I think this trick comes from Ben Klemens' "21st Century C".

If you like this, the book has some other tricks in it you might also like. (One example: you can define generic functions in C11).

> I haven't seen anything like this before in C or C++, and now I would wish for it to become standard.

I've seen people try to simulate this in C++ with tags.

Yes, this exactly - as soon as I saw that, I was like “how did he pull that off?”.

Of course, the first comment on HN has the answer.

I am going to try this trick out today.

The exact line is here:

https://github.com/boazsegev/facil.io/blob/master/lib/facil/...

I've seen other hacks to do this, but this is the cleanest that I've seen. He uses the same name for the macro and the function.

The only thing I would change is the way the struct is passed. I would pass by reference.

I really like this.

> I would pass by reference.

Interesting. I expected taking the address of a struct literal to be illegal, but GCC and Clang both accept it. It does make sense, I guess.

But may I ask why you would pass it by reference? With old ABIs, the struct would be passed on the stack (i.e., by reference) anyway, and with more recent ones, a small prefix would be in registers but the rest passed by reference as well. So why bother manually cluttering the code?

That's really cool, but don't unassigned fields have undefined values? How would the function know what was set? I imagine that in practice they're almost always 0, but that's not guaranteed, right? You'd need to set the struct as {0}, to have everything be 0. Is this a feature of a new(ish) standard?

It is guaranteed per C99. It's a compound initializer.

Also if you set a struct with {1} the first field is set to 1 and the rest to 0. AFAIR it was always like that (at least since standardized C, but probably earlier).

C99 specifies that struct literals with absent fields are assigned a zero value. (Notably unspecified is the value of any implicit padding bytes in the struct.)