postgresql Update with SUM - postgresql

Question

In the given simplified example, I would like to update ‘total_score’ for each event in the events table with sum of score of users which participated in that event.
Using cursor for this is easy to comprehend and implement, but I would like to refactor this to set-based approach, using SELECT / UPDATE on the entire column as cursor is just too slow and probably a bad practice in this case.
I would be grateful if you could not only provide required query, but also explain or link to explanations how to think in ‘set-based’ manner rather than procedural cursor one.
POSTGRESQL version = 13
SETUP:
Tables are simplified as much as possible, there are no PKs etc.
There are users, user-activities and events in which users can participate. Every event is unique, takes place at certain time on certain world.
If user activity time on certain world coincide with the whole event duration, it is considered that this user participated in the event. For each event I need to sum score of all users who participated and then update the total_score (as the calculated sum).
TABLES:
Analyzed users table (contains selected users from whole user base who meet certain criteria (eg. Age etc.). Users inserted below already meet those requirements.
CREATE TABLE analyzed_user
(
user_id bigint NOT NULL,
score numeric
);
INSERT INTO analyzed_user VALUES(100, 400);
INSERT INTO analyzed_user VALUES(200, 800);
INSERT INTO analyzed_user VALUES(300, 1500);
Events table - events in which users could participate. Events always end before 23:59 (never lap over next day)
CREATE TABLE event
(
event_id bigint NOT NULL,
date date,
start_time time without time zone,
end_time time without time zone,
world varchar,
total_score numeric NOT NULL DEFAULT 0
);
INSERT INTO event VALUES (1, '2021-07-27', '08:00:00', '09:00:00', 'Earth', 0);
INSERT INTO event VALUES (2, '2021-07-27', '12:00:00', '13:00:00', 'Earth', 0);
INSERT INTO event VALUES (3, '2021-07-27', '14:00:00', '15:00:00', 'Mars', 0);
INSERT INTO event VALUES (4, '2021-07-27', '20:00:00', '21:00:00', 'Mars', 0);
Users activity table (for simplicity also end before 23:59 (server restarts at midnight and kicks everybody out)).
CREATE TABLE activity
(
user_id bigint NOT NULL,
date date,
start_time time without time zone,
end_time time without time zone,
world varchar
);
INSERT INTO activity VALUES (100, '2021-07-27', '07:00:00', '14:00:00', 'Earth');
INSERT INTO activity VALUES (100, '2021-07-27', '23:00:00', '23:30:00', 'Earth');
INSERT INTO activity VALUES (100, '2021-07-27', '15:00:00', '22:00:00', 'Mars');
INSERT INTO activity VALUES (200, '2021-07-27', '7:30:00', '9:30:00', 'Earth');
INSERT INTO activity VALUES (200, '2021-07-27', '13:00:00', '16:30:00', 'Mars');
INSERT INTO activity VALUES (200, '2021-07-27', '18:00:00', '20:20:00', 'Mars');
INSERT INTO activity VALUES (300, '2021-07-27', '11:30:00', '14:30:00', 'Earth');
INSERT INTO activity VALUES (300, '2021-07-27', '17:00:00', '18:30:00', 'Mars');
INSERT INTO activity VALUES (300, '2021-07-27', '19:30:00', '22:30:00', 'Mars');
In short, cursor approach(bad practice here?) would be following:
Create temp table qualified_user (id, score) – to keep track on users participating in the current row event.
Open cursor for select * from event
Fetch row
Truncate table qualified_user
Insert into qualified_user (user_id, score)
select user_id, score from activity a
left join analyzed_user u ON a.user_id = u.user_id
WHERE
a.world = record.world AND
a.start_time <= record.start_time AND
a.end_time >= record.end_time
i.e. gathering users and their score whose activity coincided with current cursor record’s (event) time and world).
Select sum of score from qualified_user table
Update current of cursor with calculated sum.
Repeat steps from 3 to 7 as long as there are next rows.
So basically it is going event by event, for each event (its start, end time and world) selecting users whose activity coincide and summing their score. The logic is fine, just computing time of cursor is awful. This is something I could not properly query in set-based manner.
My attempts were multiple combinations of the following (also with WITH clause etc), but always ending in group by which I believe is the problem(?) (aggregate functions like SUM are not allowed without group by clause):
UPDATE event ee
SET total_score = gg.total
FROM (SELECT SUM(uu.score) AS total, aa.user_id, aa.world, aa.start_time, aa.end_time, aa.date
FROM activity aa
LEFT JOIN qualified_user uu ON aa.user_id = uu.user_id
GROUP BY aa.user_id, aa.world, aa.start_time, aa.end_time, aa.date
) AS gg
WHERE
gg.world = ee.world AND
gg.start_time <= ee.start_time AND
gg.end_time >= ee.end_time AND
gg.date = ee.date;
The correct results would be:
Event #1 total score = 400 + 800 = 1200
Event #2 total score = 400+1500 = 1900
Event #3 total score = 800
Event #4 total score = 400+1500 = 1900
But as you can check, the results from the above query are wrong (800/1500/800/400)
For me it is not logical to group users activities, although I can’t go through without it and getting an error.
I would be grateful if you could explain what is wrong with query above and provide the proper query.

Answer 1

Two major problems I was dealing here were:
aggregate functions are not allowed in update statements
reference problem from inner to outer queries
After failing, digging and learning about CTEs, I came up with following solution.
WITH cte AS(
SELECT ee.event_id, ee.date, ee.start_time, ee.end_time, ee.world,
(SELECT SUM(score) FROM
activity aa
LEFT JOIN analyzed_user uu ON aa.user_id = uu.user_id
WHERE
aa.world = ee.world AND
aa.start_time <= ee.start_time AND
aa.end_time >= ee.end_time AND
aa.date = ee.date) AS total
FROM event ee)
UPDATE event eee
SET total_score = cte.total
FROM cte
WHERE eee.event_id = cte.event_id