Reputation: 6890
I have a query that is somewhat equal to the following:
SELECT @address = a.Id FROM dbo.[Address] AS a, dbo.Item as u
WHERE a.Name = u.Name
AND ISNULL(a.Number, '') = ISNULL(u.Number, '')
AND ISNULL(a.Floor, '') = ISNULL(u.Floor, '')
AND ISNULL(a.Door, '') = ISNULL(u.Door, '')
AND a.Zip = u.Zip
AND u.ItemId = @id;
The whole idea is to find the addresses from dbo.Address that correspond to the dbo.Item table.
Now, the problem is that from this query I need to have the non-matching ones and determine the reason why they didn't match example: we cannot find the Name, or the Number, or the Floor etc. I have been trying to accomplish this by using consecutive SELECTs:
SELECT @addressId = a.Id FROM dbo.[Address] AS a, dbo.Item as u
WHERE a.Name = u.Name
AND u.ItemId = @id;
-- Name not found
IF @addressId IS NULL
BEGIN
SET @retval = 4
RETURN @retval;
END
ELSE
BEGIN
SELECT @addressId = a.Id FROM dbo.[Address] AS a, dbo.Item as u
WHERE a.Name = u.Name
AND a.Number = ISNULL(u.HusNr, '')
AND u.ItemId = @id;
-- Number not found
IF @addressId IS NULL
BEGIN
SET @retval = 5
RETURN @retval;
END
... and so on
but this is slow and I come to think that there's got to be a more intelligent way to accomplish this.
NOTE(on Icarus's answer):
The problem with this approach is that it tries to join several fields at once. Sql tries to join all the fields and thus will return NULL for the whole row, and not for the specific columns. If we want the specific columns we would have to do something like:
SELECT a.Id ,
case when a.Name is null then 1
when a.Number is null then 2
when a.Floor is null and a.Door is null and a.Zip is null then 3
when a.Number is not null and a. Floor is not null and a.Door is not null and a.Zip is not null then 0 end as Reason
FROM dbo.Item u1 left join dbo.[Address] a
ON a.Name = u1.Name left join dbo.Item u2
ON a.Number = u2.Number left join dbo.Item u3
ON a.Floor = u3.Floor left join dbo.Item u4
ON a.Door = u4.Door left join dbo.Item u5
ON a.Zip = u5.Zip
where u.ItemId = @id;
...but then we get a bunch of results that we don't need i.e. all the possible combinations, which again...we don't need. and if we use the above CASE
on that the first one is always true because the first result is all NULLs
Upvotes: 2
Views: 1007
Reputation: 63962
You could return the reason code in a separate column as so:
SELECT a.Id ,
case when a.Name is null then 1
when a.Number is null then 2
when a.Floor is null and a.Door is null and a.Zip is null then 3
when a.Number is not null and a. Floor is not null and a.Door is not null and a.Zip is not null then 0 end as Reason
FROM dbo.[Address] a left join dbo.Item u
on a.Name = u.Name
or a.Number = u.Number
or a.Door = u.Door
where u.ItemId = @id;
Where Reason 0
above is a perfect match. You can define your reasons as you wish, my intention was to give you an idea on how to do this using a case.
Upvotes: 6
Reputation: 4156
Is there a reason you are trying to do this one address at a time?
You could find all the address that are null by using a LEFT OUTER join between address and item adding Address = NULL to your WHERE clause.
Upvotes: 0