Probably the most common distinct aggregate is COUNT(DISTINCT). In some respects, it’s similar to the windowed function DENSE_RANK(). DENSE_RANK(), as opposed to ROW_NUMBER(), will only increment the row counter when the ordering column(s) actually change from one row to the next, meaning that we can use DENSE_RANK() as a form of windowed distinct count
This is a very interesting approach to the problem. Read the whole thing.