Question

I have a table that looks like this:

CREATE TABLE [Test].[dbo].[MyTest]
(
[Id]    INT NOT NULL,
[ColA]  VARCHAR(255) NULL,
[ColB]  VARCHAR(255) NULL,
[ColC]  VARCHAR(255) NULL,
[ColD]  VARCHAR(255) NULL
);

And let’s say I have:

Id     ColA     ColB     ColC     ColD
---------------------------------------
1       A        B        NULL     C
2       A        A        NULL     D
3       NULL     A        B        NULL
4       B        B        B        B
5       NULL     NULL     NULL     NULL

What I’m trying to do is select every row from this table, but only want the first 2 distinct non-null values from ColA-ColD, in order. In other words, If for one row ColA & ColB are both non-null and are distinct from each other, those are the 2 that I want for that row. Using the data I gave above, the wanted results from this query would be:

1, A (from ColA), B (from ColB)
2, A (from ColA), D (from ColD)
3, A (from ColB), B (from ColC)
4, B (from ColA)

Notice that if all data for ColA-D for a row is NULL that row is not selected. Also it’s ok if there is only 1 non-null distinct column (you can see from result row with Id 4) – it doesn’t have to be 2, but ideally it would be 2.

Basically I would ideally be able to get the TOP(2) DISTINCT from ColA-ColD WHERE none of those are NULL, but I realize TOP and DISTINCT work on rows, not on columns like I’m trying to do. Any help greatly appreciated.

Thanks!

Answer 1

This might be fastest with a CTE and a CASE statement:

;WITH cte AS (
    SELECT id
          ,COALESCE(ColA, ColB, ColC, ColD) AS col1
          ,ColB
          ,ColC
          ,ColD
    FROM   tbl
    )
SELECT id
      ,col1
      ,CASE WHEN ColB <> col1 THEN ColB 
            WHEN ColC <> col1 THEN ColC
            WHEN ColD <> col1 THEN ColD
            ELSE                   NULL
       END AS col2
--    ,COALESCE(NULLIF(ColB, col1)
--             ,NULLIF(ColC, col1)
--             ,NULLIF(ColD, col1)) As col2   -- alt. syntax doing the same
FROM   cte
WHERE  col1 IS NOT NULL
ORDER  BY id

The first column returned (col1) is the first non-null column and the second column (col2) is the second distinct non-null value in that order. The CASE statement works, because NULL <> value never evaluates to TRUE.

If there is only one distinct value, col2 returns NULL.
If all source columns are NULL, col1 is NULL and the row is filtered by the WHERE clause.

The commented alternative syntax with NULLIF() is shorter, but I don’t expect it to be faster.