Re: naming results in C?

[Patrik Hagglund <patha at ida dot liu dot se>]

> In C++, copying into the result often has significant costs
> (copying constructor), which is reduced with that extension. For
> C, no such need exists.

Yes, that's what the documentation tells you. :)

I thought of using it for returning arrays. Take the following
example program in C, where mult_v multiplies a vector with a
scalar (using variable length arrays):

/* printf */
#include <stdio.h>

void
print_v(int n, double v[]) {

   int i;
   for (i = 0; i < n; i++) 
      printf("%g ", v[i]);
   printf("\n");
}   

void
init_v(int n, double v[]) {

   int i;
   for (i = 0; i < n; i++) 
      v[i] = i;

}

/* multipy vector v with float f */
void
mult_v(double f, int n, double v[], double res[]) {

   int i;
   for (i = 0; i < n; i++)
      res[i] = f * v[i];
}

/* atoi */
#include <stdlib.h>

int
main(int argc, char **argv) {

   int n = atoi(argv[1]);
   double v1[n], v2[n], v3[n];

   init_v(n, v1);
   print_v(n, v1);

   mult_v(2.0, n, v1, v2);
   mult_v(3.0, n, v2, v3);

   print_v(n, v3);

   return 0;
}

The array v1 is multiplied two times and the final result is
stored in v3.

Now, if I was able to declare mult_v like (this is impossible with
the current C++ extension):

  mult_v(double f, int n, double v[]) return double res[n] 

then it would be possible to use mult_v like:

  double *v3;
  v3 = mult_v(3.0, n, mult_v(2.0, n, v1));


There have to be many situations in C (not just C++) where you
want to contruct operator-like functions (operating on objects
with non-trivial size) instead of passing a reference to the
result variable.

BTW, I think that using a result variable is more intuitive than
using a return statement (it is more symmetric).

Regards,
Patrik Hägglund