Select rows from joined tables with more than n occurrence - postgresql

Question

My problem is similar with MySQL: Select rows with more than one occurrence but I am using PostgreSQL. I have a query like:
select d.user_id, d.recorded_at, d.glucose_value, d.unit
from diary as d
join (
select u.id
from health_user as u
join (
select distinct user_id
from care_connect
where clinic_id = 217
and role = 'user'
and status = 'active'
) as c
on u.id = c.user_id
where u.is_tester is false
) as cu
on d.user_id = cu.id
where d.created_at >= d.recorded_at
and d.recorded_at < current_date and d.recorded_at >= current_date - interval '30 days'
and d.glucose_value > 0
and (d.state = 'wakeup' or (d.state = 'before_meal' and d.meal_type = 'breakfast'))
The result looks like:
+---------+---------------------+---------------+--------+
| user_id | recorded_at | glucose_value | unit |
+---------+---------------------+---------------+--------+
| 12041 | 2018-06-26 01:10:12 | 100 | mg/dL |
| 12041 | 2018-06-30 02:10:11 | 90 | mg/dL |
| 12214 | 2018-06-25 12:40:13 | 10 | mmol/L |
| 12214 | 2018-06-26 12:41:13 | 12 | mmol/L |
| 12214 | 2018-06-29 00:21:14 | 11 | mmol/L |
| 12214 | 2018-06-29 12:59:32 | 10 | mmol/L |
+---------+---------------------+---------------+--------+
As you can see that is already a long query with many conditions. Now I want to get only the records that are from users who have no less than four records (rows) in the result, so I tried:
select d.user_id, d.recorded_at, d.glucose_value, d.unit, count(d.*)
from diary as d
join (
select u.id
from health_user as u
join (
select distinct user_id
from care_connect
where clinic_id = 217
and role = 'user'
and status = 'active'
) as c
on u.id = c.user_id
where u.is_tester is false
) as cu
on d.user_id = cu.id
where d.created_at >= d.recorded_at
and d.recorded_at < current_date and d.recorded_at >= current_date - interval '30 days'
and d.glucose_value > 0
and (d.state = 'wakeup' or (d.state = 'before_meal' and d.meal_type = 'breakfast'))
group by d.user_id
having count(d.*) >= 4
My expected output is:
+---------+---------------------+---------------+--------+
| user_id | recorded_at | glucose_value | unit |
+---------+---------------------+---------------+--------+
| 12214 | 2018-06-25 12:40:13 | 10 | mmol/L |
| 12214 | 2018-06-26 12:41:13 | 12 | mmol/L |
| 12214 | 2018-06-29 00:21:14 | 11 | mmol/L |
| 12214 | 2018-06-29 12:59:32 | 10 | mmol/L |
+---------+---------------------+---------------+--------+
However, it throws an error, saying that d.recorded_at should also be added in group by, but that's not what I want. Besides grouping raw timestamps is not meaningful.
I know I can probably join another table, which are generated by the same query but only select d.user_id, count(d.*) at the first line, but the whole query would look crazy.
Would somebody please help me how to achieve this in a better way? Sorry I don't put table structures here, but I can edit and clarify things if needed.

Answer 1

Try this
Select user_id, recorded_at, glucose_value, unit
From (
select d.user_id, d.recorded_at, d.glucose_value, d.unit, count(1) over (partition by d.user_id) rcnt
from diary as d
join (
select u.id
from health_user as u
join (
select distinct user_id
from care_connect
where clinic_id = 217
and role = 'user'
and status = 'active'
) as c
on u.id = c.user_id
where u.is_tester is false
) as cu
on d.user_id = cu.id
where d.created_at >= d.recorded_at
and d.recorded_at < current_date and d.recorded_at >= current_date - interval '30 days'
and d.glucose_value > 0
and (d.state = 'wakeup' or (d.state = 'before_meal' and d.meal_type = 'breakfast'))
) x
Where rcnt >= 4

Answer 2

Try this:
Replace your_query with your actual query.
Using with clause and exists clause.
with original_query as ( your_query )
select * from original_query q1
where
exists( select q2.user_id from original_query q2 where q1.user_id = q2.user_id
group by q2.user_id
having count(q2.user_id) >= 4 )