Converting signed to unsigned

Is there a way to store the absolute value of a signed value into an unsigned type of the same size? I don't want to use a larger type, because this is a template function that will need to work with all integral types (including the largest type).

For example, this doesn't work because 128 would overflow int8_t. It would work if we allow automatic promotions, but that's not an option for larger types.

Code:

---

int8_t x = -128;
int8_t neg_x = -x;
uint8_t abs_x = neg_x;

---

Thanks

Ah actually disregard. static_cast<> would do it. I thought it was technically undefined but it isn't!

Quote:

---

Originally Posted by cyberfish

Is there a way to store the absolute value of a signed value into an unsigned type of the same size?...
For example, this doesn't work because 128 would overflow int8_t...

---

Yep, use stdlib.h abs() function:

Code:

---

#include <stdlib.h>
#include <stdio.h>
#include <stdint.h>

int main( void )
{
  int8_t x = -128;
  uint8_t y;

  // since uint8_t has more precision (8 bits) than int8_t (7 bits), this is safe:
  y = abs(x);

  printf( "abs(%hhd) == %hhu\n", x, y );

  // this will overflow!
  x = abs(y);

  printf( "abs(%hhu) == %hhd\n", y, x );
}

---

Quote:

---

$ cc -o test test.c
$ ./test
abs(-128) == 128
abs(128) == -128

---

If your type has precision bigger than 31 bits (int) use labs() (long) or llabs() (long long).

This works because of the same thing I just realized in my own reply above - the standard guarantees that casting a signed type to unsigned preserves the bit pattern in 2's complement (and architectures not using 2's complement also have to emulate this behaviour).

So

Code:

---

int8_t x = -128;
x *= -1;
uint8_t y = static_cast<uint8_t>(x);

---

works (in my case I know the number must be negative).