SQL - how to find unique values from 2 columns? - sql

Question

I'm using SQLite, and would like to get only rows with values that are unique per table. I know how to get values per each column, but can not create a single (sqlite compatible) select to get unique value from both column.
Here is data example:
col_A
col_B
111
777
222
333
222
111
444
333
222
555
333
666
333
777
My desired result from above example would be rows with unique values that include only: 444, 555, 666.
col_A
col_B
444
333
222
555
333
666
222 should not be in results as it appears multiple times in col_A.
777 should not be in results as it appears multiple times in col_B.
I need only single occurrence of value in either col_A or col_B. Unique value per table, not column.
How far i managed to get:
SELECT *
FROM my_table
WHERE
(SELECT col_A as asset from my_table GROUP BY col_A HAVING COUNT(*) = 1
UNION
SELECT col_B as asset from my_table GROUP BY col_B HAVING COUNT(*) = 1)
IN (col_A, col_B)
I have tried with UNION and few other approaches, but can not achieve desired result.

Answer 1

To Start filtering by UNION ALL considering the keyword either
would be suitable option along with HAVING clause to determine the distinct values such as
WITH t1 AS
(
SELECT col
FROM (SELECT col_A AS col FROM t UNION ALL
SELECT col_B FROM t) AS tt
GROUP BY col
HAVING COUNT(*)=1
)
SELECT t.*
FROM t1
JOIN t
ON col IN ( col_A , col_B )
Demo

Answer 2

I think you need a query like this:
;with unqs as (
select col_c
from (
select col_c, count(col_c) cnt
from (
select col_A as col_c
from my_table
union all
select col_B as col_c
from my_table) as t
group by col_c) as tt
where cnt = 1
)
select *
from my_table
where col_A in (select col_c from unqs)
or col_B in (select col_c from unqs);
SQL Fiddle

Answer 3

WITH
symetric AS
(
SELECT col_a, col_b, col_a AS lookup FROM t
UNION ALL
SELECT col_a, col_b, col_b AS lookup FROM t
)
SELECT
col_a, col_b
FROM
symetric AS t
WHERE
NOT EXISTS (
SELECT *
FROM symetric
WHERE lookup = t.lookup
AND (col_a, col_b) <> (t.col_a, t.col_b)
)
ORDER BY
col_a, col_b
Demo ; https://dbfiddle.uk/9tzUASle