cpp-reference-card.org

CPP / C++ and General Programming Ref Card

Modern C++ Style

Modern C++ code style based on C++11 and C++ core guidelines:

Miscellaneous

• Prefer usign nullptr instead of 0 or NULL for null pointer.
• Prefer enum classes (scoped enums) instead of the old C-enums as the old enums are non strongly typed and prone to name clashes.
• Prefer defining type alias with “using” keyword instead of “typedef”.
• In order to avoid unecessary copies, pass large objects by reference, const reference or by pointer instead of passing them by value.
• Prefer using C++ string (std::string or std::wstring) to C-string const char*, char* and so on.
• Use standard STL containers std::vector, std::deque, std::map, std::unordered_map instead of custom non-standard containers.
• Instead of using heap-allocated arrays (A* pArray = new A[10];) when the array size is not known at compile-time, use std::vector (most likely), std::deque, std::list or any other STL container for avoiding accidental memory leaks and boilerplate memory management code. Note: std::vector already wraps a heap-allocated C-array.

// Avoid 
typedef double Speed;
typedef double (* MathFunction) (double);

// Better 
using Speed = double;
using MathFunction = double (*) (double);

Prefer using enum class to enums

Lots of possible runtime erros hard to detect can be avoided by using C++11 enum classes instead of enums that are vulnerable to implicit conversion to integer or any other type. Enum classes avoids those implicit conversion buges by yielding compile-time errors whenever there is an enum class implicit conversion that should be made explicit with static cast. Another problem of old enums is that they are not scoped which can lead to name clashes in large code.

Avoid:

enum ErrorCode {
   ErrorCode_OK,
   ErrorCode_SYSTEM_FAILURE, 
   ErrorCode_LOW_VIOLTAGE, 
  ... 
  ..
};

ErroCode x = ::getOperationStatus();

if(error == ErrorCode_OK){
  std::cout << "Proceed to next step" << "\n";
}

int x;
// Implicit convernsion bug 
x = error; 

Better:

enum class ErrorCode {
   OK,
   SYSTEM_FAILURE, 
   LOW_VIOLTAGE, 
  ... 
  ..
};

ErroCode x = ::getOperationStatus();

if(error == ErrorCode::OK){
  std::cout << "Proceed to next step" << "\n";
}
 
int x; 
// Compile-time error !
//-------------------
 x = error; 

// Conversion only possible with static cast 
x = static_cast<int>(error);

Function Parameter Passing:

• Prefer passing parameters by value, reference or const reference rather than by pointer as in old C++ codes that looks like C with classes.

Avoid:

double vector_norm(const Vector* vec)
{
  // ... compute Euclidian norm ... 
   return value;
}

Better:

double vector_norm(Vector const& vec)
{
  // ... compute Euclidian norm ... 
   return value;
}

Function Parameter Passing of Polymorphic Objects

• Pass polymorphic objects by pointer (T*) or referece (T&) or (const T&) rather than by smart pointer. Functions that accepts referece or pointer are more flexible tha functions that accepts smart pointers. – (Core Guideline F7)

Example: Class hierarchy.

 class Shape{
    public: 
      virtual double       GetArea() const = 0;
      virtual std::string  Name() const = 0;
      virtual ~Shape() = default;
 };

class Square: public Shape  {  ...   };
class Circle: public Shape  {  ...  };

std::unique<Shape> shapeFactory(std::string cosnt& name)
{
   if(name == "square") return std::make_unique<Square>();
   if(name == "circle") return std::make_unique<Circle>();
   return nullptr;
}

Avoid:

// Avoid: 
void printShapeInfo(std::unique_ptr<Shape> const& shape)
{
  std::cout << "The shape name is " << shape->Name()  
            << " ; area is "        << shape->Area() << "\n" ;
}
 
// Or: 
void printShapeInfo(std::shared_ptr<Shape> const& shape)
{
  std::cout << "The shape name is " << shape->Name()  
            << " ; area is "        << shape->Area() << "\n" ;
}

Better:

• The previous functions only work with smart pointers, the following functions using reference or pointer works with smart pointers or stack allocated objects.

void printShapeInfoA(Shape const& shape)
{
  std::cout << "The shape name is " << shape.Name()  
            << " ; area is "        << shape.Area() << "\n" ;
}

// If the function can accept a no-shape parameter, better use pointer: 
void printShapeInfoB(Shape* pShape)
{
  if(pShape == nullptr)
     return; // Do nothing.
  std::cout << "The shape name is " << shape->Name()  
            << " ; area is " << shape->Area() << "\n" ;
}

Square shapeStack;
std::unique<Shape> shapeHeap = shapeFactory("square");
printShapeInfoA(shapeStack);
printShapeInfoA(*shapeHeap);

printShapeInfoB(&shapeStack);
printShapeInfoB(shapeHeap.get());

Function Return Value

Many old C++ codes avoided returning large objects by value due to the copy-constructor overhead in C++98. In those codes, functions returned the result by setting some parameter passed by pointer or reference.

Old C++: (Pre C++11 or C++98)

• Code afraid of returning by value due to the copy overhead.

// Code afraid of returnig by value or returning multiple-values as parameter. 
void sum(std::vector<double> const& xs, std::vector<double> const& ys, std::vector<double>& result)
{  
    // Pre-condition 
    assert(xs.size() == ys.size() && xs.size() == result.size());
    for(size_t i = 0; i < xs.size(); i++)
       result[i] = xs[i] + ys[i];
}

// Usage: 
std::vector<double> xs;
std::vector<double> ys;
xs.resize(3); 
xs.push_back(1); xs.push_back(4); xs.push_back(5);
ys.resize(3);
ys.push_back(6); ys.push_back(8); ys.push_back(9);

std::vector<double> result(xs.size());
sum(xs, ys, result);
DisplayResult(result);

Modern C++: (>= C++11)

• Returning by value is safe and efficient due to the compiler RVO (Return Value Optimization), copy elision and move semantics (move constructor and move destructor) which eliminates the copy-overhead of temporary objects. Since, C++11 all STL containers implements move semantics member functions which makes returning by value more efficient and safer.
• Remark:
  • Returning by value is safe and efficient in C++11 due to RVO (Return-value optimization) and move semantics.

vstd::vector<double> 
sum(std::vector<double> const& xs, std::vector<double> const& ys)
{  
    // Pre-condition 
    assert(xs.size() == ys.size());
    std::vector<double> result(xs.size());
    for(size_t i = 0; i < xs.size(); i++)
       result[i] = xs[i] + ys[i];

   // Copy may not happen due to move semantics (move member functions)
   // and/or Return-Value Optimization.
   return result;
}

// Usage: 
//----------------------------------//

// Uniform initialization with initializer list 
std::vector<double> xs {1, 4, 5};
std::vector<double> ys = {6, 8, 9};

std::vector<double> result = sum(xs, ys);
// Or:
auto result  = sum(xs, ys);
DisplayResult(result);    

Memory Ownership

Raw pointers should not own memory or be responsible for releasing memory due to them be prone to memory leaks which can happen due to missing call to delete operator; exceptions befored the delete operator; functions with early return multiple return paths; and shared ownership of the heap-allocated memory.

Summary:

• Avoid calling new and delete directly, instead use std::make_unique, std::make_shared from header <memory>.
• Avoid using raw pointers for memory ownership, instead use smart pointers.
• Smart pointers should only be used for heap-allocated objects (objects allocated at runtime), never stack-allocated ones.
• Rule of thumb for choosing std::unique_ptr or shared_ptr
  • If more than one objects need to refere to some heap-allocated object during their entire lifetime, the best choice is std::shared_ptr.

Avoid:

Shape* shapeFactory(std::string cosnt& name)
{
   // WARNING: new operator can throw std::bad_alloc 
   if(name == "square") return new Square();
   if(name == "circle") return new Circle();
   return nullptr;
} 

void clientCode(Shape* sh){
   std::cout << "Name = " << sh->Name() << " ; Area = " << sh->Area() << "\n";
}

// Usage: 
//-------------------------------
Shape* shape = shapeFactory("square"); 
clientCode(shape); 

// Exception happens => Memory Leak!  
// Forget to delete ==> Memory leak!
delete shape;

Better:

• Note: A factory function or any function returning a polymorphic object should preferentially return an unique_ptr smart pointer instead of shared_ptr because unique_ptr has a lower overhead than shared_ptr and it is easier to convert unique_ptr to shared_ptr, but the other way around is harder.

 std::unique_ptr<Shape> 
 shapeFactory(std::string cosnt& name)
 {
    // WARNING: new operator can throw std::bad_alloc 
    if(name == "square") return std::make_unique<Square>(300 ,400);
    if(name == "circle") return std::make_unique<Circle>();
    return nullptr;
 }

 void clientCode(Shape const& sh){
    std::cout << "Name = " << sh.Name() << " ; Area = " << sh.Area() << "\n";
 } 

// Usage: 

// Releases allocated memory automatically when out scope. 
std::unique_ptr<Shape> shape = shapeFactory("square"); 

// Or: 
auto shape = shapeFactory("square"); 
clientCode(*shape);

References and Further Reading

• C++ Core Guidelines
• How to avoid bugs using modern C++ | PVS Studio [BEST]
• C++ Core Guidelines - GSL: Guideline Support Library
• Microsft’s Guideline Support Library:
  • https://github.com/Microsoft/GSL/
• GSL Lite: Guidelines Support Library for C++98, C++11 up
  • https://github.com/martinmoene/gsl-lite
• C++ Core Guidelines: Rules for Error Handling - ModernesCpp.com
• Welcome Back to C++ (Modern C++) | Microsoft Docs
• The View from Aristeia: Effective C++11: Content and Status
• ROOT::RWhy! | ROOT a Data analysis Framework

Standard Library Reference Card

STL Components

• Containers - standard collections or data structures, they are a fundamental building block of most programming languages, in C++ the addition benefit is that most of them abstracts away the memory allocation as they can grow or shrink during the program runtime.
  • Sequential
    • vector
    • deque
    • array
    • list
    • forward list
    • valarray [DEPRECATED] - It would provide a Fortran-like fixed size array for linear algebra. But the STL implementation is incomplete.
  • Associative
    • Ordered Associative Container
      • map - key-value data structure, also known as dictionary. A map always have unique keys. hash-map, hash table and so on.
      • set - A set is data structure which cannot have any repeated values.
      • multimap - A multimap can have repeated keys.
      • multiset
    • Unordered Associative Containers
      • unordered_map
      • unordered_set
• Iterators
• Algorithms
• Adapters
  • Queue
  • Stack
• Functors - Function-objects or objects that can be called like a function. Functors have several use cases in the STL, for instance many STL containers and algorithms expects functors as arguments or optional arguments and also the STL provides many standard functors in the header <functional>
  • Standard functors:
    • plus<T>
    • minus<T>
    • multiplies<T>
    • divides<T>
    • modules<T>
    • equal_to<T>
    • not_equal_to<T>
    • greater<T>
    • greater_equal<T>
    • less<T>
    • less_equal<T>
• Allocators

Further references:

• Standard library header <functional> - cppreference.com
• STL Function Objects (Functors)

See:

• Helsinki - university - STL Containers
• Helsinki University - STL Algorithms

STL Sequential Container Methods - Cheat Sheet

Use Cases

Use Cases:

• vector
  • Operations where the vector size is known in advance and it is necessary constant access time for random access to any element. Example of use case: linear algebra and numerical algorithms. Insertion of elements at end or at the front is efficient, however it less efficient than the deque container and whenever a new element is added. Vectors are not ideal for operations where the number of elements is not known because its elements are stored in C-array allocated in the heap, as result, all elements are reallocated whenever a new element is added or removed.
  • Use cases:
    • General sequential container
    • Linear algebra and numerical algorithms
    • C++ replacement for C-arrays
    • C-arrays interoperability
• deque
  • Operations with requires fast random access time and fast insertion or deletion of elements at both ends. Unlike vectors, deque is not stored internally as a C-array and unlike vectors, whenever an element is inserted, any reallocation happens which means that deques are more efficient than vectors when the size of container is not known in advance.
  • Use Case:
    • General sequential container
    • Fast random access
    • Number of elements aren’t known in advance.

Member Functions / Methods reference table

Method of Container<T>	Return type	Description	vector	deque	list	array
Element Access
operator[](int n)	T&	return nth-element, doesn’t throw exception.	yes	yes	no	yes
at(int n)	T&	return nth-element, but throws exception.	yes	yes	no	yes
front()	T&	return first element	yes	yes	yes	yes
back()	T&	return last element	yes	yes	yes	yes
data()	T*	Return pointer to first element of container.	yes	no	no	yes
Capacity
size()	size_t	Return number of container elements.	yes	yes	yes	yes
max_size()	size_t	Return maximum container size.	yes	yes	yes	yes
empty()	bool	Return true if container is empty	yes	yes	yes	yes
reserve(size_t n)	void	Reserve a minimum storage for vectors.	yes	no	no	no
resize(size_t n)	void	Resize container to n elements.	yes	yes	yes	no
Modifiers
push_back(T t)	void	Add element at the end of container	yes	yes	yes	no
push_front(T t)	void	Add element at the beggining of container.	yes	yes	yes	no
pop_back()	void	Delete element at the end of container.	yes	yes	yes	no
pop_front()	void	Delete element at beginning of container.	yes	yes	yes	no
emplace_back	void	Construct and insert element at the end without copying.	yes	yes	yes	no
clear()	void	Remove all elements.	yes	yes	yes	no
fill(T t)	void	Fill all elements	no	no	no	yes
Iterator
begin()	iterator	Return iterator to beggining
end()	iterator	Return iterator to end
rbegin()	iterator	Return reverse iterator to beggining
rend()	iterator	Return reverse iterator to end
cbegin()	iterator	Return const iterator to beginning
cend()	iterator	Return const iterator to end
crebegin()	iterator	Return const iterator to beginning
crend()	iterator	Return const iterator to end

Constructors

Vector constructors:

// Empty vector 
>> std::vector<double> xs1
(std::vector<double> &) {}

// Intialize vector with a given size
>> std::vector<double> xs2(5, 3.0)
(std::vector<double> &) { 3.0000000, 3.0000000, 3.0000000, 3.0000000, 3.0000000 }

// Constructor with uniform initialization 
>> std::vector<double> xs4 {1.0, -2.0, 1.0, 10 }
(std::vector<double> &) { 1.0000000, -2.0000000, 1.0000000, 10.000000 }

// =========== Constructors with C++11 auto keyword =============//

>> auto xs1 = vector<double>()
(std::vector<double, std::allocator<double> > &) {}
>> 
>> auto xs2 = vector<double>(5, 3.0)
(std::vector<double, std::allocator<double> > &) { 3.0000000, 3.0000000, 3.0000000, 3.0000000, 3.0000000 }
>> 
>> auto xs3 = vector<double>{1, -2, 1, 1}
(std::vector<double, std::allocator<double> > &) { 1.0000000, -2.0000000, 1.0000000, 1.0000000 }
>> 

Deque constructors:

>> std::deque<int> ds1
(std::deque<int> &) {}
>> 
>> std::deque<int> ds2(5, 2)
(std::deque<int> &) { 2, 2, 2, 2, 2 }
>> 
>> std::deque<int> ds3 {2, -10, 20, 100, 20}
(std::deque<int> &) { 2, -10, 20, 100, 20 }
>> 
// ======== Constructors with auto type inference ========== //
>> auto ds1 = std::deque<int>()
(std::deque<int, std::allocator<int> > &) {}
>> 
>> auto ds2 = std::deque<int>(5, 2)
(std::deque<int, std::allocator<int> > &) { 2, 2, 2, 2, 2 }
>> 
>> auto ds3 = std::deque<int>{2, -10, 20, 100, 20}
(std::deque<int, std::allocator<int> > &) { 2, -10, 20, 100, 20 }
>> 

References:

• vector - C++ Reference
• Containers - C++ Reference
• Choosing the Right Container: Sequential Containers — Embedded Artistry
• STL Sequential Container Member Function Summary

Tips and tricks

Pass containers by reference or const reference

If the intent of the operation is not modify the container, it is preferrable to pass it by const reference in order to avoid copying overhead.

For instance, the function:

double computeNorm(std::vector<double> xs)
{
 // The vector xs is copied here, if it has 1GB of memory.
 // It will use 2GB instead of 1GB!
  ... ... 
}

Should be written as:

double computeNorm(const std::vector<double>& xs)
{
  ... ... 
}
double computeNorm(const std::list<double>& xs)
{
  ... ... 
}
double computeNorm(const std::deque<double>& xs)
{
  ... ... 
}

Use the member function emplace_back to avoid uncessary copies.

Example:

• file: stl-emplace.cpp

#include <iostream>
#include <ostream>
#include <iomanip>
#include <string>
#include <vector>
#include <deque>

struct Product{
        std::string  name;	
        int          quantity;
        double       price;
        Product(){
                std::cerr << " [TRACE] - Empty constructor invoked\n";
        }
        Product(const std::string& name, int quantity, double price):
                name(name),
                quantity(quantity),
                price(price){
                std::cerr << " [TRACE] - Product created as " << *this << "\n" ;
        }
        // The compiler generate an copy constructor automatically,
        // but this one was written to instrument C++ value semantics
        // and check when copies happen.
        Product(const Product& p){
                this->name		= p.name;
                this->quantity	= p.quantity;
                this->price		= p.price;
                std::cerr << " [TRACE] Copy constructor invoked -> copied = " << *this << "\n";
        }
        // Copy assignment-operator
        void operator=(const Product& p){
                this->name		= p.name;
                this->quantity	= p.quantity;
                this->price		= p.price;
                std::cerr << " [TRACE] Copy assignment operator invoked = " << *this << "\n";		
        }
        // Make class printable 
        friend std::ostream& operator<< (std::ostream& os, const Product& p)
        {
                int size1 = 10;
                int size2 = 2;
                return os << " Product{ "
                                  << std::setw(1) << " name = "       << p.name
                                  << std::setw(10) << "; quantity  = "  << std::setw(size2) << p.quantity
                                  << std::setw(size1) << "; price = "      << std::setw(size2) << p.price
                                  << " }";
        }
};


int main(){
        auto inventory = std::deque<Product>();

        // Using push_back
        std::cerr << "====== Experiment .push_back() ======\n";
        std::cerr << " [INFO] - Adding orange with .push_back\n";
        inventory.push_back(Product("Orange - 1kg", 10, 3.50));
        std::cerr << " [INFO] - Adding rice with .push_back \n";
        inventory.push_back({"Rice bag", 20, 0.80});

        // Using emlace_back
        std::cerr << "====== Experiment .emplace_back() ======\n";	
        std::cerr << " [INFO] - Adding apple with .emplace_back \n";
        inventory.emplace_back("Fresh tasty apple", 50, 30.25);
        std::cerr << " [INFO] - Adding soft drink with .emplace_back \n";
        inventory.emplace_back("Soft drink", 100, 2.50);

        std::cerr << " ====== Inventory =======\n";
        // Print inventory
        int nth = 0;
        for(const auto& p: inventory){
                std::cout << "product " << nth << " = " << p << "\n";
                nth++;
        }	
        return 0;
}

Running:

• It can be seen in the program output that .emplace_back doen’t invoke the copy constructor, so it has less overhead than .push_back which copies the passed element.

$ clang++ stl-emplace.cpp -o stl-emplace.bin -g -std=c++11 -Wall -Wextra && ./stl-emplace.bin

====== Experiment .push_back() ======
 [INFO] - Adding orange with .push_back
 [TRACE] - Product created as  Product{  name = Orange - 1kg; quantity  = 10; price = 3.5 }
 [TRACE] Copy constructor invoked -> copied =  Product{  name = Orange - 1kg; quantity  = 10; price = 3.5 }
 [INFO] - Adding rice with .push_back 
 [TRACE] - Product created as  Product{  name = Rice bag; quantity  = 20; price = 0.8 }
 [TRACE] Copy constructor invoked -> copied =  Product{  name = Rice bag; quantity  = 20; price = 0.8 }
====== Experiment .emplace_back() ======
 [INFO] - Adding apple with .emplace_back 
 [TRACE] - Product created as  Product{  name = Fresh tasty apple; quantity  = 50; price = 30.25 }
 [INFO] - Adding soft drink with .emplace_back 
 [TRACE] - Product created as  Product{  name = Soft drink; quantity  = 100; price = 2.5 }
 ====== Inventory =======
product 0 =  Product{  name = Orange - 1kg; quantity  = 10; price = 3.5 }
product 1 =  Product{  name = Rice bag; quantity  = 20; price = 0.8 }
product 2 =  Product{  name = Fresh tasty apple; quantity  = 50; price = 30.25 }
product 3 =  Product{  name = Soft drink; quantity  = 100; price = 2.5 }

Methods of C++ STL Vetor<T>

Vector Class Member	Description
Constructors
vector<a>(int size)	Create a vector of size n
vector<a>(int size, a init)	Create a vector of size n with all elements set to init
vector<a>(a [])	Intialize vector with an C-Array.
Methods
vector<a>[i]	Get the element i of a vector. i ranges from 0 to size - 1
int vector<a>::size()	Get vector size
a vector<a>::at(i)	Get the nth element of a vector and checks if the index is within the bounds
bool vector<a>::empty()	Returns true if vector is empty and false, otherwise.
void vector<a>::resize(int N)	Resize vector to N elements.
void vector<a>::clear()	Remove all elements and sets the vector size to 0.
void vector<a>::push_back(elem a)	Insert element at the end of v.
a vector<a>::begin()	Returns first element.
a vector<a>::end()	Returns last element
void vector<a>::pop_back()	Remove last element of vector.

Associative Container - Map methods

Map is a data structure similar to a hash map, also known as dictionary hash table or dictionary. However, stl std::map is not implemented as true hash table as all data inserted in std::map are sorted. Due to the implementation and sorting, std::map is less performant than std::unordered_map, which is implemented as true hash table, therefore in most cases std::unordered_map is better choice than std::map.

Documentation:

• map - C++ Reference

Method of map<K, V>	Return type
Capacity
empty()	bool	Return true if container empty
size()	size_t	Return number of elements
max_size()	sizet_t	Return maximum number of elements
Element Access
operator[](K k)	V&	Return value associated to key k. It doesn’t throw exception.
at(K k)	V&	Return value associated to key k. Note: it can throw exception.
find(const K& k)	iterator	Search for an element and returns map::end if it doesn’t find the given key.
count(const K& k)	size_t	Count number of elements with a given key.
Modifiers
clear()	void	Remove all elements.
insert(std::pair<K, V> pair)	void	Insert a new key-value pair.
emplace(Args&&& … args)	pair<iterator, bool>

Map example:

• File: map-container.cpp

#include<iostream>
#include<string>
#include<map>
#include <iomanip>

struct Point3D{
        double x;
        double y;
        double z;
        Point3D(): x(0), y(0), z(0){}
        Point3D(double x, double y, double z): x(x), y(y), z(z){}
        /* Copy constructor 
     * -> Implement redundant copy constructor for logging purposes and 
     * detect when copy happens. 
     */
        Point3D(const Point3D& p){		
                std::cerr << " I was copied" << std::endl;
                this->x = p.x;
                this->y = p.y;
                this->z = p.z;
        }
        ~Point3D() = default;
};

std::ostream& operator<< (std::ostream& os, const Point3D& p){
        os << std::setprecision(3) << std::fixed;
        return os << "Point3D{"
                          << "x = "  << p.x
                          << ",y = " << p.y
                          << ", z = "<< p.z
                          << "}";
}

int main(){	
        auto locations = std::map<std::string, Point3D>();
        locations["point1"] = Point3D(2.0, 3.0, 5.0);
        locations["pointX"] = Point3D(12.0, 5.0, -5.0);
        locations["pointM"] =  {121.0, 4.0, -15.0};
        locations["Origin"] = {}; // Point32{} or Point3D()
	
        // Invokes copy constructor
        std::cerr << "  <== Before inserting" << "\n";
        locations.insert(std::pair<std::string, Point3D>("PointO1", Point3D(0.0, 0.0, 0.0)));
        std::cerr << "  <== After inserting" << "\n";
	
        // operator[] doesn't throw exception 
        std::cout << "point1 = " << locations["point1"] << "\n";
        std::cout << "pointX = " << locations.at("pointX") << "\n";
        std::cout << "pointM = " << locations.at("pointM") << "\n";

        // Safer and uses exception 
        try {
                std::cout << "pointY = " << locations.at("pointY") << "\n";
        } catch(const std::out_of_range& ex){
                std::cout << "Error - not found element pointY. MSG = " << ex.what() << "\n";
        }

        if(auto it = locations.find("pointX"); it != locations.end())
                std::cout << " [INFO]= => Location pointX found =  " << it->second << "\n";

        if(locations.find("pointMAS") == locations.end())
                std::cout << " [ERROR] ==> Location pointMAS  not found" << "\n";
	
        std::cout << "Key-Value pairs " << "\n";
        std::cout << "-------------------------" << "\n";
        for (const auto& x: locations)
                std::cout << x.first << " : " << x.second << "\n";
        std::cout << '\n';

        return 0;
}

Running:

$ clang++ map-container.cpp -o map-container.bin -std=c++1z -Wall -Wextra  && ./map-container.bin

  <== Before inserting
 I was copied
 I was copied
  <== After inserting
point1 = Point3D{x = 2.000,y = 3.000, z = 5.000}
pointX = Point3D{x = 12.000,y = 5.000, z = -5.000}
pointM = Point3D{x = 121.000,y = 4.000, z = -15.000}
pointY = Error - not found element pointY. MSG = map::at
 [INFO]= => Location pointX found =  Point3D{x = 12.000,y = 5.000, z = -5.000}
 [ERROR] ==> Location pointMAS  not found
Key-Value pairs 
-------------------------
Origin : Point3D{x = 0.000,y = 0.000, z = 0.000}
PointO1 : Point3D{x = 0.000,y = 0.000, z = 0.000}
point1 : Point3D{x = 2.000,y = 3.000, z = 5.000}
pointM : Point3D{x = 121.000,y = 4.000, z = -15.000}
pointX : Point3D{x = 12.000,y = 5.000, z = -5.000}

Associative Container - Unordered map

The unordered map, introduced in C++11, is generally faster for insertion and deletion of elements since the unordered map is implemented as a true hash table, unlike the std::map which is implemented as tree. The downside of unordered_map this data structure is the loss of elements sorting.

• Header: <unordered_map>

Benefits:

• True hash table.
• Faster for insertion, retrieval and removal of elements that the map.

Downsides:

• Loss of elements insertion order.

Example:

Constructors:

std::unordered_map<std::string, int> m1;

auto m2 = std::unordered_map<std::string, int>{};

// Uniform initialization 
//--------------------------
>> std::unordered_map<std::string, int> m3 {{"x", 200}, {"z", 500}, {"w", 10}, {"pxz", 70}}
 { "pxz" => 70, "w" => 10, "z" => 500, "x" => 200 }

//  More readable 
>> auto m4 = std::unordered_map<std::string, int> {{"x", 200}, {"z", 500}, {"w", 10}, {"pxz", 70}}
 { "pxz" => 70, "w" => 10, "z" => 500, "x" => 200 }

Insert Elements:

>> auto m = std::unordered_map<std::string, int>{}

>> m["x"] = 100
(int) 100
>> m["x"] = 100;
>> m["z"] = 5;
>> m["a"] = 6710;
>> m["hello"] = -90;
>> m["sword"] = 190;

>> m
{ "sword" => 190, "hello" => -90, "a" => 6710, "x" => 100, "z" => 5 }

Insert element using stl::pair:

>> auto mm = std::unordered_map<std::string, int>{};

>> mm.insert(std::make_pair("x", 200));
>> mm.insert(std::make_pair("z", 500));
>> mm.insert(std::make_pair("w", 10));

>> mm["x"]
(int) 200
>> mm["w"]
(int) 10
>> 

Number of elements:

>> m.size()
(unsigned long) 6
>>

Retrieve elements:

>> m["x"]
(int) 100
>> m["sword"]
(int) 190
>>
// Doesn't  throw exception if element is not found 
>> m["sword-error"]
(int) 0
>> 

// Throw exception if element is not found
>> m.at("x")
(int) 100
>> m.at("sword")
(int) 190
>> m.at("sword error")
Error in <TRint::HandleTermInput()>: std::out_of_range caught: _Map_base::at
>> 
>> 

Find element:

// -------- Test 1 -----------//
auto it = m.find("sword");
if(it != m.end()) {
        std::cout << "Found Ok. => {"
                  << "key = " << it->first
                  << " ; value = " << it->second
                  << " }"
                  << "\n";
	
} else {
        std::cout << "Error: key not found." << "\n";
}
// Output: 
Found Ok. => {key = swordvalue = 190 }
> 

// -------- Test 1 -----------//

auto it = m.find("this key will not be found!");
if(it != m.end()) {
     std::cout << "Found Ok. => {"
               << "key = "      << it->first
               << " ; value = " << it->second
               << " }"
               << "\n";
} else {
    std::cout << "Error: key not found." << "\n";
}
// ----- Output: ----------//
Error: key not found.
>> 

Loop over container elements:

for(const auto& p: m) {
         std::cout << std::setw(5) << "key = " << std::setw(6) << p.first
                   << std::setw(8) << " value = " << std::setw(5) << p.second
                   << "\n";
}

// Output: 
key =  sword value =   190
key =  hello value =   -90
key =      a value =  6710
key =      x value =   100
key =      z value =     5

Loop with iterator and stl “algorithm” std::for_each.

std::for_each(m.begin(), m.end(),
               [](const std::pair<std::string, int>& p){
                       std::cout << std::setw(5)  << p.first
                                 << std::setw(10) << p.second
                                 << "\n";									  
               });
// Output:
sword       190
hello       -90
    a      6710
    x       100
    z         5

Associative Container - Multimap

The container std::multimap is similar to map, however it allows repeated keys.

Header: <map>

Documentation:

• std::multimap - cppreference.com

Examples:

• Initialize std::multimap

#include <iostream>
#include <string>
#include <map>

std::multimap<std::string, int> dict;

>> dict
(std::multimap<std::string, int> &) {}
>> 

// Insert pair object 
dict.insert(std::make_pair("x", 100));
dict.insert(std::make_pair("status", 30));
dict.insert(std::make_pair("HP", 250));
dict.insert(std::make_pair("stamina", 100));
dict.insert(std::make_pair("stamina", 600));
dict.insert(std::make_pair("x", 10));
dict.insert(std::make_pair("x", 20));

>> dict
{ "HP" => 250, "stamina" => 100, "stamina" => 600, "status" => 30, "x" => 100, "x" => 10, "x" => 20 }
>> 

Find all pair with a given key

// Find elements:
>> auto it = dict.find("x"); // Iterator
>> 
for(auto it = dict.find("x"); it != dict.end(); it++){ 
  std::printf(" ==> it->first = %s ; it->second = %d\n", it->first.c_str(), it->second); 
}
/** Output: 
  ==> it->first = x ; it->second = 100
  ==> it->first = x ; it->second = 10
  ==> it->first = x ; it->second = 20
 */

Count all elements with a given key

>> dict.count("x")
(unsigned long) 3

>> dict.count("stamina")
(unsigned long) 2

>> dict.count("HP")
(unsigned long) 1

>> dict.count("")
(unsigned long) 0

>> dict.count("wrong")
(unsigned long) 0
>> 

Iterate over multimap:

 for(const auto& pair : dict){ 
   std::printf(" ==> key = %s ; value = %d\n", pair.first.c_str(), pair.second); 
 }
 /** Output: 
    ==> key = HP ; value = 250
    ==> key = stamina ; value = 100
    ==> key = stamina ; value = 600
    ==> key = status ; value = 30
    ==> key = x ; value = 100
    ==> key = x ; value = 10
    ==> key = x ; value = 20
*/

Clear multimap object:

>> auto dict2 = std::multimap<std::string, int> { {"x", 100}, {"y", 10}, {"x", 500}, {"z", 5}};
>> dict2
 { "x" => 100, "x" => 500, "y" => 10, "z" => 5 }

>> dict2.size()
(unsigned long) 4

>> dict2.clear();

>> dict2
{}

>> dict2.size()
(unsigned long) 0

Associative Container - Sets

Set std::set is an associative container implementing the mathematical concept of finite set. This container stores sorted unique values and any attempt to insert a repeated value will discard the value to be inserted.

• Header: <set>
• Implementation: Binary search tree.
• Note: as this collection has sorting, its unordered version, without sorting, std::unordered_set performs better.

Example: Set constructors

• Instantiate a set object with a default constructor (constructor with empty parameters):

#include <iostream> 
#include <string>
#include <set>

std::set<int> s1;

>> s1.insert(10);
>> s1.insert(20);
>> s1.insert(20);
>> s1.insert(30);
>> s1.insert(40);
>> s1
(std::set<int> &) { 10, 20, 30, 40 }
>> s1.insert(40);
>> s1
(std::set<int> &) { 10, 20, 30, 40 }

• Instantiate a set with initializer list constructor:

>> auto s2 = std::set<std::string>{ 
    "hello", "c++", "c++", "hello", "world", "world", 
     "c++11", "c++", "c++17", "c++17"
   };
>> s2
{ "c++", "c++11", "c++17", "hello", "world" }
>> 

// Any repeated element is discarded 
>> s2.insert("c++");
>> s2
{ "c++", "c++11", "c++17", "hello", "world" }

• Instantiate a set with range constructor or iterator pair constructor:

>> std::vector<int> numbers {-100, 1, 2, 10, 2, 1, 3, 15, 3, 5, 4, 4, 3, 3, 2};

>> std::set<int> sa1(numbers.begin(), numbers.end());
>> sa1
(std::set<int> &) { -100, 1, 2, 3, 4, 5, 10, 15 }


>> auto sa2 = std::set<int>{numbers.begin() + 4, numbers.end() - 2};
>> sa2
{ 1, 2, 3, 4, 5, 15 }

• Instantiate a set with copy constructor.
  • std::set<T>(const T&)

>> std::set<int> xs{1, 1, 10, 1, 2, 5, 10, 4, 4, 5, 1};
>> xs
{ 1, 2, 4, 5, 10 }

>> std::set<int> copy1(xs);
>> copy1
(std::set<int> &) { 1, 2, 4, 5, 10 }

>> auto copy2 = xs;
>> copy2
{ 1, 2, 4, 5, 10 }

>> auto copy3 = std::set<int>{xs};
>> copy3
{ 1, 2, 4, 5, 10 }

>> if(&copy1 != &xs){ std::puts(" => Not the same"); }
 => Not the same

>> if(&copy2 != &xs){ std::puts(" => Not the same"); }
 => Not the same

>> if(&copy3 != &xs){ std::puts(" => Not the same"); }
 => Not the same

• Instantiating a set with a move constructor.
  • std::set<T>(T&&)

>> std::set<int> xs1{1, 1, 10, 1, 2, 5, 5, 6, 10, 4, 4, 5, 1, 6, 7, 7};

>> xs1
{ 1, 2, 4, 5, 6, 7, 10 }

// Move constructor:  
>> std::set<int> m1(std::move(xs1));
>> m1
(std::set<int> &) { 1, 2, 4, 5, 6, 7, 10 }
>> xs1
(std::set<int> &) {}
>>

>> std::set<int> xs2{1, 1, 10, 1, 2, 5, 5, 6, 10, 4, 4, 5, 1, 6, 7, 7};
>> xs2
(std::set<int> &) { 1, 2, 4, 5, 6, 7, 10 }

// ========  Move constructor ===================
>> auto m2 = std::move(xs2);
>> m2
{ 1, 2, 4, 5, 6, 7, 10 }
>> xs2
(std::set<int> &) {}
>> 

Operations on sets:

Instantiating sample set:

>> auto aset = std::set<int> {1, 1, 10, 1, 2, 5, 5, 6, 10, 4, 4, 5, 1, 6, 7, 7};
>> aset
{ 1, 2, 4, 5, 6, 7, 10 }   

Count number of elements:

>> aset.size()
(unsigned long) 7
>> 

Clear set (remove all elements):

>>  auto asetb = std::set<int> {1, 1, 10, 1, 2, 5, 5, 6, 10, 4, 4, 5, 1, 6, 7, 7};
>> asetb
{ 1, 2, 4, 5, 6, 7, 10 }

>> asetb.clear();
>> asetb
{}

>> asetb.empty()
(bool) true

Check whether an element is in the set without iterator:

>> aset.count(10)
(unsigned long) 1
>> aset.count(100)
(unsigned long) 0
>> aset.count(1)
(unsigned long) 1
>> aset.count(-12)
(unsigned long) 0
>> 
>> if(aset.count(10) != 0 ) { std::puts("Element in the set."); }
Element in the set.
>> if(aset.count(10)) { std::puts("Element in the set."); }
Element in the set.
>> if(aset.count(25) != 0 ) { std::puts("Element in the set."); }
>> if(aset.count(25)) { std::puts("Element in the set."); }
>> 

Check if element is in the set with iterator:

>> aset
{ 1, 2, 4, 5, 6, 7, 10 }

>> aset.find(10)
(std::set<int, std::less<int>, std::allocator<int> >::iterator) @0x22f1ff0
>> 

std::set<int>::iterator it;
>> if((it = aset.find(10)) != aset.end()) std::printf(" ==> Found element = %d\n", *it)
 ==> Found element = 10

>> if((it = aset.find(2)) != aset.end()) std::printf(" ==> Found element = %d\n", *it) 
 ==> Found element = 2

>> if((it = aset.find(-100)) != aset.end()) std::printf(" ==> Found element = %>> ", *it)

// Or: ----------------------------------------------------
>> auto itr = aset.find(7);
>> if(itr == aset.end()) std::puts("Element not found");
>> if(itr != aset.end()) std::puts("Element  found");
Element  found
>> int element = *itr
(int) 7
>> 

Remove element from set:

>> aset
{ 1, 2, 4, 5, 6, 7, 10 }

>> auto itr2 = aset.find(10);
// Remove element using iterator.
>> aset.erase(itr2);

>> aset
{ 1, 2, 4, 5, 6, 7 }

// Segmentation fault!! 
>> aset.erase(aset.find(-10));
free(): invalid pointer

Iterate over a set:

int i = 0;
for(const auto& x: aset){  std::printf(" element[%d] = %d\n", ++i, x); }

// For-range based loop
>> for(const auto& x: aset){  std::printf(" element[%d] = %d\n", ++i, x); }
 element[1] = 1
 element[2] = 2
 element[3] = 4
 element[4] = 5
 element[5] = 6
 element[6] = 7
 element[7] = 10

// Iterator based loop 
int j = 0;
for(auto it = aset.begin(); it != aset.end(); it++){  std::printf(" element[%d] = %d\n", ++j, *it); }

>> for(auto it = aset.begin(); it != aset.end(); it++){  std::printf(" element[%d] = %d\n", ++j, *it); }
 element[1] = 1
 element[2] = 2
 element[3] = 4
 element[4] = 5
 element[5] = 6
 element[6] = 7
 element[7] = 10

Bitset Container

Class template for representing a sequence of N bits.

• Header: <bitset>

Default Constructor:

#include <bitset>

 >> #include <bitset>

 >> std::bitset<4> b;

 >> std::cout << " b = " << b << std::endl;
  b = 0000

Test bits;

// Set bit 0 
>> b.set(0)
(std::bitset<4UL> &) @0x7f92db9c7010

>> b
(std::bitset<4> &) @0x7f92db9c7010
>> std::cout << " b = " << b << std::endl;
 b = 0001

// Set bit 1 and 3 
>> b.set(1).set(3)
(std::bitset<4UL> &) @0x7f92db9c7010

>> std::cout << " b = " << b << std::endl;
 b = 1011

Test bits:

// Check whether bit 0 is set  (equal to 1)
>> b.test(0)
(bool) true

// Check whether bit 1 is set
>> b.test(1)
(bool) true

// Check whether bit 1 is set
>> b.test(2)
(bool) false

// Check whether bit 1 is set
>> b.test(3)
(bool) true
>> 

// Clear bit 0 
>> b.set(0, false);
>> b.test(0)
(bool) false

Create a bitset initialized with some integer value:

>> std::bitset<8> b1{0xAE};

>> std::cout << "b1 = " << b1 << std::endl;
b1 = 10101110

// Test bits 
>> b1.test(0)
(bool) false
>> b1.test(1)
(bool) true
>> b1.test(7)
(bool) true
>> b1.test(6)
(bool) false
>> 

// Number of bits 
>> b1.size()
(unsigned long) 8
>> 

Convert to numerical value:

// Convert to numerical value 
>> b1.to_ulong()
(unsigned long) 174

>> 0xAE
(int) 174

Flip bitset:

>> b1.flip()
(std::bitset<8UL> &) @0x7f92db9c7018

>> b1.to_ulong()
(unsigned long) 81

>> std::cout << "b1 flipped = " << b1 << std::endl;
b1 flipped = 01010001
>> 

Create bitset from binary string:

>> auto bb = std::bitset<8>("01010001");

>> bb
(std::bitset<8> &) @0x7f92db9c7020

>> std::cout << " bb = " << bb << "\n";
 bb = 01010001

>> bb.to_ulong()
(unsigned long) 81
>> 

>> bb.test(0)
(bool) true

>> bb.test(1)
(bool) false

Getting individual bits:

>> std::cout << "bit0 = " << bb[0] << " ; bit1 = " << bb[1] << " ; bit2 = " << bb[2] << "\n";
bit0 = 1 ; bit1 = 0 ; bit2 = 0
>> 

>> if(bb[1]){ std::puts("bit is set"); } else { std::puts("bit is cleared"); }
bit is cleared

>> if(bb[2]){ std::puts("bit is set"); } else { std::puts("bit is cleared"); }
bit is cleared

>> if(bb[3]){ std::puts("bit is set"); } else { std::puts("bit is cleared"); }
bit is cleared

>> if(bb[5]){ std::puts("bit is set"); } else { std::puts("bit is cleared"); }
bit is cleared

>> if(bb[6]){ std::puts("bit is set"); } else { std::puts("bit is cleared"); }
bit is set
>> 

Getting reference to individual bit:

>> auto gpio0 = bb[0]
(std::bitset<8>::reference &) @0x7f92db9c7038

>> (int) gpio0
(int) 1

>> gpio0 = true;
>> (int) gpio0
(int) 1

>> gpio0 = false;
>> (int) gpio0
(int) 0

>> (bool) gpio0
(bool) false

Bitset to string:

>> auto ba = std::bitset<8>("01010101");
>> ba
(std::bitset<8> &) @0x7f92db9c7058

>> std::string repr(ba.to_string('0', '1'));
>> repr
(std::string &) "01010101"
>> 

See:

• Is there any advantage to C-style bit manipulation over std: :bitset in C++?
• One Approach to Using Hardware Registers in C++

General C++ Reference Card

Data Types and Data Models

C++ Types and Data Models

This table shows the numeric types data sizes in bits per memory model, architechture operating system and ISA - Instruction Set Architechture. Note: *ptr is the pointer size in bits.

Data	Arch. ISA	Operating System	*ptr.	short	int	long	long
Model			size_t				long
16 Bits Systems
IP16	PDP-11	Unix (1973)	16	-	16	-	-
IP16L32	PDP-11	Unix (1977)	16	16	16	32	-
LP32	x86 (16 bits)	Microsft Win16 and Apple’ MacOSX	32	16	16	32	-
32 Bits Systems
I116LP32	MC680000, x86 (16 bits)	Macintosh (1982), Windows	16		16
ILP32	IBM-370	Vax Unix	32	16	32	32	-
ILP32LL or ILP32LL64	x86 or IA32	Microsft Win32	32	16	32	32	64
64 Bits Systems
LLP64, IL32LLP64 or P64	x86-x64 (IA64, AMD64)	Microsft Win64 (x64 / x86)	64	16	32	32	64
LP64 or I32LP64	IA64, AMD64	Linux, Solaris, DEC OSF, HP UX	64	16	32	64	64
ILP64	-	HAL	64	16	32	64	64
SILP64	-	UNICOS

Sumary:

• ILP32
  • int, long and pointer are all 32 bits
• ILP32LL - Used by most compilers and OSes on 32 bits platforms. (De facto standard for 32 bits platforms)
  • int, long, and pointer are all 32 bits, but the type long long has 64 bits in size.
• LP64 - Used by most 64 bit Unix-like OSes, including Linux, BSD and Apple’s Mac OSX (De facto standard for 64 bits platforms)
  • int, long and ponter are all 64 bits.
• ILP64
  • int, long and pointer are all 64 bits.
• LLP64 (Used by Windows 64 bits)
  • pointers and long long are 64 bits and the types int and long are 32 bits.

Note:

• It is not safe to rely on the size of numeric data type or make assumptions about the numeric sizes. In cases where the size of the data type matters such as serialization, embedded systems or low level code related to hardware it is better to use fixed-width integer.
• Underflow and overflow can lead to undefined behaviors and unpredictable results.

References:

• Fundamental types - cppreference.com
• PVS Studio - Data model
• Data Models and Word Size - Nick Desaulniers
• Seven Steps of Migrating a Program to a 64-bit System
• ILP32 and LP64 data models and data type sizes
• Implementation Sins - C Language Issues (Security)
  • Discusses potential buffer overflow security vulnerability in C code due to wrong assumptions about size of numeric types.
• Chapter 6 - C Language Issues (Security)
• Vulnerabilities in C : When integers go bad! - Sticky Bits - Powered by FeabhasSticky Bits – Powered by Feabhas
• About size_t and ptrdiff_t

Float Point Numebers

Type	Size (bits)	Size (bytes)	Description
Float Points
float	32	4	Single-precision IEEE754 float point
double	64	8	Double-precision IEEE754 float point
long float	128	16	Quadruple-precision IEEE754 float point

Fixed-Width Numeric Types

Type	Size	Size	Description	Maximum number of
	(bits)	(bytes)		decimal digits
Fixed-width integer
int8_t	8	1	8-bits signed int	2
uint8_t	16	2	8-bits unisgned int (positive)	2
int16_t	16	2	16-bits signed int	4
uint16_t	32	4	16-bits unsigned int	4
int32_t	32	4	32-bits signed int	9
uint32_t	32	4	32-bits unsigned int	9
int64_t	64	8	64-bits signed int	18
uint64_t	64	8	64-bits unsigned int	18

Sample code for showing numeric limits:

File:

• src/numeric-limits.cpp
• Online Compiler: https://rextester.com/BBXAM15894

/*******************************************************************************************
 * File: numeric-limits.cpp 
 * Brief: Shows the numeric limits for all possible numerical types.  
 * Author: Caio Rodrigues
 *****************************************************************************************/

#include <iostream>
#include <limits>    // Numeric limits 
#include <iomanip>   // setw, and other IO manipulators 
#include <string>    // std::string 
#include <cstdint>   // uint8_t, int8_t, ...
#include <functional>

struct RowPrinter{
        int m_left;  // Left alignment 
        int m_right; // Right alignment  
        RowPrinter(int left, int right): m_left(left), m_right(right){
                // Print bool as 'true' or 'false' instead of 0 or 1.
                std::cout << std::boolalpha;
        }
	
        template<class A>
        auto printRow(const std::string& label, const A& value) const -> void {
                std::cout << std::setw(m_left)  << label
                                  << std::setw(m_right) << value << "\n";
        }
};

#define SHOW_INTEGER_LIMITS(numtype) showNumericLimits<numtype>(#numtype)
#define SHOW_FLOAT_LIMITS(numtype)   showFloatPointLimits<numtype>(#numtype)

template <class T>
void showNumericLimits(const std::string& name){
        RowPrinter rp{30, 25};	
        std::cout << "Numeric limits for type: " << name << "\n";
        std::cout << std::string(60, '-') << "\n";
        rp.printRow("Type:",                    name);
        rp.printRow("Is integer:",              std::numeric_limits<T>::is_integer);
        rp.printRow("Is signed:",               std::numeric_limits<T>::is_signed);
        rp.printRow("Number of digits 10:",     std::numeric_limits<T>::digits10);
        rp.printRow("Max Number of digits 10:", std::numeric_limits<T>::max_digits10);

        // RTTI - Run-Time Type Information 
        if(typeid(T) == typeid(uint8_t)
           || typeid(T) == typeid(int8_t)
           || typeid(T) == typeid(bool)
           || typeid(T) == typeid(char)
           || typeid(T) == typeid(unsigned char)
                ){
                // Min Abs - samllest positive value for float point numbers 
                rp.printRow("Min Abs:",         static_cast<int>(std::numeric_limits<T>::min()));
                // Smallest value (can be negative)
                rp.printRow("Min:",             static_cast<int>(std::numeric_limits<T>::lowest()));
                // Largest value  
                rp.printRow("Max:",             static_cast<int>(std::numeric_limits<T>::max()));	
        } else {
                rp.printRow("Min Abs:",         std::numeric_limits<T>::min());
                rp.printRow("Min:",             std::numeric_limits<T>::lowest());
                rp.printRow("Max:",              std::numeric_limits<T>::max());
        }
        rp.printRow("Size in bytes:",       sizeof(T));
        rp.printRow("Size in bits:",        8 * sizeof(T));
        std::cout << "\n";
}

template<class T>
void showFloatPointLimits(const std::string& name){
        RowPrinter rp{30, 25};	
        showNumericLimits<T>(name);
        rp.printRow("Epsilon:",        std::numeric_limits<T>::epsilon());
        rp.printRow("Min exponent:",   std::numeric_limits<T>::min_exponent10);
        rp.printRow("Max exponent:",   std::numeric_limits<T>::max_exponent10);
}

int main(){
        SHOW_INTEGER_LIMITS(bool);
        SHOW_INTEGER_LIMITS(char);
        SHOW_INTEGER_LIMITS(unsigned char);
        SHOW_INTEGER_LIMITS(wchar_t);

        // Standard integers in <cstdint>
        SHOW_INTEGER_LIMITS(int8_t);
        SHOW_INTEGER_LIMITS(uint8_t);
        SHOW_INTEGER_LIMITS(int16_t);
        SHOW_INTEGER_LIMITS(uint16_t);
        SHOW_INTEGER_LIMITS(int32_t);
        SHOW_INTEGER_LIMITS(uint32_t);
        SHOW_INTEGER_LIMITS(int64_t);
        SHOW_INTEGER_LIMITS(uint64_t);

        SHOW_INTEGER_LIMITS(short);
        SHOW_INTEGER_LIMITS(unsigned short);
        SHOW_INTEGER_LIMITS(int);
        SHOW_INTEGER_LIMITS(unsigned int);
        SHOW_INTEGER_LIMITS(long);
        SHOW_INTEGER_LIMITS(unsigned long);
        SHOW_INTEGER_LIMITS(long long);
        SHOW_INTEGER_LIMITS(unsigned long long);
	
        SHOW_FLOAT_LIMITS(float);
        SHOW_FLOAT_LIMITS(double);
        SHOW_FLOAT_LIMITS(long double);
	
    return 0;
}

Output:

$ clang++ numeric-limits.cpp -o numeric-limits.bin -g -std=c++11 -Wall -Wextra && ./numeric-limits.bin

...   ...   ...   ...   ...   ...   ...   ...   ...   ... 

Numeric limits for type: short
------------------------------------------------------------
                         Type:                    short
                   Is integer:                     true
                    Is signed:                     true
          Number of digits 10:                        4
      Max Number of digits 10:                        0
                      Min Abs:                   -32768
                          Min:                   -32768
                          Max:                    32767
                Size in bytes:                        2
                 Size in bits:                       16

Numeric limits for type: unsigned short
------------------------------------------------------------
                         Type:           unsigned short
                   Is integer:                     true
                    Is signed:                    false
          Number of digits 10:                        4
      Max Number of digits 10:                        0
                      Min Abs:                        0
                          Min:                        0
                          Max:                    65535
                Size in bytes:                        2
                 Size in bits:                       16

Numeric limits for type: int
------------------------------------------------------------
                         Type:                      int
                   Is integer:                     true
                    Is signed:                     true
          Number of digits 10:                        9
      Max Number of digits 10:                        0
                      Min Abs:              -2147483648
                          Min:              -2147483648
                          Max:               2147483647
                Size in bytes:                        4
                 Size in bits:                       32

Numeric limits for type: unsigned int
------------------------------------------------------------
                         Type:             unsigned int
                   Is integer:                     true
                    Is signed:                    false
          Number of digits 10:                        9
      Max Number of digits 10:                        0
                      Min Abs:                        0
                          Min:                        0
                          Max:               4294967295
                Size in bytes:                        4
                 Size in bits:                       32

Numeric limits for type: long
------------------------------------------------------------
                         Type:                     long
                   Is integer:                     true
                    Is signed:                     true
          Number of digits 10:                       18
      Max Number of digits 10:                        0
                      Min Abs:     -9223372036854775808
                          Min:     -9223372036854775808
                          Max:      9223372036854775807
                Size in bytes:                        8
                 Size in bits:                       64

Numeric limits for type: unsigned long
------------------------------------------------------------
                         Type:            unsigned long
                   Is integer:                     true
                    Is signed:                    false
          Number of digits 10:                       19
      Max Number of digits 10:                        0
                      Min Abs:                        0
                          Min:                        0
                          Max:     18446744073709551615
                Size in bytes:                        8
                 Size in bits:                       64

Numeric limits for type: long long
------------------------------------------------------------
                         Type:                long long
                   Is integer:                     true
                    Is signed:                     true
          Number of digits 10:                       18
      Max Number of digits 10:                        0
                      Min Abs:     -9223372036854775808
                          Min:     -9223372036854775808
                          Max:      9223372036854775807
                Size in bytes:                        8
                 Size in bits:                       64

Numeric limits for type: unsigned long long
------------------------------------------------------------
                         Type:       unsigned long long
                   Is integer:                     true
                    Is signed:                    false
          Number of digits 10:                       19
      Max Number of digits 10:                        0
                      Min Abs:                        0
                          Min:                        0
                          Max:     18446744073709551615
                Size in bytes:                        8
                 Size in bits:                       64

... ....    ... ....    ... ....    ... ....    ... .... 

Numeric Literals

Literal	Suffix	Type	Description	Sizeof Bytes
2001	-	int	signed integer	4
20u	u or U	unsingned int		4
0xFFu	u or U	unsigned int	unsingned int literal in hexadecimal (0xff = 255)	4
100l or 100L	l or L	long		8
100ul or 100UL	ul or UL	unsigned long		8
0xFAul or 0xFAUL		unsigned long	unsigned long literal in hexadecimal format (0xfa = 250)	8
100.23f or 100.23F	f or F	float	32 bits IEEE754 Float Point number mostly used in games and computer graphics.	8
20.12 (default)		double	64 bits IEEE754 Float Point number commonly used in scientific computing.	4

Types of Parameter Passing

Parameter Passing	Alternative	Parameter t passed by
Value
T t		by value
const T* t	const T* t
Pointer
T* t	T *t	pased by pointer
T t []	T* t	by pointer, this notation is used for C-array parameters
Reference
T& t	T &t	by reference or L-value reference
const T& t	const T &t	by const reference or const L-value reference.
T const& t	-	by const reference - alternative notation
T&& t	T &&t	by r-value reference
template<class T> function(T&& t)	-	Universal reference can become either L-value or R-value reference.

Notes:

• Function here means both member function (class methods) or free functions (aka ordinary functions).
• Parameters passed by value cannot be modified within the function as they are copied. It happens for all C++ types, including instances of classes what is different from most OO languages like Java, C#, Python and etc.
• When an object is passed by value, its copy constructor is invoked, as a result a copy is created.
• Prefere passing large objects such large matrices or arrays by reference or const reference when the function is not supposed to modify the parameter in order to avoid memory overhead due to copy.
• I is better to pass objects instantiated on the heap (dynamic memory) with new operator using smart pointers (unique_ptr, shared_ptr) in order to avoid memory leaks.

Operators and operator overload

Summary Table

Description	Operator	Class operator overload declaration
Equal to	a == b
Logical not	!a, !false, !true
Logical and	a && b
Logical or	a |\vert b
Pre increment (prefix)	++i
Post increment	i++
Pre decrement	++i
Post increment	i–
Addition assignment (+=)	a += b ; a <- a + b
Subtraction assignment (-=)	a -= b ; a <- a - b
Multiplication assignment (*=)	a *= b ; a <- a * b
Division assignment (/=)	a /= b ; a <- a / b
Subscript, array index	a[b]	A C::operator [](S index)
Indirection - defeference	*a	A C::operator *()
Address or reference	&a	A* C::operator &()
Structure dereference	a->memberFunction(x)
Structure reference (.)	a.memberFunction(x)	- N/A
Function call (function-object declaration)	A(p0, p1, p2)	R C::operator()(P0 p0, P1 p1, P2 p2)
Ternary conditional - similar to if x = (if cond 10 20)	a ? b : c	- N/A
Scope resolution operator	Class::staticMethod(x)	- N/A
Sizeof - returns size of type at compile-time	sizeof(type)	- N/A

For more details check out:

• Operators in C and C++ - Wikipedia

Operator Overload Snippet 1

class SomeClass{
private:
    // ---->> Private data here <------
public:
    SomeClass(){}
    SomeClass(double x, double y){
        m_x = x;
        m_y = y;
    }
    // Copy assignment operator 
    SomeClass& operator=(const SomeClass& other){
        //  ...  ......
    }
    // Equality operator - check whether current object is equal to
    // the other.
    //-----------------------------------------------
    bool operator==(const SomeClass& p){
        return this->x == p.x && this->y == p.y;
    }
    // Not equal operator - checks whether current object is not equal to
    // the other.
    //-----------------------------------------------
    bool operator!=(const SomeClass& p){
        return this->x != p.x || this->y != p.y;
    }
    // Not logical operator (!) Exclamation mark.
    // if(!obj){ ... }
    //-----------------------------------------------
    bool operator! (){
        return this->m_data != nullptr;
    }
    // Operator ++obj
    //-----------------------------------------------
    SomeClass& operator++(){
        this->m_counter += 1;
        return *this;
    }

    // Operator (+)
    // SomeClass a, b;
    // SomeClass c = a + b;
    SomeClass operator+(SomeClass other){
        SomeClass res;
        res.x = m_x + other.x;
        res.y = m_y + other.y;
        return res;
    }
    // Operator (+)
    SomeClass operator+(double x){
        SomeClass res;
        res.x = m_x + x
        res.y = m_y + x
        return res;
    }
    // Operator (*)
    SomeClass operator*(double x){
        SomeClass res;
        res.x = res.x * x;
        res.y = res.y * x;
        return res;
    }

    // Operator (+=)
    // SomeClass cls;
    // cls += 10.0;
    SomeClass& operator +=(double x){
        m_x += x;
        m_y += y;
        return *this;
    }
    // Operator index -> obj[2]
    // SomeClass cls;
    // double z = cls[2];
    //-----------------------------------------------
    double operator[](int idx){
        return this->array[idx];
    }
    // Function application operator
    // SomeClass obj;
    // double x = obj();
    //-----------------------------------------------
    double operator()(){
        return m_counter * 10;
    }
    // Function application operator
    // SomeClass obj;
    // double x = obj(3.4, "hello world");
    //-----------------------------------------------
    double operator()(double x, std::string msg){
        std::cout << "x = " << x << " msg  = " << msg;
        return 3.5 * x;                                       
    }
    // Operator string insertion, allows printing the current object 
    // SomeClass obj;
    // std::cout << obj << std::enl;
    //-----------------------------------------------
    friend std::ostream& operator<<(std::ostream &os, const SomeClass& cls){
        // Print object internal data structure 
        os << cls.m_x << cls.m_y  ;
        return os;
    }
};

Operator Overload Snippet 2

File: SomeClass.hpp - Header file.

class SomeClass{
private:
    // ---->> Private data here <------
public:
    SomeClass(){}
    SomeClass(double x, double y);
    bool operator==(const SomeClass& p);
    bool operator!=(const SomeClass& p);
    bool operator! ();
    SomeClass& operator++();
    SomeClass operator+(SomeClass other);
    SomeClass operator+(double x);
    SomeClass operator*(double x);
    SomeClass& operator +=(double x);
    double operator[](int idx);
    double operator()();
    double operator()(double x, std::string msg);
    friend std::ostream& operator<<(std::ostream &os, const SomeClass& cls);
};

File: SomeClass.cpp - implementation

SomeClass::SomeClass(){}

SomeClass::SomeClass(double x, double y){
        m_x = x;
        m_y = y;
    }
    
// Equality operator - check whether current object is equal to
// the other.
//-----------------------------------------------
bool SomeClass::operator==(const SomeClass& p){
    return this->x == p.x && this->y == p.y;
}

// Not equal operator - checks whether current object is not equal to
// the other.
//-----------------------------------------------
bool SomeClass::operator!=(const SomeClass& p){
    return this->x != p.x || this->y != p.y;
}
    
// Not logical operator (!) Exclamation mark.
// if(!obj){ ... }
//-----------------------------------------------
bool SomeClass::operator! (){
    return this->m_data != nullptr;
}
    
// Operator ++obj
//-----------------------------------------------
SomeClass& SomeClass::operator++(){
    this->m_counter += 1;
    return *this;
}

// Operator (+)
// SomeClass a, b;
// SomeClass c = a + b;
SomeClass SomeClass::operator+(SomeClass other){
    SomeClass res;
    res.x = m_x + other.x;
        res.y = m_y + other.y;
        return res;
}
// Operator (+)
SomeClass SomeClass::operator+(double x){
    SomeClass res;
    res.x = m_x + x;
    res.y = m_y + x;
    return res;
}
// Operator (*)
SomeClass SomeClass::operator*(double x){
    SomeClass res;
    res.x = res.x * x;
    res.y = res.y * x;
        return res;
}

// Operator (+=)
// SomeClass cls;
// cls += 10.0;
SomeClass& SomeClass::operator +=(double x){
    m_x += x;
    m_y += y;
    return *this;
}
    
    
// Operator index -> obj[2]
// SomeClass cls;
// double z = cls[2];
//-----------------------------------------------
double SomeClass::operator[](int idx){
    return this->array[idx];
}
    
// Function application operator
// SomeClass obj;
// double x = obj();
//-----------------------------------------------
double SomeClass::operator()(){
    return m_counter * 10;
}

// Function application operator
// SomeClass obj;
// double x = obj(3.4, "hello world");
//-----------------------------------------------
double SomeClass::operator()(double x, std::string msg){
    std::cout << "x = " << x << " msg  = " << msg;
    return 3.5 * x;                                       
}

// Operator string insertion, allows printing the current object 
// SomeClass obj;
// std::cout << obj << std::enl;
//-----------------------------------------------
friend std::ostream& SomeClass::operator<<(std::ostream &os, const SomeClass& cls){
    // Print object internal data structure 
    os << cls.m_x << cls.m_y  ;
    return os;
}

Array index operator overload

This example how to overload the operator array index to allow returning a value or performing an assignment operation.

File: array-index-overload.cpp

#include <iostream>
#include <vector>

class Container{
private:
        std::vector<double> xs =  { 1.0, 2.0, 4.0, 6.233, 2.443};
public:
    Container(){}
    double& operator[](int index){
        return xs[index];
    }
};

int main(){
    Container t;
    std::cout << "t[0] = " << t[0] << std::endl;
    std::cout << "t[1] = " << t[1] << std::endl;
    std::cout << "t[2] = " << t[2] << std::endl;
    std::cout << "\n--------\n";
    t[0] = 3.5;
    std::cout << "t[0] = " << t[0] << std::endl;
    t[2] = -15.684;
    std::cout << "t[2] = " << t[2] << std::endl;    
    return 0;
}

Running:

$ cl.exe array-index-overload.cpp /EHsc /Zi /nologo /Fe:out.exe && out.exe
t[0] = 1
t[1] = 2
t[2] = 4

--------
t[0] = 3.5
t[2] = -15.684

Conversion Operators and user-defined type conversion

Conversion operators allow to convert a class to any type implicitly or explicitly with type-cast operator static_cast<T>.

Example:

• ROOT Script File: conversion-operator.cpp

#include <iostream>
#include <string>

#define LOGFUNCTION(type)  std::cerr << "Convert to: [" << type << "] => Called: line " \
        << __LINE__ << "; fun = " << __PRETTY_FUNCTION__ << "\n"

// Or: struct Dummy { 
class Dummy{
public:
        bool flag = false;

        // Type conversion operator which converts an instance
        // of dummy to double.	
        explicit operator double() {
                LOGFUNCTION("double");
                return 10.232;
        }	
        #if 1
        // Implicit conversion to int is not allowed, it is only possible to convert
        // this object explicitly with static_cast. 	
        explicit operator int() const {
                LOGFUNCTION("int");
                return 209;
        }	
        explicit operator long() const {
                LOGFUNCTION("long");
                return 100L;
        }
        operator std::string() const {
                LOGFUNCTION("std::string");
                return "C++ string std::string";
        }
        explicit operator const char*() const {
                LOGFUNCTION("const char*");
                return "C string";
        }		
        operator bool() const {
                LOGFUNCTION("bool");
                std::cerr << " Called " << __FUNCTION__ << "\n";
                return flag;
        }
        #endif 
};

Testing:

• C-style casting

>> .L conversion-operator.cpp 
>> Dummy d;  

>> (double) d
Convert to: [double] => Called: line 15; fun = double Dummy::operator double()
(double) 10.232000

>> (int) d
Convert to: [int] => Called: line 22; fun = int Dummy::operator int() const
(int) 209

>> (long) d
Convert to: [long] => Called: line 26; fun = long Dummy::operator long() const
(long) 100

>> (std::string) d
Convert to: [std::string] => Called: line 30; fun = std::string Dummy::operator basic_string() const
(std::string) "C++ string std::string"
>> 

• C++ style casting:

>> static_cast<int>(d)
Convert to: [int] => Called: line 22; fun = int Dummy::operator int() const
(int) 209
>> 
>> static_cast<long>(d)
Convert to: [long] => Called: line 26; fun = long Dummy::operator long() const
(long) 100
>> 
>> static_cast<double>(d)
Convert to: [double] => Called: line 15; fun = double Dummy::operator double()
(double) 10.232000

>> static_cast<std::string>(d)
Convert to: [std::string] => Called: line 30; fun = std::string Dummy::operator basic_string() const
(std::string) "C++ string std::string"
>> 

>> static_cast<bool>(d)
Convert to: [bool] => Called: line 38; fun = bool Dummy::operator bool() const
 Called operator bool
(bool) false
>> 

>> d.flag = true
(bool) true

>> static_cast<bool>(d)
Convert to: [bool] => Called: line 38; fun = bool Dummy::operator bool() const
 Called operator bool
(bool) true
>> 

• Simulating implicit conversion:
  • Note: implicitly assignment type conversion is not allowed for operators annotated with explicit. So it is not possible to perform the assignment: const char* s = d

// Implicit conversion 
>> std::string message = d
Convert to: [std::string] => Called: line 30; fun = std::string Dummy::operator basic_string() const
(std::string &) "C++ string std::string"
>> 
>> std::cout << "text = " << message << "\n";
text = C++ string std::string
>> 
>> 

>> const char* s = d
ROOT_prompt_16:1:13: error: no viable conversion from 'Dummy' to 'const char *'
const char* s = d
            ^   ~
// Conversion operators marked as explicit can only casted using C-style cast or 
// or static_cast<T>
>> const char* s = static_cast<const char*>(d)
Convert to: [const char*] => Called: line 34; fun = const char *Dummy::operator const char *() const
(const char *) "C string"

>> d ? "true" : "false";
Convert to: [bool] => Called: line 38; fun = bool Dummy::operator bool() const
 Called operator bool

>> d ? "true" : "false"
Convert to: [bool] => Called: line 38; fun = bool Dummy::operator bool() const
 Called operator bool
(const char *) "true"

>> d.flag = false;

>> d ? "true" : "false"
Convert to: [bool] => Called: line 38; fun = bool Dummy::operator bool() const
 Called operator bool
(const char *) "false"
>> 

• Bool type conversion in conditional statements.

>> d.flag = true;

>> if(d) { std::cout << "Flag is true OK" << std::endl; }
Convert to: [bool] => Called: line 38; fun = bool Dummy::operator bool() const
 Called operator bool
Flag is true OK

>> d.flag = false;

>> if(!d) { std::cout << "Flag is false OK" << std::endl; }
Convert to: [bool] => Called: line 38; fun = bool Dummy::operator bool() const
 Called operator bool
Flag is false OK
>> 
>> 

• Note: The macro __PRETTY_FUNCTION__ is only available in GCC or CLANG, in MSVC use __FUNCSIG__

Further Reading:

• Cast Operator: ()
• User-Defined Type Conversions (C++) | Microsoft Docs
• How do conversion operators work in C++? - Stack Overflow
• Explicit Conversion Operators - http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2004/n1592.pdf

General C++ Terminology and Concepts

Design Principles

• Performance Oriented
  • Zero-cost abstractions.
  • Avoid runtime cost.
  • Value speed over safety.
  • Don’t pay for what you don’t use.
• Backward compatibility - avoid breaking old code.
  • Backward compatibility with C
  • Backward compatibility with old versions of C++.
• Explicit is Better than implicit (Python lemma). For instance, explicit conversion with C++-style casting operators static_cast or reinterpret_pointer are better and safer than C implicit conversion.
• Type-safety: Compile-time errors are always better than run-time errors as compile-timer errors can be caught earlier and doesn’t cause bad surprises if it is deployed elsewhere.

Pointers

• Pointer: A variable which holds the address of another variable. It is used for indirect access of variables, accessing memory mapped IO in embedded systems or in low-level software and also for referencing heap-allocated objects. All C++ ordinary pointers (not function pointers or pointer to member functions) have the same size and store a numeric address of some memory location, the only difference between pointers of different type is the type of memory location that they reference.
• Types of Pointers
  • Ordinary pointers: int*, const char*, Object* …
  • Pointer to function, aka function pointer
  • Pointer to member function (pointer to class method)
  • Pointer to member variable (pointer to class variable or field)
  • Smart “pointers”: (they are not pointers) Stack-allocated objects used for managing heap-allocated objects through RAAI and pointer emulation.
• Wild pointer
  • Non-intialized pointer.
• Dangling pointer
  • A pointer which points to an object that was deleted or to a non-valid memory address. Segmentation faults crashes can happen if one attempts to delete a dangling pointer or invoke object’s method through a dangling pointer.
• Null pointer
  • A pointer set to null (0), NULL or nullptr. It is safe to delete a null pointer, however attempting to invoke an object’s method through the pointer (null pointer deference) has undefined behavior and may cause segmentation fault.
    • c++ - How to catch the null pointer exception? - Stack Overflow
• Void pointer void*
  • A pointer without any specific type associated. A pointer to any type can be converted to void pointer and void pointer can be coverted back to any type. A void pointer also cannot be used before being casted.
  • Can point to:
    • To primitive types int, float, char and so on.
    • To class instances.
    • To functions. Function pointers can be casted to void*
  • Cannot point to:
    • member functions or class methods. So, pointers to member functions cannot be casted to void*.
    • member variables or pointer to class variables. So, pointers to member variables cannot be casted to void*.
  • Use cases:
    • Root class. C++ doesn’t have a root class from which all classes inherites like Java’s Object class. A root class allows unrelated types to be stored in the same data structure or collection and perform type erasure. Void* pointer can work as “pseudo” root class as the pointer to any class can be coverted to it.
    • Type erasure of pointer to primitive types, pointer to classes and pointer to member functions.
    • Type erasure in C-APIs, for instance, malloc and C-API GetProcAddress from Windows which returns a function pointer to a function exported by a DLL casted as void*.
• Owning X Non-owning pointers
  • An owning pointer is responsible to release some allocated memory for a heap-allocated object. In general, raw pointers should not be used as owning pointer as they provide no indication if they point to a heap-allocated object or stack-allocated object or to an heap-allocated array. Another problem, is that every Type ptr* = new Type statement needs to be matched by an delete statement and it is easy to forget to track all possible execution paths. Besides that, raw pointers aren’t exception safe since a matching delete statement may not be executed if an exception occurs. In modern C++, only smart pointers should be used as owning pointers.
• Opaque pointer, also called handler
  • This technique is widely used in C for emulating object-oriented programming features such as abstraction and encapsulation. This approach consists of a pointer to forward declared class or struct and a set of functions which takes the pointer to struct or class as argument. The client code is only allowed to manipulate the data structure by using a set of defined functions since the class or struct declaration is not exposed in the header files.
  • Uses cases:
    • C++ programming
      • => Pimpl - Pointer to Implementation, idiom for improving binary compatibility, reducing compilation time or obfuscating the implementation in the header.
      • => Creating C-APIs for C++ classes with extern “C” for allowing calling the C++ code from C or any foreign function-interface based on LibFFI (example: Java’s JNA, Python’s ctypes or Ruby’s Ruby-FFI) or programming the language C-API such as Python C-API.
    • C Programming
      • => Decrease compile-time.
      • => Improving binary compatibility. The client code doesn’t have to be recompiled if the implementation changes as all pointers in C have the same size and the client program doens’t have any access to the implementation.
      • => Implementation obfuscation.
      • => Emulating object oriented programming: As the client code cannot access the data structure directly without predefined functions, the implementation can be changed without disrupted the client code. In addition, this approach improves binary compatibility.
        • Examples:
          • C’s FILE data structure is not exposed and it is only possible to manipulate pointers to FILE (FILE*) with the functions fopen (works as a constructor), fread (works as a accessor method), fprintf and fclose (works as a destructor) and so on. Those functions emulates object-oriented methods and the opaque pointers works as an object or class instance.
          • Windows API and Windows’ Kernel “Objects” with the type HANDLE.
    • See:
      • Give Me a Handle, and I’ll Show You an Object
      • Object Oriented Programming in C
      • Opaque Data Pointers | Ruminations
      • Object-oriented design patterns in the kernel, part 2
      • What defines an opaque type in C, and when are they necessary and/or useful? - Stack Overflow
      • Object-oriented programming in C - Florian octo Forster’s Homepage
      • Opaque Type Oriented Programming in C | Alejandro Segovia Azapian
      • Legato: C APIs
      • https://www.cs.princeton.edu/courses/archive/fall04/cos217/lectures/06adts.pdf
      • https://en.wikipedia.org/wiki/Libffi
      • Book: C interfaces and implementation techniques: Techniques for creating reusable software
• Pointer “this”
  • Every class has a pointer this of type Class* which points to the current object. The pointer this is similar to Java’s this keyword inside classes.
  • Use cases:
    • Return a reference or pointer to current object.
    • Ambiguity resolution, for instance, if a function has a parameter named count, and a class member has the same name, the ambiguity in assignment operation can be solved with this->count = count;
    • Make it explicit and indicate that a class method is being invoked, for instance, this->method(arg0, arg1, arg2) is more explicit than using method(arg0, arg1, arg2), which could be an external function instead of a class’ member function.

Classes

• struct SomeName { … }
  • A struct is just a class and equivalent to a class with member variables and member public by default. So, unlike in C, structs can have methods, constructor and destructor in C++.
• Polymorphic class
  • A class with at least one virtual member function.
• Abstract class
  • A class with at least one pure virtual member function (abstract method.)
• Pure abstract class or interface class
  • For short: a class with only
  • As C++ doesn’t have the interface keyword, it can be emulated using with only pure virtual member functions (abstract method).
• POD - Plain Old Data
  • Any C-compatible type that can serialized or copied with the C-function std::memcpy (in the header <cstring> or <string.h>). A POD can be a int, double, char, pointer, array, union, struct or class without any constructor, destructor, virtual functions and so on.
    • See:
    • C++ named requirements: PODType - cppreference.com
    • std::is_pod - cppreference.com
    • What are POD types in C++? - Stack Overflow
    • Passive data structure - Wikipedia
    • https://snipplr.com/view/10148/
    • http://www.math.ucla.edu/~wgoakley/Spring2018/04-12.pdf

Member Functions

• member function
  • C++ terminology for class method.
• virtual member function (aka virtual function or virtual method)
  • For short: Method that can be overriden, in other words, derived classes can replace the base class implementation.
  • Any class’ member function (aka method) which can be overriden by derived classes. Only methods annotated with virtual can be overriden.
• pure virtual member function
  • For short: Abstract method. A derived class must provide an implementation.
  • A member function from base class annotated as virtual, however without any implementation. It is the same as an abstract method that should be implemented by derived classes.
• static member function
  • For short: static method.
  • A class method that can be called without any instance.
• special member functions
  • Destructor
  • Constructors
    • Default constructor
    • Copy constructor
    • Move constructor
  • Copy assignment operator
  • Move assignment operator
• Common constructors
  • Default / Empty constructor
    • Signature: CLASS()
    • Constructor without arguments used for default initialization. If this constructor is not defined, the compiler generates it by default. Without this constructor, it is not possible to store a instances of a particular class by value in STL containers.
  • Conversion Constructor
    • Signature: Class(T value)
    • Constructor with a single argument or callable with a single argument. This type of constructor instantiates an object with implicit conversion by assignment or when an instance of type T is passed to a function expecting an object of the underlying class. For instance, this constructor allows intialization as:
      • Class object = value; // Value has type T
      • Class object = 100; // Calls constructor Class(int x).
    • To forbid this implicit conversion use the keyword explicit.
      • explicit Class(T value)
  • List initializer constructor
    • Signature: CLASS(std::intializer_list<T>)
    • Constructor which takes an initializer list as argument. This constructor makes possible to initialize an object with:
      • CLASS object {value0, value1, value2, value3 … };
      • auto object = CLASS {value0, value1, value2, value3 … };
  • Range constructor
    • Signature: CLASS(beginIterator, endInterator)
    • Constructor which takes an iterator pair as arguments. It allows to instantiate objects from STL container iterators.

• Types of polymorphism in C++
  • Dynamic - Resolution at runtime
    • AKA: subtyping polymorphism.
    • Inheritance and virtual functions.
  • Static - Resolution at compile-time
    • Function overload - multiple functions with different signatures sharing the same name.
    • Templates (Parametric polymorphism)
• Polymorphism Binding
  • Early Binding
    • The class method (aka member function) to be called is resolved at compile-time.
  • Late Binding
    • The calss method to be called is resolved at runtime, rather than at compile-time. Late binding is only possible with inheritance and member functions marked as virtual.
    • Drawbacks:
      • Performance cost.
      • Compilers cannot inline virtual member functions.

Linkage

• External Linkage (Default)
  • Variables and functions are accessible from all compilation units (source files) through the whole program. All global variables and functions definitions without the static keyword or outside an anonymous namespace have external linkage.
  • Multiple symbols (variable or function) cannot have the same name.
• Internal Linkage
  • Global variables or functions only acessible in the compilation unit (source file) they are defined. Such variables and functions are defined with static (C-style) keyword annotatation or are defined inside an anonymos namespace (preferable in C++).
  • Multiple symbols can have the same.
  • Symbols with default internal linkage:
    • const objects, constexpr objects, typedefs and objects annoated with static keyword.
• No linkage
  • Local variables in functions or member functions. They are only accessible in the scope they are defined or stack-allocated variables.
• References:
  • Programs and Linkage (C++) | Microsoft Docs
  • What is the point of internal linkage in C++ - Stack Overflow
  • IBM Knowledge Center - Language linkage (C++ only)

Undefined Behavior X Unspecified Behavior

• Undefined Behavior: The C++ ISO Standard provides no gurantees about the program behavior under a particular condition. It means that anything can happen such as runtime crashing, returning an invalid or random value and so on. Undefined behavior should be avoided in order to ensure that the program can work with all possible compilers and platforms.
  • Note: The best case for an UB is a runtime crash as when it doesn’t happen the application may run with a bug which is hard to detect and reproduce, as a result the applicaiton can generate invalid and unpredictable results.
  • Example:
    • Delete a pointer to heap-allocated object twice.
    • Dereference or access a null, wild (non initialized) or dangling pointer.
    • Go out of bounds of a std::vector, std::deque or an C-array.
    • Arithmetic erros, for instance, division by zero.
    • Deference a pointer to a non-initialized heap-object.
    • Signed integer overflow.
    • Call pure virtual-functions from destructors or constructor.
  • See:
    • Undefined behavior - cppreference.com
    • What are all the common undefined behaviours that a C++ programmer should know about? - Stack Overflow
    • A Guide to Undefined Behavior in C and C++, Part 2 – Embedded in Academia
    • 4 Undefined Behaviors in C that You’ve Run into without Knowing It
• Unspecified Behavior
  • It is basically “implementation defined behavior”, the C++ ISO standard requires the behavior to be well defined by a compliant compiler.

Compilation

• Cross-compilation -> Compiling a source code for a different processor architecture or operating system than the compiler was run (host operating system). Cross compilation is common for embedded systems, example: compiling an a program/app or firmware on Windows / x64 for an ARM 32 bits processor.

ABI - Application Binary Interface and Binary Compatibility

The ABI - Application Binary Interface is are a set of specifications about how a source coede is compiled to object-code (machine code). As C++ does not have a standard and stable ABI, it is not possible to static link object codes generated by different compilers or reuse a shared library without a C interface built with a different compiler. Due to the mentioned ABI issues, binary reuse of a C++ code becomes almost impossible, as a result, in general, most C++ codes are only reused as source.

The ABI is defined by the compiler and the operating system and it the binary interface is not specified by the ISO C++ standard. Among other things, the ABI specifies:

• Class layout: VTable Layout, padding, member function-pointer, RTTI and so on.
• Exception implementation and exception hanling
• Linkage information
• Name decoration schema (name mangling)
  • The schema or rules used to encode symbols in a unique way. In C, every symbol in an object code has the same name as the function that it refers to. As the object code must have a unique symbol for every function and C++ supports templates, classes or function overloading, the compiler must generate a unique name for every symbol. This process is called name mangling or name decoration. This name encoding is compiler dependent and one of the sources of ABI incompatibilities.
  • Note: the statement (extern “C”) disables name mangling specifying to the compiler that the function has C-linkage and the function symbol is the same as its name.

Notes:

• The ABI incompatibility can also happen even between different versions of the same compilers.
• Due to the ABI problems, it is almost impossible to distribute pre-compiled C++ code as static or shared libraries. As a result, unlike C shared libraries, it is hard to find pre-compiled C++ libraries available as shared libraries.
• The only way to build binary componets with C++ which can be reused by other codes in C, C++ or other programming languages via FFI (Foreign-Function Interface) is by defining a C-interface (extern “C”) for all C++ classes and functions.
• Newer verions of GCC and Clang on Unix-like operating systems are adopting the Itanium ABI which mitigates the ABI problem, however it is not guaranteed by the C++ standard.

References:

• Defining a Portable C++ ABI - Herb Sutter
  • https://isocpp.org/files/papers/n4028.pdf
• Stability of the C++ ABI: Evolution of a Programming Language
  • https://www.oracle.com/technetwork/articles/servers-storage-dev/stablecplusplusabi-333927.html
• GCC5 and the C++11 ABI - Red Hat Develper
  • https://developers.redhat.com/blog/2015/02/05/gcc5-and-the-c11-abi/
• Binary Compatibility for library developers
  • https://events.static.linuxfound.org/sites/events/files/slides/Binary_Compatibility_for_library_devs.pdf
• Some thoughts on binary compatibility
  • https://blog.qt.io/blog/2009/08/12/some-thoughts-on-binary-compatibility/
• Application binary interface
• Itanium C++ ABI (IA-64 ABI) - Itanium processor ABI is becoming a standard among compiler vendos. This ABI was adopted by Clang and GCC. MSVC, aka VC++ (“Visual” Studio Compiler) does not use this ABI and breaks its ABI compatibility on every release.
• Macintosh C/C++ ABI - Standard Specification 1993

ABI - Fragile Base Class or Fragile Binary Interface

• C++ has the fragile base class problem that happens when changes in a base class break its ABI requiring recompilation of all derived classes, client code or third-party code. This issue is specially important for large projects, SDKs osftware development kit, libraries or plugin-systems where third-party a code is dynamically loaded at runtime.

What can keep or break a base class ABI compatibility:

• DO Changes which that do not break the base class ABI: (KDE Guide)
  • Append new non-virtual member functions.
  • Add Enumeration to class.
  • Changet the implementation of virtual member functions (overridable methods) without changing its signature (interface).
  • Create new static member functions (static methods)
  • Add new classes
  • Append or remove friend functions
  • Rename class private member variables
• DONT Changes that breaks the class ABI and disrupts binary compatibility: (KDE Guide)
  • Change the order of existing virtual member functions
  • Add virtual member function (method) to a class without any virtual member function or virtual base class.
  • Add or remove virtual member functions
  • Addition or removal of member variables
  • Change the order of member variables
  • Change the type of member variables

Techniques for keeping the ABI compatibility:

• PIMPL - Use the PIMP (Pointer to implementation) technique for encapsulatiing the member variables into a opaque pointer which the implementation is not exposed in the header file. The opaque pointer becomes the unique class member variable exposed in the header file, as a result any change of the encapsulated member variables no longer breaks the class ABI.
• Interface Class - An interface class has only virtual member functions, virtual constructor and no member variables.
• Extend, but not modify, do not change interfaces or base classes relied by external codes, libraries or client code. If a new functionality is needed, it is better to create a new class extending the base class instead of modifying it what would break extenal codes relying on it.
• Prefer composition to inheritance
• C-interface (extern “C”) with opaque pointer - C-interface or C-wrapper with C-linkage functions and opaque pointers. The classes and functions are not exposed and the client code can only access the library using the C-API or functions with C-linkage. This is the only reliable way to share compiled code between different compilers.

References:

• Binary Compatibility for library developers
  • https://events.static.linuxfound.org/sites/events/files/slides/Binary_Compatibility_for_library_devs.pdf
• TECH : ABI bugs are a NIGHTMARE!
  • https://www.snsystems.com/tech-blog/2014/12/18/abi-bugs-are-a-nightmare/
• http://wiki.c2.com/?FragileBinaryInterfaceProblem
• KDE - Policies/Binary Compatibility Issues With C++
  • https://community.kde.org/Policies/Binary_Compatibility_Issues_With_C%2B%2B
• https://www.gamedev.net/forums/topic/282869-working-around-fragile-base-class-syndrome/
• java - What is the fragile base class problem? - Stack Overflow
• Why is base-for-all-objects discouraged in C++ - Software Engineering Stack Exchange
• programming practices - Is it bad habit not using interfaces? - Software Engineering Stack Exchange
• http://www.vincehuston.org/cpp/inheritance.html
• MacTech - C++ Versus Objective C - Shows how ObjectiveC solves the fragile base class problem.
• If the shell is written in C++, why not just export its base classes? (The Old-new thing)
• A Very Short Note on Why C++ Is Not Suitable for Plug-In Architecture

Processes and Operating Systems

Protection mode, Kernel and User Spaces

Real Mode

• Old operating systems like Microsft MSDOS and Windows 95 ran in real mode, which means that any programs can access the physical memory (RAM memory), memory mapped IO and hardware directly without any restriction which could result in security and stability problems as any process could take down the whole operating system. Summary: no separating between kernel and user spaces.

Protected Mode

• Modern operating systems such as Windows, MacOSX and Linux run in protected mode, which has the kernel space and user space.
• User Space - Programs running in user space, runs with less privilege, they are not allowed to run some CPU machine instructions and to access hardware devices or physical memory directly. Applications in user space, can only a restricted portion of the physical memory assigned by the operating system, called virtual memory. This protection is enforced both by the operating system and the processor.
• Kernel Space - Only programs running in kernel space can access the whole physical memory, any process memory and execute all CPU instructions.

Processes ad Virtual Memory

Process

• A unique instance of a running program with its own PID (Process Identifier), address-space, virtual memory and threads. Any application, executable or program can have multiple processes running on the same machine with different states.

Process State (PCB - Process Control Block) Every process has the following states.

• CPU Registers (IP Instruction pointer and stack pointer). A CPU core only has a single IP Instruction pointer. However every process has its own IP pointer because the operating system switches between processes in a very fast way performing context switch, saving and restoring the CPU register for every process giving the illusion that multiple processes are running simultaneously.
• => PID - Unique Process ID (Identifier) number.
• => Command line arguments used to start the process.
• => Current directory.
• => Environment variables
• => One or more threads
• => File descriptors associated with the process.

Virtual Memory

• Portion of physical memory assigned to a process by the operating system’s kernel. In most operating systems, a process cannot access the physical memory, all the memory that it can see and referece is its virtual memory. For instance, the address of a pointer to some variable is not the address of the variable in the physical memory, instead it is the address of the variable in the current process virtual memory.
  • => C++ => Pointers to variables stores the numerical value of a virtual memory address. (Note: only for programs that runs on operating systems, not valid for firmwares.)
  • => The C++ standard does not define whether pointer addresses refer to virtual or physical memory, this behavior is platform-dependent.
• Physical Address
• Virtual Address
• Process Isolation: One of the purposes of the virtual memory is to not allow a user-space process to read the memory of another process.
  • Note: Operating systems provide APIs for reading and writing process memory, otherwise debuggers would not exist.
• Virtual Memory Segments: Every process, no matter the programming language it was written, has the following memory segments in its virtual memory:
  • Stack segment => Stores stack frames, functions local variables and objects and return addresses.
  • Heap segment (ak free store) => Dynamically allocated variable with C++ operator new or C function malloc.
  • Data Segment => Stores initialized and non-initialized global variables.
  • Text Segment => Stores the program machine code that cannot be modified. (read-only)

Other Virtual Memory Segments

• Memory Mapped Files (Inter process communication)
  • Allows a disk file to be mapped into the virtual memory and be accessed just as an ordinary memory through pointer manipulation. This segment can be mapped the virtual memory of many processes without incurring on copying overhead.
• Shared Memory - allows processes to shared data without copying.
• Dynamic Library Loading (DLLs)
• Thread Stack

System Calls and Operanting Systems C-APIs

Operating System APIs - Most operating systems are written in C and processor-specific assembly. Their APIs (Application Programming Interfaces) and services are exposed in C language, this API can be:

• System Calls
  • => Documented on Linux, BSD and etc. Undocumented on Windows. Note: Linux has fixed number for every system call which is documented and standardize. On Windows, the system calls may change on every release, so it is only safe to rely on the Win32 API encapsulating them.
• Basic C APIs that encapsulates system calls. Some those APIs are:
  • Win32 API - Windows API
  • POSIX API - Standardized UNIX API shared by most Unix-like operating systems, Linux, BSD, MacOSX and so on.

References

• Linux System-calls table
  • https://filippo.io/linux-syscall-table/
• Linux System Call Table for x86 with parameters
  • https://thevivekpandey.github.io/posts/2017-09-25-linux-system-calls.html
• Lecture 2: System Calls &API Standards
  • https://cw.fel.cvut.cz/old/_media/courses/ae3b33osd/osd-lec2-14.pdf
• Difference Between Reald and Protected Mode
  • https://www.geek.com/chips/difference-between-real-mode-and-protected-mode-574665/
• OS Segmentation
  • https://alex.dzyoba.com/blog/os-segmentation/
• PC Hardware & Booting - IIT Madras
  • http://www.cse.iitm.ac.in/~chester/courses/15o_os/slides/3_Hardware.pdf
• Processes, Addresses Spaces and Context Switches - IIT Madras
  • http://www.cse.iitm.ac.in/~chester/courses/15o_os/slides/6_Processes.pdf
• Context Switch – Software vs Hardware Approach
  • https://shobhitsharda.wordpress.com/2011/05/31/context-switch-software-vs-hardware-approach/
• Instrumentation usign Free/Open Code
  • https://www.uio.no/studier/emner/matnat/fys/FYS3240/v16/lecture-notes/linux.pdf
• 10 Operating System Concepts Software Developers Need To Remember
  • https://medium.com/cracking-the-data-science-interview/the-10-operating-system-concepts-software-developers-need-to-remember-480d0734d710

Common Acronyms, abbreviations and technologies

• Note: Technical standards aren’t laws, they are specifications, recommendations for standardization and set of good practices.

Acronym, name or technology	Description
Organizations
ANSI	American National Standards Institute
NIST	National Institute of Standards and Technology
ISO	International Organization for Standardization
IEEE	Institute of Electrical and Electronics Engineers
IEC	International Electrotechnical Commission
CERN	European Organization for Nuclear Research
MISRA	Motor Industry Software Reliability Association
Technical Standards
ISO/IEC 14882 - C++	C++ Programming Language Standard and Specification used by most compiler vendors.
ISO/IEC 14882:2003	C++03 Standard
ISO/IEC 14882:2011	C++11 Standard
ISO/IEC 14882:2014	C++14 Standard
ISO/IEC 14882:2017	C++17 Standard
ANSI X3.159-1989	C-89 - C programming language standard
ISO/IEC 9899:1990	C-90 standard
ISO/IEC 9899:1999	C-99 standard
IEE754	Floating Point technical standard
ISO 8601	Date and time standard widely used on computers and internationalization.
Technical Standards for Embedded Systems
IEC 61508	Standards for funcitonal safety of Electrical/Electronic/Programmable Safety-Related System
ISO 26262	IEC 61508 Applied to automotives up to 3.5 tons - comprises electronic/electrical safety (includes firmware)
IEC 62304	International standard for medical device software life cycle
General - C++
CPP	C++ Programming Language
TMP	Template Meta Programming
STL	Standard Template Library
ODR	One Definition Rule
ADL	Argument Dependent Lookup
ASM	Assembly
GP	Generic Programming
CTOR	Constructor
DTOR	Destructor
RAAI	Resource Acquisition Is Initialization
SFINAE	Substitution Is Not An Error
RVO	Return Value Optmization
EP	Expression Template
CRTP	Curious Recurring Template Pattern
PIMPL	Pointer to Implementation
RTTI	Runtime Type Identification
MSVC	Microsoft Visual C++ Compiler
VC++	Microsoft Visual C++ Compiler
AST	Abstract Syntax Tree
RPC	Remote Procedure Call
rhs	right-hand side
lhs	left-hand side
Operating Systems Technologies
IPC	Interprocess Communication
COM	Component Object Model - (Microsoft Technology)
OLE	Object Linking and Embedding (Windows/COM)
IDL	Interface Description Language
MIDL	Microsft Interface Description Language - used for create COM components
DDE	Dynamic Data Exchange - Windows shared memory protocol
RTD	Real Time Data (Excel)
U-NIX like	Any operating based on UNIX (Opengroup trademark) such as Linux, Android, BSD, MacOSX, iOS, QNX.
BLOB	Binary Large Object
GOF (Gang of Four)	Book: Design Patterns: Elements of Reusable Object-Oriented Software
POSIX	Portable Operating System Interface (POSIX)
Network Protocols
RFC	Internet Taskforce - Request for Comment
ARP	Address Resolution Protocol
DHCP	Dynamic Host Configuration Protocol
IP	Internet Protocol (Sockets)
TCP	Transmissiion Control Protocol (Sockets)
UDP	User Datagram Protocol
DNS (UDP Protocol)	Domain Name System
ICMP (ping)	Internet Control Message Protocol - Ping Protocol
HTTP	Hyper Text Transfer Protocol
FTP	File Transfer Protcol
Modbus	Network protocol used by PLCs
CAN Bus (not TCP/IP)	Controller Area Network - distributed network used in cars and embedded systems.
Executable Binary Formats
PE, PE32 and PE64	Portable Executable format - Windows object code format.
ELF, ELF32 and ELF64	Executable Linkable Format - [U]-nix object code format.
MachO	Binary format for executables and shared libraries used by the operating systems iOS and OSX.
DLL	Dynamic Linked Library - Windows shared library format.
SO	Shared Object - [U]-nix, Linux, BSD, AIX, Solaris shared library format.
DSO	Dynamic Shared Object, [U]-nix shared library format.
Cryptography
HMAC	Keyed-Hash Message Authentication Code
MAC	Message Authentication Code
AES	Advanced Encryption Standard
Crypto Hash Functions
MD5
SHA1
SHA256
Processor Architectures
CISC	Complex Instruction Set Computer
RISC	Reduced Instruction Set Computer
SIMD	Single Instruction, Multiple Data
Havard Architechture	Used mostly in DSPs, Microcontrollers and embedded systems.
Von-Neumann Architechture	Used mostly in conventional processors.
IBM-PC Architecture Components
BIOS	Basic Input/Output System - Firmware used to initialize and load OS in IBM-PC arch.
UEFI	Unified Extensible Firmware Interface - BIOS replacement on new computers.
DMA	Direct Memory Access
MMU	Memory Management Unit - Hardware that translates physical memory to virtual memory.
PCI	Peripheral Component Interconnect Express - BUS used in IBM PCs
NIC	Network Interface Controller/Card
RAID (storage)	Redundant Array of Independent Disks
Hardware and processors
CPU	Central Processing Unit
MPU	Micro Processor Unit
FPU	Floating Point Unit
DSP	Digital Signal Processor
MCU	Microcontroller Unit
SOC	System On Chip
GPU	Graphics Processing Unit
FPGA	Field Programmable Gate Array
ASIC	Application-Specific Integrated Circuit
ECU	Engine Control Unit or Electronic Control Unit - Car’s embedded computer.
Peripherals
RAM	Random Access Memory
ROM	Read-Only Memory
EPROM	Erasable Programmable Read-only Memory
EEPROM	Electrically Erasable Programmable Read-Only Memory
GPIO	General Purpose IO
ADC	Analog to Digial Converter
DAC	Digital to Analog Converter
PWM	Pulse Width Modulation
Serial interface I2C
Serial interface SPI	Seria Peripheral Interface
Serial interface UART	Serial communication similar to the old computer serial interface RS232
Serial interface Ethernet
CAN bus	Controller Area Network - Widely used BUS in the automotive industry.
DSI	Display Serial Interface
MEMs	Microelectromechanical Systems - mechanical sensors implemented in silicon chips.

Bits, bytes, sizes, endianess and numerical ranges

Bits in a byte by position

 MSB                                LSB 
(Most significant bit)         (Least significant bit)
   |                                  |
 | b7 | b6 | b5 | b4 | b3 | b2 | b1 | b0 |         Bit Decimal    Bit shift    Multiplier
 +----|----+----+----+----+----+----+----+           Value         Operation   DEC   HEX
   |    |    |    |    |   |     |     |             ........     ........     ........
   |    |    |    |    |   |     |     \---------->> b0 x 2^0   =  b0 << 0     1    0x01
   |    |    |    |    |   |      \--------------->> b1 x 2^1   =  b1 << 1     2    0x02 
   |    |    |    |    |    \--------------------->> b2 x 2^2   =  b2 << 2     4    0x04
   |    |    |    |     \------------------------->> b3 x 2^3   =  b3 << 3     8    0x08 
   |    |    |    \------------------------------->> b4 x 2^4   =  b3 << 4    16    0x10 
   |    |    \------------------------------------>> b5 x 2^5   =  b5 << 5    32    0x20 
   |    \----------------------------------------->> b6 x 2^6   =  b6 << 6    64    0x40
   \---------------------------------------------->> b7 x 2^7   =  b7 << 7   128    0x80

Example:

Binary number: 0b10100111 = 0b1010.0111 = 167 = 0xA7
1010 => Upper nibble in the hex table is equal to 'A'
0111 => Lower nibble in the hex table is equal to '7'

| b7 | b6 | b5 | b4 | b3 | b2 | b1 | b0 |  
+----+----+----+----+----+----+----+----+
| 1  |  0 | 1  | 0  | 0  | 1  | 1  |  1 |

Decimal Value of a Bit of Order N = 2^N

Decimal Value of 0xA7 = Σ b[i] x 2^i 
              = b0 x 2^0 + b1 x 2^1 + b2 x 2^2 + b3 x 2^3 + b4 x 2^4 + b5 x 2^5 + b6 x 2^6 + b7 x 2^7
              = 1  x 2^0 + 1  x 2^1 + 1  x 2^2 + 0  x 2^3 + 0  x 2^4 +  1 x 2^5 + 0  x 2^6 + 1  x 2^7
              = 1  x 1   + 1  x 2   + 1  x 4   + 0  x 8   + 0  x 16  +  1 x 32  + 0  x 64  + 1  x 128
              = 1 + 2 + 4 + 0 + 0 + 32 + 0 + 128
              = 167 OK

Decimal Value of 0xA7 = Σ H[i] x 16^i  where H[i] is a hexadecimal digits 
                      = 16^0 * 7  + A * 16^1  
                      = 1 * 7     + 10 * 16  
                      = 7         + 160 
                      = 167 OK

Decimal - Hexadecimal and Binary Conversion Table

Decimal	Hexadecimal	Binary
Base 10	Base 16	Base 2
0	0	0000
1	1	0001
2	2	0010
3	3	0011
4	4	0100
5	5	0101
6	6	0110
7	7	0111
8	8	1000
9	9	1001
10	A	1010
11	B	1011
12	C	1100
13	D	1101
14	E	1110
15	F	1111

Bits and byte bitmak by position

Bit N	Binary	Decimal	Hex
0	0b0000.0001	1	0x01
1	0b0000.0010	2	0x02
2	0b0000.0100	4	0x04
3	0b0000.1000	8	0x08
4	0b0001.0000	16	0x10
5	0b0010.0000	32	0x20
6	0b0100.0000	64	0x40
7	0b1000.0000	128	0x80
All bits set	0b1111.1111	255	0xFF

Binary - Octal Conversion

Octal	Binary
Base 8	Base 2
0	000
1	001
2	010
3	011
4	100
5	101
6	110
7	111

Ascii Table - Character Enconding

Ascii Table

Special Characters and New Line Character(s)

Dec	Hex	Char		Dec	Hex	Char		Dec	Hex	Char		Dec	Hex	Char
0	00	NUL ‘\0’		32	20	SPACE		64	40	@		96	60	`
1	01	SOH		33	21	!		65	41	A		97	61	a
2	02	STX		34	22	”		66	42	B		98	62	b
3	03	ETX		35	23	#		67	43	C		99	63	c
4	04	EOT		36	24	$		68	44	D		100	64	d
5	05	ENQ		37	25	%		69	45	E		101	65	e
6	06	ACK		38	26	&		70	46	F		102	66	f
7	07	BEL ‘\a’		39	27	’		71	47	G		103	67	g
8	08	BS ‘\b’		40	28	(		72	48	H		104	68	h
9	09	HT ‘\t’		41	29	)		73	49	I		105	69	i
10	0A	LF ‘\n’		42	2A	*		74	4A	J		106	6A	j
11	0B	VT ‘\v’		43	2B	+		75	4B	K		107	6B	k
12	0C	FF ‘\f’		44	2C	,		76	4C	L		108	6C	l
13	0D	CR ‘\r’		45	2D	-		77	4D	M		109	6D	m
14	0E	SO		46	2E	.		78	4E	N		110	6E	n
15	0F	SI		47	2F	/		79	4F	O		111	6F	o
16	10	DLE		48	30	0		80	50	P		112	70	p
17	11	DC1		49	31	1		81	51	Q		113	71	q
18	12	DC2		50	32	2		82	52	R		114	72	r
19	13	DC3		51	33	3		83	53	S		115	73	s
20	14	DC4		52	34	4		84	54	T		116	74	t
21	15	NAK		53	35	5		85	55	U		117	75	u
22	16	SYN		54	36	6		86	56	V		118	76	v
23	17	ETB		55	37	7		87	57	W		119	77	w
24	18	CAN		56	38	8		88	58	X		120	78	x
25	19	EM		57	39	9		89	59	Y		121	79	y
26	1A	SUB		58	3A	:		90	5A	Z		122	7A	z
27	1B	ESC		59	3B	;		91	5B	[		123	7B	{
28	1C	FS		60	3C	<		92	5C	\		124	7C
29	1D	GS		61	3D	=		93	5D	]		125	7D	}
30	1E	RS		62	3E	>		94	5E	^		126	7E	~
31	1F	US		63	3F	?		95	5F	_		127	7F	DEL
Char	Caret	Name	Hex	Dec	Observation
	Notation
‘\0’		Null character	0x00	00	-
‘\t’		Tab	0x09	09	-
’ ’		Space	0x20	32	-
‘\r’	^M	(CR) - Carriage Return	0x0D	13	Line separator for text files on Old Versiosn of MacOSX
‘\n’	^J	(LF) - Line Feed	0x0A	10	Line separator for text files on most Unix-like OSes, Linux and MacOSX
‘\r\n’	^J^M	(CR-LF)	-		Line separator for text files on Windows

Bits, bytes, Megabytes and information units

Unit	In bits	In bytes	In Kbytes	In Mega Bytes	In Gigabytes
bit	1	-	-	-	-
byte	8	1	-	-	-
Kbyte (kb)	1024 x 8	1024	1	-	-
Mega Byte (MB)	1024 x 1024 x 8	1024 x 1024	1024	1	-
Giga Bytes (GB)	-	-	-	1024	-
Tera Bytes	-	-		-	1024

Summary:

• Basic unit 1 bit = (0 or 1), (True or False), (On or Off)
• 1 Nibble = 4 bits
• 1 byte = 8 bits
• 1 kb (kbyte) = 1024 bytes
• 1 Mb (Mega byte) = 1024 Kbytes
• 1 Gb (Giga byte) = 1024 Megabytes
• 1 TB (Tera Byte) = 1024 Giga bytes
• 1 PT (Penta Byte) = 1024 Tera bytes

Bit Manipulation for Low Level and Embedded Systems

The following bit manipulation idioms are widely used in legacy C code, embedded systems code, device driver code or for manipulating arbitrary bits of some variable:

Memory Mapped IO

The following code simulates a MMIO memory mapped IO in a embedded system (a microcontroller), more specifically a 8-bits GPRIO - General Purpose IO a digital IO located at the fixed address 0xFF385A (defined in the device’s datasheet or memory map). Setting the first bit (bit 0) of this IO device, makes the LED attached to the first pin be turned ON, clearing this bit makes the led to be turned off.

• volatile keyword => Tells the compiler to disable optimization for this variable and indicates that it can be changed any time.
• reinterpret_cast => Indicates that it is a memory reinterpretation cast, indicates that the memory at address 0xFF385A is being reintreted as an 8-bit unsigned integer.
• constexpr => Compile-time constant, has no storage space. Costs any program memory (ROM, flash) space. The value GPRIO_ADDRESS is replaced where it is used.
• The hypothetical program (firmware) runs without any operating system, therefore, it has access to all physical memory.

#include <cstdint> 

// address taken from device's datasheet supplied by manufacturer. 
constexpr uintptr_t GPRIOA_ADDRESS = 0xFF385A;
 
// Access memory mapped IO register at 0xFF385A using pointer. 
volatile const std::uint8_t>* pGPRIOA = reinterpret_cast<std::uint8_t*>(GPRIOA_ADDRESS);    

// Access memory mapped IO register at 0xFF385A using reference. 
volatile std::uint8_t& GPRIOA = *reinterpret_cast<std::uint8_t*>(GPRIOA_ADDRESS);    

Bitwise Operators Reminder

(|) => X_or_Y   = a | b; => bitwise OR 
(&) => X_and_Y = a & b;  => bitwise AND 
(^) => X_xor_Y = X ^ Y;  => bitwise XOR 
(~) => not_x = ~X;       => bitwise NOT => Invert all bits 

Left shift => bitshift Operator: 
X << Y = X * 2^Y    => Shift Y bits to the left.

Right shift => bitshift Operator:
X >> Y = X / 2^Y    => Shift Y bits to the right.

Read/Get the N-th bit

bit_value = (GPRIOA >> N) & 0x01;

// Check if bit 4 is set 
if((GPRIOA >> 4) & 0x01)
{ 
  ... 
}

// Check if 0-th bit is set 
if((GPRIOA >> 0) & 0x01 == 1)
{ 
   ... 
}

  // Check if 6-th bit is set 
if((GPRIOA >> 6) & 0x01 == 1)
{ 
   ... 
}

Setting the Nth-bit

Set the N-th bit (turn bit into 1) of a general variable:

// Verbose way 
<VARIABLE> = <VARIABLE> | (1 << N); 
// Short way 
<VARIABLE> |= (1 << N); 

Set the 4-th bit - (turn on the 4th LED in this case)

// Verbose way 
GPRIOA = GPRIOA | (1 << 4); 
// Short way 
GPRIOA |= 1 << 4; 

Clear the Nth-bit

Clear the N-th (turn the bit into zero) bit of general variable:

// Verbose way 
<VARIABLE> = <VARIABLE> & ~(1 << N); 
// Short way 
<VARIABLE> &= ~(1 << N); 

Clear the 5-th bit (turn on the 4th LED in this case)

// Verbose way 
GPRIOA = GPRIOA & ~(1 << 5); 
// Short way 
GPRIOA &= ~(1 << 5); 

Analysis:

         Bitshift operation 
         1 << 5 = 2^5 = 32 = 0x20 = 0b00010000

                      B7  B6  B5  B4  B3  B2  B1  B0   BITS 
                      --------------------------------  
           1 << 5  =>  0   0   1   0   0   0   0   0    => Equivalent value to 1 << 5 
         ~(1 << 5) =>  1   1   0   1   1   1   1   1    => Invert all bits of (1 << 5)
          GPRIOA   =>  b7  b6  b5  b4  b3  b2  b1  b0   => Bits of GPRIOA 
                   -----------------------------------
GPRIOA & ~(1 << 5) =>  b7  b6  0   b4  b4  b3  b1  b0   => Result of AND (&) bitwise operation 

Invert all bits

VARIABLE = ~VARIABLE; 

Invert all bits of GPIOA:

GPIOA = ~GPIOA;

Toggle the Nth-bit

Toggle operation: if the bit is 1, turn it into 0, if it is 0, turn it into 1.

// Verbose way 
<VARIABLE> = <VARIABLE> ^ (1 << N); 
// Short way 
<VARIABLE> ^= (1 << N); 

Toggle the bit 6 of GPIOA register:

// Verbose 
GPIOA = GPIOA ^ (1 << 6); 
// Short 
GPIOA ^= (1 << 6); 

Numerical Ranges for Unsigned Integers of N bits

N bits	Min	Max	Max in Hexadecimal	Number of values
8	0	255	0x00FF	256
10	0	1023	0x03FF	1024
12	0	4095	0x0FFF	4096
16	0	65535	0xFFFF	65536
32	0	1E9 =~ 10 billions	-	2^32
64	0	1E19	-	2^64

Formula:

Maximum Unsigned NumberOf N bits  = 2^(n - 1) 
Max Unsigned 8 bits  =  2^8 - 1  = 256 - 1  = 255 
Max Unsigned 10 bits =  2^10 - 1 = 1024 - 1 = 1023 

Numerical Ranges for Signed Integers of N bits

N bits	Min	Max
8	-128	127
10	-512	511
12	-2048	2047
16	-32768	3767
32	-2147483648	+2147483647
64	~ -1E19 = -1 x 10^19	~ 1E19 = 1 x 10^19

minNumberOfNbits = -2^(N - 1)
maxNumberOfNbits = 2^(N - 1) - 1

minNumberOfNbits[N = 8] = -2^(8 - 1)    = -2^7     = -128 
maxNumberOfNbits[N = 8] =  2^(8 - 1) - 1 = 2^7 - 1 = +127

minNumberOfNbits[N = 16] = -2^(16 - 1)    = -2^15     = -32768
maxNumberOfNbits[N = 16] =  2^(16 - 1) - 1 = 2^15 - 1 = +32767

Endianess - Big Endian X Little Endian

The endinaess is the order in which bytes are stored in which the bytes of some data are encoded in the memory, disk, file or network protocol.

The endianess matters in:

• Embedded Systems
• Dealing with raw binary data
• Data Serialization
• Processor memory layout
• Network data transmission

Little Endian - LE

The least significant byte is stored first. In a big-endian processor or system, the number 0xFB4598B2 (bytes 0xFB 0x45 0x98 0xB2 ) would be stored as:

• LSB - Least Significant Byte
• MSB - Most Significant Byte

Memory Address	Order	Data	Tag
0x100	0	0xB2	LSB
0x101	1	0x98
0x102	2	0x45
0x103	3	0xFB	MSB

Endianess and C++:

• This session in CERN’s REPL shows the memory layout endianess of the number 0xFB4598B2 in a Intel x64 processor (Vanilla Desktop - IBM-PC processor). Note: In a bing-endian processor the byte order display in the next code block would be in reverse order.

>> int k = 0xFB4598B2
(int) -79325006
>> 

// Print integers in hex formats 
std::cout << std::hex; 

>> std::cout << "k = " << k << "\n";
k = fb4598b2
>> 

>> *p
(char) '0xb2'

>> *(p + 1)
(char) '0x98'

// Print bytes using pointer offset
>> std::cout << "p[0] = 0x" << (0xFF & (int) *(p + 0)) << "\n";
p[0] = 0xb2
>> 
>> std::cout << "p[1] = 0x" << (0xFF & (int) *(p + 1)) << "\n";
p[1] = 0x98
>> std::cout << "p[2] = 0x" << (0xFF & (int) *(p + 2)) << "\n";
p[2] = 0x45
>> std::cout << "p[3] = 0x" << (0xFF & (int) *(p + 3)) << "\n";
p[3] = 0xfb

// Print bytes using array notation 
>> std::cout << "p[0] = 0x" << (0xFF & (int) p[0]) << "\n";
p[0] = 0xb2
>> std::cout << "p[1] = 0x" << (0xFF & (int) p[1]) << "\n";
p[1] = 0x98
>> std::cout << "p[2] = 0x" << (0xFF & (int) p[2]) << "\n";
p[2] = 0x45
>> std::cout << "p[3] = 0x" << (0xFF & (int) p[3]) << "\n";
p[3] = 0xfb
>>  

Big Endian - BE

The most signficant byte is stored first, the bytes of some data are stored in reverse order than the little endian (LE) encoding.

Detect Edianess in C++ at runtime

Memory Address	Order	Data	Tag
0x100	0	0xFB	MSB
0x101	1	0x45
0x102	2	0x98
0x103	3	0xB2	LSB

Check whether current system is little endian:

bool isLittleEndian()
{
    int n = 1;
    return *(reinterpret_cast<unsigned char*>(&n)) == 1;
}

Check whether current system is big endian:

bool isBigEndian()
{
    int n = 1;
    return *(reinterpret_cast<unsigned char*>(&n)) == 0;
}

Processors Endianess

Processor / CPU Family	Endianess	Note:
Intel x86, x86-x64 and IA-32	Little Endian	Default processor of IBM-PC architechture
ARM	Little Endian	Default endianess, can also be Big-Endian
Sparcs	Big Endian
Motorola 68000	Big Endian
*JVM - Java Virtual Machine	Big-Endian	*Not a processor.
MIPS	Supports Both
PowerPC	Supports Both

References

• QNX - Freedom from Hardware and Platform Dependencies
• ECE2049 Lecture 2: Endianness and Memory Organization
• Endianness - Wikipedia
• CDA-4101 Lecture 4 Notes
• Computer Systems and Network - Endianess

Quotes

Bjarne Stroustrup

• Bjarne Stroustrup

C++11 feels like a new language: The pieces just fit together better than they used to and I find a higher-level style of programming more natural than before and as efficient as ever.

• Bjarne Stroustrup - A brief look at C++

C++is a multi-paradigm language. In other words, C++was designed to support a range of styles. No sin-gle language can support every style. However, a variety of styles that can be supported within the frame-work of a single language. Where this can be done, significant benefits arise from sharing a common type system, a common toolset, etc. These technical advantages translates into important practical benefits suchas enabling groups with moderately differing needs to share a language rather than having to apply a num-ber of specialized languages.

• Bjarne Stroustrup - C++ Programming Language

There are only two kinds of languages: the ones people complain about and the ones nobody uses.

• Bjarne Stroustrup – http://www.stroustrup.com/bs_faq.html#really-say-that

C makes it easy to shoot yourself in the foot; C++ makes it harder, but when you do it blows your whole leg off.

• Bjarne Stroustrup – Mit tech review

C++ has indeed become too “expert friendly” at a time where the degree of effective formal education of the average software developer has declined. However, the solution is not to dumb down the programming languages but to use a variety of programming languages and educate more experts. There has to be languages for those experts to use– and C++ is one of those languages.

What I did do was to design C++ as first of all a systems programming language: I wanted to be able to write device drivers, embedded systems, and other code that needed to use hardware directly. Next, I wanted C++ to be a good language for designing tools. That required flexibility and performance, but also the ability to express elegant interfaces. My view was that to do higher-level stuff, to build complete applications, you first needed to buy, build, or borrow libraries providing appropriate abstractions. Often, when people have trouble with C++, the real problem is that they don’t have appropriate libraries or that they can’t find the libraries that are available.

• Bjarne Stroustrup - Slashdot Interview

The technical hardest problem is probably the lack of a C++ binary interface (ABI). There is no C ABI either, but on most (all?) Unix platforms there is a dominant compiler and other compilers have had to conform to its calling conventions and structure layout rules - or become unused. In C++ there are more things that can vary - such as the layout of the virtual function table - and no vendor has created a C++ ABI by fiat by eliminating all competitors that did not conform. In the same way as it used to be impossible to link code from two different PC C compilers together, it is generally impossible to link the code from two different Unix C++ compilers together (unless there are compatibility switches).

Alexander A. Stepanov

• Alexander A. Stepanov

I still believe in abstraction, but now I know that one ends with abstraction, not starts with it. I learned that one has to adapt abstractions to reality and not the other way around.

• Alexander A. Stepanov - From Mathematics to Generic Porgramming.

To see how to make something more general, you need to start with something concrete. In particular, you need to understand the specifics of a particular domain to discover the right abstractions.

• Alexander A. Stepanov, From Mathematics to Generic Programming

When writing code, it’s often the case that you end up computing a value that the calling function doesn’t currently need. Later, however, this value may be important when the code is called in a different situation. In this situation, you should obey the law of useful return: A procedure should return all the potentially useful information it computed.

• Alexander A. Stepanov

Object-oriented programming aficionados think that everything is an object.... this [isn’t] so. There are things that are objects. Things that have state and change their state are objects. And then there are things that are not objects. A binary search is not an object. It is an algorithm

• Alexander A. Stepanov

You cannot fully grasp mathematics until you understand its historical context.

• Alexander A. Stepanov and David R. MUsser – Generic Programming

By generic programming, we mean the definition of algorithms and data structures at an abstract or generic level, thereby accomplishing many related programming tasks simultaneously. The central notion is that generic algorithms, which are parameterized procedural schemata that are completely independent of the underlying data representation and are derived from concrete, efficient algorithms.

Alan Kay

• Alan Kay

Simple things should be simple, complex things should be possible.

• Alan Kay

It’s easier to invent the future than to predict it.

• Alan Kay

Normal is the greatest enemy with regard to creating the new. And the way of getting around this is you have to understand normal not as reality, but just a construct. And a way to do that, for example, is just travel to a lot of different countries and you’ll find a thousand different ways of thinking the world is real, all of which are just stories inside of people’s heads. That’s what we are too. Normal is just a construct, and to the extent that you can see normal as a construct in yourself, you have freed yourself from the constraints of thinking this is the way the world is. Because it isn’t. This is the way we are.

Edsger W. Dijkstra

• Edsger W. Dijkstra

Program testing can be used to show the presence of bugs, but never to show their absence!

• Edsger W. Dijkstra (1970) “Notes On Structured Programming” (EWD249), Section 3 (“On The Reliability of Mechanisms”), p. 6.

The art of programming is the art of organizing complexity, of mastering multitude and avoiding its bastard chaos as effectively as possible.

John Von Neumman

• John von Neumann

If you tell me precisely what it is a machine cannot do, then I can always make a machine which will do just that;

• John von Neumann - The Role of Mathematics in the Sciences and in Society (1954)

A large part of mathematics which becomes useful developed with absolutely no desire to be useful, and in a situation where nobody could possibly know in what area it would become useful; and there were no general indications that it ever would be so. By and large it is uniformly true in mathematics that there is a time lapse between a mathematical discovery and the moment when it is useful; and that this lapse of time can be anything from 30 to 100 years, in some cases even more; and that the whole system seems to function without any direction, without any reference to usefulness, and without any desire to do things which are useful.

• John von Neumann, The Computer and the Brain

Any computing machine that is to solve a complex mathematical problem must be ‘programmed’ for this task. This means that the complex operation of solving that problem must be replaced by a combination of the basic operations of the machine.

• John von Neumman

Problems are often stated in vague terms… because it is quite uncertain what the problems really are.

• John von Neumman

The sciences do not try to explain, they hardly even try to interpret, they mainly make models. By a model is meant a mathematical construct which, with the addition of certain verbal interpretations, describes observed phenomena. The justification of such a mathematical construct is solely and precisely that it is expected to work - that is correctly to describe phenomena from a reasonably wide area. Furthermore, it must satisfy certain esthetic criteria - that is, in relation to how much it describes, it must be rather simple.

• John von Neumman

The calculus was the first achievement of modern mathematics and it is difficult to overestimate its importance. I think it defines more unequivocally than anything else the inception of modern mathematics; and the system of mathematical analysis, which is its logical development, still constitutes the greatest technical advance in exact thinking.