SQL onderscheidt meerdere kolommen

greatest-n-per-group vragen worden meestal opgelost met behulp van vensterfuncties:

select *
from (
   select book_id, author_id, mark, year, 
          row_number() over (partition by author_id order by case mark when 'GREAT' then 1 when 'MEDIUM' then 2 else 3 end) as rn
   from books
) t
where rn = 1;

Het bovenstaande is standaard ANSI SQL, maar in Postgres met de (eigen) distinct on is meestal veel sneller:

select distinct on (author_id) book_id, author_id, mark, year, 
from books
order by author_id, 
         case mark when 'GREAT' then 1 when 'MEDIUM' then 2 else 3 end