Question

I’m trying to solve the following: the data is organized in the table with Column X as the foreign key for the information (it’s the ID which identifies a set of rows in this table as belonging together in a bundle, owned by a particular entity in another table). So each distinct value of X has multiple rows associated with it here. I would like to filter out all distinct values of X that have a row associated with them containing value “ABC” in Column Q.

i.e.

data looks like this:

Column X   Column Q
--------   ---------
123        ABC
123        AAA    
123        ANQ
456        ANQ
456        PKR
579        AAA
579        XYZ
886        ABC

the query should return “456” and “579” because those two distinct values of X have no rows containing the value “ABC” in Column Q.

I was thinking of doing this with a minus function (select distinct X minus (select distinct X where Q = “ABC”)), as all I want are the distinct values of X. But i was wondering if there was a more efficient way to do this that could avoid a subquery? If for example I could partition the table over X and throw out each partition that had a row with the value “ABC” in Q?

Answer 1

This should work:

SELECT DISTINCT t.ColX
FROM mytable t
  LEFT JOIN mytable t2 on t.colx = t2.colx and t2.colq = 'ABC'
WHERE t2.colx IS NULL

And here is the SQL Fiddle.

Good luck.