Sql server combine multiple data sets without duplicate data - sql

Question

Given three tables Ta, Tb, Tc:
Ta(ID, Field1)
Tb(ID, Field2)
Tc(ID, Field3)
Given data example:
Ta
ID Field1
---------
1 A
1 B
Tb
ID Field2
---------
1 C
1 D
2 E
Tc
ID Field3
---------
1 F
2 G
2 H
Question:
How can I join this data to return:
ID Field1 Field2 Field3
-----------------------
1 A C F
1 B D NULL
2 NULL E G
2 NULL NULL H
I thought I could achieve this with outer joins but that doesn't seem to be the case. The order of the groupings doesn't really matter, as long as I bring back all information without duplicate rows.
Just to clarify. I don't really mind which combination I get as long as the result set returns all data in the minimum number of rows. Here's a more realistic example of what I am trying to do:
Given a person, call him John. He has two phone numbers and three email addresses:
PID Email
---------
John john#test.com
John john#mail.com
John john#john.com
PID Tel
--------
John 011
John 022
I want to return:
PID Email Tel
----------------------
John john#test.com 011
John john#mail.com 022
John john#john.com NULL

Answer 1

You can come close with the following:
select coalesce(ta.id, tb.id, tc.id), ta.field1, tb.field2, tc.field3
from (select ta.*, row_number() over (partition by id order by (select NULL)) as seqnum
from ta
) ta full outer join
(select tb.*, row_number() over (partition by id order by (select NULL)) as seqnum
from tb
) tb
on ta.id = tb.id and
ta.seqnum = tb.seqnum
(select tc.*, row_number() over (partition by id order by (select NULL)) as seqnum
from tc
) tc
on coalesce(ta.id, tb.id) = tc.id and
coalesce(ta.seqnum, tb.seqnum) = tc.seqnum
group by coalesce(ta.id, tb.id, tc.id),
coalesce(ta.seqnum, tb.seqnum, tc.seqnum)
order by 1, 2
As I said, though, in my comment, the ordering of rows in a table is not guaranteed, so these may not come out in the order you expect. With your sample data, you could use:
over (partition by id order by field<n>)
If the fields define the ordering

Answer 2

Here's an alternative, using CTE's and a Union, with MIN to exclude the nulls. It doesn't guarantee the ordering, but as since you say you don't care as long as the ID's are all present.
SQL Fiddle here
WITH TaRanked AS
(
SELECT ROW_NUMBER() OVER (PARTITION BY ID ORDER BY Field1) as Rnk, ID, Field1
FROM Ta
),
TbRanked AS
(
SELECT ROW_NUMBER() OVER (PARTITION BY ID ORDER BY Field2) as Rnk, ID, Field2
FROM Tb
),
TcRanked AS
(
SELECT ROW_NUMBER() OVER (PARTITION BY ID ORDER BY Field3) as Rnk, ID, Field3
FROM Tc
),
TUnion AS
(
SELECT Rnk, ID, Field1, NULL AS Field2, NULL AS Field3
FROM TaRanked
UNION ALL
SELECT Rnk, ID, NULL, Field2, NULL
FROM TbRanked
UNION ALL
SELECT Rnk, ID, NULL, NULL, Field3
FROM TcRanked
)
SELECT ID, MIN(Field1), MIN(Field2), MIN(Field3)
FROM TUnion
GROUP BY ID, Rnk
ORDER BY ID, Rnk
The result is
1 A C F
1 B D (null)
2 (null) E G
2 (null) (null) H