Question

Which would be the preferred way of passing information back, created in a function?

A)

create_b_from_a(... a, struct b *my_b) { ... }

int main() {
  struct b my_b;
  ... define and create a ...
  create_b_from_a(a, &my_b);
  ... work on b ...
}

B)

struct st *create_b_from_a(... a)
{
  struct st *my_b = malloc(sizeof *my_b);
  ...
}

int main()
{
  struct b *my_b = create_b_from_a(a);
  ... work on b ...
  free(b);
}

I think I remember something like ‘avoid malloc when possible’ but on the other hand B looks more modern to me. Maybe there are pros/cons to one of the approaches that I am not aware of?

Answer 1

Don’t forget that structs are values, so you are not restricted to those two. You can also do:

struct b create_b_from_a(struct a an_a)
{
  struct b new_b;

  new_b.field = value; /* and so on ... */

  return new_b;
}

This can also lead to cleaner code, since it’s more straight-forward, not spending so much code on (micro) management.

And of course, if the compiler decides that on the target architecture actually passing a structure on the stack is horribly in-efficient, there are optimizations that target this pattern.