Question

The Problem

I’m trying to write a stored procedure in SQL Server to find the best matching record. Given 5 input parameters @A, @B, @C, @D, and @E (all varchar50) that correspond to the 5 columns A, B, C, D, and E in my table, I’d like to find the record with the most matching columns. Each column that doesn’t match in the chosen record should contain a space ‘ ‘.

For example, if I have the input “Sony”, “PlayStation”, “Controller”, “Black”, “Damaged”, and my table contains the following columns:

"Sony"  "Playstation"  "Unit"   "Black"  "Damaged"
"Sony"  "Playstation"  " "      " "      " "

It should return the second row, because 2 parameters match and for the 3 that don’t, there are spaces. I don’t want to return the first row because even though 4 parameters match, the middle one does not match and it is not a space. If it had been a space, the first row would have been the winner.

My Approach

There are various specifics that I cannot reveal, but my basic approach (note that I’m an SQL novice) was to test every combination from MOST specific to LEAST specific. So my query would look something like this:

-- start with most specific
SELECT * FROM dbo.Items WHERE
   A = @A
   B = @B
   C = @C
   D = @D
   E = @E

-- if no matches, try next
IF @@ROWCOUNT = 0
SELECT * FROM dbo.Items WHERE
   A = @A
   B = @B
   C = SPACE(1)
   D = @D
   E = @E

... etc.

In my case, I only need to really test 16 configurations, because some of the permutations will never exist. Even so, this seems like a very inefficient way to achieve what I want. On top of that, it’s not even working. It seems like comparing against spaces is problematic because of some auto-trimming that’s going on. In any case, my current approach seems inefficient and it doesn’t work – so I turn to you for help.

Answer 1

Something like this?

-- Temp table to play with
SELECT 1 AS Id, 'Sony' AS A, 'Playstation' AS B, 'Unit' AS C, 'Black' AS D, 
    'Damaged' AS E
INTO #Items
UNION SELECT 2, 'Sony', 'Playstation', ' ', ' ', ' '

-- The query
DECLARE @a nvarchar(50), @b nvarchar(50), @c nvarchar(50), @d nvarchar(50), @e 

nvarchar(50)
SET @a = 'Sony'
SET @b = 'Playstation'
SET @c = 'Controller'
SET @d = 'Black'
SET @e = 'Damaged'

SELECT TOP 1 I.*
FROM (
    SELECT ID,
        CASE WHEN @a = a THEN 1 WHEN a = ' ' THEN 0 ELSE NULL END AS AResult,
        CASE WHEN @b = b THEN 1 WHEN b = ' ' THEN 0 ELSE NULL END AS BResult,
        CASE WHEN @c = c THEN 1 WHEN c = ' ' THEN 0 ELSE NULL END AS CResult,
        CASE WHEN @d = d THEN 1 WHEN d = ' ' THEN 0 ELSE NULL END AS DResult,
        CASE WHEN @e = e THEN 1 WHEN e = ' ' THEN 0 ELSE NULL END AS EResult
    FROM #Items
) IW
INNER JOIN #Items I ON I.ID = IW.ID
WHERE AResult IS NOT NULL AND BResult IS NOT NULL AND CResult IS NOT NULL 
    AND DResult IS NOT NULL AND EResult IS NOT NULL
ORDER BY AResult + BResult + CResult + DResult + EResult DESC

That should return this value:

"Sony"  "Playstation"  " "      " "      " "

If you change the temp table I played with to this:

SELECT 1 AS Id, 'Sony' AS A, 'Playstation' AS B, ' ' AS C, 'Black' AS D, 
    'Damaged' AS E
INTO #Items
UNION SELECT 2, 'Sony', 'Playstation', ' ', ' ', ' '

Then you should get

"Sony"  "Playstation"  " "   "Black"  "Damaged"

And finally, if you have this for your temp table example:

SELECT 1 AS Id, 'Sony' AS A, 'Playstation' AS B, 'Unit' AS C, 'Black' AS D, 
    'Damaged' AS E
INTO #Items
UNION SELECT 2, 'Sony', 'Playstation', 'Unit', ' ', ' '

Nothing will get returned, since both have ‘Unit’ in the third column.