Use ptrdiff_t or size_t for indexing

Question

I got into the habit of declaring for loop indices with size_t instead of int . That however already bit me multiple times when I was iterating an array backwards, ie checking for the index to be greater or equal to zero:

for (size_t i = n-1; i >= 0; i--) {
    // ...
}

When the body has been run for i == 0 , it gets decremented and wraps around, to SIZE_T_MAX probably. That makes the break condition a tautology. There might be some clever bit manipulations to check for possible underflows, but wouldn't it be simpler to use ptrdiff_t instead?

What is the proper idiomatic C way to solve this? size_t plus bit twiddling or ptrdiff_t and feeling uncomfortable about semantics?

Answer 1

A backwards loop should look like this:

for (size_t i = 0; i != n; ++i) {
    arr[n - i - 1] = foo();
}

The n - i - 1 is the "reverse iterator" corresponding to i , if you will.

Answer 2

Here is a safer way to write your downward loop:

• it works with both signed or unsigned index types,
• it uses the array length unmodified just once. You could also write i = sizeof(arr) / sizeof(*arr) if arr is defined as an array.
• i takes all valid index values in decreasing order.

for (size_t i = n; i-- > 0;) {
    arr[i] = compute_value(i);
}

---

_{One could write i --> 0 as an eye catcher idiom, but some experts seem to disagree on this alternative.}