Question

class CheckPointer {
public:
    CheckPointer(int * mbeg, int * mend) :
        beg(mbeg), end(mend), curr(mbeg) {}

    // subscript operator
    int & operator[] (const size_t pos) {
        if (beg + pos < beg) {
            throw out_of_range("ERR: before beg!");
        }
        if (beg + pos >= end)
            throw out_of_range("ERR: end or past end!");
        return *(const_cast<int *>(beg + pos));
    }
private:
    const int * beg;
    const int * end;
    int * curr;
};

I have defined a subscript operator for the class CheckPointer. As the @param pos is of type size_t, I cannot check whether the user has passed a positive or negative value. Howerver I try to write the code to do the bound-check instead and it works :

        if (beg + pos < beg) {
            throw out_of_range("ERR: before beg!");
        }

I don’t know why it works… Could any one help me?

Thank you for considering my question!

---

For more information:

Environment: eclipse CDT, Ubuntu 10.04

test-code:

int iarr[6] = {1, 2, 3, 4, 5, 6};
CheckPointer cp(iarr, iarr+6);
// subscript
cout << cp[2] << endl;
cout << cp[5] << endl;
cout << cp[-2] << endl; // error: before beg

test-code_output:

terminate called after throwing an instance of 'std::out_of_range'
  what():  ERR: before beg!
3
6

Answer 1

This works because the negative value is cast to an unsigned value. Your machine architecture probably uses two’s-complement for negative values, so a value of -2 results in:

0xfffffffe (assuming 32-bit)

When you add this to beg, it wraps around and provided beg is greater than or equal to 2, has the effect of subtracting 2 from it. Thus, despite the index being extremely large, it has essentially performed a subtraction. If the value of beg was 1 or 0, the error would be “past end” instead.