<algorithm>: `contains_subrange` should not use `std::distance`

[hewillk]

llvm-project/libcxx/include/__algorithm/ranges_contains_subrange.h

Lines 72 to 80 in c335acc

	operator()(_Range1&& __range1, _Range2&& __range2, _Pred __pred = {}, _Proj1 __proj1 = {}, _Proj2 __proj2 = {}) {
	auto __n2 = 0;
	if constexpr (sized_range<_Range2>) {
	__n2 = ranges::size(__range2);
	} else {
	__n2 = std::distance(cbegin(__range2), cend(__range2));
	}
	if (__n2 == 0)
	return true;

...because it is based on the C++17 iterator model rather than the C++20 Ranges iterator model.
In addition, __n2 is an integer type and may not accommodate difference_type.
I believe the above can be just simplified to:

if (ranges::distance(__range2) == 0)
  return true;

testcase:

#include <ranges>
#include <algorithm>

static_assert(
  std::ranges::contains_subrange(
    std::views::iota(0, 5),
    std::views::iota(0, 5) | std::views::filter([](int i) { return true; }) )
);

https://godbolt.org/z/G6oMoM5Wz

[hewillk]

If we only want to know whether __range2 is empty, we only need to check first == last or use ranges::empty, which is what MSVC-STL does.
If __range2 is very large, the ranges::distance will only be an unnecessary bottleneck.

[frederick-vs-ja]

Currently __n2 is of type int here, which can lead to wrong results with uint64_t (Godbolt link). This was pointed out during review (#66963 (comment)) but not fully resolved.

@ZijunZhaoCCK Did you think there is some case in which ranges::empty doesn't work?

Edit: oops, perhaps we shouldn't call ranges::empty as a member empty function can return an undesired value (which doesn't seem UB or IFNDR)?

[hewillk]

Edit: oops, perhaps we shouldn't call ranges::empty as a member empty function can return an undesired value (which doesn't seem UB or IFNDR)?

That's why I wrote P3156 even though it was closed. But I have submitted a LWG for this.

[ZijunZhaoCCK]

@ZijunZhaoCCK Did you think there is some case in which ranges::empty doesn't work?
I didn't think that ranges::empty doesn't work before 😢 . Maybe it can be replaced by empty_checkable_range.

[hewillk]

But I have submitted a LWG for this.

https://cplusplus.github.io/LWG/issue4062