C# language official

size_t is unsigned, so this loop will never terminate. Here's my question:

(side note but the fact C can't abstract those concerns into a single construct and the for-loop expects you to spell them out every time is embarrassing)

2 and 3 use constructs that are not very common. In 2 ssize_t is technically not standard C (it is POSIX, however). 1 uses ptrdiff_t which is not intended for indexing.

Because of lack of this clean separation, if I were using options 4 or 5, I would be using them with the "while" loop instead.

Possible alternative to consider:

// everybody is familiar with this form // for n in 0, ... SIZE-1 for (size_t n = 0; n < size; n++) { // map the 0, 1, 2, ..., SIZE-1 range into SIZE-1, ..., 2, 1, 0 range size_t i = size - n - 1; printf("%" PRIu8 "\n", some_array[i]); }

The expressed for-loop invariant prevents n from being ever being anything else other a value in range [0; SIZE).

Therefore the correctness of the mapping expression can be verified against corner cases of n == 0 and n == size-1.

to keep all our loop parameters in one place. On godbolt, with gcc 16.1 -O2, this ends up being equivalent to #4 and #5. (Though, weirdly, with the operands of the cmp instruction the other way around.)

I'll also note that #3 could be rewritten like this:

If you prefer that separation of concerns.

One correction: ptrdiff_t absolutely is intended for indexing, at least according to POSIX:

I'd say as far as POSIX is concerned, its assertion that ptrdiff_t is for indexing is correct. And I'd also say that the reason why C itself seems to be hesitant in doing the same is

potential theoretical concern of say, allocating a 3GB char array on a 32-bit machine with 4GB of RAM (32-bit ptrdiff_t finds itself unable to index beyond 2147483648, 64-bit ptrdiff_t would be larger than the pointer)

segmented memory model with 16-bit size_t and 32-bit ptrdiff_t for far pointers maybe? also the same concern as above except hitting smaller sizes for arrays.

embedded platforms doing weird shit?

size_t is unsigned, so this loop will never terminate. Here's my question:

(side note but the fact C can't abstract those concerns into a single construct and the for-loop expects you to spell them out every time is embarrassing)

2 and 3 use constructs that are not very common. In 2 ssize_t is technically not standard C (it is POSIX, however). 1 uses ptrdiff_t which is not intended for indexing.

Because of lack of this clean separation, if I were using options 4 or 5, I would be using them with the "while" loop instead.

Possible alternative to consider:

// everybody is familiar with this form // for n in 0, ... SIZE-1 for (size_t n = 0; n < size; n++) { // map the 0, 1, 2, ..., SIZE-1 range into SIZE-1, ..., 2, 1, 0 range size_t i = size - n - 1; printf("%" PRIu8 "\n", some_array[i]); }

The expressed for-loop invariant prevents n from being ever being anything else other a value in range [0; SIZE).

Therefore the correctness of the mapping expression can be verified against corner cases of n == 0 and n == size-1.