Convert Unsigned Char * to uint16_t *

Question

I have unsigned char * data like below:

unsigned char * data = ...;
data[0]=0x24;
data[1]=0x8A;
data[2]=0x07;
data[3]=0x75;

I want to convert it to uint16_t with below code:

uint16_t *res= (uint16_t*)(data);

The output will be:

res[0]= 0x8A24; res[1]=0x7507 

But I want res become like this:

res[0]=0x248A; res[1]=0x0775;

How can I change my code?

Answer 1

If you were to use operating system's header, you have the ntohs and htons function that can swap the bytes.

Else, it's trivial to add:

template <T> T swap_endianness(T in) {
  unsigned char array[sizeof(in)] = {};
  std::memcpy(array, &in, sizeof(in));
  // Byte swap now
  for (size_t i = 0; i < sizeof(in)/2; i++) 
    std::swap(array[i], array[sizeof(in) - i - 1]);
  std::memcpy(&in, array, sizeof(in));
  return in;
}

Usage is:

uint16_t *res= (uint16_t*)(data); // This is undefined behavior. Use memcpy instead.
res[0] = swap_endianness(res[0]); // and so on

This only works for Plain Old Data type, such as uint64 , uint32 , uint16 , int64 , int32 , int16 , float , double .

As @KamilCuk pointed out, the like (uint16_t *)(data) might lead to issues if data happens not to be aligned at 16-bit boundary and will likely cause a memory access error when you'll dereference it on platfrom like ARM. The correct solution is to use memcpy like this:

uint16_t res[2];
std::memcpy(res, data, sizeof(res));
res[0] = swap_endianness(res[0]);
res[1] = swap_endianness(res[1]);

Answer 2

But I want res become like this:

How can I change my code?

It's just:

res[0] = data[0] << 8 | data[1];
res[1] = data[2] << 8 | data[3];

I want to convert it to uint16_t with below code:

 uint16_t *res= (uint16_t*)(data);

The output will be:

Such code will be invalid and dereferencing res results in undefined behavior. This is a strict alias violation and data may not be aligned to alignof(uint16_t) . If the compiler happens to generate code that doesn't result in a hard fault because of unaligned access, the result depends on the endianess of the target machine. But anyway, this is just undefined behavior.