根据列和SQL中的两个条件选择不同的行

Question

I would like to retrieve rows from this data set where a T1/T3 value exists, but no T2/T3 value exists for a corresponding ID.

ID  sample1 sample2 value
A_000002    T2  T3  -0.934119
A_000002    T1  T3  -0.866637
A_000029    T2  T3  -1.07677
A_000037    T2  T3  -0.76506
A_000057    T1  T3  -5.34988

I'd like to say something like:

SELECT * FROM table
WHERE DISTINCT ID
AND sample_1 == "T1"
AND sample_2 == "T3"

...and return only the following because it has no corresponding T2/T3 row for that ID:

A_000057    T1  T3  -5.34988

If I use sample_1 and sample_2 conditions, I get distinct values anyway because it filters out the the "T2" values before checking if the ID is distinct.

The closest I've come is to make 3 tables with the possible T1/T2/T3 combinations and screen for NOT EXISTS T1T2.ID = T2T3.ID

select * from T1T2
where not exists (select * from T2T3 where T2T3.id = T1T2.id)
and not exists (select * from T1T3 where T1T3._id = T1T2.id)
order by id

I'm not sure I trust the code yet though.

Answer 1

You can use not exists :

select t.*
from table t
where (sample1 = 'T1' and sample2 = 'T3') and
       not exists (select 1 
                   from table t1 
                   where t1.id = t.id and 
                         t1.sample1 = 'T2' and t1.sample2 = 'T3'
                  );

Answer 2

You can use this technique based on an outer join:

select t1.*
from table t1
left join table t2
on t2.id = t1.id
  and t2.sample1 = 'T2' and t2.sample2 = 'T3'
where t1.sample1 = 'T1' and t1.sample2 = 'T3'
and t2.id is null

It works because outer joins return nulls if there's no join, and you return only those via the t2.id is null condition in the where clause.

The big advantage of this technique is the efficient use of the index on the id column; this technique will generally out-perform other apporaches. Plus IMHO it's a neater looking query.

Note that the condition for the sample columns must be in the join condition for t2 , otherwise you'd effectively get an inner join, defeating the required outer join.

Answer 3

I would use not exists :

SELECT t.*
FROM table t
WHERE t.sample_1 = 'T1' AND t.sample_2 = 'T3' AND
      NOT EXISTS (SELECT 1 FROM table t2 WHERE t2.id = t.id);

Answer 4

Here will surly work:

Select distinct ID, * from table Where sample1 = 't1' and sample2 = 't3'