Casting one C structure into another

Question

I have two identical (but differently named) C structures:

typedef struct {
      double x;
      double y;
      double z;
} CMAcceleration;


typedef struct {
    double x;
    double y;
    double z;   
} Vector3d;

Now I want to assign a CMAcceleration variable to a Vector3d variable (copying the whole struct). How can I do this?

I tried the following but get these compiler errors:

vector = acceleration;           // "incompatible type"
vector = (Vector3d)acceleration; // "conversion to non-scalar type requested"

Of course I can resort to set all members individually:

vector.x = acceleration.x;
vector.y = acceleration.y;
vector.z = acceleration.z;

but that seems rather inconvenient.

What's the best solution?

Answer 1

That's your only solution (apart from wrapping it into a function):

vector.x = acceleration.x;
vector.y = acceleration.y;
vector.z = acceleration.z;

You could actually cast it, like this (using pointers)

Vector3d *vector = (Vector3d*) &acceleration;

but this is not in the specs and therefore the behaviour depends on the compiler, runtime and the big green space monster.

Answer 2

You could use a pointer to do the typecast;

vector = *((Vector3d *) &acceleration);

Answer 3

memcpy(&vector, &acceleration, sizeof(Vector3d));

Please note that this works only, if the physical layout of the structs in memory are identical. However, as @Oli pointed out, the compiler is not obliged to ensure this!

Answer 4

You use an utility function for that:

void AccelerationToVector( struct CMAcceleration* from, struct Vector3d* to )
{
     to->x = from->x;
     to->y = from->y;
     to->z = from->z;
}

Answer 5

Another version of the utility function making use of C99:

static inline struct Vector3d AccelerationToVector(struct CMAcceleration In)
{
    return (struct Vector3d){In.x, In.y, In.z};
}

With the compiler optimization turned up (e.g., -Os), this should turn into absolutely no object code when invoked.

Answer 6

Why dont you use.

typedef CMAcceleration Vector3d;

(instead of creating a whole new structure)

in that case vector = acceleration; compiles just fine.

Answer 7

This is achieved easily through a union:

typedef struct {
      double x;
      double y;
      double z;
} CMAcceleration;

typedef struct {
    double x;
    double y;
    double z;
} Vector3d;

typedef union {
    CMAcceleration acceleration;
    Vector3d vector;
} view;

int main() {
    view v = (view) (Vector3d) {1.0, 2.0, 3.0};
    CMAcceleration accel = v.acceleration;
    printf("proof: %g %g %g\n", accel.x, accel.y, accel.z);
}

Answer 8

A safe (albeit somewhat convoluted) way to do it would be to use a union:

union { CMAcceleration a, Vector3d v } tmp = { .a = acceleration };
vector = tmp.v;

Values are reinterpreted (since C99) when the accessed member is not the last set one. In this case, we set the acceleration and then we access the vector, so the acceleration is reinterpreted.

This is the way the NSRectToCGRect function is implemented, for example.

Answer 9

You could create a union with pointers that way you avoid copying data.

example :

struct Ocp1SyncHeader
{
    aes70::OCP1::OcaUint8 syncVal;          
    aes70::OCP1::Ocp1Header header;
};

struct convertToOcp1SyncHeader
{
    union
    {
        Ocp1SyncHeader* data;
        const uint8_t* src;
    };
    convertToOcp1SyncHeader& operator=(const uint8_t* rvalue) 
    { 
      src = (const uint8_t*)rvalue; 
      return *this; 
    }
};

** access like this:

convertToOcp1SyncHeader RecvData;
RecvData = src;  // src is your block of data that you want to access

** access members like this :

RecvData.data.header