How to handle exceptions when a number might overflow the range of specific data type in C++?

You can use numeric_limits to do some checks before doing the arithmetic operation.

For instance a function that adds two int. You could do something similar to this:

int AddInt(int a, int b)
{
    if (a>0 && b>0)
    {
        // May overflow
        if (a > std::numeric_limits<int>::max() - b)
        {
            // Will overflow
            throw ....something....
        }
    }
    else if (a<0 && b<0)
    {
        // May overflow
        if (a < std::numeric_limits<int>::min() + b)
        {
            // Will overflow
            throw ....something....
        }
    }

    // We are good - no overflow
    return a+b;
}

SafeInt library does something like that. It provides template classes that act like regular integers but have checks on all operations for overflow etc. You can read an article on codeguru: Improve Microsoft Visual C++ Application Security and Robustness with SafeInt. It's maintained by microsoft, but it's not windows-only and should most likely be portable.

With SafeInt basically you write regular code and replace int with safeint and all mathematical operations will be automatically checked for overflows. Not sure if there are specializations for doubles though. Perhaps you may take similar idea and write your own wrappers to suite your needs, or you may simply use safeint with 128-bit integers and allocate 64 bits for fractional part and 63 for integer part and have very precise calculations that are always checked.

Beside these answers, you may be interesting in validating the input itself.

In this case you can:

• read the input as string
• validate it
  (make sure that only numerical characters included like '0'-'9' , '+' , '-', and the decimal-point)
• check it against the limits.
• if all of the above succeed, convert the string into the numerical value.

These requires a lot of functions and logic to accomplish. I think it is not a trivial task.

Alternatively, you can use <sstream> header functions to do it for you.

Here is an example code:

#include <iostream>
#include <sstream>

using namespace std;

template <typename T>
bool is_valid_numerical_input(T& num)
{
    string str_num;

    cout << "\n Enter a number: ";

    cin >> str_num;

    stringstream sn(str_num);

    if (!(sn >> num))
        return false;

    return true;
}

int main()
{

    short number;

    if (is_valid_numerical_input(number))
        cout << "\n Your number is: " << number;
    else
        cout << "\n Invalid input";

    cout << "\n\n\n";
}