It's a unary operator, kind of like ~ or !, which can apply to parenthesis (), until it meets a type, then the parenthesis are mandatory:
int x;
sizeof x; // okay
sizeof (x); // okay
sizeof int; // not okay
sizeof (int); // okay
It's a weird double rule to parse, but even worse, the strongest binding precedence typically doesn't like an alpha identifier at that point in the recursive descent, so you get weird rules like:
read_prefix():
is peak() '!': ...
is peak() '~": ...
is alpha(peak())
maybe alpha is sizeof?
wait, it wasn't sizeof? it was an identifier? long rewind
is peak() '+': ...
Something like @ would've been a good pick for sizeof.
\ramble on
(actually, better yet, it should've been used instead of & for taking the address). Imagine:
> (everything here also applies to c2y's newly introduced _Countof)
Stop the press, they're adding what? So I don't have to keep adding #define count_of(arr) (sizeof(arr) / sizeof(*arr)) to every codebase and living with its failure modes?
I'm really loving C's character arc at the moment, where they keep going back to fix old problems instead of piling everything and the kitchen sink into the language.
> I'm really loving C's character arc at the moment, where they keep going back to fix old problems
Valid, but it does seem a bit Stockholm Syndrome to have this much rejoicing that your language added a builtin for getting the length of an array in 2026.
Yeah, despite puritanical hand-wringing over C23, I do like both the discipline and the selectivity with WG14 has shows (and I say this as a card-carrying C++ hater).
I do wish they will start looking into proper arrays since there is already a syntax for this can can be augmented semantically:
That was discussed, but consistency with sizeof was considered more important. And if you figured out how to parse sizeof, parsing countof is not an additional burden.
It's a unary operator, kind of like ~ or !, which can apply to parenthesis (), until it meets a type, then the parenthesis are mandatory:
It's a weird double rule to parse, but even worse, the strongest binding precedence typically doesn't like an alpha identifier at that point in the recursive descent, so you get weird rules like: Something like @ would've been a good pick for sizeof.\ramble on
(actually, better yet, it should've been used instead of & for taking the address). Imagine:
\ramble off> (everything here also applies to c2y's newly introduced _Countof)
Stop the press, they're adding what? So I don't have to keep adding #define count_of(arr) (sizeof(arr) / sizeof(*arr)) to every codebase and living with its failure modes?
I'm really loving C's character arc at the moment, where they keep going back to fix old problems instead of piling everything and the kitchen sink into the language.
> I'm really loving C's character arc at the moment, where they keep going back to fix old problems
Valid, but it does seem a bit Stockholm Syndrome to have this much rejoicing that your language added a builtin for getting the length of an array in 2026.
Yeah, despite puritanical hand-wringing over C23, I do like both the discipline and the selectivity with WG14 has shows (and I say this as a card-carrying C++ hater).
I do wish they will start looking into proper arrays since there is already a syntax for this can can be augmented semantically:
Can you explain what you mean exactly regarding this augmenting this syntax?
I guess they are saying sizeof(...) should return the annotated static size rather than sizeof(int*) in this case
I feel like what you struggle with is expression parsing, not sizeof itself?
Wait, they added a new operator and didn't fix that parsing stupidity by always requiring parentheses?
I assume that there would be a zillion of legacy C source files that would be broken by such a change.
There are no legacy C source files using _Countof.
That was discussed, but consistency with sizeof was considered more important. And if you figured out how to parse sizeof, parsing countof is not an additional burden.