C++ can be a confusing language. Pointers, ownership, memory management -- there is a lot going on in even a simple C++ program and it is often difficult for beginners to approach. As students entering the workforce we want to learn how to write industry grade C++ that would impress our interviewers, coworkers, and employers. This guide gives an introduction to doing just this, covering some of the most important topics in object-oriented design in C++.
Object-oriented programming or OOP are the fundamental principles that most modern languages are optimized to follow. We can summarize OOP with 4 primary concepts.
- Encapsulation is the principle of bundling data and the methods that operate on that data under a common slass. This allows us to hide the internal details of an object and restrict access to its state.
- Inheritance is the mechanism that allows a class to inherit the properties and behaviors of another, known as the parent class. This promotes code reusability and facilitates the creation of a hierarchical structure for classes.
- Polymorphism means "many forms" and is concept that allows objects of different classes to be treated as objects of a common parent class, enabling code to work with objects of various types in a uniform way. In particular, mechanisms like method overloading, defining multiple methods with the same name but different parameters, and method overriding, providing a specific implementation of a parent classes' method in a child class, make polymorphism possible.
- Abstraction involves simplifying complex systems by modeling classes based on the essential properties and behaviors that they share, focusing on what an object does rather than how it does it. This make it easy for code to be utilized by different developers and reduces complexity by hiding unnecessary details.
C++ is a versatile and powerful programming language known for its efficiency and performance. It's used in a wide range of applications across various domains. Here are some common use cases for C++:
-
System Software: C++ is commonly used in developing system software like operating systems (Windows, Linux), device drivers, and firmware.
-
Game Development: Many parts of game development, especially engines and resource-intensive components like physics engines, are often written in C++ due to its performance.
-
High-Performance Applications: C++ is used in applications where high performance is critical, such as real-time simulations, high-frequency trading systems, and scientific computing.
-
Embedded Systems: Its ability to directly interact with hardware makes C++ a preferred choice for embedded systems, like those in automotive systems, IoT devices, and robotics.
-
Graphics and Animation Software: Tools for graphics, animation, and visual effects often use C++ for their performance-critical components. Software like Adobe Photoshop and Autodesk Maya utilize C++.
-
Web Browsers: Components of web browsers like Chrome and Firefox are implemented in C++ to optimize performance, especially in rendering engines and parts handling complex tasks.
C++ has a variety of features that make object-oriented design easy. For example, it supports the creation of classes and objects, the fundamental building blocks of OOP, allowing users to define classes to encapsulate data and methods and then create instances(objects) of those classes. Furthermore, access specifiers like public, private, protected allow programmers to control the visibility of class members further enforcing the principle of encapsulation. C++ also supports both single and multiple inheritance, allowing classes to inherity properties and behaviors from one or more parent classes, giving us flexibility in our class hierarchies. Additionally, C++ has both compile-time (method overloading) and runtime polymorphism (method overriding); these tools not only enable polymorphism but also abstraction as they allow us to define abstract classes, facilitating the implementation of abstraction.
In C++, header files (.h) are used to provide an interface to other parts of the program. Think of it as the bridge that connects .cpp or .cc files (the files containing our C++ source code). In them, we declare the structure and characteristics of entities without defining how they are implemented. Usually, we want to have one header file per source file.
While not enforced semantically, header files should ONLY include what you use. If your source file has a helper function foobar() that will only be called within the source file, don't include it in your header file! This goes both ways. If your source file isn't using at least one of the declarations within a header file, don't include that header file! Additional #include statements force the compiler to open more files and process more input, slowing down compile time.
Moreover, ALL header files should begin with the #define guard. For big projects, it can be easy to lose track of where a file has already been included. These guards help prevent circular dependencies by restricting the inclusion of a file if it has already been defined.
#ifndef FOO_BAR_BAZ_H
#define FOO_BAR_BAZ_H
...
#endif
If you have learned C, you may be familiar with the concept of structs. C++ also has structs, however, in addition to that, we also have classes. Although the two keywords behave basically the same in C++, we should ONLY use a struct for passive objects that carry data. Everything else should be a class. All fields of a struct should be public and as a result, should not have rules that imply relationships between different members, since direct access can break those rules.
Imagine a scenario where you want to represent a basic geometric point:
struct Point {
float x;
float y;
};Here a struct is a great design choice because the members are simple and benefit from having a name instead of just being referenced by indexes.
As Google says, when in doubt, make a class.
Nevertheless, when elements can have meaningful names (but not require some high level of functionality that would suit a class), we should prioritize a struct over pairs and tuples.
You can redefine what certain operators mean in C++. However, don't go crazy and make + actually perform *. An operator should only be overloaded when their meaning is obvious and consistent with built-in operators. For example, using + to concatenate two strings.
This is a big topic that can get quite complex, but in the end it really boils down to one question: who should own this object? For example, when creating a banking system it makes sense that the Bank should own all instances of Account. Don't make the bank store references to the Account, this makes it confusing when handling the dynamic memory associated with the object. The Bank's scope should control the existence of the Account in memory. If the Bank disappears simultaneously the Account disappears (and your money!). Additionally, the Account should only be erasable through the Bank. Imagine if the contract your signed owns your Account and ripping that contract simultaneously deletes your Account (and your money!), yikes.
To handle ownership, modern C++ has this fancy mechanism called smart pointers. Smart pointers automatically handle memory allocation and deallocation. The only decision you need to make is who owns the smart pointer (in our prior example of Bank and Account, the Bank should own the smart pointer to Account). More on smart pointers below.
There are a few types of smart pointers: unique_ptr, shared_ptr, and weak_ptr. I will leave it up to you to learn the differences. Below are some basic guidelines that will (hopefully) help you utilize smart pointers effectively and correctly, preventing memory leaks and/or loose pointers.
- NEVER use new or delete. This makes it unclear whether a pointer owns the object or doesn't.
- Use unique_ptr by default because it has basically no overhead. If you know for sure that there will be several owners, use shared_ptr.
- When a pointer will not own the object, use raw pointers. For example, passing an object to a function should be done using a raw pointer. Going back to our example, it never makes sense to transfer ownership of an Account to anything other than the Bank. To pass a unique_ptr as a raw pointer to a function, the best way is to dereference it and pass it by reference using .get().
- Do NOT use delete on a raw pointer, assume that you have followed the previous steps correctly in which case a raw pointer should never own the object.
- When you create an object and want to return it, return it as a smart pointer. This will transfer the ownership of that object to whatever is receiving the function's return value. If you don'tdo this, you're probably going to get a memory leak.
- When you need to transfer the ownership of a smart pointer. Use std::move(). This basically does the same thing as the previous point.
If you can manage your memory well. C++ will be a piece of cake.
Now let's consider the following object-oriented design problem:
Design and implement a simple bank management system that allows the creation and management of different types of accounts in a bank.
// Account.h
#ifndef ACCOUNT_H
#define ACCOUNT_H
#include <iostream>
#include <memory>
class Account {
public:
virtual void deposit(double amount) = 0;
virtual void withdraw(double amount) = 0;
virtual void display() const = 0;
virtual ~Account() = default; // Virtual destructor to ensure proper cleanup
// Factory method to create different account types
static std::unique_ptr<Account> createAccount(int accountType);
};
#endif // ACCOUNT_H
// Account.cpp
#include "Account.h"
class SavingsAccount : public Account {
private:
double balance;
public:
SavingsAccount() : balance(0) {}
void deposit(double amount) override {
if (amount > 0) {
balance += amount;
std::cout << "Deposited: $" << amount << std::endl;
}
}
void withdraw(double amount) override {
if (amount > 0 && balance >= amount) {
balance -= amount;
std::cout << "Withdrawn: $" << amount << std::endl;
} else {
std::cout << "Insufficient balance or invalid amount!" << std::endl;
}
}
void display() const override {
std::cout << "Savings Account Balance: $" << balance << std::endl;
}
};
class CheckingAccount : public Account {
private:
double balance;
public:
CheckingAccount() : balance(0) {}
void deposit(double amount) override {
if (amount > 0) {
balance += amount;
std::cout << "Deposited: $" << amount << std::endl;
}
}
void withdraw(double amount) override {
if (amount > 0 && balance >= amount) {
balance -= amount;
std::cout << "Withdrawn: $" << amount << std::endl;
} else {
std::cout << "Insufficient balance or invalid amount!" << std::endl;
}
}
void display() const override {
std::cout << "Checking Account Balance: $" << balance << std::endl;
}
};
std::unique_ptr<Account> Account::createAccount(int accountType) {
switch (accountType) {
case 1:
return std::make_unique<SavingsAccount>();
case 2:
return std::make_unique<CheckingAccount>();
default:
return nullptr;
}
}
// Bank.h
#ifndef BANK_H
#define BANK_H
#include "Account.h"
#include <vector>
class Bank {
private:
std::vector<std::unique_ptr<Account>> accounts;
public:
void addAccount(std::unique_ptr<Account> account);
void displayAllAccounts() const;
void makeDeposits();
void makeWithdrawals();
};
#endif // BANK_H
// Bank.cpp
#include "Bank.h"
void Bank::addAccount(std::unique_ptr<Account> account) {
accounts.push_back(std::move(account));
}
void Bank::displayAllAccounts() const {
for (const auto &account : accounts) {
account->display();
}
}
void Bank::makeDeposits() {
for (auto &account : accounts) {
account->deposit(100); // Deposit $100 to each account (for example)
}
}
void Bank::makeWithdrawals() {
for (auto &account : accounts) {
account->withdraw(50); // Withdraw $50 from each account (for example)
}
}