4

How do you add extra types to c11 _Generic Functions?

Do you have to #undef/re-#define it?(if so would the following work) or is there a nicer way?

#define to_str(X) _Generic((X), \
    long double: ld_str, \
    double: d_str, \
    float: f_str, \
    )(X)

#undef to_str

#define to_str(X) _Generic((X), \
    long double: ld_str, \
    double: d_str, \
    float: f_str, \
    int: i_str, \
    )(X)
Roman A. Taycher
  • 18,619
  • 19
  • 86
  • 141

2 Answers2

8

I am not sure that I understand your question completely. You mean that you have a type generic macro that is given by some library and you want to amend it with a new type of your own?

What you always could do is to give it another name and use the default case to obtain the provided behavior:

#define to_str2(X) _Generic((X), default: to_str(X), int: i_str(X))

Edit:

This will not work perfectly because you'd have to put the function argument evaluation inside the _Generic. This means in particular that the type of X has to be compatible with all branches of the nested generic expressions.

It would be easier if the library in question had a macro that would just return the function itself, without the (X), say to_strGen, and that never would evaluate X. Then you could do

#define to_str2Gen(X) _Generic((X), default: to_strGen(X), int: i_str)
#define to_str2(X) to_str2Gen(X)(X)
Jens Gustedt
  • 76,821
  • 6
  • 102
  • 177
  • Caution: if the original macro doesn’t define a `default` selection, then wrapping won’t work. See the thread [Nested _Generic selections](https://groups.google.com/d/topic/comp.lang.c/1RQvT2a5DF0/discussion) on comp.lang.c. – J. C. Salomon Apr 17 '12 at 21:23
  • @J.C.Salomon, the lack of a `default` isn't the problem, I think, but the compability of `X` with all branches of the type generic expression is, I think. See my edit. – Jens Gustedt Apr 18 '12 at 07:45
3

If it's your code, you would have to #undef it and re#define it, yes. There's no way to extend a type-generic expression (AFAIK).

If it's not your code I'd introduce a second expression with the extension like Jens suggested.

hroptatyr
  • 4,702
  • 1
  • 35
  • 38