Group only if there are rows to be grouped - mysql

Question

I would like your assistance in solving an issue which I am battling now for days without even coming close to a solution. Unfortunately, I have already posted my issue and was not able to make any improvement with the suggestion delivered.
What I would like to achieve is somewhat attained by GROUP BY and HAVING with the possibility of CASE WHEN but whatever I do I am not getting to what I desire.
What I want to achieve is a GROUP BY only when the contents of the group exceed 3 rows and leave the individual items i.e. not grouped when group is less than or equal to three.
EXAMPLE
ID DESC VAL1 VAL 2 VAL 3
1 DESC1 2 2 4
2 DESC2 2 2 4
3 DESC3 2 2 4
4 DESC4 2 2 4
5 DESC5 1 1 2
6 DESC6 1 1 2
GROUP BY will be through VAL1, VAL2, VAL 3 through the following
SELECT * FROM TABLE1 GROUP BY VAL1,VAL2,VAL3
This will yield the following:
ID DESC VAL1 VAL 2 VAL 3
1 DESC1 2 2 4
5 DESC5 1 1 2
However what I need is the following:
ID DESC VAL1 VAL 2 VAL 3
1 DESC1 2 2 4
5 DESC5 1 1 2
6 DESC6 1 1 2
Can this be achieved with GROUP BY, what I think of is subquery but I cannot manage. Your assistance will be very much appreciated.
DBMS is MySQL.

Answer 1

Try this one. It might require some minor tweaks, as I didn't test it. But i think you will get the idea.
Select *
from table1
where md5(concat(val1,val2,val3)) in (
SELECT md5(concat(val1,val2,val3))
FROM TABLE1
GROUP BY VAL1,VAL2,VAL3
having count(*) > 3)
group by VAL1,VAL2,VAL3
union
Select *
from table1
where md5(concat(val1,val2,val3)) not in (
SELECT md5(concat(val1,val2,val3))
FROM TABLE1
GROUP BY VAL1,VAL2,VAL3
having count(*) > 3)

Answer 2

With UNION ALL, for the 2 different cases:
select t.* from tablename t inner join (
select min(id) minid
from tablename
group by val1, val2, val3
having count(*) > 3
) g on g.minid = t.id
union all
select * from tablename t
where (
select count(*) from tablename
where val1 = t.val1 and val2 = t.val2 and val3 = t.val3
) <= 3
See the demo

Answer 3

If you are using MySQL 8.0, you can achieve this simply with window functions COUNT() and ROW_NUMBER():
SELECT id, descr, val1, val2, val3
FROM (
SELECT
t.*,
COUNT(*) OVER(PARTITION BY val1, val2, val3) cnt,
ROW_NUMBER() OVER(PARTITION BY val1, val2, val3 ORDER BY id) rn
FROM mytable t
) x WHERE cnt < 3 OR rn = 1
ORDER BY id
In the inner query, cnt indicates how many records have the same val1, va2, val3 as the current one. rn assigns a rank to each record within groups of records having the same val1, va2, val3. The outer query then uses these two pieces of information to filter the relevant records.
Demo on DB Fiddle:
| id | descr | val1 | val2 | val3 |
| --- | ----- | ---- | ---- | ---- |
| 1 | DESC1 | 2 | 2 | 4 |
| 5 | DESC5 | 1 | 1 | 2 |
| 6 | DESC6 | 1 | 1 | 2 |