选择具有相同外键但另一列具有一组值的行

Question

I have designed a database that stores data in a format like this where both columns are foreign keys leading to different tables. This is a simplified version.

RNA_id | Experiment_id |
   1   |       a       |
   1   |       b       |
   2   |       a       |
   2   |       b       |
   2   |       c       |
   3   |       b       |
   4   |       a       |
   4   |       c       |

I want to select rows that have all three experiment IDs. In this example the results should be

RNA_id | Experiment_id |
   2   |       a       |
   2   |       b       |
   2   |       c       |

I tried

GROUP BY RNA_id HAVING COUNT (DISTINCT Experiment_id)=3

but that just resulted in me having a seemingly random assortment of rows.

The database is already quite large and my query as is is fairly complex already, additionally, I may possibly want to expand to 4 or more Experiment_ids.

Here is a sanitized version of my query as it is complicated and I don't want to explain my whole data structure

My existing query is:

SELECT RNA_id, Experiment_id, <data values>
FROM data
LEFT JOIN ref1
LEFT JOIN ref2
LEFT JOIN ref3
LEFT JOIN ref4
WHERE <required data parameters>
ORDER BY RNA_id
LIMIT 0,5000;

Returns about 700 values, as it should, but when I change ORDER BY to that GROUP BY command earlier, it returns 9 values all of which have unique RNA_ids and there should be about 100 that get returned.

My RNA_id is actually 3 columns so I may just have to redo the entire database to get any of these solutions to work.

My Solution

I figured out how to make it work with the 3 column identifier

AND (RNA_id1, RNA_id2, RNA_id3) IN (SELECT RNA_id1, RNA_id2, RNA_id3
FROM data
WHERE <parameter>
GROUP BY RNA_id1, RNA_id2, RNA_id3
HAVING COUNT (DISTINCT Experiment_id)=3)

This works with different numbers of Experiment_ids too

Answer 1

I would recommend just returning RNA_id that have all three:

select RNA_id
from t
where Experiment_id in ('a', 'b', 'c')
group by RNA_id
having count(*) = 3;

If you can have duplicates, then use count(distinct experiment_id) .

If you want the original rows, then in MySQL 8+, you can adapt this using window functions:

select t.*
from (select t.*, count(*) over (partition by RNA_id) as cnt
      from t
      where Experiment_id in ('a', 'b', 'c')
     ) t
where cnt = 3;

Answer 2

SQL DEMO

SELECT *
FROM Table1 t1
WHERE ( SELECT COUNT(DISTINCT `Experiment_id`)
        FROM Table1 t2
        WHERE t2.`RNA_id` = t1.`RNA_id`
          AND t2.Experiment_id in ('a', 'b', 'c') -- if you have more than 3 experiment
      )  = 3

Answer 3

If a , b and c are the only possible values for Experiment_id , then you need to put your query in the WHERE clause like this:

select *
from tablename
where RNA_id in (
  select RNA_id from tablename
  group by RNA_id
  having count(distinct Experiment_id) = 3
)

If there are other values also:

select *
from tablename
where
  Experiment_id in ('a', 'b', 'c') 
  and RNA_id in (
  select RNA_id from tablename
  where Experiment_id in ('a', 'b', 'c')
  group by RNA_id
  having count(distinct Experiment_id) = 3  
)

These queries can be easily changed if you want to expand to more than 3 values.