C++: How do I cast an int to an unsigned long and not change any bits?

C++: How do I cast an int to an unsigned long and not change any bits? I want to pack and unpack values into memory. The word size is 64 bits.

This snippet illustrates the problem:

int v1 = -2; // 0xfe
unsigned long v2=(unsigned long)v1; // 0xfffe, I want 0x00fe

The simple solution is:

unsigned long v2=(unsigned int)v1; // 0x00fe

However, this code is in a template where the targe开发者_如何学Pythont type is a parameter, so I had to resort to this:

uint64 target = mem[index] & mask;
uint64 v;
if (value < 0) {
    switch (bits) {
    case 8:
        v = (uint8)value;
        break;
    case 16:
        v = (uint16)value;
        break;
    case 32:
        v = (uint32)value;
        break;
    }
} else {
    v = value;
}
v = v << lShift;
target |= v;
mem[index] = target;

Assume, for example, the type for "value" is an int (16 bits) and bits=16. The goal is to mask the bits in memory for value and replace them.

Does anyone know an easier way?

Assuming you have C++0x support:

#include <type_traits>
v= static_cast<std::make_unsigned<decltype(value)>::type>(value);

I'm assuming that you are parameterizing on the type of value, otherwise this doesn't make any sense.

EDIT: making it more C++-ish by using static_cast instead of a C cast. I suppose that's what got me a downvote.

If you don't mind the typing, a trait class comes to mind:

template <typename IType> struct ToULong;

template <> struct ToULong<signed char>
{
  static inline unsigned long int get(signed char c) { return (unsigned char)(c); }
};

template <> struct ToULong<signed short int>
{
  static inline unsigned long int get(signed short int c) { return (unsigned short int)(c); }
};

/* ... signed int, signed long int, signed long long int ... */

Usage:

template <typename IType>
struct Foo
{
  unsigned lont int get_data() const { return ToULong<IType>::get(m_data); }
private:
  IType m_data;
}

Update: Even simpler, you could just make a bunch of overloads:

unsigned long int toULong(            char c) { return (unsigned      char)(c); }
unsigned long int toULong(signed      char c) { return (unsigned      char)(c); }
unsigned long int toULong(signed short int c) { return (unsigned short int)(c); }
unsigned long int toULong(signed       int c) { return (unsigned       int)(c); }
unsigned long int toULong(signed  long int c) { return (unsigned  long int)(c); }

2nd update: You should probably say static_cast<T>(x) rather than (T)(x) if you want to be even more C++-like.

How about union?

union u1 {
    short int si;
    unsigned long int uli;

    unsigned long int stub;

    operator unsigned long int () {return uli;};
public:
    u1(short int nsi) : stub(0) {si = nsi;}

};

I believe that you could use a bitwise-AND to get the desired result.

unsigned long v2 = 0;
v2 = v2 | v1;

Using the idea put forth by "Kerrek SB" I came up with a solution.

template <typename Tint> uint64 ToMemdata(Tint value) {
    return (uint64)value;};
template <> uint64 ToMemdata<int8>(int8 value) {
    return (uint64)((uint8)value);};
template <> uint64 ToMemdata<int16>(int16 value) {
    return (uint64)((uint16)value);};
template <> uint64 ToMemdata<int32>(int32 value) {
    return (uint64)((uint32)value);};
template <> uint64 ToMemdata<int64>(int64 value) {
    return (uint64)((uint64)value);};

template <typename Tint> void packedWrite(Tint value, int vectorIndex, uint64* pData) {

    uint64 v = ToMemdata(value);
    // This call eliminates a run time test for minus and a switch statement
    // Instead the compiler does it based on the template specialization

    uint64 aryix, itemofs;
    vectorArrayIndex(vectorIndex, &aryix, &itemofs); // get the memory index and the byte offset
    uint64 mask = vectorItemMask(itemofs); // get the mask for the particular byte
    uint64 aryData = pData[aryix]; // get the word in memory
    aryData &= mask; // mask it
    uint64 lShift = (uint64)(itemofs * sizeof(Tint) * 8); 
    uint64 d = v << lShift; // shift the value into the byte position
    aryData |= d; // put the value into memory
    pData[aryix] = aryData;
}

Using this concept I was able to make other improvements to the code. For example, the call to vectorItemMask() is now templateized also.

To cast without changing bits, take a reference and then dereference with appropriate type:

int v1 = -2; // 0xfe
unsigned long v2=*(unsigned long *)&v1;

This assumes the sizes are the same. It has undefined behavior if sizeof(int) != sizeof(unsigned long). You probably want unsigned int.

Edit: realized i answered wrong question.

Boost type_traits has something (i believe it is make_unsigned) to convert an int type to the unsigned version (if it is signed) and do nothing if it is unsigned.