Memory allocation, polymorphism, textures in SFML

Changing to C++17 standard

From this classes we will be using some elements of C++ standard that are available only in the newest version of the standard. In time of this script creation a C++17 is the newest one, however Qt Creator creates all projects using C++11 standard. In order to change the standard it is needed to edit, a .pro file and change:

CONFIG += console c++11

, to:

CONFIG += console c++17

, than right click on project name and choose Run qmake. Do this after each new project creation!

Pointers

Variables that we have created through a simple variable declaration had their life limited to one scope - a fragment of code, most often separated by braces, e.g. the body of a function or the interior of a loop. These variables were created in the memory section called stack. Variables and such objects are automatically deleted when they leave the scope where they were allocated, e.g. when they leave a function. In most cases, this is desirable and relieves the programmer of the need for manual memory management.

However, there are cases where it is advisable to separate the lifetime of an object from the scope in which it was created. For this purpose, we use the allocation on the heap.

Historically, in order to allocate variables on the heap, the operator new was used. The variable created in this way had to be removed with the delete operator. Since dealocation operations are easy to forget, many programs still have a problem called memory leaks - THIS APPROACH IS NOT RECOMMENDED.

🛠🔥 Assignment 🔥🛠

Perform the following code and observe the use of memory in the Task Manager (Ctrl + Alt + Delete):

for (int i = 0; i < 10000000000; i++) {
    int *number = new int;
}
std::getchar();

In order to avoid memory leaks, C++11 introduced new types of so-called smart pointers, including std::unique_ptr and std::shared_ptr. In this instruction, we will omit std::shared_ptr, because it is used to store objects that have multiple owners (for example, they are simultaneously in more than one vector). std::unique_ptr cannot be copied - an object to which it points has only one owner.

std::unique_ptr is a class in which the constructor allocates an object with new, and the destructor destroys it with delete. By replacing the use of ordinary C-style pointers with std::unique_ptr you can easily prevent memory leaks. A variable of the std::unique_ptr type will call delete on the stored pointer when it goes out of the scope in which it was declared.

🛠🔥 Assignment 🔥🛠

Change the previous code so that instead of the new operator it will use the std::unique_ptr type. Use the std::make_unique function to create std::unique_ptr. Observe memory usage.

Tip: as we pass the type of variable we want to create on the heap to the std::make_unique function, we can optionally use the keyword auto to bypass declaring an object with a long type name: std::unique_ptr<int>. This will not result in a loss of readability of the code.

After reading the description above and completing the task, you may get the impression that the allocation on the stack and the allocation on the heap using std::unique_ptr are functionally identical. In both cases, the object is destroyed after leaving the range in which it was declared. So why use std::unique_ptr?

In C++11/17, in addition to the type std::unique_ptr, there is also a move semantics - the ability to "transfer" the object ownership to another scope or container. While in the case of objects declared on the stack, this transfer will copy at least some of the fields of the transferred object, in the case of std::unique_ptr, the transfer will not copy the object on the heap.

🛠🔥 Assignment 🔥🛠

Try to compile the code below:

std::vector<std::unique_ptr<int>> some_pointers;
for (int i = 0; i < 10000000000; i++) {
    auto number = std::make_unique<int>();
    some_pointers.emplace_back(number);
}

Consider why this program does not work.

Bearing in mind that the object pointed to by std::unique_ptr can have only one owner, pass the ownership to the vector by calling std::move on the number variable.

Polymorphism

Using pointers to the objects from the heap instead of objects on the stack allows you to easily interpret a pointer to a derived class object as a pointer to a base class object.

🛠🔥 Assignment - Introduction 🔥🛠

Using example classes from the previous instruction, try to compile the code below. This short code shows, that a derived class object (Car), can be stored using a pointer to the base class (Vehicle). The object will not loose any additional information stored or functionality provided by derived class, however access to this parts is different.

std::unique_ptr<Vehicle> skoda_superb_as_vehicle = std::make_unique<Car>(
    "Skoda Superb", "Gasoline", 200, true);

std::cout << "Name: " << skoda_superb_as_vehicle->name() << std::endl;
std::cout << "Has ABS: " << skoda_superb_as_vehicle->has_abs() << std::endl;

Knowing that we treat our car as a Vehicle class object and analyzing the methods that are implemented in this class, consider why this program does not work. We will try to fix this in next assignment.

Note the difference in a way we are accessing object fields (-> instead of .). Why is that?

For the base class to be polymorphic it must contain at least one virtual method (virtual keyword). Virtual methods allow to overload them in the derived class in such a way that when we call them through the pointer to the base class, the proper implementation of the derived class will be called. If the function did not have the keyword virtual, calling it through the pointer to the base class would call the implementation from the base class, regardless of the fact that there is an overloaded version of the function in the derived class.

It is very important that every base polymorphic class contains a declaration of virtual destructor. If such a declaration is not present in the base class, deleting the derived class object through the pointer to the base class is an undefined behavior and may lead to memory leak, as only the base class destructor will be called. The simplest declaration of a virtual destructor is as follows:

virtual ~ClassName() = default;

🛠🔥 Assignment - Introduction cont. 🔥🛠

In public space of Vehicle class add virtual default destructor:

virtual ~Vehicle() = default;

This however still does not solve our problem, we made a base class polymorphic, but the compiler still does not know we want to treat skoda_superb_as_vehicle as a Car type object.

In order to reinterpret the base class pointer as a derived class pointer, you should dereference the pointer ("pull" the object the pointer points to) and use a special dynamic_cast function. A method std::unique_ptr::get can be used in order to get a standard pointer to the managed object. Than, dynamic_cast has to be used in order to cast the pointer to derived class pointer. E.g.:

Car *car = dynamic_cast<Car *>(skoda_superb_as_vehicle.get());

🛠🔥 Assignment - Introduction cont. 🔥🛠

Modify the code in that it is finally possible to access the has_abs method of skoda_superb_as_vehicle object.

The upside of using dynamic_cast in this case (instead of using static_cast) is that dynamic_cast will check if casting of the types is possible, if not will return a nullptr.

🛠🔥 Assignment - Introduction cont. 🔥🛠

Add a Bike class definition from the previous classes to your project and try the code below:

std::vector<std::unique_ptr<Vehicle>> vehicles;
vehicles.emplace_back(std::make_unique<Car>("Skoda Superb", "Gasoline", 200, true));
vehicles.emplace_back(std::make_unique<Bike>());

for (auto &v : vehicles) {
    Car *some_car = dynamic_cast<Car *>(v.get());
    if (some_car != nullptr) { // cast successful
        std::cout << v->name() << ": abs=" << some_car->has_abs() << std::endl;
    } else { // nope
        std::cout << v->name() << ": not a Car" << std::endl;
    }
}

Note: note that in this case, due to the direct transfer of the result of the function std::make_unique to the emplace_back method (without using the variable to which we assign the value returned by std::make_unique), it was not necessary to use std::move.

Polymorphism in SFML

All "drawable" classes from the SFML library inherit the polymorphic class sf::Drawable. This means that we can store all scene objects in one container as std::unique_ptr<sf::Drawable>.

🛠🔥 Assignment - vector of sf::Drawable 🔥🛠

Modify an entry code from the first SFML classes to operate on std::vector vector of std::unique_ptr<sf::Drawable>. Create a vector to store multiple object of different types shapes:

std::vector<std::unique_ptr<sf::Drawable>> shapes;

Than create a function that will accept a reference to the vector and fill the collection with shapes inside the function. Remember to use std::move when emplacing objects to vector:

void create_shapes(std::vector<std::unique_ptr<sf::Drawable>> &shapes_vec) {}

Now in the main loop of the program, you can draw all the objects stored in the vector independent on their specific type, as sf::RenderWindow::draw method accepts a reference to sf::Drawable:

for(auto &s : shapes) {
    window.draw(*s);
}

Textures/sprites in SFML

Texturing is the application of a flat, two-dimensional image to the geometric shapes in order to give the object a look closer to the real one. In the case of two-dimensional graphics, the term sprite is used - a flat bitmap that is rendered in whole or in part at a specific point on the screen.

How to obtain textures?

There are many websites with free multimedia resources such as textures, sprites and sounds - especially for non-commercial use. Databases usually have a well organised directory, which makes it easy to find the right resource. Remember - if you are using graphics downloaded from the Internet, check the license under which resources have been made available. Often, the authors expect only the so-called attribution - small mention in program information, where the resources come from and who created them.

The links to a few textures that you can start with are: grass, wall, guy.

In SFML, texture is the image that is stored in memory, while sprite is the shape that will be displayed on the screen and that can be associated with the texture. An additional description of the sprites can be found here

The simplest way to use texture is to load a bitmap from a file into the sf::Texture object, then create the sf::Sprite object and associate it with the loaded texture.

sf::Texture texture;
if (!texture.loadFromFile("grass.png")) {
    std::cerr << "Could not load texture" << std::endl;
    return 1;
}

sf::Sprite sprite;
sprite.setTexture(texture);

sf::Sprite class inherits from sf::Drawable and sf::Transformable classes, so it contains methods that allow geometric transformations (shifting, rotating, scaling). Moreover, thanks to polymorphism sprites can be stored in one container with shapes such as sf::RectangleShape.

🛠🔥 Assignment - Textures 🔥🛠

Load the supplied textures into the program. Create one sprite for each of them, display them on the screen in different places and with different scales (HINT: sf::Transformable::setScale).

If needed a sprite can display only a fragment of texture associated with it. In order to limit the texture area displayed sf::Sprite::setTextureRect(sf::RectInt) method can be used. E.g. in order to display only face of our guy we can do:

sf::Texture texture_guy;
if(!texture_guy.loadFromFile("guy.png")) { return 1; }

sf::Sprite guy;
guy.setTexture(texture_guy);
guy.setTextureRect(sf::IntRect(10, 20, 20, 15)); //left, top, width, height

It is also possible to make a texture cover a larger area. In order to do that we have to set a Repeated property of the texture by calling sf::Texture::setRepeated(bool) method resulting in tilling (the texture will be repeated to fill the area). Than a sprite rectangle has to be set to be larger in order for texture to fill the area. E..g.:

sf::Texture texture_wall;
if(!texture_wall.loadFromFile("wall.png")) { return 1; }
texture_wall.setRepeated(true);

sf::Sprite wall;
wall.setTexture(texture_wall);
wall.setScale(0.3, 0.3);
wall.setTextureRect(sf::IntRect(0, 0, 500, 500));

🛠🔥 Assignment - Textures cont. 🔥🛠

Use tiling and grass texture to create a background in the program. Determine the appropriate TextureRect for the sprite representing the background so that it covers the entire window size. HINT: Use window.getSize().x and window.getSize().y to get the windows size.

Important: Linking a texture to a sprite does not copy its contents - if the sf::Texture object is removed and we try to display the sprite to which it was linked, a white fill will appear in place of the texture.

It's also a good idea to keep the number of textures and operations on them to a minimum - for example, use the same texture for multiple objects. For simple games that don't have many resources, it's best to load all the necessary textures at the start of the program and keep them in memory until the end of the program.

Final assignment 🔥🛠

1. Labyrinth

Use the supplied textures to build a Labyrinth game.

• Set the background of the game
• Add several objects to the scene - walls with appropriate texture.
• Add a character that you can move with the cursor keys.
• Add collision detection so that it is impossible to pass through walls.

The final effect should be similar to that in the film.

Authors: Dominik Pieczyński, Jakub Tomczyński