About D1, you can reconcile the idea and the code like this: the real constraint isn’t simply “you can’t insert in the middle,” but that any insertion must not break an already satisfied prefix w condition (mex = k). The loop from i = 0 to n − 2 is effectively counting how many insertion positions are safe before the first position where inserting would destroy the existing prefix property. When m = 1, a full segment with mex = k must remain intact, so only the two ends work; when m = 0, as long as the prefix hasn’t reached mex = k yet, inserting won’t affect it, giving i valid positions. From this perspective, the code is implicitly checking the previous w-state rather than explicitly reasoning about “middle vs ends,” so the implementation and the intended logic do line up—just from different angles.

it is crazy for me

难:)