Full Outer Join self with null values - sql-server

Question

I want to do a full outer self-join that includes nulls. For example, if the table Data looks like:
N Name Val
--------------
1 ABC 8
1 DEF 7
2 ABC 9
2 XYZ 6
(where N is a general index column to enable a self-join on sequential groups) and I do:
SELECT COALESCE(a.n, b.n) as n, COALESCE(a.Name, b.Name) as Name, a.Val as A, b.Val as B
FROM Data a
FULL OUTER JOIN Data b on a.N = b.N - 1 and a.Name = b.Name
I want:
N Name A B
---------------
1 ABC 8 9
1 DEF 7 NULL
1 XYZ NULL 6
but what I get is more like a cross-join:
n Name A B
--------------
1 ABC 8 9
1 DEF 7 NULL
2 ABC 9 NULL
2 XYZ 6 NULL
1 ABC NULL 8
1 DEF NULL 7
2 XYZ NULL 6
How do I perform this full outer join in order to get the condensed self-join results?
(Note: In practice column N is a generalized index, so solutions that require naming the values of N aren't practical.)

Answer 1

So far I've been only able to see doing this as a union. and a left and right join since the criteria of what you're after changes.
SELECT COALESCE(a.Name, b.Name) as Name, a.Val as A, b.Val as B
FROM Data a
LEFT JOIN Data b on a.Name = b.Name
and B.N = 2
WHERE A.N = 1
UNION
SELECT COALESCE(a.Name, b.Name) as Name, a.Val as A, b.Val as B
FROM Data a
RIGHT JOIN Data b on a.Name = b.Name
and A.N = 1
WHERE B.N = 2
Giving us:
+------+---+----+
| NAME | A | B |
+------+---+----+
| ABC | 8 | 9 |
| DEF | 7 | |
| XYZ | | 6 |
+------+---+----+
However this relies on a hardcoded N value which I don't think is that useful... working on better.

Answer 2

Since we want to handle a generalized self-join index column N let's extend the sample set a little further:
create table #Data (n int, name char(3), val int)
insert into #Data values (1, 'ABC',8)
insert into #Data values (1, 'DEF',7)
insert into #Data values (2, 'ABC',9)
insert into #Data values (2, 'XYZ',6)
insert into #Data values (3, 'ABC',9)
insert into #Data values (3, 'DEF',5)
insert into #Data values (3, 'XYZ',4)
For this sample we want the SQL to produce this output:
N Name A B
---------------
1 ABC 8 9
1 DEF 7 NULL
1 XYZ NULL 6
2 ABC 9 9
2 DEF NULL 5
2 XYZ 6 4
The following code works on the general case:
SELECT COALESCE(a.n, b.n-1) as i, COALESCE(a.Name, b.Name) as Name, a.Val as A, b.Val as B
FROM #Data a
FULL OUTER JOIN #Data b ON a.N = b.N - 1 AND a.Name = b.Name
WHERE a.n < (SELECT MAX(n) FROM #Data) -- Deals with end index case
OR (a.n is null AND b.n-1 IN (SELECT DISTINCT n FROM #Data))
ORDER BY COALESCE(a.n, b.n-1), Name
To see why this works, a good intermediate step is to note that when a.N = 1 we want the rows where n = 1 from:
SELECT COALESCE(a.n, b.n - 1) as n, COALESCE(a.Name, b.Name) as Name,
a.Val as A, b.Val as B
FROM #Data a
FULL OUTER JOIN #Data b ON a.N = b.N - 1 AND a.Name = b.Name

Answer 3

Please see the code below:
create table Data (n int, name char(3), val int)
insert into data values (1, 'ABC',8)
insert into data values (1, 'DEF', 7)
insert into data values (2 , 'ABC' , 9)
insert into data values (2 , 'XYZ', 6)
SELECT COALESCE(a.Name, b.Name) as Name, a.Val as A, b.Val as B
FROM Data a
FULL OUTER JOIN Data b on a.N = b.N - 1 and a.Name = b.Name
The output is this:
There are nulls on both sides.

Answer 4

Maybe this:
SELECT [Name]
,[1]
,[2]
FROM [table]
PIVOT
(
MAX([val]) FOR [N] IN ([1], [2])
) PVT;