Postgresql left join without duplicates - postgresql

Question

DB structure:
CREATE TABLE page
(
id serial primary key,
title VARCHAR(40) not null
);
CREATE TABLE page_rating
(
id serial primary key,
page_id INTEGER,
rating_type INTEGER,
rating INTEGER
);
CREATE TABLE user_history
(
id serial primary key,
page_id INTEGER
)
Data:
INSERT INTO page (id,title) VALUES(1,'Page #1');
INSERT INTO page (id,title) VALUES(2,'Page #2');
INSERT INTO page (id,title) VALUES(3,'Page #3');
INSERT INTO page (id,title) VALUES(4,'Page #4');
INSERT INTO page (id,title) VALUES(5,'Page #5');
INSERT INTO page_rating VALUES (1,1,60,100);
INSERT INTO page_rating VALUES (2,1,99,140);
INSERT INTO page_rating VALUES (3,1,58,120);
INSERT INTO page_rating VALUES (4,1,70,110);
INSERT INTO page_rating VALUES (5,2,60,50);
INSERT INTO page_rating VALUES (6,2,99,60);
INSERT INTO page_rating VALUES (7,2,58,90);
INSERT INTO page_rating VALUES (8,2,70,140);
Purpose - select unique values for rating_type ​​in a table "page" sorted by "rating_page.rating". And exclude table user_history from the result
My query:
SELECT DISTINCT ON(pr.rating_type) p.*,pr.rating,pr.rating_type FROM page as p
LEFT JOIN page_rating as pr ON p.id = pr.page_id
LEFT JOIN user_history uh ON uh.page_id = p.id
WHERE
pr.rating_type IN (60, 99, 58, 45, 73, 97, 55, 59, 70, 43, 74, 97, 64, 71, 46)
AND uh.page_id IS NULL
ORDER BY pr.rating_type,pr.rating DESC
Result:
ID TITLE RATING RATING_TYPE
1 "Page #1" 120 58
1 "Page #1" 100 60
2 "Page #2" 140 70
1 "Page #1" 140 99
Duplicate values ( Ideal:
ID TITLE RATING RATING_TYPE
1 "Page #1" 120 58
1 "Page #2" 50 60
Thx for help!

Answer 1

You almost certainly need a UNIQUE constraint on {page_id, rating_type} in the table "page_rating". You're also missing every necessary foreign key constraint. The primary key on "user_history" is suspicious, too.
Purpose - select unique values for rating_type ​​in a table "page"
sorted by "rating_page.rating".
You can select distinct values for rating_type without referring to any other tables. And you should, at first. Let's look at the data.
select page_id, rating_type, rating
from page_rating
order by page_id, rating_type;
page_id rating_type rating
--
1 58 120 *
1 60 100
1 70 110
1 99 140
2 58 90
2 60 50 *
2 70 140
2 99 60
You seem to want one row per page_id. Those rows are marked with an asterisk in the table above. How can we get those two rows?
Those rows have different values for rating_type, so we can't just use rating_type in the WHERE clause. The values in rating are neither the max nor the min for both values of rating_type, so we can't use GROUP BY with max() or min(). And we can't use GROUP BY with an aggregate function, because you want the unaggregated value of "rating" for an arbitrary value of "rating_type".
So, based on what you've told us, the only way to get the result set you want is to specify rating_type and page_id in the WHERE clause.
select page_id, rating_type, rating
from page_rating
where (page_id = 1 and rating_type = 58)
or (page_id = 2 and rating_type = 60)
order by page_id, rating_type;
page_id rating_type rating
--
1 58 120
2 60 50
I'm not going to follow through with the joins, because I'm 100% confident that you don't really want to do this.