3. The C programmers guide to C++¶

In this article you will learn to use C++, if you previously have coded in Arduino or C. We will go through to fundamental C++ constructs and concepts.

Who should read this?¶

I assume that you are familiar with basic C, nothing fancy - just plain main() with function calling and maybe - just maybe, some pointer juggling. Also, if you have coded Arduino and are used to the concepts there - this article is for you!

Read on, and we shall take a tour of the basic features and headaches of C++.

Overview¶

Allow me to start off with a ultra brief history lesson in C++. Back in the days it started its life as an extension to C. This means you still have access to all of C from C++. Originally C++ was named C with Classes, later the named was changed to C++. The name comes from the increment by one operator in C: ++. The symbolism is to indicate that C++ is C incremented or enhanced.

C++ introduces a set of new language constructs on top of C, and also tightens some type casting rules that are looser in C.

In this article we will examine the following C++ features:

Classes
Inheritance
Contructors
Namespaces
References
The rule of 3

Let’s dive in.

Classes¶

If you ever heard of object oriented programming, you might have heard about classes. I believe the idea behind the class concept, is best explained by an example.

Let’s say we want to programmatically represent a rectangle (the geometric shape, that is). Our rectangle will have the following properties:

X offset (X coordinate of upper left corner)
Y offset (Y coordinate of upper left corner)
Width
Height

In C code you should normally create a struct type that represents the collection of properties like this:

struct Rect {
    int x;
    int y;
    int width;
    int height;
};

If we want to create a Rect variable in C, we now do:

struct Rect windowFrame;
windowFrame.x = 10;
windowFrame.y = 10;
windowFrame.width = 500;
windowFrame.height = 500;

A struct in C provides a great way of grouping properties that are related. In C++ we achieve the same with the class keyword:

class Rect {
public:
    int x;
    int y;
    int width;
    int height; 
};

Note that apart from the word change from struct to class, the only difference is the public keyword. Do not mind about this know, we shall get back to it.

Now in C++ we have declared the class Rect, and we can use it like this:

Rect winFrm;
winFrm.x = 10;
winFrm.y = 10;
winFrm.width = 500;
winFrm.height = 500;

Notice we just declare the type Rect, no need for the extra keyword struct, like in C.

What we have in fact created now are an instance of our class Rect. An instance is also called an object.

Adding functions to classes¶

Let’s say we want a function to calculate the area of our rectangle. Simple in C:

int calcArea(struct Rect rct)
{
    return rct.width * rct.height;
}

If I were to write the same function in C++, I could then use the class Rect and not the the struct. To rewrite the function to handle the Rect, I need only to remove the struct keyword and the function would work with C++ class types. However, the concept of object oriented programming teaches us to do something else.

We should group functionality and data. That means our Rect class should itself know how to calculate its own area. Just like the Rect has widthand height properties, it should have an area property.

We could define an extra variable in the class, like this:

class Rect {
public:
    int x;
    int y;
    int width;
    int height; 
    int area; // a new area variable
};

This would be highly error prone though. Since we have to remember to update this variable everytime we change widthor height. Let us instead define area as a function that exists on Rect. The complete class definition will look like this:

class Rect
{
public:
    int x, y;
    int width, height;

    int area()
    {
        return width * height;
    }
};

Now our Rect class consists of the 4 variables and a function called area(), that returns an int. A function that is defined on class like this, is called a method.

We can use the method like this:

Rect winFrm;
winFrm.x = 10;
winFrm.y = 10;
winFrm.width = 500;
winFrm.height = 500;

int area = winFrm.area();

This is the idea of object oriented coding - where data and related functionality are grouped together.

Access levels¶

As promised earlier let us talk briefly about the public keyword. C++ lets you protect variables and methods on your classes using 3 keyword: public, protected and private.

So far we only seen public in use, because it allows us to access variables and methods from outside the class. However, we can use the other keywords to mark variables or methods as inaccessible from outside the class. Take an example like this:

class CreditCard
{
private:
    const char *cardNumber;

protected:
    const char *cardType;
    const char * cardholderName;
    int expirationMonth;
    int expirationYear;

public:
    const char *cardAlias;

    int getAmount();
};

In this example we created the class CreditCard, that defines a persons credit card with all the data normally present on a credit card.

Some of the variables are sensitive and we don’t want developers to carelessly access them. Therefore we can use access protection levels to block access to these from code outside the class itself.

Private members¶

The variable cardNumber is marked as private. This means it is visible only from inside the CreditCard class itself. No outside code or class can reference it. Not even a subclass of CreditCard have access to it. (We will get to subclasses in the next section.)

You should use private properties only when you actively want to block future access to variables or methods. Don’t use it if you just can’t see any reason not to. The paradigm should be to actively argue that future developers shall never access this variable or method.

Unfortunately in C++ this is the default access level. If you do not mark your members with public or protected, they become private by default.

Protected members¶

All protected variables and methods are inaccessible from outside the class, just like private variables. However, subclasses can access protected variables and methods.

If you have a variable or method that should be not be accessible from outside code, you should mark it as protected.

Public members¶

Public variables and methods are accessible both from outside the class and from subclasses. When a method is public, we can call it from outside the class, as we saw done with the area() method.

Inheritance¶

Inheritance is classes standing on the shoulders of each other. If classes are one leg of object oriented programming, inheritance are the other.

Let us continue the rectangle example. I have heard about an exciting new trend called 3D graphics! I really want my Rect shape to support this extra dimension. At the same time I already use my existing class Rect many places in my code, so I cannot modify it.

My first thought is to just reimplement the class as a 3D shape. Unfortunately my code is open source and I do not want to loose any street cred in the community, by not following the DRY (Don’t Repeat Yourself) paradigm.

It is now inheritance comes to the rescue. We can use it to create a new class Rect3D that builds upon the existing Rect class, reusing the code and extending the functionality:

class Rect3D : public Rect
{
public:
    int z;
    int depth;

    int volume()
    {
        return width * height * depth;
    }
}

See the colon at the first line? It defines that our Rect3D class inherits from Rect. We say that Rect3D is a subclass of Rect and that Rect is the parent class of Rect3D.

The magic here is the variables x, y, width and height now exists on Rect3D through inheritance. The same is true for the method area().

Inheritance takes all methods and variables (also called properties) and makes them present on subclasses.

The public keyword instructs that public members on the parent class are still public on the subclass.

Unfortunately in C++ the default inheritance access level is private. This means you almost always need to declare the inheritance as public.

Let’s try our new 3D class:

Rect3D cube;
cube.x = 10;
cube.y = 10;
cube.z = 10;

cube.width = 75;
cube.height = 75;
cube.depth = 75;

int area = cube.area(); // gives 5625
int volume = cube.volume(); // gives 421875

Now we have two classes, where one (Rect3D) inherits from the other (Rect). Our code is kept DRY and existing code that uses Rect is not affected by the existence of Rect3D.

Overloading methods¶

Luckily for us, we denote area and volume different. This means that the method volume() in Rect3D, does not conflict with the exitsing method area(). However, we are not always that lucky.

Say we had added a method that calculated the surface area of the shape. In the two dimensional Rect the surface and area are the same, so a surface() method is trivial:

int surface()
{
    return area();
}

Our method simply calls the existing area() method and returns its result. But now Rect3D inherits this behaviour - which is incorrect in three dimensions.

To get the surface area of a cube we must calculate the area of each side, and sum for all sides. We use method overloading to re-declare the same method on Rect3D:

int surface()
{
    return width  * height * 2
         + width  * depth  * 2
         + height * depth  * 2;
}

Now both classes declare a method with the same name and arguments. The effect is Rect3D‘s surface() method replaces the method on its parent class.

A complete exmaple of the code is:

class Rect {
public:
    int width, height;

    int area()
    {
        return width*height;
    }

    int surface()
    {
        return area();
    }
};

class Rect3D : public Rect
{
public:
    int depth;

    int surface()
    {
        return width  * height * 2
             + width  * depth  * 2
             + height * depth  * 2; 
    }
};

Multiple inheritance¶

Classes in C++ can, as in nature, inherit from more than one parent class. This is called multiple inheritance. Let us examplify this by breaking up our Rect into two classes: Point and Size:

class Point
{
public:
    int x, y;
};

class Size
{
public:
    int width, height;

    int area()
    {
        return width * height;
    }
};

If we now combine Point and Size, we get all the properties needed to represent a Rect. Using multiple inheritance we can create a new Rect class that build upon both Point and Size:

class Rect : public Point, public Size
{

};

Our new Rect class does not define anything on its own, it simply stands on the shoulders of both Point and Size.

Inbreeding¶

When using multiple inheritance you should be aware of what is called the diamond problem. This occurs when your class inherits from two classes with a common ancestor.

In our geometry example, we could introduce this diamond problem by letting both Point and Size inherit from a common parent class, say one called: Shape.

In C++ there are ways around this issue called virtual inheritance, it is an advanced topic though. In this article we will not go into detail about this - you should just know that the problem is solvable.

Constructors¶

A constructor is a special method on a class that gets called automatically when the class in created. Constructors often initialize default values of member variables.

When we develop for embedded systems, we cannot assume variable values are initialized to 0, upon creation. For this reason we want to explicitly set all variables of our Rect class to 0:

class Rect
{
public:
    int x, y;
    int width, height;

    Rect()
    {
        x = y = 0;
        width = height = 0;
    }

    int area()
    {
        return width*height;
    }
};

Notice the special contructor method Rect() has no return type - not even void! Now we have created a constructor that sets all member variables to zero, so we ensure they are not random when we create an instance of Rect:

Rect bounds;
int size = bounds.area(); // gives 0

Our constructor is executed upon creation of the bounds variable.

When a constructor takes no arguments, as our, it is called the default constructor.

Non-default Constructors¶

We can declare multiple constructors for our class in C++. Constructors can take parameters, just as functions can.

Let us add another constructor to Rect that takes all the member variables as parameters:

class Rect
{
public:
    int x, y, width, height;
    
    // The default constructor
    Rect()
    {
        x = y = width = height = 0;
    }

    // Our special contructor
    Rect(int _x, int _y, int _w, int _h)
    {
        x = _x;
        y = _y;
        width = _w;
        height = _h;
    }

    int area() { return width*height; }
};

Now we have a special constructor that initializes a Rect object with a provided set of values. Such contructors are very convenient, and make our code less verbose:

Rect bounds; // default constructor inits to 0 here
bounds.x = 10;
bounds.y = 10;
bounds.width = 75;
bounds.height = 75;

// now the same can be achived with a single line
Rect frame(10,10, 75, 75);

When you call a special constructor like Rect(10,10,75,75) the default constructor is not executed! In C++ only one constructor can be executed, they can not be daisy chained.

Namespaces¶

When developing your application you might choose class names that already exists in the system. Say you create a class called String, changes are that this name is taken by the system’s own String class. Indeed this is the case in OpenMono SDK.

To avoid name collisions for common classes such as Array, String, Buffer, File, etc. C++ has a feature called namespaces.

A namespace is a grouping of names, inside a named container. All OpenMono classes provided by the SDK is defined inside a namespace called mono. You can use double colons to reference classes inside namespaces:

mono::String str1;
mono::io::Wifi wifi;
mono::ui::TextLabelView txtLbl;

Here we define instances (using the default constructor) that are declared inside the namespace mono and the sub-namespaces: io and ui.

Declaring namespaces¶

So far in this guide, we have only seen classes declared in global space. That is outside any namespace. Say, we want to group all our geometric classes in a namespace called geo.

Then Rect would be declared as such:

namespace geo {
    class Rect
    {
    public:
        int x,y,width,height;

        // ...
    };
}

Now, any code inside the namespace geo { ... } curly braces can reference the class Rect by its name. However, any code outside the namespace must define the namespace as well as the class name: geo::Rect.

Namespaces can contains other namespaces. We can create a new namespace inside geo called threeD. Then, we can rename Rect3D to Rect and declare it inside the threeD namespace:

namespace geo {
    namespace threeD {
        class Rect : public geo::Rect
        {
            int z, depth;

            // ...
        };
    }
}

The using directive¶

If you are outside a namespace (like geo) and often find yourself referencing geo::Rect, there is a short cut. C++ offers a using directive much like C# does.

The using directive imports a namespace into the current context:

using namespace geo;

Rect frame;

Now you do not have to write geo::Rect, just Rect - since geo has become implicit.

If you look through OpenMono SDK’s source code, you will often see these using statement at the beginning of header files:

using namespace mono;
using namespace mono::ui;
using namespace mono::geo;

This reduces the verbosity of the code, by allowing referencing classes without namespace prefixes.

Using a single class¶

Another less invasive option is to import only a specific class into the current context - not a complete namespace. If you now you are only going to need the geo::Rect class and not any other class defined in geo, you can:

using geo::Rect;

Rect frame;

This imports only the Rect class. This allows you to keep your context clean.

Tip

On a side note, remember that importing namespaces has no effect on performance. C++ is a compiled language, and namespaces does not exist in binary. You can declare and import as many namespaces as you like - the compiled result is not affected on performance.

References¶

C++ introduces an alternative to C pointers, called references. If you know C pointers, you are familiar with the * syntax. If you don’t, just know that in C you can provide a copy of a variable or a pointer to the variable.

In C you denote pointer types with an asterisk (*). C++ introduces references denoted by an ampersand (&), which are somewhat like pointers.

A reference in C++ is constant pointer to another object. This means a reference cannot be re-assigned. It is assigned upon creation, and cannot be changed later:

Rect frame = Rect(0,0,25,25);
Rect& copy = frame;
Rect frm2;
copy = frm2; // AArgh, compiler error here!!

The copy variable is a reference to frame - always. In contrast to pointers in C, you do not have to take the address of an variable to assign the reference. C++ handles this behind the scenes.

Reference in functions¶

A great place to utilze references in C++ is when defining parameters to functions or methods. Let us declare a new method on Rect that check if a Point is inside the rectangles interior. This method can take a reference to such a point, no reason to copy data back and forth - just pass a reference:

class Rect
{
public:
    // rest of decleration left out

    bool contains(const Point &pnt)
    {
        if (   pnt.x > x && pnt.x <= (x + width)
            && pnt.y > y && pnt.y <= (y +height))
            return true;
        else
            return false;
    }
}

Our method takes a reference to a Point class, as denoted by the ampersand (&). Also, we have declared the reference as const. This means we will not modify the pnt object.

If we left out the const keyword, we are allowed to make changes to pnt. By declaring it const we are restraining ourselves from being able to modify pnt. This help the C++ compiler create more efficient code.

The rule of 3¶

I shall briefly touch the Rule of Three concept, though it is beyond the scope of this article.

When you assign objects in C++ its contents is automatically copied to the destination variable:

Rect rct1(5,5,10,10); // special constructor
Rect rct2; // default constructor
rct2 = rct1; // assignment, rct1 is copied to rct2

All of Rect member variables are automatically copied by C++. This is fine 90% of the time, but there are times when you need or want special behaviour. Often in these cases a advanced behaviour is needed, for example to implement reference counting or similar.

As an example here, we just want to modify our Rect class to print to the console everytime it is copied.

To achieve this, we must overwrite two implicit defined methods in C++. These are the copy constructor and the assignment operator.

The Copy Constructor¶

The copy constructor is a special constructor that takes an instance of an object and initializes itself as a copy. C++ calls the copy constructor when creating a new variable from an existing one. These are common examples:

Rect frame(0,0,100,100); // special constructor
Rect frame2 = frame; // copy constructor
someFunction(frame); // copy constructor again

When we create a new instance by assigning an existing object the copy constructor is used. Further, if we have a function or method and takes a class type as parameter, the function is provided with a fresh copy of the object by the copy contructor.

To create your own copy constructor you define it like this:

class Rect
{
public:
    // copy constructor
    Rect(const Rect &other)
    {
        x = other.x;
        y = other.y;
        width = other.width;
        height = other.height;

        printf("Hello from copy constructor");
    }
};

We left out the other constructors, and members in this example. The copy constructor is a constructor method that takes a const reference to another instance of its class.

In the Rect class we copy all variables (the default behaviour of C++, if we had not defined any copy constructor) and prints a line to the console.

This serves to demonstrate that you can define exactly what it means to assign your class to a new variable. You can make your new object a shallow copy of the original or change some shared state.

The Assignment Operator¶

There is still the other case: assignment oprator. It is where the default assignment operator is used. The default assignment operator occurs when:

Rect view(10,10,100,100); // convenient constructor
Rect bounds;   // default constructor
bounds = view; // assignment operator

Here we create a instance with some rectangle view, and a zeroed instance bounds. If we want the same behaviour as with the copy constructor, we need to declare the assignment operator on Rect:

class Rect
{
public:
    // rest of class content left out

    // assignment operator
    Rect& operator=(const Rect &rhs)
    {
        x = rhs.x;
        y = rhs.y;
        width = rhx.width;
        height = rhs.height;

        printf("Hello from assignment operator");
        return *this;
    }
};

Just like the copy constructor, the assignment operator takes a const reference to the object that need to be assigned (copied). But its assignment must also return a reference of itself, as defined by Rect&. This is also why we have to include the return *this statement. In C++ this is a pointer to the instance of the class - the object itself.

A C pointer juggling champ, will recognize that we dereference the pointer by adding the asterisk (*) in front.

Just as is the case with the copy constructor, we can now define the assignment behavior of Rect. Here (again), it is illustrated by printing to the console upon assignment.

The Deconstructor¶

This is the last part of the Rule of Three in C++.

THe deconstructor is the inverse of the constructor - it is called when an object dies or rather - is deallocated. Objects get deallocated when they go out of scop. As is the case when a function or method returns.

To follow our previous examples we want the deconstructor to just print to the console.

class Rect
{
public:
    // rest of class content is left out

    //the deconstructor
    ~Rect()
    {
        printf("Hello from the de-constructor");
    }
};

The deconstructor is defined as the class’ name with a tilde (~) in front. A deconstructor takes no arguments.

Now this is the rule of three. Defining the:

copy constructor : Class(const Class &other)
assignment operator : Class& operator=(const CLass &rhs)
deconstructor : ~Class()

When you create your own C++ classes, think about these three. Mostly you don’t have to implement them, but in some cases you will.