattern, the Observer is dependent on the derived Subject class (in this case the
Player), but the Update method is more flexible. Additionally, this Observer could be observing many
different Subjects. The Update method could have a series of if statements to determine which of the
many possible Subjects performed the call. The classes are more tightly coupled to specific objects.
However, since an Observer can now observe multiple Subjects, the same Observer class can be used
for a wider range of objects. For example, a single Observer class could be used to keep track of how
many objects of each type in the game have died by registering itself with every game object that gets
created and monitoring the health of all Subjects.
Benefits of using the Observer pattern
At the start of this chapter we saw three problems with interacting gameplay code. As we said before,
these problems aren't that big, but they creep up all over the place and can lead to inflexible code as
the project moves forward. The Observer pattern solves these problems in a very simple way.
The biggest benefit of using the Observer pattern is that we can reduce dependency and coupling. By
using the Push version of the Observer pattern, our classes can interact completely through interfaces,
so they don't depend on each other at all. In the preceding example, the Player and Player Display are
completely decoupled. This means that changes to one won't affect the other. For starters, this makes
each class easier to test and debug because they can be worked on separately. However, this also
means that as the game changes, these classes can change independently. This means the individual
class can easily be reused within the current project or in separate projects. The only problem with
the Push version is that we are stuck with a single Update method signature.
Using the Pull version of the Observer pattern increases the dependencies; however, the
Subject/Observer system is much more flexible. An Observer can now listen to multiple Subjects,
and a Subject can share different data with each of its Observers. Even though the dependencies are
increased, this is only on the Observer side, since Subjects still don't need to know about their
Observers. These increased dependencies are still better than the alternative, because they are limited
to only the two classes that need to interact. Without using the Observer pattern, a third class would
need to be used to link these two classes together.
The second benefit of this pattern is that we no longer need to hardcode methods or requirements.
Since the event is broadcast to any object that registers, there is no need to recompile if a new object
needs the event. This reduces compile times as well as limits the chance of breaking code.
This isn't just limited to connecting systems to other systems. Since Observers can register and
unregister at runtime, game objects can register themselves with other game objects. Imagine an
Enemy Space Station that continuously spawns Raiders. By registering Raiders with the station that
spawned them, each station can act as a mini commander, easily coordinating units for attack and
defense.
The last benefit is that there is no need to poll objects. In the preceding example, we saw two while
loops--one using the Observer pattern, and one polling for data every frame. We saw how much
cleaner the code looked when using the Observer pattern. Besides just looking cleaner, the first
example is less likely to have bugs because once registered, the Observer will always receive the
updated data. In a larger project, the alternative requires that every object constantly asks for the data
all over the code base. Ensuring that every object that needs the data asks every frame in many
locations can be a Herculean task.
In addition to being very difficult to find and maintain each poll occurrence, this method leads to less
efficient code. We shouldn't try to prematurely optimize our code, and most likely this won't be the
bottleneck of our game, so that is why we mentioned it last. However, the fastest code is the code that
never runs. When using the Observer pattern, if the data never changes and the event never occurs, the
Update methods never get called. So, using the Observer method has the chance to boost our
performance compared to polling dozens or hundreds of times per frame.
Problems using the Observer pattern
Of course, the Observer pattern isn't a perfect solution. There is no perfect solution, if there was, this
would be a very short book. As with every pattern, it is important to learn how to use it but also how
not to use it. It is equally as important to learn when to use a pattern as it is to learn when not to use a
pattern.
The Observer pattern has some things that you need to look out for when using it. The goal of all
patterns is to simplify the development cycle. However, if a pattern isn't fully understood, lots of time
can be wasted debugging layers and layers of code that were meant to save time.
Dangling references
The first major problem that everyone should be aware of is the problem of dangling references. We
must ensure that the pointers held by our objects never become invalidated. The best way to
understand this problem is to look at the Command pattern and compare it with the Observer.
The biggest difference between the Command pattern and the Observer pattern is a difference in
ownership. In the Command pattern, the client or user of the command owns a pointer. This means that
the client is responsible for deleting the command. The reason this is important is because no other
class owns the command pointer, so it should never become invalidated. In the Mach5 Engine, the
UIButtonComponent owns a command. It deletes that command in the destructor or when it is given a new
command pointer.
Contrast this with the Observer pattern. The Subject contains a pointer to an Observer, and the
Observer can hold a pointer to the Subject. Neither object owns the other, and either object can be
deleted at any time. They are separate objects that just happen to communicate with each other via
pointers.
Before an Observer is deleted, it must be unregistered from every Subject it is observing. Otherwise,
the Subjects will keep trying to Update it via the invalid pointer, causing undefined behavior. In the
example, we do this by having the Observer hold a pointer to the Subject and unregistering itself in
the destructor. This causes a problem if the Subject gets deleted first. In the example, there is no way
for the Subject to tell the Observer that it will be deleted. If the Subject does get deleted first, the
Observers will still try to unregister in their own destructor, causing undefined behavior.
There are two solutions to this problem. First, we could ensure the Subject never gets deleted before
the Observers. Second, we could remove the Subject pointer from the Observer and ensure that the
Observer still gets unregistered from the correct Subject before it is deleted. Both solutions can be
difficult to guarantee. If you implement the Observer pattern in your game, you must consider this
problem.
Overuse
The second problem with the Observer pattern is the danger of overuse. We stated before that the
Observer pattern is a great choice when the requirements of a method aren't completely known or are
likely to change. While it is true that anything in a game can change, taking this to the extreme would
create a project that is impossible to program in and that will never be completed. In the most
extreme case, every object could be both a Subject and an Observer. Every method would be
ultimately flexible because we could change everything about it at runtime. Every call would notify
Observers, which in turn would notify more Observers.
Of course, no one would implement their engine to that extreme, but overuse of the Observer pattern
is still a problem. It can be difficult to debug a method that notifies a long list of Observers. This
becomes even worse if there are multiple layers to traverse before finding any real code. It can be
difficult to find the right balance. If you feel like your code is becoming a spider web of
interconnected pointers, it might be time to look for a better solution.
Implementing interfaces
The next thing to consider is the implementation of the Observer pattern. In the above example, the
Subject and the Observer were designed to be interface classes. C++ doesn't have a language feature
that supports interfaces but that doesn't mean we can't write code that uses them.
An interface is a class that contains no data members and has no method implementations. C++ does
allow for the concept of abstract base classes. This is any class that marks at least one method a pure
virtual, but using the = 0 notation. This means that the class can't be instantiated and instead must be
derived from. This is different from an interface because abstract base classes can contain member
data as well as implementations for methods (including pure virtual methods).
This distinction is important to us because C++ allows for multiple inheritance, although it is strongly
recommended that you do not use this feature. Inheriting from multiple base classes means that you
could be inheriting data members and methods with the same name. Since each of the parents could
also be inheriting from multiple classes, it can be difficult to fully understand the entire tree that any
derived class gets the implementations from. This will almost certainly lead to unexpected behavior.
That is why you should avoid multiple inheritances and instead stick with a single parent class that
your implementation derives from.
Interfaces are a little different. Many languages do not let you inherit from multiple classes that
contain implementations, but they do let you implement multiple interfaces. Since interfaces do not
contain method implementations or data members, there is no chance of having name clashes or
inheriting unexpected behavior. By implementing an interface, you are simply promising that your
class will respond to a method call. It is the responsibility of the derived class to write the behavior.
The Subject and Observer were written as interface classes in case they need to be used with Stages,
Components, or any other class that already needs to derive from a base class. Of course, instead of
using multiple inheritance, you could always find another solution. For example, you could create a
new class that derives from the Observer and have your component contain, as opposed to inherit,
that new class.
When to notify
The final thing to consider when using the Observer pattern is when to notify the Observers. The first
option is to notify the Observers after any data has changed. In the above example, the Player called
Notify within the AdjustHealth method. This means that the Observers will have the most up to date
information immediately. However, if the AdjustHealth method is called many times in one frame, time
is wasted updating the Observers many times. This problem becomes worse if the Observers were
observing more than just the Player's health.
The other option would be to notify the Observers after all data has changed. This solution is
obviously more efficient but a lot depends on how your system works. In the preceding example, it
could be possible to update all the Player information (in this case, just the health) first, then call
Notify before displaying anything.
In a game engine with many different Subjects and Observers that interact in different ways every
frame, it can be difficult to know when all data has been updated. There might not be a way to sort the
game objects such that all Subjects are updated before the Observers. It is possible that a HUD object
that is an Observer has already been updated before the Subject gets updated. This becomes even
more difficult since other systems such as physics can affect the data inside game objects.
In the end, there isn't a correct time to notify the Observers. It can be more efficient to only notify
them once. However, if the system is complex, it probably isn't worth the programming trouble trying
to batch all Notify calls together.
Summary
In this chapter, we focused on the communication between reusable engine code and game specific
gameplay code. We learned that it can be difficult to make a clear distinction between these two parts
because the gameplay code has the potential to creep into every system. This makes sense because to
make a game, you must write code that interacts with every other system. However, this means
reusing code is a little difficult.
We saw that the solution to this was to decouple our engine and gameplay code by having all
communication go through interface classes. These interface classes were the basis for what is known
as the Observer pattern. By using this pattern, we can make our code much cleaner and easier to use,
and reuse.
The Observer pattern is as easy as it gets when it comes to design patterns. Few patterns are as
simple to implement or understand. Once you start using it you will wonder how you ever
programmed without it. However, we also learned there are some things to watch out for when using
the pattern, the worst of which is dangling references.
Now that we have a great tool for separating gameplay from our engine, let's move onto another
problem related to code reuse. In the next chapter, we will talk about a pattern that allows us to reuse
objects and save memory. Along the way, we will learn about making great particle systems.
Sharing Objects with the Flyweight Pattern
We previously learned about object pools in Chapter 7, Improving Performance with Object Pools, and
that they are great for avoiding slowdowns in our game due to dynamic memory allocation. But, there
are still other steps that we can take to reduce the amount of memory that we use to begin with.
When creating projects, you'll often run into times where you want to have many objects on the screen
at once. While computers have become much more powerful over the past few years, they still can't
handle thousands of complex game objects on the screen by themselves.
In order to accomplish this feat, programmers need to think of ways to lighten the memory load on
their program. Using the Flyweight pattern, we abstract the common parts of our object and share
them with only the data that's unique to each instance (such as position and current health) being
created.
Chapter overview
In this chapter, we will construct a particle system consisting of two parts: the particle itself, which
will be a simple struct, as well as a particle system class that contains the system's data.
We will construct two different types of particle system: an explosion that moves on its own, and a
static one that spawns at the position of our player's ship. We will also explore two ways to deal with
the system data. The first will be for each particle system to contain its own copy of the system data.
Then, after learning about the Flyweight pattern, we will use it to construct separate system data
classes that we can assign using files or code. Then, each particle system will simply reference an
instance of the system data that it needs.
Your objectives
Introduction to particles
Introductions to particles
In game development, you may have heard of particles. They are typically small 2D sprites or simple
3D models that are created in order to simulate fuzzy things such as fires, explosions, and smoke
trails to add visual flair to your projects. This visual flair is sometimes referred to as juiciness. Made
popular by indie developers Martin Jonasson and Petri Purho, making a game juicy makes it more
enjoyable to play and increases the feedback the player receives by playing the game.
This is usually something worked on more toward the end of development of titles in order to polish
the project and add more feedback, but it's a good example of how we can want to have many things
on the screen at one time.
For more information on juiciness and to watch their Martin and Petri's GDC talk on
the subject, check out http://www.gamasutra.com/view/news/178938/Video_Is_your_game_juicy_enough
.php .
The reason that these objects are so simple is because they are spawned hundreds and sometimes
thousands of times, and this is done over and over again.
Now that we know what particles are, let's put them into Mach5 so we can get an example of how
they work. We will be creating particles to follow our ship while it moves in a similar fashion to a
smoke trail. This will be a great way to show an example of particles on the screen but, to have
something to show, we will first need to bring a new archetype into the game.
To do that, open up the Example Code folder for this chapter and bring the particle.tga file into the
EngineTest/Textures folder of your Visual Studio project.
After that, open up the EngineTest/ArcheTypes folder, create a new text file called Particle.ini, and fill it
with the following info:
posX = 0
posY = 0
velX = 0
velY = 0
scaleX = 2.5
scaleY = 2.5
rot = 0
rotVel = 0
components = GfxComponent ParticleComponent
[GfxComponent]
texture = particle.tga
drawSpace = world
After that, we need the Mach5 engine to support our new object, so go to the EngineTest folder and then
double-click on the PreBuild.bat file. The M5ArcheTypes.h file will be updated to include our particle:
//! AutoGenerated enum based on archetype ini file names
enum M5ArcheTypes {
AT_Bullet,
AT_Particle,
AT_Player,
AT_Raider,
AT_Splash,
AT_INVALID
};
Nice! Now that we have the object in the game, there's still the issue of putting in the Particle
component. Since this component is not exclusive to our game, let's move over to the
Core/Components filter and create a new filter called ParticleComp. From there, create two new files,
ParticleComponent.h and ParticleComponent.cpp, making sure their locations are set to the Mach5-
master\EngineTest\EngineTest\Source\ folder.
In the .h file, use the following code:

/******************************************************************************/
/*!
\file ParticleComponent.h
\author John Doran
\par email: [email protected]
\date 2016/12/06
Used to display a single particle on the screen.
*/
/******************************************************************************/
#ifndef PARTICLE_COMPONENT_H
#define PARTICLE_COMPONENT_H
#include "Core\M5Vec2.h"
class ParticleComponent : public M5Component

{
public:
ParticleComponent();
bool activated;
float lifeTime;
float endScale;
private:
M5Vec2 startScale;
float lifeLeft;
float Lerp(float start, float end, float fraction);
};
#endif // !PARTICLE_COMPONENT_H
This class looks similar to other components that we've added in the past, but this time we've added a
startScale property to keep track of what scale our object had at the start of its life, and an endScale
property to be a modifier on how to change the scale. We also have lifeTime, which will be how long
this object should live before we remove it, and lifeLeft, which will be how much longer this object
has to live. Finally, since we are going to change our scale, we added another function, Lerp, to
linearly interpolate between a starting and ending value.
In the .cpp file, use the following code:

/******************************************************************************/
/*!
\file ParticleComponent.cpp
\author John Doran
\par email: [email protected]
\date 2016/12/06
Particle system component. Allows you to draw many particles on the screen.
*/
/******************************************************************************/
#include "ParticleComponent.h"
#include "EngineTest\M5ObjectPool.h"
#include "Core\GfxComponent.h"
#include "Core\M5IniFile.h"
/******************************************************************************/
/*!
Construtor for ParticleSystem component. Sets default values
*/
/******************************************************************************/
ParticleComponent::ParticleComponent() :
M5Component(CT_ParticleComponent)
{
}
/******************************************************************************/
/*!
Takes care of the particle system, decrease lifetime and adjust scaling.
Will mark for destruction if needed.
\param [in] dt
The time in seconds since the last frame.
*/
/******************************************************************************/
void ParticleComponent::Update(float dt)
{
// Decrease our life by the change in time this frame
// (dt stands for delta time)
lifeLeft -= dt;
// Change our size based on where we want it to be

float currentPercentage = 1 - (lifeLeft / lifeTime);
m_pObj->scale.x = Lerp(startScale.x,
startScale.x * endScale, currentPercentage);
m_pObj->scale.y = Lerp(startScale.y,
startScale.y * endScale, currentPercentage);
// If there is no life left, destroy our object

if (lifeLeft <= 0)
{
}
This code will modify the object's scale by using the Lerp function to interpolate between the starting
and ending scale. We also will modify how much life the particle has left, and if it has none, mark the
particle for deletion:
/******************************************************************************/
/*!
Will give you the percentage of the fraction from start to end
\param [in] start

What value to start from
\param [in] end

What value to end from
\param [in] fraction

What percentage of the way are we are from start to finish
*/
/******************************************************************************/
float ParticleComponent::Lerp(float start, float end, float fraction)
{
return start + fraction * (end - start);
}
Linear interpolation (Lerp) allows us to obtain a value between start and end using the fraction
property for how far along the transition it should be. If fraction is 0, we would get the value of start. If
we give 1, we will get the value of end. If it's .5, then we would get the half-way point between start
and end.
For more information on interpolation including linear interpolation, check out

Keith Maggio's notes on the topic at https://keithmaggio.wordpress.com/2011/02/15/math-magician-
lerp-slerp-and-nlerp/.
/******************************************************************************/
/*!
Clones the current component and updates it with the correct information.
\return
A new component that is a clone of this one
*/
/******************************************************************************/
M5Component * ParticleComponent::Clone(void)
{
ParticleComponent * pNew = new ParticleComponent;
pNew->startScale = m_pObj->scale;
pNew->lifeTime = lifeTime;
pNew->lifeLeft = lifeTime;
pNew->endScale = endScale;
return pNew;
}
The Clone function allows us to create a copy of this object. It will create a new version of this
component, and we will initialize the values of the new component with the values we currently have.
This is used by the Mach5 engine in the creation of new game objects:
/******************************************************************************/
/*!
Reads in data from a preloaded ini file.
\param [in] iniFile

The preloaded inifile to read from.
*/
/******************************************************************************/
void ParticleComponent::FromFile(M5IniFile& iniFile)
{
// Get our life time value
std::string lifeTimeText;
iniFile.SetToSection("ParticleComponent");
iniFile.GetValue("lifeTime", lifeTimeText);
// Convert the string into a float

lifeTime = std::stof(lifeTimeText);
lifeLeft = lifeTime;
// Then do the same for endScale

std::string endScaleText;
iniFile.GetValue("endScale", endScaleText);
endScale = std::stof(endScaleText);
Just like before, the FromFile function will read in our ini file we created previously and will use the
values from it to set the properties of this component. In our case, here we set lifeTime, lifeLeft, and
endScale.
Finally, let's start putting these objects into our game. Open up the PlayerInputComponent.cpp file and add
the following to the top of the Update function:
M5Object* particle = M5ObjectManager::CreateObject(AT_Particle);
particle->pos = m_pObj->pos;
This will cause a particle to get spawned in every single frame and have the same position as our
ship. Now, if we run the game, we should see some cool stuff! We can see this in the following
screenshot:
As you can see, our ship now has a trail following behind it. Each part is a particle!
The particle system that we are currently showing is probably running well enough on some
computers, but note that a large number of the variables that we have created hold data that will never
change once we've initialized them. Now, generally in programming we would mark a variable that
wouldn't change as const, but we don't set the variable until we read from a file. We could potentially
make the variables static, but there's also the chance that we may want to have more particle systems
in the future and I don't want to create an archetype for each one.
If we continue to spawn many particles, the memory that it takes up will increase and we will be
wasting valuable space in memory that we could be using for other purposes. To solve this issue, we
will employ the Flyweight pattern.
The Gang of Four states that a Flyweight is a shared object that can be used in multiple contexts
simultaneously. Similarly to flyweight in boxing, which is the lightweight boxing category, we can
have a lighter object that can be used in different places in our system simultaneously.
While not used terribly often nowadays, the Flyweight pattern can be very helpful in scenarios when
memory is constrained.
A Flyweight will consist of two parts: the intrinsic state and the extrinsic state. The intrinsic state is
the part that can be shared. The extrinsic state is modified based on the context it's being used in and,
as such, cannot be shared.
Let's take a look at a UML diagram to see a closer look:
We have the FlyweightFactory class, which is used to manage the Flyweights. Whenever we request
one, we will either give one that's been created or create a new one ourselves.
The Flyweight object itself has data that is of whatever type is needed, as long as it won't change
depending on the object that we're working with.
Finally, we have the ConcreteFlyweight, which acts as our extrinsic information that can access and
use our Flyweight via the FlyweightFactory.
So with that in mind, what we will do is separate the information that will be shared by each particle,
which we will call a ParticleSystem:
// Abstract class for us to derive from
class ParticleSystem
{
public:
float lifeTime;
M5Vec2 startScale;
float endScale;
// Pure virtual functions

virtual void Init(M5Object * object) = 0;
virtual void Update(M5Object * object, float dt, float lifeLeft) = 0;
float Lerp(float start, float end, float fraction);
};
The class acts as our intrinsic state, which is shared. Since the starting scale, end scale, and lifetime
of our object never change, it makes sense for these variables to be shared instead of each object
having one. In our previous example, we only had one particle system, but we may want the ability to
have more as well, and it's when we start using it that some of the benefits of the Flyweight pattern
become even more apparent. That's why we gave this class two virtual functions: Init and Update. We
can have our extrinsic state call these functions, giving the function information about the particular
object we're dealing with, and then we can modify it using these properties.
Creating different system types
Let's add a new type of particle system in addition to our current one that doesn't move. Let's call it
Moving and our previous one, Static. To differentiate between the two, let's add an enum:
enum ParticleType
{
PS_Static,
PS_Moving
};
We can now modify the original ParticleComponent class, by removing the previously created variables
and instead including a reference to the kind of ParticleSystem we wish to use:
class ParticleComponent : public M5Component
{
public:
ParticleComponent();
bool activated;
float lifeLeft;
private:
ParticleType particleType;
};
The ParticleComponent class acts as our extrinsic state, holding information about how much time it has
left and the properties from the M5Component class, such as a reference to the object we want to create.
At this point, we need to create two classes to refer to each of these:

class StaticParticleSystem : public ParticleSystem
{
void Init(M5Object * obj);
void Update(M5Object *, float, float);
};
class MovingParticleSystem : public ParticleSystem

{
void Init(M5Object * obj);
void Update(M5Object *, float, float);
};
Developing the ParticleFactory
We need some way for our ParticleComponent to access this information. With that in mind, we will make
use of the Factory design pattern that we learned about in Chapter 5, Decoupling Code via the Factory
Method Pattern, and create a ParticleFactory class:
class ParticleFactory
{
public:
static int objectCount;
static std::map<ParticleType, ParticleSystem *> particleSystems;
// Getting our Flyweight

static ParticleSystem & GetParticleSystem(ParticleType type);
~ParticleFactory();
};
This ParticleFactory class is what we use to manage the creation of these Flyweights and to ensure that,
if the object is already located in our map, we will return it. Otherwise, we will create a new object
to be able to access it. I also added an objectCount variable to help us know how many objects
currently exist and to verify that no memory leaks are occurring.
The ParticleSystems variable is of type map, which is actually one of my favorite containers in the stl
and can be considered an associative array. By that, I mean instead of memorizing numbers in order
to access certain indexes of an array, you can use a different type, such as a string, or in this case, an
enum.
For more information on the map container, check out http://www.cprogramming.com/tutorial/s

tl/stlmap.html.
After this, we will need to define the two static variables:

#include <map>
// Define our static variables

int ParticleFactory::objectCount = 0;
std::map<ParticleType, ParticleSystem *> ParticleFactory::particleSystems;
Using the ParticleFactory
Next, we will need to adjust our previously created Particle archetype and component to reflect these
changes.
First, we want to change our .ini file. Since the Particle object is meant for all particle types, instead
of having the properties being set there, we will instead set a base type for us to use:
posX = 0
posY = 0
velX = 0
velY = 0
scaleX = 2.5
scaleY = 2.5
rot = 0
rotVel = 0
components = GfxComponent ParticleComponent
[GfxComponent]
texture = particle.tga
drawSpace = world
[ParticleComponent]
type = Moving
This simplifies the particle object itself, but it's for a good cause. We will now update the code of the
ParticleComponent class as follows:
/******************************************************************************/
/*!
Construtor for ParticleSystem component. Sets default values
*/
/******************************************************************************/
ParticleComponent::ParticleComponent() :
M5Component(CT_ParticleComponent)
{
}
/******************************************************************************/
/*!
Takes care of the particle system, decrease lifetime and adjust scaling.
Will mark for destruction if needed.
\param [in] dt
The time in seconds since the last frame.
*/
/******************************************************************************/
void ParticleComponent::Update(float dt)
{
// Decrease our life by the change in time this frame (delta time, dt)
lifeLeft -= dt;
ParticleFactory::GetParticleSystem(particleType).Update(m_pObj, dt, lifeLeft);
// If there is no life left, destroy our object

if (lifeLeft <= 0)
{
}
In this instance, you'll notice that instead of modifying the scale and/or movement being done here, we
use the ParticleFactory to update our code based on the particleType property:
/******************************************************************************/
/*!
Clones the current component and updates it with the correct information.
\return
A new component that is a clone of this one
*/
/******************************************************************************/
M5Component * ParticleComponent::Clone(void)
{
ParticleComponent * pNew = new ParticleComponent;
pNew->particleType = particleType;
ParticleSystem & system =

ParticleFactory::GetParticleSystem(particleType);
system.Init(pNew->m_pObj);
pNew->lifeLeft = system.lifeTime;
return pNew;
}
Here, we call the Init function for our particle system based on its type from the factory:
/******************************************************************************/
/*!
Reads in data from a preloaded ini file.
\param [in] iniFile

The preloaded inifile to read from.
*/
/******************************************************************************/
void ParticleComponent::FromFile(M5IniFile& iniFile)
{
// Get our initial particle type
std::string particleTypeText;
iniFile.SetToSection("ParticleComponent");
iniFile.GetValue("type", particleTypeText);
if (particleTypeText == "Static")
{
particleType = PS_Static;
}
else if(particleTypeText == "Moving")
{
particleType = PS_Moving;
}
We are now going to set our particle type based on what is marked on the ini file.
But, of course, now that we are using the GetParticleSystem function, we need to implement it for our
code to compile:
/******************************************************************************/
/*!
Used to get our Flyweight object and access the shared properties of the
particles.
\param type
What kind of particle we want to get access to
*/
/******************************************************************************/
ParticleSystem & ParticleFactory::GetParticleSystem(ParticleType type)
{
// If our object exists, return it
if (particleSystems.find(type) != particleSystems.end())
{
return *particleSystems[type];
}
ParticleSystem * newSystem = nullptr;
// Otherwise, let's create one

switch (type)
{
case PS_Static:
newSystem = new StaticParticleSystem();
newSystem->endScale = 0;
newSystem->lifeTime = 1.5;
newSystem->startScale = M5Vec2(2.5, 2.5);
particleSystems[PS_Static] = newSystem;
objectCount++;
break;
case PS_Moving:
newSystem = new MovingParticleSystem();
newSystem->endScale = 0;
newSystem->lifeTime = 1.5;
newSystem->startScale = M5Vec2(2.5, 2.5);
particleSystems[PS_Moving] = newSystem;
objectCount++;
break;
}
return *newSystem;
In this script, we make use of the particleSystems map that we talked about earlier. The first thing that
we do is check if there is an object in the map that has our ParticleType in it. If not, then we need to
create one. In this case, I added a switch statement that will assign different values depending on the
value mentioned in the case statement, but you could easily read these values from a text file in a
similar manner to how files are read for archetypes. You'll notice that we are calling new in order to
create these, so we will need to call delete on them as well in order to avoid any memory leaks. To
accomplish this, I've added in a destructor for the ParticleFactory class:
/******************************************************************************/
/*!
Deconstructor for the ParticleFactory. Removes all of the elements in ourparticleSystems map
*/
/******************************************************************************/
ParticleFactory::~ParticleFactory()
{
for (auto iterator = particleSystems.begin();
iterator != particleSystems.end();
iterator++)
{
// iterator->first = key
// iterator->second = value
delete iterator->second;
}
Finally, we need to write the implementations for our different ParticleSystems:

/******************************************************************************/
/*!
Will give you the percentage of the fraction from start to end
\param start
What value to start from
\param end
What value to end from
\param fraction
What percentage of the way we are from start to finish
*/
/******************************************************************************/
float ParticleSystem::Lerp(float start, float end, float fraction)
{
return start + fraction * (end - start);
}
The Lerp function does the same for either particle type, so it's fine the way it was:
/******************************************************************************/
/*!
Used to initialize the particle system and set any parameters needed
\param obj
A reference to the object
*/
/******************************************************************************/
void StaticParticleSystem::Init(M5Object * obj)
{
obj->vel = M5Vec2(0, 0);
}
/******************************************************************************/
/*!
Used to update the particle system. Called once per frame
\param m_pObj
\param dt
Amount of time that has passed since the previous frame
\param lifeLeft
The amount of lifetime the object has left
*/
/******************************************************************************/
void StaticParticleSystem::Update(M5Object * m_pObj,
float /*dt*/, float lifeLeft)
{
}
The static version of the Init and Update functions will just set our velocity to 0 so we don't move:
/******************************************************************************/
/*!
Used to initialize the particle system and set any parameters needed
\param obj
*/
/******************************************************************************/
void MovingParticleSystem::Init(M5Object * obj)
{
obj->vel = M5Vec2(M5Random::GetFloat(-1, 1),
M5Random::GetFloat(-1, 1)) * 10;
}
/******************************************************************************/
/*!
Used to update the particle system. Called once per frame
\param m_pObj
\param dt
Amount of time that has passed since the previous frame
\param lifeLeft
The amount of lifetime the object has left
*/
/******************************************************************************/
void MovingParticleSystem::Update(M5Object * m_pObj, float /*dt*/, float lifeLeft)
{
}
For our moving particle system, we will set our velocity to a random number in the x and y axis,
causing a nice explosion effect!
Now, instead of creating a copy of this data each time, we will have one copy that we will access, as
shown in the following screenshot:
As we play, you'll notice that we now have a new particle system working and it's doing its job quite
well.
Summary
Over the course of this chapter, we learned about particles and how they can be used in order to
improve the polish of our game project. We learned how we can implement a particle system inside
of the Mach5 engine, and then learned about the Flyweight pattern and how it can be used effectively
in order to reduce the memory usage on your projects. We saw how to do this by making use of the
Factory pattern too, while making it a lot easier for us to create new particle system types as well.
Keeping this in mind, it will be a lot easier in the future to break apart pieces of your programs that
stay consistent and only create additional variables when you need to!
Moving forward, in the next chapter we will dive into graphics and the concepts needed to understand
how our code will affect moving and animating game objects.
Understanding Graphics and Animation
Over the last 10 chapters, we have dived deeply into some of the most popular design patterns. The
goal of each chapter was to understand and solve some common problems that everyone encounters
when creating games. Along the way, we have created component-based game objects with flexible
State-based, decision-making capabilities. We have created core engines such as the StageManager and
ObjectManager using the Singleton pattern, so that communication between game objects, components,
and engines is incredibly simple. We also looked at Object Pools and the Flyweight pattern, which
allow our game to use memory more efficiently.
In this chapter, we will focus on graphics. However, we will not be focusing on how to implement a
graphics engine. That would require more than a single chapter. Instead we will focus on concepts
that need to be understood, regardless of which graphics Application Programming Interface (API)
you use.
Graphics is a large part of any game engine and is very likely the performance bottleneck of the game.
However, whether we are using DirectX, OpenGL, or some other graphics API, we must understand
what is happening behind the scenes. We shouldn't fall into the trap of thinking that just because we
didn't write the graphics API, there are no design decisions to be made.
Chapter overview
This chapter is a little different than the ones before because it isn't focused on a design pattern.
Instead we will focus on the low-level details of graphics so we can better understand how our code
affects moving and animating our game objects.
First, we will look at how a computer monitor works. We will dive into the details of pixels and
screen resolutions. We will look at how pixels are drawn on screen as well as understand the concept
of tearing, learn why we hear so much about frames per second (fps), and why games try to achieve
30 or 60 frames per second.
Next, we will look at timing in games. We will learn why we want a consistent frame rate. We will
also look at what happens when our frame rate isn't consistent and how we can ensure that our frame
time stays consistent throughout the game.
Your objectives
Learn how computer monitors work and what a refresh rate is
Learn about double buffering and why it is used for graphics
Learn about time-based movement and animation, and why we want a consistent frame rate
Introduction to monitor refresh rates
These days, flat screen Liquid Crystal Display (LCD) monitors are very common. However, to
understand refresh rates and double buffering, we need to understand how older monitors display an
image. Along the way, we will learn about common graphics terms, such as pixels and screen
resolution:
Figure 11.1 - Simplified cathode ray tube diagram
Cathode Ray Tube (CRT) monitors contain screens with millions of tiny red, green, and blue
phosphor dots. These dots glow for a short time when struck by an electron beam that travels across
the screen to create an image. The cathode is a heated filament inside a vacuum sealed glass tube.
The ray is a stream of electrons generated by an electron gun, which is directed by magnetic
deflection plates. By adjusting the magnetic field of the plates, the electron beam can be moved
around and adjusted to strike every part of the screen.
The screen is coated with phosphor, an organic material that glows for a short time when struck by the
electrons. It contains many groups of red, green, and blue dots. By varying the intensity of the electron
ray on each dot, different colors can be generated. For example, with red, green, and blue fired at
maximum strength, the color white is produced.
Since one color is produced by a group of red, green, and blue dots, the maximum number of these dot
groups limits how many colors can be displayed horizontally or vertically. The diagonal distance
between two dots of the same color is known as the dot pitch.
Figure 11.2 Close of view of the Pixel on the screen

What is a pixel?
A pixel, or picture element, is the basic unit of programmable color on a computer image or display.
It is better to think about a pixel as a logical unit, rather than a physical unit. This is because the size
of a pixel depends on the current resolution of the display screen. At the maximum resolution of the
screen, one pixel maps exactly to one dot group. This means the size of a pixel at maximum resolution
is equal to the dot pitch. Smaller resolutions will use more than one dot group to create a single color.
The resolution of the screen is the number of horizontal pixels multiplied by the number of vertical
pixels, and is usually written as width x height. For example, a 640 x 480 resolution means that the
screen is 640 pixels wide and 480 pixels high for a total of 307,200 pixels. Of course, the color data
for each pixel must be stored in computer memory, so a higher resolution uses more pixels and more
memory. For example, if each pixel used one byte of memory, our 640 x 480 display would need 300
kilobytes of memory. A 1280 x 1024 display would need 1.25 megabytes. Let's have a look at the
following screenshot:
Figure 11.3 - An example of 800 x 600 screen resolution
The location in RAM that stores pixel color information is called the framebuffer. The framebuffer
is written by a program and then transmitted to the monitor. The cathode ray interprets the pixel color
and fires the electron ray at the proper intensity. The deflection plates direct the electron beam to the
proper dot groups on the phosphor screen.
In the examples above, the size of each pixel was only 1 byte. However, a pixel can be, and usually
is, more than 1 byte. As computers get faster and memory gets cheaper, we can use more bits per
pixel. With 8-bit color, red and green each use 3 bits or a total of 8 levels of color each, while blue
only uses 2 bits or 4 levels. This totals 256 possible colors for each pixel.
16-bit color, or high color, offers a few different options. One possibility is 4 bits for each red, green,
and blue. These 4 bits provide 16 levels for each color for a total of 4,096 (16 x 16 x 16) colors,
with an optional 4 bits for transparency. Another possibility is 5 bits per color and 1 bit for
transparency for a total of 32,768 colors. Finally, a total of 65,536 can be achieved with 5 bits for
both red and blue, and 6 bits for green.
True color is defined at 8 bits per color. This means that red, green, and blue all have 8 bits, or 256
possible color levels. If 24 bits are used, we get a total of 16,777,216 possible colors. These days,
32 bits per pixel are often used. The last 8 bits are for transparency. The transparency allows for
different levels of blending with the background colors. This allows for a total of 4,294,967,295
colors per pixel.
The size of the framebuffer is calculated by multiplying the resolution and the numbers of bytes per
pixels (color depth). For a game using a 1280 x 1024 display, we need 1280 x 1024 x 4 bytes, or 5
megabytes for the framebuffer. This might not seem like much considering that modern computers
often have 8 to 12 gigabytes of RAM. However, it is worth remembering that if we are updating every
pixel on screen, we are updating 1,310,720 pixels, or 5,242,880 bytes of data. This is assuming we
only fill in each color one time and don't need to blend with overlapping colors.
The horizontal and vertical blank
The display is updated by reading data from the framebuffer and updating in sequence, starting from
the left to right and top to bottom. We can think about this exactly as if we were iterating through a 2D
array in C or C++. At the end of each scan line, the electron gun is adjusted to point at the start of the
next scan line. Another adjustment must be made after the last pixel is lit so the beam can start again
at the top.
The time it takes for the electron gun to move from the rightmost pixel of scan line X to the leftmost
pixel of scan line X + 1 is called the Horizontal Blank Interval. This is because the electron gun is
blanked, meaning that it is outputting zero electrons during this interval. This is to prevent pixels from
being lit up while traveling from scan line to scan line. Similarly, the time it takes for the electron gun
to move from the end of the last scan line back up to the first scan line is called the Vertical Blank
Interval. Again, the electron gun is blanked to prevent pixels from being lit up while traveling back
to the top scan line. The Vertical Blank Interval is a short period where the entire display has been
updated and the framebuffer is not currently being read by the display:
//Example code drawing 640x480 display
//Including H Blank Interval and V-Blank Interval
for(int h = 0; h < 480; ++h)
{
for(int w = 0; w < 640; ++w)
{
//Sets pixel color and moves to next pixel
SetPixel(framebuffer[h][w]);
}
//Resets to start of scan line and moves down one row
ResetHorizontal();
}
//Resets to first pixel of first scan line
ResetVerticle();
Figure 11.4-Showing the movement pattern of the electron gun including the Horizontal and Vertical Blank Intervals
The phosphor dots on the screen are only lit up for a short period, so the electron gun must constantly
relight them. The gun moves from left to right and top to bottom many times per second to refresh each
pixel and display the correct image. If this process is too slow, the display will appear to flicker.
Refresh rate
The number of times per second a monitor refreshes the display is known as it's vertical refresh rate,
or just refresh rate. The refresh rate of a monitor is measured in hertz (Hz). So, a monitor that can
refresh the display 30 times per second has a refresh rate of 30 Hz. Many monitors refresh at 60 Hz,
however, it is becoming common to see monitors with refresh rates of 120 Hz or even 240 Hz.
It is important to realize that the refresh rate of a monitor has nothing to do with the performance of a
game or program. Having a monitor with a higher refresh rate will not improve the frame rate of a
game, unless the game itself can support higher frame rates. The number of times a program updates
the framebuffer is measured in frames per second or fps, and is completely independent from the
number of times the monitor refreshes. When these two numbers are out-of-sync, the display will not
look correct. Let's look at the problem of when the frames per second is less than the refresh rate:
Figure 11.5 Comparing FPS versus Refresh Rate
These days, TV shows and movies are often displayed at 24 frames per second while the TV sets and
monitors have a typical refresh rate of 60 Hz. In Figure 11.5, we have taken one second and split it
into 60 red bars on the top to represent our refresh rate and 24 blue bars on the bottom to represent
the frames of a movie. Each red bar represents the 1/60th of a second that the frame will be on screen.
Each vertical dark red line represents when the monitor is refreshed.
As you can see, the refresh periods do not line up properly with the frames of the movie. In Figure
11.5 we can clearly see that Frame 1 is on screen for 3/60th (or 1/20th) of a second, while Frame 2
is only on screen for 2/60th (or 1/30th) of a second. Frame 1 is on screen longer than the original
1/24th of a second and Frame 2 is on screen less. Since the frames are not on screen for an equal
amount of time, the video seems jittery. While this may not seem like much of a difference, some
people are sensitive enough to notice a slight speeding up and slowing down effect:
Figure 11.6 - Example of refresh rate and frames per second Out-Of-Sync
TVs and movies have ways of adjusting for this problem, such as interpolating between frames. Since
Frame 1 and Frame 2 are already known, it is easy to blend, interpolate, and generate intermediate
frames before displaying them. However, that doesn't help us when our games are out-of-sync with
the monitor.
Since games are interactive, the player's actions determine what will be displayed next on screen.
The details of Frame 2 are determined by player input in response to Frame 1. This means that the
contents of the framebuffer must be generated each frame, it can't be known ahead of time.
In the above example, we were simplifying a little by pretending that writing to the framebuffer was
instantaneous. Of course, this isn't true. Even though the next frame of a movie is already known, it
still takes time to write to the framebuffer. Unless we can copy to the framebuffer completely during
the Vertical Blank Interval, we will need to write to the framebuffer while the display is being drawn
on screen.
If we can time everything correctly, we can always write a pixel just after the electron gun has read it.
However, if we are out-of-sync with the electron gun, eventually there will be a point when we
haven't written the pixel for the current frame and the electron gun reads the old value. Let's look at
this problem up close:
Figure 11.7 - Start position (Left) and End position (Right) in the framebuffer
Figure 11.7 shows what we would like to see on screen. The left image is the position of a game
object in frame 1. The right image is the ending position after the object has moved. These are two
discrete points in time that we want to show. However, if the display is reading the pixels as we are
writing them to the framebuffer, we can see a tearing effect if we haven't finished writing the current
frame:
Figure 11.8 - Example of tearing in the framebuffer
As we can see in Figure 11.8, the first image is correct. We are writing to a pixel after the display has
read it. In the second image, the pixel being read has almost caught up to the pixel that is being
written. At this point, the framebuffer contains half of each image. The third image shows that the
display has overtaken the write pixel. The correct pixels have been written but they were written too
late. The fourth image shows what the user would see on screen. Since the display read faster than the
pixels could be written, the image looks as if it has been torn in half. This effect is known as tearing.
Tearing occurs when our frames per second and our refresh rate are out-of-sync. Unfortunately, it can
be very difficult to get these two values perfectly aligned, and being a little off will cause some
tearing. To solve this problem, we need to write out an entire frame of pixels before the display reads
them.
Double buffering
The solution to our read/write problem is double buffering. Double buffering is exactly what it
sounds like. Instead of using only one framebuffer, we will use two: one for reading and one for
writing. Of course, since we now have two framebuffers, we need twice the memory. For a 1280 x
1024 display using 4 bytes per pixel, we need 5 megabytes per framebuffer for a total of 10
megabytes.
Everything up to this point could have been implemented in software by using operating system
commands. However, as displays started requiring more memory and more complex images, special
hardware was created. Modern graphics cards can contain gigabytes of memory used for
framebuffers, textures, 3D triangle meshes, and much more. They can also contain hundreds or even
thousands of cores to transform 3D points into pixel data simultaneously.
It is important to understand this because, as a programmer, you don't need to implement double
buffering yourself. It is implemented at a hardware level and our games will be double buffered
automatically by using a 3D graphics API such as DirectX or OpenGL. However, we still need to
understand how this process works so we can understand the implications to our frame rate.
The back buffer
As we said, double buffering works by using two framebuffers so that we are never setting pixels on
the same buffer that is being used for the display. The framebuffer that is currently being displayed is
called the front buffer or primary buffer, and the framebuffer that we are drawing to is called the back
buffer or secondary buffer.
When drawing is completed on the back buffer, the buffers are swapped so that the back buffer is now
the front buffer and the front buffer is now the back buffer. Of course, the buffers themselves are not
swapped. Instead, pointers to the buffers are swapped. The monitor has a pointer to one buffer and is
currently reading it. The graphics card has a pointer to another buffer that it uses for all draw
operations. This pointer swap, or page flipping as it is sometimes called, is much faster than copying
data from the back buffer to the front buffer:
Figure 11.9 - Example of double buffering
As you can see in the next two images, the display can read the front buffer while the graphics
processing unit or GPU draws to the back buffer. However, there is nothing about double buffering
that has prevented tearing. We have prevented drawing to the same buffer that is being displayed but,
if we swap buffers in the middle of a screen refresh, we will still have tearing. This is an important
first step in solving the tearing problem. However, before we solve it, let's talk about what happens
when we are out-of-sync with the monitor refresh rate because we are generating too many frames per
second:
Figure 11.10 - Double buffering after the page flip

Imagine the situation in which our game is generating 90 frames per second. This is 30 frames more
than we need to achieve 1 frame per refresh. This means we are wasting time creating frames that
will never be seen. At 90 frames per second, one third of the frames will never be seen by the player.
As you can see in Figure 11.11, every third frame, which is highlighted in green, will be skipped
because it falls between two refresh intervals. Even more frames will be skipped if we are updating
at 120 fps or more:
Figure 11.11 - Comparing 60 Hz refresh with 90 fps
At 90 frames per second, every frame takes 1/90th of a second to complete. From the player's
perspective, the third frame that gets shown (which is really our fourth frame) has been updated for a
total of 1/45th of a second, or double the amount of time. Just like with the movie at 24 fps, this can
cause a jittery effect that some players may notice because all objects will appear to have moved
twice as far.
When updating this many times per second, these time slices are very small. It is possible that they
are so small that they may go unnoticed by the player. The real problem with skipping frames is that
our game is doing work that it just doesn't need to. Since the frame will never be seen by the player,
there is no point in wasting game time generating it.
It is worth pointing out that only the graphics part of the update is wasted. It is perfectly fine to update
input, AI, or physics faster than the monitor can refresh. In fact, physics will be more accurate if the
time step is smaller. We just want to really emphasize that there is absolutely no reason to draw
frames faster than the monitor's refresh rate.
For most games this is never a problem. The usual solution to having a high frame rate is to do more
work and make your game more interesting. If your game is too fast, draw more particles, have more
enemies, calculate more accurate physics, or just do more physics effects. If you find that your game
runs faster than 60 fps, your game probably isn't the best it could be. Again, there is no reason to draw
more frames per second than can appear on screen.
VSync
So now we have seen both situations where our frame count is out-of-sync with our refresh rate. In
both cases we can have tearing, and in the case of a very high frame rate, we are wasting CPU/GPU
cycles that could be used to improve our game. We want to avoid tearing. We want to be in-sync with
the monitor refresh. How can we solve this problem?
The solution is the Vertical Blank Interval. Remember that the Vertical Blank Interval is when the
electron gun is repositioning itself from the last pixel on the display back to the first pixel. During this
short period, the entire display has been drawn and the front buffer isn't being used. This period is too
short to copy all the contents of the back buffer to the front buffer. However, it is long enough to swap
the pointers, which is how the page flipping mechanism works.
Modern computer monitors and TVs can send a signal back to the computer when the Vertical Blank,
Interval, or V-Blank occurs. The graphics card waits until a V-Blank before swapping the buffers after
the back buffer is completely drawn. This guarantees that there will never be tearing because parts of
different frames can never be read in a single refresh.
As we said, double buffering is implemented at the hardware level. This is also true of synchronizing
with the Vertical Blank Interval. By using a 3D graphics API such as DirectX or OpenGL, you get this
for free. This V-Blank syncing or VSync is an option that must be enabled when initializing graphics.
Once VSync is turned on, we don't need to worry about tearing. Additionally, the game will never
generate more than one frame per monitor refresh because the graphics card will always wait for the
V-Blank signal before swapping the buffers.
Syncing with the refresh rate of the monitor is a great way to prevent tearing. If a game can update 60
or more times a second, the back buffer and the front buffer will always be able to swap and we will
have a smooth, tear-free game. However, we haven't talked about what happens if the back buffer isn't
ready to be swapped because the frame is taking longer than 1/60th of a second to complete.
It is important to understand that the front and back buffers do not swap automatically every time the
V-Blank signal arrives. Instead, the swap is performed at the request of the programmer. A function
call is made to tell the graphics card that a new frame is prepared and the buffers should be swapped.
This swap will happen immediately if VSync is turned off. However, if the VSync is turned on, the
graphics card will wait for the V-Blank signal to arrive, no matter how long that is. If our update is
just 1/100th slower, meaning a frame takes 1/59th of a second to complete, we will miss the V-Blank
and need to wait for the next one.
Since the current frame wasn't ready, the monitor displays the previous frame again. The same frame
will be on screen for 1/30th of a second. Since the back buffer must wait to be swapped until the next
V-Blank, our game can't start working on the next frame. Our game is idle while waiting for the V-
Blank. This means that if our game is using VSync and it can't achieve 60 frames per second, our
frame rate will drop down to 30 frames per second. If our game can't achieve 30 frames per second,
our frame rate will take 3/60th of a second, or 20 fps.
For some programmers and some games, achieving 30 frames per second is perfectly fine. To
implement more beautiful effects or more accurate physics, dropping to 30 frames per second might
be an important trade off. Everyone must decide for their own game what is right. However, many
players simply do not like 30 frames per second. Players often say they can notice more jittery
movement and more importantly, they notice input lag.
Remember, if we can't achieve our goal of 60 fps, the graphics card must wait for the next V-Blank
before returning from the swap call. This means that our game can't process physics, AI, or even
input. The player is seeing half as many frames on screen, which means objects are moving more each
frame. Additionally, input is now being gathered from the player once every 1/30th of a second
instead of once every 1/60th of a second. While this may not seem like much, for fast, twitch response
games such as First Person Shooters, this can be too long.
Figure 11.12 shows an example of the contents of both the front and back buffers in a VSync scenario
where a game can't update at the same rate as the monitor. The display refreshes every 1/60th of a
second or every 0.0167 seconds. The game can update every 1/50th of a second or every 0.02
seconds. In the image below, the monitor refreshes are colored red or green. The red refreshes are
when the game frame isn't ready and so the previous frame is displayed. The green refreshes are when
the game frame is ready and so the buffers are swapped.
The blue represents when the game frame is completed. This doesn't mean that the new frame is
instantly displayed. This is because the graphics card waits until the next refresh to swap buffers. It is
important to understand that the game doesn't update every 1/50th of a second for the same reason.
Instead, each game update is 1/50th of a second after the last buffer swap:
Figure 11.12 - Showing contents of back buffer and front buffer when using VSync
Triple buffering
Turning on VSync in our games can improve the look of our graphics because we are guaranteed that
tearing will never occur. Unfortunately, if our game frame isn't completed in time for the next refresh,
the graphics card waits until the next V-Blank to swap buffers. This is true even if our game misses
the refresh by only 1/100th of a second. If our frame is off by this short amount, our fps drops to 30.
This is because the content of the back buffer hasn't been swapped yet, so we can't start drawing the
next frame.
It would be nice if we could start drawing the next frame while still waiting for the V-Blank signal.
To do this, we would need an extra framebuffer to draw to while we are waiting. This is exactly how
triple buffering works.
For triple buffering we have a total of three framebuffers. For a 1280 x 1024 display with 4 bytes per
pixel we would need a total of 15 megabytes. However, by using this extra memory, we will always
have a framebuffer to draw to, so we should always be able to hit our fps goal.
In triple buffering, we have our primary and secondary buffers just like before but now we have a
tertiary buffer as well. We start out drawing to the back buffer. If we finish the current frame before
the refresh, we can immediately move on to drawing on the tertiary buffer. If we don't finish in time
for the refresh we will need to wait for the next refresh, just as we did in the double buffer scenario.
However, we only need to miss a refresh this one time. Once the back buffer is filled, we can
immediately start working on the tertiary buffer. Either way, once the back buffer has been filled, we
will be forever one frame ahead. The primary buffer will be used for display, the secondary buffer
will be ready and waiting for the swap, and the graphics card will use the tertiary buffer for drawing:
Figure 11.13 - Drawing to the tertiary buffer before the V-Blank
When the V-Blank occurs, all the buffers can swap if they are ready. The secondary buffer becomes
the primary buffer for display. The tertiary buffer becomes the secondary buffer and waits to be
displayed. Finally, the primary buffer becomes the new tertiary buffer used for drawing:
Figure 11.14 - Scenario 1 - Swapping all buffers after V-Blank
If the tertiary buffer isn't ready at the time of the V-Blank, it continues drawing until the frame is
completed and can swap with the original primary buffer without waiting for the V-Blank:
Figure 11 15 - Scenario 2A - Swap primary and secondary buffers after V-Blank
By using triple buffering, we solve the problem of suddenly dropping from 60 fps to 30 fps when
there are a few slow frames in our game. This also allows us to avoid the drop to 30 fps and (almost)
consistently achieve 60 fps in cases which are just below 60, since we don't need wait until a V-
Blank before we can start the next frame:
Figure 11.16 - Scenario 2B - Swap secondary and tertiary buffers when drawing is complete
However, as you can see in Figure 11.17, triple buffering still has the chance to miss a refresh. Using
the same case as before where our game updates at 1/50th of a second and the monitor refreshes at
1/60th of a second, we still miss one out of every six refreshes. It should come as no surprise that the
lower our frame rate, the more refreshes we will miss. If each of our frames takes 1/30th of a second
or more to complete, we can't expect to achieve 60 fps:
Figure 11.17 - Example of missing the refresh when using triple buffering
Even though triple buffering allows us to avoid tearing while maintaining 60 fps, there is an important
factor that you must consider before deciding to use it. There is a two-frame delay between what has
just appeared on screen and the current frame being processed. Triple buffering allows us to process
another frame in the tertiary buffer while the next frame in the secondary buffer is waiting to be
displayed, all while the primary buffer is currently being displayed. This puts us one frame ahead so
we can avoid frame rate drop, but lowers response time for the player.
If the jump button is pressed, the player avatar won't appear to jump until the current frame and the
next frame have been displayed. This is because the frame being processed, including game object
responses to input, is being put into the tertiary buffer. If the primary and secondary buffers are on
screen for 1/60th of a second each, there will be an effective lag of 1/30th of a second for player
input. The game will look like it is running at 60 fps (because it is), physics will behave like the game
is running at 60 fps (because it is), but because of the input delay, the game will feel like it is running
at 30 fps.
It is up to you to decide what is best for your game. Many players may not even notice the input lag
because the time slice is so small. However, for fast twitch games such as first person shooters, or
games that require precision control for jumping or steering, this may be unacceptable.
LCD monitors
We spent a lot of time talking about how Cathode Ray Tube monitors work. This was important so
that we could understand the Vertical Blank interval and how it relates to double buffering. However,
it seems a little outdated since LCD and LED monitors are much more common. We aren't going to
discuss how these two types of monitors work, because it has no effect on our frame rate. What is
important is that these monitors fake the V-Blank signal. Even though they don't need to refresh and
have no electron gun, they still send a fake signal to the graphics card. This way, your program can
still be locked with the refresh rate of the monitor.
Time-based movement and animation
We have covered a lot so far in this chapter. We have been looking at the frame rate and refresh rate
so we can understand how it relates to what is displayed on screen. However, the frame rate of a
game has the chance to impact upon every engine of the game. It can even affect testing and debugging
during development.
At the start of a game's development, the game logic isn't very complicated and the unit count is very
low. For this reason, it is common to see thousands of frames per second. As development continues,
this frame rate will slowly drop to hundreds and then (hopefully) settle around 60 frames per second.
Imagine if there was some game logic to spawn an enemy once every 10 frames. Depending on where
we are in the development cycle, we might be spawning six or sixty enemies every second. This
makes the game very hard to test and debug because it is not consistent.
What makes this problem even more interesting is that, even within a single play session, nothing is
guaranteed to be consistent. At the start of the game, there are no enemies, so the frame rate might be
as high as 600 frames per second. This means we are spawning 60 enemies every second. After five
seconds, there are 300 enemies on screen which makes both physics and graphics very slow.
Suddenly our frame rate might drop to 30 frames per second, slowing down enemy creation. As the
player kills more enemies, the frame rate will rise, causing enemies to spawn faster and dropping the
frame rate again.
This problem isn't just limited to affecting game logic. It is a problem that affects anything that
changes between frames. In particular, it is going to affect animation of game objects. Here, animation
refers to anything that changes within a game object. This includes changing a game object's texture
over time, which is how we typically think of animation. However, it also includes moving, scaling,
rotating, or changing the color and transparency of an object over time. To understand this problem
more, let's look at how moving a unit every frame can adversely affect game development.
Frame-based movement
When we want to move a game object, the simplest way to do it is to update the position at a constant
rate. This can work fine for some games. However, for simulating cars, spaceships, or even gravity, it
won't look correct. For now, we will use this method as an example then look at physics-based
movement a little later. Since we are just updating the object's position, the code to move the player
to the right will look like this:
//Move the player to the right
//This is just an example
//A game should not be hard-coded like this
pos.x += 5;
It is worth noting that the value 5 is not measured in inches or meters; it is in game units. Game units
are completely arbitrary and are dependent on the size of objects in the game. 3D modeling programs
will allow you to set up units of scale for your models. If the same units are used for every model,
and the models are not scaled in the game world, it would be possible to think about the game world
in terms of those units. However, it is more common to consider everything as arbitrary game units.
The amount an object moves on the screen depends on the size of the object and how far it is from the
camera. If everything looks and feels correct in relation to everything else, it is OK.
In the preceding example, assuming the size and camera distance are fixed, the distance on screen that
the player moves will be completely dependent on our frame rate. For example, if we are getting
1,000 frames per second, as we might early in the development, our player will move 5,000 game
units in the x direction. Later in development, when we are getting 100 frames per second, our player
will only be moving 500 units in the x direction. To get the same amount of movement, we would need
to change the speed of the player:
//Move the player to the right
//This is just an example
//A game should not be hard-coded like this
pos.x += 50;
As we get closer to finishing the game, we might only be getting 60 frames per second. This would
mean the speed of the player would need to be changed again. To get the same amount of movement,
the player speed would need to be 83.333 fps. Unfortunately, even after the game is released, we still
have the same problem. As graphics cards and CPUs become faster, our game's frame rate will
increase, meaning the player will be moving too fast. The gameplay experience is completely
dependent on the computer hardware.
This problem could be solved by enabling VSync. As we saw earlier, using VSync will effectively
lock our frame rate to the refresh rate of the monitor. This would guarantee that our frame rate has a
maximum value. However, the player will be very confused when they upgrade to a 120 Hz monitor
and the player moves twice as fast. Additionally, if the game runs slow for a few frames, VSync
causes our frame rate to drop down to 30 fps. Suddenly, the player is moving at half speed. Even
when using VSync, our gameplay experience is completely dependent on the hardware.
Obviously using frame-based movement is not the way to go. As our frame rate rises and falls, we
must change the movement speed of the player. As we said, this same problem occurs in all
animation. Rotational speed of spinning objects, scale, fade speed of particles, and the number of
frames to display a texture before changing it all must be constantly modified during development,
and still won't be consistent across different hardware.
Time-based movement
Instead of using frame-based movement, it is much better to base our movement on time. Time is
guaranteed to be consistent throughout development and on all hardware, now and in the future.
Whether our game is updating at three frames per second or 3,000 frames per second, one second will
always be equal to one second.
Using time-based movement is great because it allows us to use equations that have existed for
hundreds of years. We don't need to reinvent the wheel to simulate velocity and acceleration. As we
said before, using a constant speed to move the player is OK for some games, but that isn't how real
physics works. Cars and spaceships do not accelerate instantly. Gravity pulls you down faster and
faster the longer you fall.
When you toss a ball to a ten year old child, they don't need to do complex calculations to catch the
ball; they can just do it. Similarly, in a semi-realistic simulation, players will expect the physics to
behave normally. To simulate realistic or semi-realistic physics in our game, we should understand
how to incorporate velocity and physics into our movement.
We can calculate an object's velocity by subtracting the initial position of an object from the final
position of an object, and dividing that by how long the displacement took. Another way to say this is
that velocity is equal to the change in the position divided by the change in time. We can use this to
create an equation that we can put in our code:
We can also calculate an object's acceleration by subtracting the initial velocity from the final
velocity and dividing by the change in time. Acceleration is change in velocity over change in time:
We also know from Newton's second law of motion that force equals mass times acceleration. This
also means that acceleration is equal to force divided by mass:
What this means is that, if we know the current position, velocity, and the forces acting on an object,
we can find the position and velocity at some point in the future. In our games, we can use these three
equations to simulate motion:
The first two of these three equations are called Euler (pronounced Oiler) Integration. Specifically, it
is called explicit Euler. In code, it will look something like this:
float EulerIntegration(float pos, float vel, float accel,
float totalTime, float dt)
{
float time = 0.0f;
while (time < totalTime)
{
pos += vel * dt;
vel += accel * dt;
time += dt;
}
return pos;
}
The inner loop of this code is a close example of how we will use this in a game. Every frame, we
will update our position and velocity based on dt and the acceleration. The velocity calculated in this
frame will be used to update the position of an object in the next frame. Outside of a loop, these
equations make perfect sense. If we are traveling at 55 mph down the freeway, after one hour we
expect to be 55 miles further down the road. Similarly, if we are accelerating at 8 miles per hour per
second then after ten seconds we expect to have a velocity of 80 mph.
However, inside of the loop we will have some errors. Euler Integration is only accurate if
acceleration and velocity are held constant. In the preceding code example, velocity is changing every
time through the loop, so it is inaccurate proportional to the square of the step size. That means that
the larger the step, the more error it has.
Let's compare Euler Integration with one of the Kinematics Equations of motion to see how this error
effects our results. The Kinematics Equation that we will test with is:
Where p is our new position, p0 is our initial position, v0 is our initial velocity, a is our acceleration
and t is our time.
Let's assume that our starting position is 0 and our starting velocity is 0. Typically, in physics
equations, acceleration is in units of seconds per second instead of hours per second. So, let's say we
are accelerating at 20 feet per second for 10 seconds. After 10 seconds, our car will have traveled
500 feet:
So, the Kinematics equations say we will be 1000 feet from where we started after 10 seconds. Using
this same data, let plug it into our Euler Integration function. We will be integrating every second for
10 seconds:
Time = 0 pos = 0.00 vel = 0.00
Time = 1 pos = 0.00 vel = 20.00
Time = 2 pos = 20.00 vel = 40.00
Time = 3 pos = 60.00 vel = 60.00
Time = 4 pos = 120.00 vel = 80.00
Time = 5 pos = 200.00 vel = 100.00
Time = 6 pos = 300.00 vel = 120.00
Time = 7 pos = 420.00 vel = 140.00
Time = 8 pos = 560.00 vel = 160.00
Time = 9 pos = 720.00 vel = 180.00
Time = 10 pos = 900.00 vel = 200.00
Euler Integration says we will be 900 feet from where we started. The Kinematics Equation and
Euler Integration are 100 feet off. This is after only 10 seconds. The longer we integrate for, the more
the error. Of course, we already explained why we have this problem. The error is proportional to the
time step. If we use a smaller time step, we will have a smaller error. Luckily our game will be
updating more than one frame per second. Let's integrate again, but let's use some more realistic time
steps. Let's choose values for 30 fps, 60 fps, and 120 fps. These give us time steps of .0333, .0167,
and .008 respectively:
dt = 1.000000 pos = 900.00
dt = 0.033333 pos = 996.67
dt = 0.016667 pos = 998.33
dt = 0.008333 pos = 1000.83
As you can see, by using a smaller time step, we come closer to the matching result. At 120 fps, we
are pretty accurate but, even at 60 fps, we have calculated a few feet off. Unfortunately, even the
Kinematics equations are not accurate unless acceleration is held constant.
For many games, Euler Integration will be all you need. The errors are small enough that players may
not notice. This of course depends on gameplay and the frame rate. Creating an extremely accurate
physics integrator is beyond the scope of this book.
If your game requires very accurate physics, check out Verlet Integration or RK4
Integration at:
https://en.wikipedia.org/wiki/Verlet_integration ,
https://en.wikipedia.org/wiki/Runge-Kutta_methods
Whichever integration method you choose, it will be better and more reliable than using frame-based
movement. The important thing to remember is that anything that changes in the game must use time.
This includes rotations, which can use similar rotational velocity and rotational acceleration if you
want. It also includes scaling over time, animating a texture over time, and even changing color and
transparency. This will give our game a very consistent look and feel, as well as making it easier to
test and debug throughout development.
Summary
We have really covered a lot in this chapter. Now, you know more about how computer monitors
work than you probably ever wanted. In this chapter, we dived into the nitty gritty details of
framebuffers and how pixels are colored on screen. We saw that having a frame rate that was out-of-
sync with the monitor can cause tearing. We also looked at how double buffering and using VSync can
fix this problem. Unfortunately, we also saw that VSync can cause problems of its own. We also
looked at triple buffering and saw the pros and cons there. In the end, there is no perfect answer.
There will always be some trade-off. You must either accept tearing or the possibility of a drastic
drop in frame rate due to VSync.
Finally, we finished this chapter by looking at how our frame rate affects the rest of our gameplay
code. Specifically, we looked at physics and animation, and learned that we must use time-based
physics and animation for a more consistent look and feel in our game.
In the next chapter, we are going to get away from the low-level details and look at the big picture of
programming. This includes our coding philosophy and why we care about high quality code. We will
look at a few tips and tricks that can help make developing a game less of a headache, as well as
covering a few specific things in the Mach5 engine that aren't patterns but can still make your coding
life much easier.
Best Practices
Learning to program is difficult for many reasons, but learning to program games is even more
difficult, specifically because there are so many different systems and object types that need to
interact with each other. Throughout this book, we have covered some of the most important design
patterns to make those interactions as simple as possible. Each chapter focused explicitly on a design
pattern to help make coding easier. However, buried within each paragraph and code sample are core
ideas and techniques to help make our design easier to read and maintain.
These Best Practices can sometimes be found in other books; however, programming books often
strive to teach you the grammar of a language, as opposed to style, design, and organization. Even
books about design patterns may gloss over these fundamental techniques. Since they are so
fundamental, it is easy to forget they are not necessarily explicitly discussed everywhere. That leaves
you, the reader, forced to read dozens of books and scour the Internet for blog posts that talk about
these fundamentals. Worse, you are required to spend hours and hours writing code with the feeling
that it could be better, but you just don't understand why it isn't.
Of course, all those things will happen anyway. Part of being a programmer is constantly reading
books like this. You should be looking for ways to improve by reading blogs and you will write code
that you consider garbage after six months. This is all part of becoming better. It is our hope in writing
this book that you can understand and incorporate these fundamentals into your programs sooner
rather than later.
Chapter overview
In this chapter, we will be focusing on fundamental ideas and techniques that will improve your code
and improve your game. These ideas come from years of programming as well as years of teaching. If
these seem simple and obvious, that is great. However, we chose these topics because they are things
that we, the authors, struggled with early on, or that our students struggled with.
Your objectives
Over the course of this chapter we will be discussing a number of topics:
Learn fundamental code quality techniques
Learn and understand the uses of the const keyword
Learn how iteration can improve your game and code design
Learn when to use scripting in a game
Learning fundamental code quality techniques
The process of moving from beginner to expert programmer can be challenging. In the beginning, you
must learn not only the rules of the language, but also how to use a compiler and understand the error
messages. Moreover, you are trying to solve increasingly difficult programming problems while
following what might seem like arbitrary rules for writing good code. Most novice programmers are
focused on solving a given problem, as opposed to making the code look nice. To many, it seems
worthless to spend time making the code look clean, because it will almost certainly be deleted after
writing it. Even more experienced programmers may ignore code style in the rush to complete an
assignment or finish a project.
This is bad for a few reasons. First, well-written code is easier to read and understand. It almost
certainly has fewer errors, and is more efficient than code that is carelessly mixed together and never
polished. Just as we discussed in earlier chapters, the time you spend upfront making sure the code is
bug free, is time you won't need to use debugging it later. The time you spend making sure the code is
readable and easy to maintain is time you won't need to spend later making changes or deciphering
old code.
Second, good programming style is a matter of habit. Taking time to read and debug your code is slow
at first. However, the more you improve your code, the easier and quicker it becomes. Eventually, you
will develop habits and writing quality code will happen naturally. Without this habit, it can be easy
to push style aside and worry about it later. However, the code that is being written is almost always
sloppy, and it can be hard to find the time later to go back and improve it, since there is always
another deadline approaching. With good habits, you will be able to write clean, readable code in
even the most time-constrained situations, such as an interview or a fast-approaching deadline.
Finally, at some point in the future you will almost certainly be working with other programmers. This
could be a small team of two or three, or perhaps in a company with dozens of programmers all over
the world. Even if you understand what your code is doing, it is not guaranteed that your teammates
will. Writing code that is hard to understand will lead to people using your code incorrectly. Instead,
strive to make your code easy to use and hard to break. Take pride in how much other people love to
use your code, your teammates and boss will thank you. You will appreciate it if your teammates do
the same thing. At some point, you will be required to maintain other programmers code after they
have left the job. You will find it much easier to do if they wrote high quality code, so write code that
will be easy to work with after you have left.
Over the next few pages, we will cover some very basic but extremely important code quality tips.
As we said, these come from years of reading programming, as well as teaching. Use these techniques
for every single line of code that you write. Think about these techniques for every piece of code that
you read. Doing so will help you form great habits.
Avoid magic numbers
Hard coding number literals into code is generally considered a bad idea. The problem with using a
number literal instead of a named constant is that the purpose of that number is unknown to the reader.
The number appears in code as if by magic. Consider the following code:
M5Object* pUfo = M5ObjectManager::CreateObject(AT_Ufo);
pUfo->pos.x = M5Random::GetFloat(-100, 100);
pUfo->pos.y = M5Random::GetFloat(-60, 60);
It is hard to know why those four numbers were chosen. It is also hard to know how the program will
be changed if the values are modified. Code like this would be much more readable and maintainable
if the named constants or variables were used:
M5Object* pUfo = M5ObjectManager::CreateObject(AT_Ufo);
pUfo->pos.x = M5Random::GetFloat(minWorldX, maxWorldX);
pUfo->pos.y = M5Random::GetFloat(minWorldY, MaxWorldY);
After the change, it is much easier to understand that the position of the new UFO is being randomly
placed within the world. We can understand that if we change the values, the possible starting
location of the UFO would either be outside of the world, or constrained to a tighter rectangle around
the center of the world.
Besides being hard to read and understand, using magic numbers makes the code hard to maintain and
update. Let's say we have an array of size 256. Every loop that needs to manipulate the array must
hardcode the value 256. If the array size needs to be larger or smaller, we would need to change
every occurrence of 256. We can't simply do a find and replace because it is very possible that 256
is used elsewhere in the code for a completely different reason. Instead we must look through every
occurrence of the number and make sure that we are changing the code correctly. If we miss even one,
we could create a bug. For example, if we are changing the size of the array to something smaller,
such as 128. Any loops that still treat the array as if it had size 256 will cause undefined behavior:
int buffer[256];
//Some function to give start values

InitializeBuffer(buffer, 256);
for(int i = 0; i < 256; ++i)

std::cout << i " " << std::endl;
As before, it is better to use a named constant instead of a magic number. The constant is more
readable and easy to change because it only needs to be changed in one place. It is also less likely to
cause bugs because we are only changing the values associated with the array. We won't accidentally
change a value we shouldn't or miss a value that we should change:
const int BUFFER_SIZE = 256;
int buffer[BUFFER_SIZE];

InitializeBuffer(buffer, BUFFER_SIZE);
for(int i = 0; i < BUFFER_SIZE; ++i)

Another important reason we don't want to use magic numbers is that they are inflexible. Throughout
this book, we try to emphasize the benefits of reading data from a file. Obviously, if you hardcode a
value, it can't be read from a file. In the preceding example, if BUFFER_SIZE ever needs to be changed,
the code needs to be recompiled. However, if the size of the buffer is read from a file at runtime, the
code only needs to be compiled one time and the program will work for buffers of all sizes:
int bufferSize = GetSizeFromFile(fileName);
//we can Dynamically allocate our buffer

int* buffer = new int[bufferSize];

InitializeBuffer(buffer, bufferSize);
for(int i = 0; i < bufferSize; ++i)

delete [] buffer;//We must remember to deallocate
In the preceding example, we must remember to deallocate the buffer. Remember, that probably won't
be the usual case because for arrays, we could always use an STL vector. The more general case is
that we are reading ints or floats from a file. These could be used for anything from screen resolution,
to player speed, or even the time interval between spawning enemies.
As with all rules, there are a few exceptions or special cases where it might be fine to hardcode
numbers. The numbers 0 and 1 are generally consider OK. These might be used as the initialization
values for ints or floats, or just the starting index for an array.
Your goal is to make your code as readable and flexible as possible, so a named constant is almost
always going to be better than a hardcoded number. Do your best to make sure that your code can be
understood by others. Your code isn't more readable if you simply have a variable named ZERO or TWO,
so you should use your best judgment and perhaps ask another programmer if you think the meaning is
unclear.
White space
When thinking about high quality code, white space is often forgotten. Maybe this is because white
space isn't code you write, but is instead the empty space between your code. However, if you don't
use white space correctly, your code will be unreadable. When we say white space, we are talking
about the spaces, tabs, newlines, and blank lines within your program. How you use these can make
the difference between code that is easy to read and maintain, and code that will give you nightmares.
Here is a piece of code with very little thought about white space:
RECT rect={0};
int xStart= 0,yStart = 0;
rect.right=s_width;rect.bottom=s_height;
s_isFullScreen = fullScreen;
if (fullScreen) {DEVMODE settings;
settings.dmSize = sizeof(settings);
EnumDisplaySettings(0, ENUM_CURRENT_SETTINGS, &settings);
settings.dmPelsWidth=(DWORD)s_width;
settings.dmPelsHeight = (DWORD)s_height;
settings.dmFields = DM_BITSPERPEL|DM_PELSWIDTH|DM_PELSHEIGHT;
s_style = FULLSCREEN_STYLE;
if (ChangeDisplaySettings(&settings
,CDS_FULLSCREEN) !=DISP_CHANGE_SUCCESSFUL) {
s_isFullScreen = false;s_style = WINDOWED_STYLE;
ChangeDisplaySettings(0, 0);M5Debug::MessagePopup(
"FullScreen is not supported. "
"You are being switched to Windowed Mode"); }
}
else {ChangeDisplaySettings(0, 0); s_style = WINDOWED_STYLE;}
The preceding code is perfectly acceptable to the compiler. For humans, however, the code above is
hard to read because there is no line spacing, no indentation, and no consistency. Of course, this is an
extreme example, but throughout our years of teaching, we have seen examples of code with just as
little thought to style and formatting. When code looks like the example above, the quality of
comments and identifier names doesn't matter because the entire block is difficult to read. Compare
the preceding code with the following version, which has tried to code for human readability:
/*Set window rect size and start position*/
RECT rect = { 0 };
rect.right = s_width;
rect.bottom = s_height;
int xStart = 0;
int yStart = 0;
/*save input parameter to static var*/

s_isFullScreen = fullScreen;
/*Check if we are going into full screen or not*/

if (fullScreen)
{
/*Get the current display settings*/
DEVMODE settings;
settings.dmSize = sizeof(settings);
EnumDisplaySettings(0, ENUM_CURRENT_SETTINGS, &settings);
/*Change the resolution to the resolution of my window*/

settings.dmPelsWidth = static_cast<DWORD>(s_width);
settings.dmPelsHeight = static_cast<DWORD>(s_height);
settings.dmFields = DM_BITSPERPEL | DM_PELSWIDTH |
DM_PELSHEIGHT;
/*Make sure my window style is full screen*/

/*If we can't change, switch back to desktop mode*/

if ( ChangeDisplaySettings(&settings, CDS_FULLSCREEN) !=
DISP_CHANGE_SUCCESSFUL )
{
s_isFullScreen = false;
s_style = WINDOWED_STYLE;
ChangeDisplaySettings(0, 0);
M5Debug::MessagePopup("FullScreen is not supported. "
"You are being switched to Windowed Mode");
}
}
else /*If we are already fullscreen, switch to desktop*/
{
/*Make sure I am in windows style*/
s_style = WINDOWED_STYLE;
ChangeDisplaySettings(0, 0);
}
While the preceding example is by no means the perfect example of high quality code, it is certainly
more readable and maintainable than the first example. When it comes to real-world programs and
programmers, there is no such thing as perfect. Every programmer has their own style, which really
means that every programmer believes their style is the easiest to read. However, as you read more
code, you will notice that there are certain common elements to code that is readable. Let's look at
some of those elements now.
Indentation
Block statements such as loops and conditionals should have the sub statements indented. This easily
shows the program intent to the reader. The number of spaces to indent is less important than the
indentation itself. Most programmers agree that 2 to 4 spaces are sufficient for readability. The most
important thing is to be consistent with your spacing. Likewise, the placement of the starting curly
brace isn't important, (although you can find some interesting arguments online), but it is important to
consistently place it in the same location:
//This shows the purpose of the statement
if (s_isFullScreen)
{
SetFullScreen(true);
}
//So does this

if (s_isFullScreen) {
}
//This does not shows the intent of the statement

if (s_isFullScreen)
{
}
It is important to remember that in C++, indentation has no meaning to the compiler. Without the curly
braces, loops and conditionals will execute only one statement. For this reason, some programmers
will always use the curly braces, regardless of how many sub statements are needed:
/*Single statement in the loop*/
while (i++ < 10)
printf("The value of i is %d\n", i);
/*After adding another statement*/

while (i++ < 10)
printf("The value of i is %d\n", i);
printf("i squared is %d\n", i*i);
The preceding example is misleading, because only the first statement will be part of the loop. It can
be a common mistake when writing a loop or conditional to forget to add curly braces after adding a
second statement. For this reason, some programmers will use curly braces even for single statement
loops and conditionals. The idea is that the code is more readable and easier to maintain, and so it is
less prone to error.
Blank lines and spaces
As we said before, how you use white space will determine how readable your code is. Using
indentation for code blocks is one way to show the logical structure of a program. Another good way
to show this is by using blank lines. Just as a good piece of writing is separated into paragraphs, a
good piece of code should be separated into some form of logical grouping. Group statements
together that logically go together. Put blank lines between the groups to improve the readability:
//Save position and scale to variables for readability.
const float HALF = .5f;
M5Vec2 halfScale = m_pObj->scale * HALF;
//Get world extents
M5Vec2 botLeft;
M5Vec2 topRight;
//If object is outside of world, mark as dead
if (pos.x - halfScale.x > topRight.x || pos.x +
halfScale.x < botLeft.x || pos.y - halfScale.y
> topRight.y || pos.y + halfScale.y < botLeft.y)
{
}
The preceding code has no blank lines so the code seems to run together. It is hard to look at it and
understand what the code is doing because your brain tries to understand all of it at once. Even though
there are comments, they don't really help because they blend in to the rest of the code. The if
statement is also hard to read because the conditions are separated by how they fit on the line instead
of a logical alignment. In the following code, we have added some blank lines to separate the logic
grouping of the statements:
//Save position and scale to variables for readability.
const float HALF = .5f;
M5Vec2 halfScale = m_pObj->scale * HALF;
//Get world extents

M5Vec2 botLeft;
M5Vec2 topRight;

if ( pos.x - halfScale.x > topRight.x ||
pos.x + halfScale.x < botLeft.x ||
pos.y - halfScale.y > topRight.y ||
pos.y + halfScale.y < botLeft.y )
{
}
By using line breaks to group related statements together, the code is separated into easy to
understand chunks. These chunks help the reader understand which statements should logically go
together. In addition, the comment at the beginning of each chunk sticks out much more and explains in
English exactly what will happen will contained in the chunk.
Complicated conditional statements should be separated and aligned based on the conditions so they
are easier to understand. In the preceding code, each of the four conditions are aligned in the same
way. This gives a clue to the reader as to how the condition will be executed. Using parentheses
along with aligning the code further adds to the readability:
if (( (pos.x - halfScale.x) > topRight.x ) ||
( (pos.x + halfScale.x) < botLeft.x ) ||
( (pos.y - halfScale.y) > topRight.y ) ||
( (pos.y + halfScale.y) < botLeft.y ))
{
}
Using parentheses isn't just helpful in conditionals. All complicated expressions should be grouped
together with parentheses. Of course, everyone's definition of complicated differs, so a good general
rule is *, /, and % are executed before + and -; use parentheses for everything else. In addition to being
clearer to the reader, this will ensure that the code is executing exactly as you expect. Even if you
understand all the C++ rules of precedence and associativity, your teammates may not. Parentheses
don't cost anything but can improve the readability, so feel free to use them as much as possible to
show the intent of the code.
Comments and self-documenting code
Comments and documentation seem to be more controversial than they should be. On the one hand,
many people think comments and documentation are a waste of time. Writing documentation literally
takes time away from writing code, and reading comments takes time away from reading the code. In
addition, some people think that comments just don't help because they can become out of date, and
don't explain anything that isn't already in the source code. The worst situation with comments is
when they are flat out wrong. In that case, the code would be better off without the comments at all.
However, there are few things more frustrating than debugging code written by someone that didn't
add comments. Even debugging your own code just a few months after you have written it can be
difficult. In the end, the time spent writing and updating comments is time that you and your teammates
won't need to spend deciphering code.
While the use of comments may be controversial, writing clean, high-quality code is important to
everyone. As we have already seen earlier, using white space well can improve readability.
However, white space alone doesn't make code readable. We really want our code to be self-
documenting. Here is an example of that, which is difficult to read, even though it has proper white
space:
void DoStuff(bool x[], int y)
{
for(int i = 0; i < y; ++i)
x[i] = true;
x[0] = x[1] = false;
int b = static_cast<int>(std::sqrt(y));
for(int a = 2; a <= b; ++a)

{
if(x[a] == false)
continue;
for(int c = a * 2; c < y; c += a)
x[c] = false;
}
}
Can you tell what the algorithm is doing? Unless you happen to know this algorithm already, chances
are that you may not understand the intent of the function. Comments would help here, but the bigger
problem is the low quality of the identifiers. Good variable names provide a hint as to what they will
be used for. The idea is that, with good variable names, you should be able to understand the code
without the need for comments. This is how you make your code self-documenting:
void CalculateSievePrimes(bool primes[], int arraySize)
{
for(int i = 0; i < arraySize; ++i)
primes[i] = true;
primes[0] = primes[1] = false;
int upperBound = static_cast<int>(std::sqrt(arraySize));

for(int candidate = 2; candidate <= upperBound; ++candidate)
{
if(primes[candidate] == false)
continue;
int multiple = candidate * 2;

for(; multiple < arraySize; multiple += candidate)
primes[multiple] = false;
}
}
Even if you don't understand every single line of the preceding code example, you can at least use the
function name as a guide. The name CalculateSievePrimes is a big clue as to what the function is doing.
From there you should be able to piece together what each line is doing. Names such as candidate,
arraySize, and multiple are much more meaningful than a, b, and c. The best part about self-documenting
code is that it can never be wrong, and never become out of date. Of course, the code can still contain
bugs. It just can't be out of sync with the documentation, because the code is the documentation.
There are a few things that you can do to attempt to make the code self-documenting, as we said
before. Good variable names are a start. The variable name should explain the exact purpose of the
variable and they should only be used for that purpose. For Boolean variables, give a name that
makes it obvious what true means. A name such as isActive is much better than simply active or
activeFlag because the names gives a hint as to what true means for that variable.
Often there will be naming conventions that distinguish between types, local variables, constants, and
static or global variables. Some of these naming conventions, such as using all capital letters for const
variables, are very common and used by most programmers. Other naming conventions, such as
beginning all static variable names with an s_, or adding a p at the beginning of pointers, are less
common. Whether you think these styles look ugly or not, understand that they are there to help
readability and make wrong code look wrong. The compiler will already catch some of the problems
that these naming conventions aim to solve, but since they can still help readability they are worth
considering.
When giving names to methods and functions, similar rules apply. Give a clear name that explains the
purpose of the function. Make sure the function or method only has one purpose. Typically, the name
should be an action. CalculateSievePrimes is a much clearer name than SeivePrimes or even just Calculate.
As with Boolean variables, methods or functions that return Boolean values are often named with a
hint. The name IsEmpty or IsPowerOfTwo are much more clear than Empty or PowerOfTwo.
Commenting
If the code is self-documenting, then why do we need to add comments at all? This is certainly the
feeling of some programmers. When the comments aren't doing more than simply repeating the code,
or when comments are out of date and difficult to update, then it is easy to understand why they feel
that way. However, this is the opposite of what good comments should be.
Good comments should explain things that the code cannot. For example, copyright information, or
author and contact information are things that can't be represented in code, but may be of use to the
reader. Additionally, good comments do not simply repeat the code. The comment below is
completely worthless. It adds nothing to the code:
//Assign START_VALUE to x
int x = START_VALUE;
Instead, good comments should explain the intent and purpose of the code. Even if you understand
what a block of code is supposed to do, you can't know what the author was thinking when they wrote
it. Knowing what the author was trying to accomplish can save you a lot of time when debugging
someone else's code:
/****************************************************************/
/*!
Given an array of "arraySize" mark all indices that are prime as true.
\param [out] primes

The array to modify and Output.
\param [in] arraySize

The number of elements in the array
\return
None. Indices that are prime will be marked as true
*/
/****************************************************************/
void CalculateSievePrimes(bool primes[], int arraySize)
{
/*Ensure array is properly initialized */
for(int i = 0; i <size; ++i)
primes[i] = true;
/*Zero and One are never prime*/

primes[0] = primes[1] = false;
/*Check values up to the square root of the max value*/

int upperBound = static_cast<int>(std::sqrt(arraySize));
/*Check each value, if valid, mark all multiples as false*/

for(int candidate = 2; candidate <= upperBound; ++candidate)
{
if(primes[candidate] == false)
continue;
int multiple = candidate * 2;

for(; multiple < arraySize; multiple += candidate)
primes[multiple] = false;
}
}
The comments above explain what the author was thinking while writing the code. They do not just
repeat what the code is doing. Notice that some comments explain a single line, while other comments
summarize entire code blocks. There aren't any hard rules about how many comments should be in
your code. A rough suggestion is that every code block should have a comment explaining its purpose,
with additional comments for more complex lines.
The comment blocks such as the one at the top of the method are the least likely to be used but they
can serve an important purpose as well. Just as the section headers of this book help when you are
scanning for something specific, function headers can help you when scanning a source code file
looking for a specific function.
Functions headers can be very helpful because they summarize everything about the function without
the need to look at the code. Anyone can easily understand the purpose of the parameters, return
values, and even any exceptions that may be thrown. The best part is, by using a tool such as
Doxygen, the header blocks can be extracted to make external documentation.
Check out the Doxygen tool and documentation at http://www.stack.nl/~dimitri/doxygen/.
Of course, these are the most difficult to write and maintain. It is comment blocks like these that often
become out of date or are flat out wrong. It is up to you and your team if you want to use them. It takes
discipline to keep up with them, but they can be worth it if you happen to be working on another
programmer's code after they have left the team.
Learning and understand the uses of the const
keyword
Using const is another area of programming that seems to be a little controversial. Some programmers
argue that they have never had a bug where using const would have helped. Others feel that, since you
can't guarantee that a const object won't be modified, it is completely worthless. The fact is that const
objects can be modified. const is not magic. So, is const correctness still a good thing? Before we get
into that, let's have a look at what const is.
When you create a const variable, you must initialize it. All const variables will be checked at compile
time to make sure that the variable is never assigned a new value. Since it happens at compile time, it
doesn't have an influence on the performance. These are a few benefits that we should consider. First,
it improves readability. By marking a variable as const, you are letting the reader know that this
variable is not supposed to change. You are sharing your intent about the variable and making your
code self-documenting. const variables are also usually named in all capital letters, which further
helps with readability. Second, since the variable is checked at compile time, there is no way for the
user to accidently change the value. If someone tries to modify the variable, it will cause a compiler
error. This is great for you if you were expecting the value to stay the same. This is great for the user
if the modification was truly an accident.
Compiler errors should always be preferred over runtime. Anytime we can use the compiler to help
us find problems, we should. This is the same reason that many programmers choose to set their
compiler warnings to maximum and treat those warnings as errors. Time spent fixing a known
compiler issue is time you won't have to spend finding the runtime error that it might cause.
Additionally, const variables should be preferred over C style #define macros. Macros are a blunt tool.
Sometimes they may be the only tool for the job, but they are overkill for simple symbolic constants.
Macros do a blind find and replace. Anywhere the symbolic constant is in the source code, the value
will replace it. While these situations may be rare, they can also be frustrating. Since the values are
replaced in the pre-processing phase, the source code is unchanged when you go to fix the problem.
The const variables, on the other hand, are part of the language. They follow all the normal language
rules for types and operators. There is nothing mysterious happening. They are just variables that can't
be reassigned:
int i1; //No initialization, OK
int i2 = 0; //Initialization, OK
const int ci1; //ERROR: No initialization

const int ci2 = 0; //Initialization, OK
i1 = 10; //Assignment, OK
i2 += 2; //Assignment, OK
ci1 = 10; //ERROR: Can't Assign

ci2 += 2; //ERROR: Can't Assign
Const function parameters
Creating const variables as symbolic constants makes code more readable because we are avoiding
magic numbers. However, const correctness is more than just creating symbolic constants. It is
important to understand const in relation to function parameters.
It is important to understand the difference between these different function signatures:

void Foo(int* a); //Pass by pointer
void Foo(int& a); //Pass by reference
void Foo(int a); //Pass by value
void Foo(const int a); //Pass by const value
void Foo(const int* a);//Pass by pointer to const
void Foo(const int& a);//Pass by reference to const
The default behavior of C and C++ is to pass by value. This means that, when you pass a variable to a
function, a copy is made. Changes made to function parameters do not modify the original variable.
The function author has freedom to use the variable however they want, while the original variable
owner can be sure that the values will remain unchanged.
That means, as far as the original variable owner is concerned, these two function signatures behave
the same. In fact, the compiler doesn't make a distinction between these two when considering
function overloading:
void Foo(int a); //Pass by value
void Foo(const int a); //Pass by const value
Since it is impossible for a pass by value variable to be modified when it gets passed to a function,
many programmers do not mark these parameters as const. It can still be a good idea to mark them as
const, because it signifies to the reader that the variable value shouldn't be changed. However, this
type of parameter is less important to mark as const since it can't be changed.
What about when you want to pass an array to a function? Remember that one of the little quirks of C
and C++ is the fact that arrays and pointers are sometimes treated similarly. When you pass an array
to a function, a copy of the array is not made. Instead, a pointer to the first element is passed. One
side effect of this default behavior is that the function can now modify the original data:
//A Poorly named function that unexpectedly modifies data
void PrintArray(int buffer[], int size)
{
for(int i = 0; i < size; ++i)
{
buffer[i] = 0; //Whoops!!!
std::cout << buffer[i] << " ";
}
}
//Example of creating an array and passing it to the function
int main(void)
{
const int SIZE = 5;
int array[SIZE] = {1, 2, 3, 4, 5};
PrintArray(array, SIZE);
return 0;
}

0 0 0 0 0
As you can see, there is nothing preventing the function from modifying the original data. The size
variable in the function is a copy of SIZE in main. However, the buffer variable is a pointer to array.
The PrintArray function is short, so finding this bug might be easy, but in a longer function that may pass
the pointer to additional functions, this problem can be difficult to track down.
If the user wanted to prevent the function from modifying the data, they could mark array as const.
However, they won't be able to use the PrintArray function, and they won't be able to modify the data
either:
int main(void)
{
const int SIZE = 5;
const int array[SIZE] = {1, 2, 3, 4, 5};//Marked as const
array[0] = 0;//ERROR: Can't modify a const array
PrintArray(array, SIZE);//Error: Function doesn't accept const

return 0;
}
Of course, sometimes the purpose of the function is to modify data. In that case, the user would just
have to accept that if they wanted to use the function. With a name like PrintArray, the user probably
expects that the data will be unchanged after the function call. Is the data modification on purpose or
an accident? The user has no way of knowing.
Since the problem is with the poor function name, it is up to the function author to make the change.
They can choose to make the name clearer, perhaps something such as ClearAndPrintArray, or fix the
error. Of course, fixing the error doesn't prevent something like this from happening again, and it
doesn't clarify the intent of the function.
A better idea would be for the author to mark buffer as a const parameter. This will allow the
compiler to catch any accidents like the one above, and it will signal to the user that the function
promises not to modify the data:
//Const prevents the function from modifying the data
void PrintArray(const int buffer[], int size)
{
for(int i = 0; i < size; ++i)
{
//buffer[i] = 0; //This would be a compiler error
std::cout << buffer[i] << " ";
}
}
int main(void)
{
const int SIZE = 5;
int array[SIZE] = {1, 2, 3, 4, 5};
array[0] = 0;//Modifying the array is fine
PrintArray(array, SIZE);//OK. Can accept non-const

return 0;
}
As we said before, the size variable could also be marked as const. This would more clearly
demonstrate that the variable shouldn't change, but it isn't necessary because it is a copy. Any
modification of size will not change the value of SIZE in main. For this reason, many programmers,
even the ones that strive for const correctness, do not mark pass by value parameters as const.
Const classes as parameters
We have now discussed the default behavior when passing arrays to functions. The compiler will
automatically pass in a pointer to the first element of the array. This is good for both speed and
flexibility. Since only a pointer is passed in, the compiler doesn't need to spend time copying a
(possibly) large array. This is also more flexible, because the function can work with arrays of all
sizes instead of just a specific size.
Unfortunately, when passing structs or classes to a function, the default behavior is pass by value. We
say unfortunately, because this will automatically evoke the copy constructor, which may be
expensive and unnecessary if the function is only reading data from the data type. A good general rule
to follow is when passing structs or classes to functions, do not pass by value, pass by pointer, or
reference. This avoids the possibly expensive copy of the data. There are certainly exceptions to this
rule, but 99% of the time, passing by value is the wrong thing to do:
//Simplified GameObject struct
struct GameObject
{
M5Vec2 pos;
M5Vec2 vel;
int textureID;
std::list<M5Component*> components;
std::string name;
};
void DebugPrintGameObject(GameObject& gameObject)

{
//Do printing
gameObject.textureID = 0;//WHOOPS!!!
}
We would like to avoid the expensive copy constructor call when passing GameObjects to functions.
Unfortunately, when we pass by pointer or reference, the function has access to our public data and
can modify it. As before, the solution is to pass by pointer to const or reference to const:
void DebugPrintGameObject(const GameObject& gameObject)
{
//Do printing
gameObject.textureID = 0;//ERROR: gameObject is const
}
When writing functions, if the purpose is to modify the data, then you should pass by reference.
However, if the purpose is not to modify the data, then pass by reference to const. You will avoid the
expensive copy constructor call, and the data will be protected from accidental modification.
Additionally, by making a habit of passing by reference or reference to const, you code will be self-
documenting.
Const member functions
In the previous example, we kept the struct very simple. Since the struct didn't have any member
functions, we only needed to worry about when non-member functions want to modify the data.
However, object-oriented programming suggests that we shouldn't have data be public. Instead, all
data should be private and accessed through public member functions. Let's look at a very simple
example to understand this concept:
class Simple
{
public:
Simple(void)
{
m_data = 0;
}
void SetData(int data)
{
m_data = data;
}
int GetData(void)
{
return m_data;
}
private:
int m_data;
};
int main(void)
{
Simple s;
const Simple cs;
s.SetData(10); //Works as Expected

int value = s.GetData();//Works as Expected
cs.SetData(10); //Error as expected

value = cs.GetData(); //Error: Not Expected
return 0;
}
As expected, when our class is not marked as const, we can use both the SetData and GetData member
functions. However, when we mark our class as const, we expect that we won't be able to use the
SetData member function, because it modifies the data. However, unexpectedly, we can't use the GetData
member function, even though it doesn't modify the data at all. To understand what is going on, we
need to understand how member functions are called and how a member function modifies the correct
data.
Whenever a non-static member function is called. The first parameter is always the hidden this
pointer. It is a pointer to the instance that is calling the function. This parameter is how the SetData and
GetData can act upon the correct data. The this pointer is optional within the member function, as
programmers, we can choose to use it or not:
//Example showing the hidden this pointer. This code won't //compile
Simple::Simple(Simple* this)
{
this->m_data = 0;
}
void Simple::SetData(Simple* this, int data)
{
this->m_data = data;
}
int Simple::GetData(Simple* this)
{
return this->m_data;
}
This is completely true. The this pointer is actually a const pointer to a Simple class. We didn't talk
about const pointers, but it just means the pointer can't be modified, but the data it points to (the Simple
class) can be. This distinction is important. The pointer is const, but the Simple class is not. The actual
hidden parameter would look something like this:
//Not Real Code. Will Not Compile
Simple::Simple(Simple* const this)
{
this->m_data = 0;
}
When we have code like this that calls a member function:

Simple s;
s.SetData(10);
The compiler is really turning it into code that looks like this:
Simple s;
Simple::SetData(&s, 10);
This is the reason we get errors when we try to pass a const Simple object to the member function. The
function signature is not correct. The function does not accept a const Simple object. Unfortunately,
since the this pointer is hidden, we cannot simply make the GetData function accept a const Simple
pointer. Instead we must mark the function as const:
//What we would like to do but can't
int Simple::GetData(const Simple* const this);
//We must mark the function as const

int Simple::GetData(void) const;
We must mark the function as const within the class as well. Notice that SetData is not marked const,
because the purpose of that function is to modify the class, but GetData is marked const because it only
reads data from the class. So, our code would look something like the following. To save space, we
didn't include the function definitions again:
class Simple
{
public:
Simple(void);
void SetData(int data);
int GetData(void) const;
private:
int m_data;
};
int main(void)
{
Simple s;
const Simple cs;
s.SetData(10); //Works as Expected
int value = s.GetData();//Works as Expected
cs.SetData(10); //Error as expected

value = cs.GetData(); //Works as Expected
return 0;
}
As you can see, by marking the GetData member function as const, it can be used when the variable
instance is marked const. Marking member functions as const allows the class to work correctly with
non-member functions that may be attempting const correctness. For example, a non-member function
(possibly written by another programmer) trying to display a Simple object by using the GetData member
function:
//Const correct global function using member functions to access
//the data
void DisplaySimple(const Simple& s)
{
std::cout << s.GetData() << std::end;
}
Since DisplaySimple isn't intending to change the data in the class, the parameter should be marked as
const. However, this code will only work if GetData is a const member function.
Being const-correct takes a little work and may seem difficult at first. However, if you make it a
habit, it will eventually become the natural way that you program. When you are const-correct, your
code is cleaner, more protected, self-documenting, and more flexible, because you are prepared for
const and non-const instances. As a rule, if your function isn't going to modify the data, mark the
parameter as const. If the member function isn't going to modify the class data, mark the member
function as const.
Problems with const
As we said before, const isn't magic. It doesn't make your code 100% secure and protected. Knowing
and understanding the rules related with const parameters and const member function will help
prevent mistakes. However, failing to understand the rules and behaviors of const can lead to bugs.
The biggest problem with const in C++ is a misunderstanding of bitwise const versus logical const.
This means that the compiler will try to ensure that the bits and bytes will not change through that
specific variable. This doesn't mean that those bits won't change through another variable, and it
doesn't mean that the data you may care about won't change. Consider the following code:
//Example of modifying const bits through different variables.
int i = 0;
const int& ci = i;
ci = 10; //ERROR: can't modify the bits through const variable

i = 10; //OK. i is not const
std::cout << ci << std::endl;//Prints 10
In the preceding example, i is not const, but ci is a reference to a const int. Both i and ci are
addressing the same bits. Since ci is marked const, we cannot change the value through that variable.
However, i is not const, so we are free to modify the value. The fact that we can have a multiple const
and non-const variables pointing to the same address has consequences for const member functions:
class Simple
{
public:
Simple(void);
int GetData(void) const;
private:
int m_data;
Simple* m_this;
};
Simple::Simple(void):m_data(0), m_this(this)
{
}
int Simple::GetData(void) const
{
m_this->m_data = 10;
return m_data;
}
int main(void)
{
const Simple s;
std::cout << s.GetData() << std::endl;
return 0;
}
In the preceding code, we have given the Simple class a pointer to itself. This pointer can be used to
modify its own data in a const member function. Remember, in a const member function, the this
pointer is marked const, so data can't be changed through that variable. However, as in this case, the
data can still be changed through another variable. Even if we didn't use another variable, the use of
const_cast could also allow us to change the data:
int Simple::GetData(void) const

{
const_cast<Simple*>(this)->m_data = 10;
m_data;
}
It is very important to understand that you should never write code like this. It is undefined behavior
to try to modify a const variable with const_cast or using a non-const pointer to a const object. The
original data could be placed in read-only memory, and code like this could cause the program to
crash. It is also possible that the compiler could optimize away multiple reads to memory that
shouldn't change. Therefore, the old value may be used for any future calculations. Casting away const
with const_cast is meant for backwards compatibility with old C++ libraries. It should NEVER be used
to modify const values. If there is a piece of data that needs to be modified even when the class is
const, use the mutable keyword.
Even when avoiding undefined behavior, bitwise constness can get us into trouble with const member
variables. Consider a simple class that will contain some dynamic memory. Since it contains dynamic
memory and pointers, we should add a copy constructor, destructor, and other things to prevent
memory leaks and memory corruption, but we will omit those for now since they aren't important for
our discussion on const:
class LeakyArray
{
public:
LeakyArray(int size)
{
m_array = new int[size];
}
void SetValue(int index, int value)
{
m_array[index] = value;
}
int GetValue(int index) const
{
//function is const so we can't do this
//m_array = 0;
//but we can do this!!!!!!!

m_array[index] = 0;
return m_array[index];
}
private:
int* m_array;
};
As you can see, bitwise const only prevents us from modifying the actual bits inside the class. This
means we can't point m_array to a new location. However, it doesn't prevent us from modifying data in
the array. There is nothing preventing GetValue from modifying the array in the const function, because
the array data isn't part of the class, only the pointer is. Most users aren't concerned about where the
data is located, but they would expect a const array to remain unchanged.
As you can see, being const, correct is certainly not a guarantee that the data will never be modified.
If you are diligent about using const, as well as understand and avoid the problems that can arise, the
benefits are worth it.
Learning how iteration can improve your game
and code design
While it's nice to imagine it to be this way, games never come fully-formed from the mind of a
designer/developer. A game is made up of many different ideas from many different people. While in
the past people could develop games with only a single person, now it's much more common for
teams to be made up of many different disciplines and every game developer on your team has ideas,
many of them good that can contribute to the final product that gets made. But with that in mind you
may be wondering how does a game get to the final point with all those different changes made? The
answer to that is iteration.
The game development cycle
Game development is a process, and different people have different names and/or phrases for each of
these steps, but most people can agree that, for commercial game development, there are three main
phases:
Pre-Production
Production
Post-Production
Each of these states has their own steps within them as well. Due to page constraints, I'm unable to
write about the entire process, but we will be focusing on the Production aspect of development,
since that's where the most relevant content will be for our readers.
If you are interested in learning more about the different aspects of the game
development process, check out https://en.wikipedia.org/wiki/Video_game_development#Development
_process.
During game development, you'll see a lot of companies use an agile development process, which is
based on iterative prototyping, using feedback and refinement of the game's iterations while gradually
increasing the game's feature set. A lot of companies enjoy this method as there is always a version of
the game playable every couple of weeks, and you can make adjustments as the project is being
worked on. If you have heard of Scrum, it is a popular method of agile software development, and
one that I use with my students as well as what I used in the game industry.
Production phase
Upon entering the production phase, we have come up with our base idea for our project and have
created our proposal and game design document. Now that we have this information, we can start
doing the following three steps:
Prototyping
Playtesting
Iterating
Each of these steps serves a valuable process and will be completed in this order. We will be
repeating these steps repeatedly until release, so it's a good idea to have an understanding of them.
Prototyping
Prototyping is when you make the simplest version of your ideas in a quick manner to prove if your
concept works well or not. For some people, they will do this via index cards, paper, chits, and
boards, which is called a paper prototype. This can be quite useful, as you don't have to think about
the code side of things to begin with, and instead it allows you to experience the core of the game
without having all the nice art and polish. A game that is fun to play with bad graphics will only be
improved when you add to it.
Of course, assuming you have purchased this book, you are probably already a developer, but it's still
a good idea to think of it as an option. Jesse Schell wrote about paper prototypes in his book The Art
of Game Design: A Book of Lenses, where he explained how you could create a paper prototype of
the game Tetris. To do so, you could cut out cardboard pieces and then put them into a pile and pick
them out randomly, and then slide them down the board, which would be a piece of paper. Once you
completed a line, you would grab an X-acto knife and then cut the pieces. While it doesn't give you
the same experience entirely, it captures enough of the experience to see if you are using the right
kinds of shapes and how fast the pieces should drop. The big advantage would be that you could
create this prototype in 10 to 15 minutes, while programming it may take a lot longer.
It's a lot easier to justify 30 minutes of work rather than a day for something that doesn't work out.
This can be done with 3D games as well, such as First Person Shooters, by creating maps in a similar
way to how you create combat encounters in a pen and paper role playing game such as Wizards of
the Coast's Dungeons and Dragons (which is a great thing to learn how to play as a designer, as you
learn about telling a story and developing interesting encounters).
A prototype's job is to prove if your game works and, specifically, how it works. Don't invest in only
one idea specifically, but rather create a number of small prototypes that are quick to make, without
worrying if it is polished or you've made it the best possible.
For some more information on building a prototype and an example of a prototype

that was created in seven days about Tower of Goo, which was the prototype for their
indie hit World of Goo, check out http://www.gamasutra.com/view/feature/130848/how_to_prototype_
a_game_in_under_7_.php?print=1 .
One of the most vital skills as a game developer is being able to create rapid prototypes, to see how
it works, and then test it. We call this process of testing out a game idea playtesting.
Playtesting
Once we have a prototype, we can then begin the playtesting process. Playtest as soon as possible in
your development, and do it often. As soon as you have something playable, get people in front of it.
First play the game yourself and see how you feel about it. Then invite some friends over to your
house and have them play it as well.
Oftentimes, I find my students have difficulty at first actually doing playtests and may be hesitant to
show people their projects as it's not ready yet or because they won't get it. Or they know that their
project isn't finished so they think they know what they should be working on already, so there's no
need to do playtesting. I find this usually comes from them being shy, and one of the first major
obstacles you'll need to overcome as a developer is being able to show your ideas to the world.
If the play testers of your game aren't your close friends and/or family, it is likely that people will say
negative things about the game. This is a good thing. They will also mention many things that you
already know your game doesn't have yet or that you don't have the budget for. This isn't a time for
you to defend any points, or explain why things are the way they are, but instead a time for you to
accept the points and note them, as you can take them into consideration in the future.
As a game developer, one thing that's important to note is that you are one of the worst judges of your
games, especially when just starting out. A lot of times I see starting developers trying to justify the
problems in their games, stating that it is their vision and people don't get it because it's not in a final
game. It's a very important skill of a game developer to be able to get feedback, take criticism, and
evaluate if things are worth changing or not.
Conducting a playtest
Now that we know how valuable it is to do a playtest, you may wonder how to go about doing one.
First, I do want to note that it is vitally important that you are in the room while your project is being
playtested. While they are playing the game, you can see not only what someone thinks, but also how
they react to things as well as what they do with your game. This is the time where you find out what
works well and what doesn't. If for some reason, you are unable to be there in person, have them
record themselves playing, both on the PC and from the webcam if possible.
When someone comes to a computer to playtest your game, you may be tempted to tell them things
about your project, such as the controls and the story, and the mechanics, and anything else, but you
should resist these urges. See what the player does without any prompting to begin with. This will
give you an idea of what players will want to do naturally given the environment that you created and
what needs to be explained more. Once they play for a while and you've gotten the information you
need on that end, then you can tell them some things.
When playtesting, it's a good idea to get as much information from the player as you can. When they
finish, ask them what they liked, what they disliked, if they found anything confusing, where they got
stuck, and what was the most interesting thing to them. Do note that what a player says and what they
actually do are two different things, so it's important that you are there and watching them. Getting
your game played and seeing what those players do is where you will start to see the cracks in your
design's form, and seeing what people do will show how they experience the things you have created.
While doing this testing, I've seen a lot of people do the exact opposite of what I expected and not
understand something that I thought was quite simple. The player isn't wrong in this regard though, I
was. The player can only do what they know from prior playing or from what is taught in the game.
All of the information that you get during a playtest is important. Not just what they say, but also what
they don't say. Once they finish playing, give them a survey to fill out. I find using Google Sheets does
a good job at storing this information and isn't too difficult to set up, and you have hard data from
which you can make decisions without having to remember what people say. Plus, it's a lot easier for
people to select a number from 1-10 for how they enjoyed different aspects of the game than to ask
their thoughts for everything and doesn't require them to write paragraphs of info (unless they want to
in a comments section at the end).
In case you'd like to see an example playtesting form, while this form is for board
games, I feel it does a good job of making it easy for play testers to give information
that can be useful: https://www.reddit.com/r/boardgames/comments/1ej13y/i_created_a_streamlined_playt
esting_feedback_form/.
If you're looking for some ideas of questions to ask, Wesley Rockholz wrote some
examples of questions that may be useful for you to use: http://www.gamasutra.com/blogs/Wesl
eyRockholz/20140418/215819/10_Insightful_Playtest_Questions.php .
In addition, the order in which players give their feedback is also important as it communicates how
important different things are to them. You may find that what was intended to be the main mechanic
isn't as engaging and/or fun as something else. That is valuable feedback, and you may decide to focus
on that secondary mechanic instead for the title, as I've seen happen repeatedly on projects. It's better
to do this as early as possible so you will waste as little of your time as possible.
Iteration
At this point, we've playtested our project and have gotten our player's feedback, and, if we have set
it up, gotten data and analytics from which we can build upon. Now we need to take this information
into consideration, make some changes to our current prototype, and then bring it back to be play
tested again. This is what's known as the iteration stage of development.
In this stage, you'll need to take this feedback and then decide how to incorporate it into your design.
You'll need to decide what should be changed as well as what shouldn't. While doing so, keep in
mind the scope of the project, realistically evaluating how long it will take to make the changes, and
be willing to cut features, even those that you love, to have the best project.
After making these decisions again, we will once again create a new prototype, which you will then
playtest again. Then iterate again. Then build another prototype, where you will continue to test,
removing prototypes that don't work and features that don't work well for the project. You'll also try
adding new features using feedback and remove previous features that would no longer fit the game in
its current state. You will keep doing this cycle repeatedly until you reach your final release!
If you're waiting for your game to be perfect before releasing it, you'll never release it. Games are
never finished, they're only abandoned. If the project is at a point where it is good enough, you should
ship, because it's only when you ship a project that you can finally say you've developed a title.
If you'd like to see an example of this process and how it can benefit a title, check
out: http://www.gamasutra.com/blogs/PatrickMorgan/20160217/265915/Gurgamoth_Lessons_in_Iterative_G
ame_Development.php .
Meeting milestones
When working on a commercial game project, especially when you have a publisher, you often times
will have a schedule to keep and milestones to meet. Milestones are ways for everyone to know if the
game is on track or not, because certain things need to be accomplished by them. Not making a
milestone can often be a horrible thing, because your publisher usually will only pay your team if the
milestone has everything in it that was agreed upon. There's no standard milestone schedule, as every
company is different, but some of the most common ones are as follows:
First-playable: This is the first version of the game that can be played. Has the main mechanic
of the game in and can demonstrate how it works.
Alpha: When all the features of your game are in, known as being feature complete. Features can
change slightly, and make revisions based on feedback and testing, but at this point,
unimplemented things may be dropped to finish the title on time.
Beta: The game is complete, with all assets and features completed and finished. At this point
you are only doing bug testing and fixes for potential problems that would prevent your game
from being shipped.
Gold: This is the final version of the game, which you'll either be releasing, or sending to your
publisher to create copies on disks, cartridges, or whatever your device uses.
Do note that every company is different and these milestones may mean different things to different
people, so be sure to clarify before diving into development.
Learning when to use scripting in a game
Scripting languages are something that can be quite beneficial to developers when working on a team
with multiple disciplines in it. But before we dive into what they are and how they work, and the pros
and cons of using a scripting language, it's best to get a bit of a history lesson in terms of how code
executes.
Introduction to assembly
Underneath the hood, all the code that we have written over the course of this book is ones and zeroes
indicating what switches should be marked as on and off by our computer's processor. Low-level
programming languages such as machine language use these switches to execute commands. This was
the only way to program to begin with, but we have developed more readable languages for us to
work with instead.
Starting with assembly languages, low-level languages have a very strong connection between the
language's instructions and the machine code's instructions. While more readable than a sequence of
0s and 1s, it was still quite difficult to write code. For example, here is some assembly code used to
add two numbers in Assembly Language:
push rbp
mov rbp, rsp
mov DWORD PTR [rbp-20], edi
mov DWORD PTR [rbp-24], esi
mov edx, DWORD PTR [rbp-20]
mov eax, DWORD PTR [rbp-24]
add eax, edx
mov DWORD PTR [rbp-4], eax
nop
pop rbp
ret
Each computer architecture has its own assembly language, so writing code in low-level languages
has the disadvantage of not being portable as they are machine dependent. Back in the day, people
would have to learn many different languages in order to port your program to another processor.
Program structures became more complicated, as the demands of functionality increased over time,
making it quite difficult for programmers to implement efficient programs that were robust enough.
Moving to higher-level programming languages
As programmers, we are inherently lazy, so we seek ways to make our job easier or rather to find the
best use of our time. With that in mind, we have developed other higher level languages, which are
even easier to read. By higher level, we mean closer to the way that humans think or closer to the
problem we are trying to solve. By abstracting the machine details from our code, we simplify
programming tasks.
Introducing the compiler
Once we have finished our code, we use a compiler to translate that high-level code into assembly,
which in turn will get turned into machine language that our computer can execute. Afterwards, it will
translate that program into an executable file that the user can run. Functionally, it looks something
like this:
This has several advantages, as it provides abstraction from the hardware details. For example, we
don't need to work directly with registers, memory, addresses, and so on anymore. It also makes our
code portable in that we can use the same program and have it translated by different assemblers for
the different machines that use it. This was one of the reasons C took off and became so popular,
because it allowed people to write code once and then have it work everywhere. You may notice that
Unity has taken the same thought process to game development, and that's one of the reasons that I feel
they've been so successful as of this writing.
This is a much more efficient use of time than writing assembly language code as it allows us to
create more complex programs and machines, and modern compilers such as Microsoft's produce
some highly efficient assembly code in most circumstances. This is what we have been using over the
course of this book.
Writing code in assembly can still have its benefits though. For instance, after you've written your
game in a higher-level language, you can start profiling it and seeing what aspects of the game are the
bottlenecks, and then determine if rewriting it in assembly will give you a speed boost. The point of
using a lower-level language is the fact that you can gain some substantial speed advantages.
For a real-life example as to how assembly was used to optimize a game engine,
check out the following article from Intel: https://software.intel.com/en-us/articles/achieving-perfo
rmance-an-approach-to-optimizing-a-game-engine/.
One of the issues with writing code that needs to be compiled before running is the fact that as
projects increase in size, so do compile times. Recompiling a full game can take minutes to hours
depending on the size of the project, and while that's going on, you are unable to work on your
project, or else you'd need to recompile again. This is one of the reasons why scripting languages can
be useful.
Introduction to scripting languages
A scripting language is a programming language that allows scripts to be written for it. A script is a
program that can be executed without being compiled in one of a few different ways. Scripting
languages are also sometimes referred to as very high-level programming languages, as they operate
at a high level of abstraction, being very fast to learn how to write in.
Scripting languages also have the advantage of taking care of a lot of things that programmers would
need to take care of, such as garbage collection, memory management, and pointers, which often
confuse non-developers. Even a visual editor such as Unreal 4's blueprints is still a scripting
language, as it accomplishes the same thing as a written one does.
Most games use scripting languages in some form, but other games may use it a lot more, such as how
GameMaker uses Game Maker Language (GML) for their logic.
Using interpreters
To use a scripting language, we need to be able to execute new code on the fly. However, unlike a
compiler, there's also another way to convert code into something that can be understood by a
machine that is called an interpreter. Interpreters do not produce a program themselves, and rather
stay around for the execution of a program. This program will either:
Execute the source code directly

Translate the source code into some other efficient intermediate representation (code) and then
immediately execute it
Explicitly execute stored precompiled code made by a compiler that is part of the interpreter
system
The interpreter translates one line at a time, as opposed to the compiler doing it all at once.
Visually, it looks a bit like the following:
As you can see here, the Interpreter takes in the Source Code and any Input that has been received,
and will then Output what is expected of it.
Just in time compilation
There is yet another way to run code, using what's called a just in time compiler, or JIT for short. A
JIT caches the instructions, that have been previously interpreted to machine code, and reuses those
native machine code instructions thus saving time and resources by not having to re-interpret
statements that have already been interpreted.
Visually, it looks similar to this:
Now, Unity uses a JIT and AOT compiler to convert code into machine code, which can then be read
by the machine. The first time that a function is called, the game will convert that code into machine
language and then the next time it is called, it will go directly to the translated code, so you only need
to do conversions to things that are happening. Due to this happening at runtime, this may cause your
project to lag when using a lot of new things.
A great talk about how scripting works inside of the Unity game engine can be found
here: https://www.youtube.com/watch?v=WE3PWHLGsX4.
Why use a scripting language?
C++ is often overkill when it comes to building things for your game, such as tools, or for high-level
game tasks that could be handled by your technical designers. It has some definite advantages for ease
of development. Specifically, that you don't have to worry about a lot of the low-level things, as the
language takes care of it for you; there's also fewer errors by the programmer due to the limited
options available. There's less of a technical programming knowledge required, and it can be
customized to fit your game's needs. This also makes the game more data driven instead of hardcoding
things into the game engine, and allows you to patch the game without having to send over the entire
project.
Often in game development, game logic and configurations can be found in script files. This way, it's
very easy for scripts to be modified and adjusted by people on the team that aren't programmers (such
as designers), allowing them to playtest the game and tweak the gameplay without having to
recompile the game.
Many games also have a console window, which uses a scripting language to do things like this even
at runtime. For instance, Unreal Engine has a console window open by default when you press Tab
and in Source engine, pressing the ~ button at the pause menu will open one up.
Scripting languages are also often used in areas that have level design as well, such as having triggers
for when you enter certain areas, or for controlling cinematics. It also allows you to let players of
your game mod the game as well, which could potentially increase the lifespan of your game and help
foster your game's community.
When to use C++
C++ is a great language to be using, as performance is a crucial first step. This used to be all aspects
of game engines, but now is used primarily with graphics and AI code. Scripting languages also have
the problem of being slower than C++, sometimes even 10x slower than what it would be otherwise.
Due to scripting languages having memory management taken care of for them automatically, there
could be times when commands are interrupted or take a while to complete garbage collection,
causing lag and other problems.
C++ also has the advantage of much better IDEs and Debuggers, making it a lot easier for you to find
mistakes and fix them when they do occur.
There's also the possibility that you are working with a legacy code base. Most game companies
aren't starting with a fresh slate. It can also be useful to make use of the middleware libraries that
C++ has, such as FMOD and AntTweakBar.
Compiled versus scripting
For some game engines, the game engine itself is programmed in C++, but the game logic is entirely
done in a scripting language, like most development for Unity, for example. This allows you to iterate
on gameplay much faster, and allows technical designers and artists to modify behaviors without
having to bother a programmer. Also, depending on the language, it can allow people to use a more
appropriate language for the problem's domain (for example, AI may not be the easiest thing to
implement in C++).
Different companies handle working with languages differently. When I was working at a AAA
(pronounced triple-A) studio, we would have designers prototype ideas for mechanics and get it
working in the best possible way making use of scripting languages. Upon approval from the leads as
something to add to the project, if the script had performance issues, programmers would use the
scripting language code as a base and then create a super-efficient version using C++ that would
work in all levels. However, when I was working on an indie title, all of the code for the project was
written in a scripting language (C#), since we didn't have access to the engine's source code (Unity).
Also, if you are wanting to target devices that are limited in memory and processing power (such as
the Nintendo 3DS), you are likely wanting to care about performance a lot more so it's more
important to use more optimized code. It's a good idea to be familiar with both options and be
comfortable working either way.
If you are interested in using a scripting language for your project, Lua is very widely used in the
game industry because it's very easy to learn and fairly easy to incorporate into your engine. Lua
started off as a configuration language. This has some nice quirks, in that it's great for creating and
configuring things - which is what you want to do in a game. It's important to note, though, that it's not
object-oriented but uses a small amount of memory.
A list of games that use Lua as a scripting language can be found here: https://en.wikiped
ia.org/wiki/Category%3aLua-scripted_video_games.
If you are interested in integrating Lua into your project or would like to see how it works, I highly
suggest checking out http://www.lua.org/start.html.
Summary
In this chapter, we covered a lot of best-practice information, which we hope will give you a good
foundation when building your own projects in the future. We touched on why hardcoding values is a
bad idea, in addition to making a number of other code-quality suggestions, to ensure that your code is
easy to understand and easy to extend from in the future, when it needs to be.
We also learned how iteration is useful in game development, talking about the traditional game
development cycle, with tips and tricks about playtesting and how it can be immensely useful when
developing your projects.
We also looked into low-level and high-level programming languages, learning about how scripting
languages run inside another program that we have to build into our project. They are not compiled
but rather interpreted, and are generally easier to use and write code for than a compiled language,
but come at the cost of performance. Depending on how complex your game is, it may be a good idea
to stick to just C++, but if you are working with designers, it can be quite useful to give them the tools
to do things on their own.
With that, we've reached the end of the book. We hope that you found the information both interesting
and useful. As you go out there and build your own projects, make use of the design patterns and best
practices that we've talked about over the past 12 chapters, and make the best games that you can!

Game Development Patterns Best Practices

Uploaded by

Copyright:

Available Formats

Game Development Patterns Best Practices

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Game Development Patterns Best Practices

Uploaded by

Copyright:

Available Formats

What are some of the advantages and disadvantages of using scripting languages versus compiled languages for game development?

What are some of the advantages and disadvantages of using scripting languages versus compiled languages for game development?

What are some reasons a game developer may want to use C++ over a scripting language?

What are some reasons a game developer may want to use C++ over a scripting language?

Title Page

Game Development Patterns and Best Practices

Better games, less hassle

Copyright 2017 Packt Publishing

First published: April 2017

Production reference: 1180417

Published by Packt Publishing Ltd.

John P. Doran Safis Editing

Commissioning Editor Proofreader

Amarabha Banerjee Safis Editing

Larissa Pinto Francy Puthiry

Content Development Editor Graphics

Arun Nadar Jason Monteiro

Technical Editor Production Coordinator

Sushant S Nadkar Arvindkumar Gupta

You can find him at www.francescosapio.com.

Understanding Graphics and Animation, provides a low-level explanation of how graphics

A block of code is set as follows:

static void StaticFunction()

// Increase the value of enemyCount

std::string toDisplay = "\n Value of enemyCount: " +

Warnings or important notes appear in a box like this.

You can download the code files by following these steps:

WinRAR / 7-Zip for Windows

To view the previously submitted errata, go to https://www.packtpub.com/books/content/support and enter the

Please contact us at [email protected] with a link to the suspected pirated material.

What are design patterns?

Converted to UML, it would look something like this:

Inside of UML, it would look like this:

void Purr(void) const //unrelated function

for(int i = 0; i < SIZE; ++i)

The following is the output of the preceding code:

Here's a simple example of an interface, and an implementation of it:

class FakeEnemy: public Enemy

virtual void Attack(void)

virtual void Move(void)

And here's how it looks in UML:

An overview of class access specifiers

For more information on friend functions, check out http://en.cppreference.com/w/cpp/languag

// Increase the value of enemyCount

std::string toDisplay = "\n Value of enemyCount: " +

And when we call it, the following would be displayed:

// Increase the value of enemyCount

std::string toDisplay = "\n Value of enemyCount: " +

Not knowing what a local variable is

Through code, it will look a little something like this:

// Disable usability of silently generated functions

static Singleton& GetInstance()

// Spawn the enemy

Here's an example of a simple templated class:

template <class T>

template <class T> TemplateExample<T>::~TemplateExample()

template <class T> T TemplateExample<T>::TemplatedFunction(T obj)

// Once destroyed, remove access to instance

// Get a reference to our instance

// Creates an instance of our instance

/Call These in Main/

/This must be called first, before the game is started/

/Call these to control or get info about the application/

/Use this to change to fullscreen and back/

M5InitData initData; /Declare my InitStruct/

/Pass InitStruct to Function. This function must be called first!!!/

/Make sure to add what stage we will start in/

/Start running the game/