Ten C++ Advantages over C in Embedded Systems

I have been using C++ to program microcontrollers for over 10 years now. I have found some really big advantages of C++ over C. So much so, that I claim the jump from C to C++ is almost as substantial as from assembly to C.

Here are 10 advantages of C++ over C for everyday microcontroller programming.

  1. Object-Oriented Programming (OOP)
  2. Stronger Types
  3. Template Programming
  4. RAII
  5. Stronger Types
  6. Polymorphism
  7. constexpr
  8. References
  9. namespaces
  10. Symbol Mangling

1. Object Oriented Programming

Object-oriented programming can be done in C. But it is tedious. If we look at a POSIX mutex we can see this. This pattern of passing an object pointer is extremely common (and tedious) in C.

//C
int pthread_mutex_lock(pthread_mutex_t *mutex);
int pthread_mutex_unlock(pthread_mutex_t *mutex);

//usage
pthread_mutex_t mutex0;
pthread_mutex_init(&mutex0, NULL);
//passing the argument leaves room for error and bugs
pthread_mutex_lock(&mutex0); 
// do some mutually exclusive stuff
pthread_mutex_unlock(&mutex0); 
pthread_mutex_destory(&mutex0);

In C++, the compiler will manage the data object for you. Consider this Mutex class.

class Mutex {
  Mutex(){
    pthread_mutex_init(&m_mutex);
  }

  ~Mutex(){
    pthread_mutex_destroy(&m_mutex);
  }


  int lock(){
    return pthread_mutex_lock(&m_mutex);
  }
  int unlock(){
    return pthread_mutex_unlock(&m_mutex);
  }
private:
  //the mutex is tied to the methods that use it
  //this means it can't be passed to the wrong function
  pthread_mutex_t m_mutex;
};

//usage
Mutex mutex0; //automatically initialized
mutex0.lock(); //impossible to pass the wrong value
mutex0.unlock(); //unlock before destroying
//automatically destroyed when Mutex goes out of scope

C++ is more concise and less prone to mistakes. If you multiply this simple approach by many thousands of lines of code. You get a HUGE improvement.

2. Stronger Types

When it comes to types, C is much more permissive than C++. This is great if you are lazy. But horrible if you want to eliminate bugs in your code. Where C will give you a warning (or just be silent), C++ will produce and error and force you to inspect the issue. This moves debugging from runtime (in C) to compile time in C++. The code below will compile without errors in C, but will fail compilation in C++.

int square(int * num){
  return *num * *num;
}

int main(){
  void * x = 0;
  square(x);
}

This is a subtle way to make the compiler do more of the work to ensure your program is correct.

3. Template Programming

In C++, you can use templates to have variable types for a function. Consider this digital filter in C:

//C
typedef struct {
  int average;
  int weight;
} filter_int_t;
int filter_int(filter_int_t * filter, int input){
  //int based implementation
}

typedef struct {
  float average;
  float weight;
} filter_float_t;
int filter_float(filter_float_t * filter, int input){
  //float based implementation
}

//and so on

C++ is more concise, the exact some code can be used for a variety of types. The compiler then handles generating the function for different types.

template<typename NumberType> class Filter {
public:
  NumberType filter(NumberType input){
    //implement here
  }

private:
  NumberType m_average;
  NumberType m_weight;
};

Filter<float> f_filter;
Filter<int> i_filter;
Filter<long int> li_filter;
Filter<double> d_filter;

4. RAII

This is one of my favorites parts of using C++ over C. The terribly named “Resource Allocation is Initialization” means that every class has a constructor that allocates resources and a destructor that cleans them up. Both are called automatically by the compiler. The constructor is called when the class variable is declared, and the destructor is called when it goes out of scope.

Here is an example in C (using the mutex again).

int make_magic_happen(int value){
  pthread_mutex_lock(&mutex0); 

  if( value < 0 ){
    //for magic, we return if value < 0
    //don't forget the mutex
    pthread_mutex_unlock(&mutex0); 
    return -1;
  }

  if( value == 10 || value == 100 ){

    pthread_mutex_unlock(&mutex0); 
    return 1;
  }
  //now really make some magic

  //don't forget the mutex!
  pthread_mutex_unlock(&mutex0); 
  return 0;
}

In C, you always need to unlock the mutex manually. So if you have multiple exit points (in this simple case), it can be a pain. This C++ class enables auto lock/unlock using RAII.

class MutexScope {
public:
  MutexScope(Mutex & mutex) : m_mutex(mutex){
    m_mutex.lock();
  }
  ~MutexScope(){
    m_mutex.unlock();
  }
private:
  Mutex & m_mutex;
}

int make_magic_happen(int value){
  //mutex is locked here and unlock automatically when the function returns
  MutexScope ms(mutex); 

  if( value < 0 ){
    //for magic, we return if value < 0
    return -1;
  }

  if( value == 10 || value == 100 ){
    return 1;
  }

  //now really make some magic
  return 0;
}

I have a whole list of examples where I have used RAII in embedded applications.

5. enum class

The enum class in C++ is a strong type that the compiler enforces. It is stronger than an enum in C++ which is stronger than an enum in C.

//C
enum value {
  //these have global scope so they need the prefix
  value_one,
  value_two,
  value_three
};

//in C this will compile (and is terrible)
int x = value_one;
enum value v = x + 5;

//C++
enum class Value {
  //scope to Value so no prefix
  one, two, three
};


//in C++ these won't compile
int x = Value::one;
Value value = x + 5;

If you have a variable with a small number of possible values, the enum class is a compiler enforced way to make sure variables of that type are only assigned allowed values. The enum class also prohibits operators (unless you create some).

6. Polymorphism

Unlike RAII, polymorphism is an awesome name. It literally means multiple forms and is used frequently in genetics to refer to genes that can manifest in different forms. In C++, it means code can take multiple forms based on the context. For example, == can mean different things in C++ depending on what is being compared. Strings are a really good example of polymorphism.

int x = 0;
int y = 0;
const char * hello = "hello";
const char * world = "world";

//in C == always means a numerical comparison
if( x == y ){}

//To compare strings we convert the compare to a number
if( strcmp(hello, world) == 0 ){
  //to make is worse, you shoudl use strncmp()
}

//C++ comparing == is polymorphic (is takes a different form in diffent places)
if( x == y ){
  // == means compare numbers
}

if( std::string(hello) == std::string(world) ){
  // == means compare strings
}

Changing what operators (like ==, +, -, etc) do is also referred to as overloading an operator. This is just one type of polymorphism in C++. When you re-implement a virtual method, that is another way C++ uses polymorphism. Polymorphism goes a long way to make code more concise and readable. It also means less room for mistakes (like not using strncmp()).

7. constexpr

In C, the use of pre-processor macros is abundant. And for C, it is generally good practice. However, it can be a big pain to debug and causes scoping problems.

C++ has a constexpr keyword that allows you to tell the compiler that the value must be evaluated at compile time.


//C
#define X_VALUE 10

//C++
static constexpr int x_value = 10;

The advantages:

  1. x_value has a type, so the compiler can enforce type compatibility. X_VALUE has no type.
  2. x_value can be scoped as narrowly as needed (inside a function, class namespace or global). X_VALUE cannot be scoped.
  3. x_value can be used in a more complex constexpr evalution where the compiler can give specific errors. Using X_VALUE in a complex macro can result in difficult-to-debug compiler errors.

8. References

In C, you can pass by value or pass a pointer (by value). C++ has references which are basically more strict pointers. The big difference is that pointers can be null and references cannot. So the compiler can do a much better job stopping the build if you have not referred to a value correctly. In C, you need to catch null or other bad pointers during runtime.

typedef struct {
  int x[3];
  int y[3];
  int z[3];
} my_data_t;  

my_data_t my_data;

//C pointer passing
void process(my_data_t * data);
process(&my_data); //passed as a pointer

//c++ refererence passing
void process(my_data_t & data);
process(my_data); //passed as if it was a pointer

9. Symbol Scoping

In C, you end up doing a lot of manually prefixing. I usually do something like this for example, logic.h.

//C
int logic_initialize(void * logic_object);
int logic_finalize(void * logic_object);
int logic_read(void * logic_object, void * buffer, int nbyte);
int logic_write(void * logic_object, const void * buffer, int nbyte);

//C++
class Logic {
public:
  Logic(){
    //initialize here
  }
  ~Logic(){
    //finalize here
  }

  //these names are scoped to the Logic class
  //so they can be re-used without prefixing in other classes
  int read(void * buffer, int byte);
  int write(const void * buffer, int nbyte);
private:
  void * m_object;

  //only this class can access this method
  int read_internal(void * buffer, int nbyte);

}

You will notice the use of public and private keywords allow you to control which variables and methods are accessible outside the scope of Logic. This type of selective access is much easier in C++ than C (you have to use static file scoping).

Namespaces are a way of creating a scope without creating a class. These are great for large projects or frameworks to ensure there are no naming collisions.

10. Symbol Mangling

In C, object level symbols are generated by function name, but C++ accounts for both the name of the function and its arguments. It creats a mangled symbol that is unique. This is fantastic for designing APIs. Consider these examples:

//C
int write(const void * buffer, int nbyte);
int write_location(int loc, const void * buffer, int nbyte);

//C++
int write(const void * buffer, int nbyte);
int write(int loc, const void * buffer, int nbyte);