Thread: Retypecasting in C

Originally Posted by gunner4life

It is only known at run time. So to implement at element level it would have to be something like this??

Yes, I think so. I don't see what would be wrong with that.

I can't promise you anything tho. It is hard to tell why you want to do this; maybe you should explain further.

For example, what determines whether it is one way or another at runtime? Can it change during runtime? If not, you could define a custom type using if/else

Code:

if (condition)
   typedef  myuint uint16;
else 
   typedef  myuint unint32;

Then just use (myuint*) in your funcition. You might even be able to redefine a typedef during runtime -- I have never tried.

Originally Posted by gunner4life

p++ will give different results for different type of array. I want p++ to automatically increment by element and not by a preset no. of bytes calculated with conditions.

Moot point. If the pointer is of a particular type, the compiler increments "by element". If you do it by hand, you have to compute the offsets. The only difference is whether the compiler does it or you do it. I wouldn't assume a performance advantage either way, without evidence through testing.

Originally Posted by gunner4life

Basically its ok if while accessing its a pointer/array with another name but as an alias to the original array. I hope its not the same thing as accessing through the same name.

OK; then do this;

Code:

uint8 data[2048];

uint16 *data16 = (uint16 *)(&data[0]);

data16[42] = 100;  /*  will treat data[84] and data[85] as if they are collectively a uint16 */

It is not possible to add an offset to a void pointer: adding an offset to a pointer means knowing the size of whatever is pointed at and a void pointer, by definition, points at something of unknown size.

Originally Posted by MK27

For example, what determines whether it is one way or another at runtime? Can it change during runtime? If not, you could define a custom type using if/else

Code:

if (condition)
   typedef  myuint uint16;
else 
   typedef  myuint unint32;

Then just use (myuint*) in your funcition. You might even be able to redefine a typedef during runtime -- I have never tried.

This is not possible. Typedefs only exist at compile time, not at run time.

Originally Posted by grumpy

This is not possible. Typedefs only exist at compile time, not at run time.

Ah...could be done with an #ifdef I think then, depending on the requirements. The OP does say runtime but possibly using two compiled version may work?

#ifdef is also compile-time.
Ifdef with typedef would work as a pair, but only if the conditions were known at compile time.

Otherwise, the best and perhaps only solution I can think of is to manually do the offset calculation.

I just have a correction to my first post:

Code:

uint8 data[2048];
uint16 *two = (uint16 *)&data;
two[42] = 2;

The second line there shouldn't have had an ampersand:

Code:

uint16 *two = (uint16 *)data;

But enough subsequent posters have said that that it shouldn't be hard to figure out.

Code:

if (within limits)
    val = *((uint16 *)void_ptr+offset);
else
   val = *((uint32 *)void_ptr+offset);

Yeah, that's pretty much how you have to do it. If you like, you can implement some wrapper functions to make the job easier.

Code:

/** Sets data[element] = value, where the size of each element in @a data is given by
    @a size .
*/
void set(char *data, size_t element, uint32 value, enum element_size_t size) {
    switch(size) {
    case BITS_16:
        ((uint16 *)data)[element] = (uint16)value;
        break;
    case BITS_32:
        ((uint32 *)data)[element] = value;
        break;
    default:
        /* report error */
        break;
    }
}

Really, it's not so bad once you add another layer of abstraction.

Sounds like the need to do these weird casts indicates a poor design in general.