This and the SpaceState::Invisible branch now continue before the if styles != current && !buf.is_empty() { business as opposed to continuing afterwards previously. I'm not sure whether that could have any unintended consequences, but might be worth double checking if it's not intentional.

This change also gave me a lot of grief reasoning through it, but I think it's correct.

First, the Invisible case already continued at the beginning of the loop, so nothing's changed there. I'm not certain, but I believe tag elements can never have different styles than the element they precede, so this should never trigger the if-statement on its own. Another non-invisible element with different styles would need to be present, and that element's presence would trigger the if-statement itself.

Second, spaces will only continue if they're being collapsed or discarded. If it's being discarded, that space shouldn't affect regex searching, and if it's being collapsed, its styles don't matter because we keep the styles of the first space in a run.

Also, I'm quite unsure, but I believe that interior spaces with different styles than adjacent text aren't ever included in a group passed to find_regex_match_in_elems because of the general text grouping rules, but I don't have a strong-enough understanding of those rules to be certain. This is part of what the show-text-styled-space test is for.

I see. The behavior tested by the two new tests is indeed confusing and I would consider it a bug.