Question

I’m solving http://www.db-class.com exercises. Even though it’s late the questions are still interesting. I stumbled upon last task in extras and cannot figure out the solution.

The SQL scheme is as follows:

create table Movie(mID int, title text, year int, director text);
create table Reviewer(rID int, name text); 
create table Rating(rID int, mID int, stars int, ratingDate date);

The whole database can be obtainted here

The question is the following:

Q12 For each director, return the director’s name together with the
title(s) of the movie(s) they directed that received the highest
rating among all of their movies, and the value of that rating.
Ignore movies whose director is NULL.

To give more specific details what’s the problem here:

One query

select m.mid, m.title, max(r.stars) as stars 
from rating r natural join movie m 
where m.director is NOT NULL group by m.mid

returns IDs of most rated movies:

101 Gone with the Wind       4
103 The Sound of Music       3
104 E.T.                     3
107 Avatar                   5
108 Raiders of the Lost Ark  4

Another query

select m.director, max(r.stars) as stars 
from rating r natural join movie m 
where m.director is NOT NULL group by m.director 

returns names of the directors with most rated movies for them:

James Cameron       5
Robert Wise         3
Steven Spielberg    4
Victor Fleming      4

How to join the queries and get the name and stars for most rated movie for the director:

James Cameron      Avatar                   5
Robert Wise        The Sound of Music       3
Steven Spielberg   Raiders of the Lost Ark  4
Victor Fleming     Gone with the Wind       4

Answer 1

In my opinion this is just a gretest-n-per-group problem. There is even a tag for that 🙂

The only difference with the typical situation is that the grouping data (director) is in one table and the data used to make the comparison is in another table (stars). So, if you look in the question tagged as such you should find similar examples.

I’d recommend you to solve it first assuming all your data is in one table (awful, but easier to solve… at least for me). And then switch to the splitted data.

Spoiler:

I think this is how to solve this so take a look at it only if you have already proposed a solution.

select distinct m1.director, m1.title, r1.stars from movie m1
join rating r1 on m1.mID = r1.mID
left join (
    select m2.director, r2.stars from movie m2
    join rating r2 on m2.mID = r2.mID
) s on m1.director = s.director and r1.stars < s.stars
where s.stars is null and m1.director is not null