Ranges Algorithms Modernization - Return

[mitaclaw]

In C++20, algorithms from the standard library evolved to feature new niceties and additional compile-time checks. A few of these updated algorithm functions (found in the std::ranges namespace) have new return values, and thus code beyond the expression being modernized must also be updated. I have isolated these changes so that they may be recognized and more easily understood.

[mitaclaw]

This PR will remain a draft until I am confident there are no mistakes in my changes.

Special shout-out to the PR that inspired me to attempt this modernization: #12992

[BhaaLseN]

Those std::erase changes seem a lot more verbose, is that really worth it?

[mitaclaw]

It's definitely subjective. The impression that the standard library's algorithm design gives me is that the returned subrange is the intended way to obtain the first and last iterator for erasing. It was a little unfortunate that this required a new variable in three places, but I preferred the newer design for being straightforward. I think it's worth it aesthetically and for consistency in usage, but if it's an eyesore to everyone else I can revert to obtaining the end iterator from the original container (thus making these three examples a one-liner again).

Also, I failed to notice this before (I guess I was a little sleep deprived?), but I actually fixed UB in OpenModeToAndroid with my changes. Previously, any openmode string that didn't have a 'b' character in it would erase the end iterator, which is no good.

[mitaclaw]

Rebase with no changes; the buildbots only failed the first time because they timed out.

[sepalani]

If you prefer, you can also use the PairLike operator for the unique results instead of calling begin() and end():

  const auto [begin, end] = std::ranges::unique(result);
  result.erase(begin, end);

[sepalani]

Wouldn't C++20's std::erase be a better fit here?

[BhaaLseN]

I don't have any strong feelings either way, it's just that most of the time, longer code isn't always the better code; idiomatic or not.
That being said, don't let me block you from this.

However, I was curious and tried sepalani's suggestion with std::erase. It works for the string, but I couldn't find a way to make the two vectors work with std::ranges::unique while keeping it fairly concise.