How do I join an unknown number of rows to another row? - sql

Question

I have this scenario:
Table A:
---------------
ID| SOME_VALUE|
---------------
1 | 123223 |
2 | 1232ff |
---------------
Table B:
------------------
ID | KEY | VALUE |
------------------
23 | 1 | 435 |
24 | 1 | 436 |
------------------
KEY is a reference to to Table A's ID. Can I somehow join these tables so that I get the following result:
Table C
-------------------------
ID| SOME_VALUE| | |
-------------------------
1 | 123223 |435 |436 |
2 | 1232ff | | |
-------------------------
Table C should be able to have any given number of columns depending on how many matching values that are found in Table B.
I hope this enough to explain what I'm after here.
Thanks.

Answer 1

You need to use a Dynamic PIVOT clause in order to do this.
EDIT:
Ok so I've done some playing around and based on the following sample data:
Create Table TableA
(
IDCol int,
SomeValue varchar(50)
)
Create Table TableB
(
IDCol int,
KEYCol int,
Value varchar(50)
)
Insert into TableA
Values (1, '123223')
Insert Into TableA
Values (2,'1232ff')
Insert into TableA
Values (3, '222222')
Insert Into TableB
Values( 23, 1, 435)
Insert Into TableB
Values( 24, 1, 436)
Insert Into TableB
Values( 25, 3, 45)
Insert Into TableB
Values( 26, 3, 46)
Insert Into TableB
Values( 27, 3, 435)
Insert Into TableB
Values( 28, 3, 437)
You can execute the following Dynamic SQL.
declare #sql varchar(max)
declare #pivot_list varchar(max)
declare #pivot_select varchar(max)
Select
#pivot_list = Coalesce(#Pivot_List + ', ','') + '[' + Value +']',
#Pivot_select = Coalesce(#pivot_Select, ', ','') +'IsNull([' + Value +'],'''') as [' + Value + '],'
From
(
Select distinct Value From dbo.TableB
)PivotCodes
Set #Sql = '
;With p as (
Select a.IdCol,
a.SomeValue,
b.Value
From dbo.TableA a
Left Join dbo.TableB b on a.IdCol = b.KeyCol
)
Select IdCol, SomeValue ' + Left(#pivot_select, Len(#Pivot_Select)-1) + '
From p
Pivot ( Max(Value) for Value in (' + #pivot_list + '
)
)as pvt
'
exec (#sql)
This gives you the following output:
Although this works at the moment it would be a nightmare to maintain. I'd recommend trying to achieve these results somewhere else. i.e not in SQL!
Good luck!

Answer 2

As Barry has amply illustrated, it's possible to get multiple columns using a dynamic pivot.
I've got a solution that might get you what you need, except that it puts all of the values into a single VARCHAR column. If you can split those results, then you can get what you need.
This method is a trick in SQL Server 2005 that you can use to form a string out of a column of values.
CREATE TABLE #TableA (
ID INT,
SomeValue VARCHAR(50)
);
CREATE TABLE #TableB (
ID INT,
TableAKEY INT,
BValue VARCHAR(50)
);
INSERT INTO #TableA VALUES (1, '123223');
INSERT INTO #TableA VALUES (2, '1232ff');
INSERT INTO #TableA VALUES (3, '222222');
INSERT INTO #TableB VALUES (23, 1, 435);
INSERT INTO #TableB VALUES (24, 1, 436);
INSERT INTO #TableB VALUES (25, 3, 45);
INSERT INTO #TableB VALUES (26, 3, 46);
INSERT INTO #TableB VALUES (27, 3, 435);
INSERT INTO #TableB VALUES (28, 3, 437);
SELECT
a.ID
,a.SomeValue
,RTRIM(bvals.BValues) AS ValueList
FROM #TableA AS a
OUTER APPLY (
-- This has the effect of concatenating all of
-- the BValues for the given value of a.ID.
SELECT b.BValue + ' ' AS [text()]
FROM #TableB AS b
WHERE a.ID = b.TableAKEY
ORDER BY b.ID
FOR XML PATH('')
) AS bvals (BValues)
ORDER BY a.ID
;
You'll get this as a result:
ID SomeValue ValueList
--- ---------- --------------
1 123223 435 436
2 1232ff NULL
3 222222 45 46 435 437

Answer 3

This looks like something a database shouldn't do. Firstly; a table cannot have arbitrary number of columns depending on whatever you'll store. So you will have to put up a maximum number of values anyway. You can get around this by using comma seperated values as value for that cell (or a similar pivot-like solution).
However; if you do have table A and B; i recommend keeping to those two tables; as they seem to be pretty normalised. Should you need a list of b.value given an input a.some_value, the following sql query gives that list.
select b.value from a,b where b.key=a.id a.some_value='INPUT_VALUE';