How to optimize huge left join in mysql? - mysql

Question

I have two tables in my mysql database:
1) Videos
+----+--------+----------+------+----------+
| id | title | category | year | director |
+----+--------+----------+------+----------+
| 1 | Title1 | Cat1 | 2021 | A.K. |
| 2 | Title2 | Cat2 | 2020 | B.C. |
| 3 | Title3 | Cat3 | 2000 | E.A. |
+----+--------+----------+------+----------+
2) Videos_insights
+----------+------------+-------+-------+----------+--------+
| video_id | date | views | likes | dislikes | shares |
+----------+------------+-------+-------+----------+--------+
| 1 | 2021-03-20 | 13 | 2 | 3 | 1 |
| 1 | 2021-03-19 | 35 | 1 | 3 | 3 |
| 1 | 2021-03-18 | 68 | 5 | 6 | 5 |
| 1 | 2021-03-15 | 86 | 3 | 0 | 1 |
| 2 | 2021-02-13 | 234 | 15 | 1 | 34 |
| 2 | 2021-02-12 | 55 | 15 | 2 | 4 |
| 2 | 2021-02-10 | 331 | 255 | 0 | 0 |
+----------+------------+-------+-------+----------+--------+
And I want to get videos that had between watched 2021-03-01 to 2021-03-31. So the result table should look like this:
+--------+-------------------------------------------+
| title | date_range |
+--------+-------------------------------------------+
| Title1 | ["2021-03-20 - 2021-03-18", "2021-03-15"] |
+--------+-------------------------------------------+
In my MySQL database, I have about 100 000 videos and each video has about 100 video_insight.
What is the best way to achieve the result table?
How to optimize? I mean I do not want to make every GET request left join? It would take too long and my server will burn out.

Answer 1

I would express the dates individually:
select v.id, v.title,
group_concat(date) as dates
from videos v join
video_insights vi
on vi.video_id = v.id
where vi.date >= '2021-03-01' and
vi.date < '2021-04-01'
group by v.id;
Note that a left join is not appropriate because you are filtering the values.
If you really want to get the ranges, then you can use window functions with a gaps-and-islands approach:
select v.id, v.title, group_concat(date_range)
from videos v join
(select vi.video_id,
concat_ws(' - ', min(vi.date), nullif(max(date), min(date))) as date_range
from (select vi.*,
dense_rank() over (partition by vi.video_id order by vi.date) as seqnum
from video_insights vi
where vi.date >= '2021-03-01' and
vi.date < '2021-04-01'
) vi
group by vi.video_id, date - interval seqnum day
) vi
on vi.video_id = v.id
group by v.id;