How do I simplify this complex query? - postgresql

Question

The referred to table definitions are:
CREATE TABLE message
(
id BIGINT PRIMARY KEY,
user_id BIGINT NOT NULL,
guild_id BIGINT NOT NULL,
content TEXT NOT NULL,
created_at TIMESTAMP NOT NULL,
);
CREATE TABLE d_user
(
id BIGINT PRIMARY KEY
);
CREATE TABLE vcsession
(
id BIGINT PRIMARY KEY,
user_id BIGINT NOT NULL,
guild_id BIGINT NOT NULL,
duration INTEGER NOT NULL,
began_at TIMESTAMP NOT NULL,
last_active TIMESTAMP NOT NULL
);
The expected result set of this query should consist of a row for each user in the guild provided with columns for:
user_id: The user id
message_count: The number of messages sent by each user within an interval defined by two datetimes (this should be 0 if no messages were sent)
voice_time: Sum of each voice session's duration last active within an interval defined by two datetimes (this should be 0 if no voice sessions were active)
active_days: Days in which the user either sent a message or had an active voice session (this should be 0 if user wasn't active in the time interval provided)
This is the query I wrote:
select
activity.user_id,
message_count,
voice_time,
coalesce(active_days, 0) as active_days
from (
select
d_user.id as user_id,
coalesce(messages.count, 0) as message_count,
coalesce(vcsessions.duration, 0) as voice_time
from d_user left join (
select
user_id,
count(*) as "count"
from message where (
(guild_id = $1) and
(created_at >= $2) and
(created_at < $3)
) group by user_id
) as messages on messages.user_id = d_user.id left join (
select
user_id,
sum(duration) as "duration"
from vcsession where (
(guild_id = $1) and
(last_active >= $2) and
(last_active < $3)
) group by user_id
) as vcsessions on vcsessions.user_id = d_user.id
) as activity left join (
select user_id, count(*) as active_days from (
select * from (
select
user_id,
(cast(extract(EPOCH from message.created_at) as int) - cast(extract(EPOCH from $2) as int)) / 86400 as day_offset
from message where (
(created_at >= $2) and
(created_at < $3)
) group by user_id, day_offset
) as message_days union (
select
user_id,
(cast(extract(EPOCH from vcsession.last_active) as int) - cast(extract(EPOCH from $2) as int)) / 86400 as day_offset
from vcsession where (
(last_active >= $2) and
(last_active < $3)
) group by user_id, day_offset
)
) as active_days group by user_id
) as active_days on active_days.user_id = activity.user_id
And this is what the result set looks like:
|user_id |message_count |voice_time |active_days |
|--------------------|--------------------|--------------------|--------------------|
|1 |752 |694 |1 |
|2 |12 |543 |2 |
|3 |323 |7163 |4 |
|4 |56 |870 |3 |

Answer 1

It looks reasonably readable to me.
Maybe you could pull the two subselects in the forst FROM clause into the main query:
SELECT ...
FROM ((SELECT ...) AS messages
LEFT JOIN
(SELECT ...) AS vcsessions
) AS ...
LEFT JOIN ...
could become
SELECT ...
FROM (SELECT ...) AS messages
LEFT JOIN
(SELECT ...) AS vcsessions
LEFT JOIN ...