Merge a two way relation in the same table in SQL Server - sql

Question

Current Data
ID | Name1 | Name2
<guid1> | XMind | MindNode
<guid2> | MindNode | XMind
<guid3> | avast | Hitman Pro
<guid4> | Hitman Pro | avast
<guid5> | PPLive | Hola!
<guid6> | ZenMate | Hola!
<guid7> | Hola! | PPLive
<guid8> | Hola! | ZenMate
Required Output
ID1 | ID2 | Name1 | Name2
<guid1> | <guid2> | XMind | MindNode
<guid3> | <guid4> | avast | Hitman Pro
<guid5> | <guid7> | PPLive | Hola!
<guid6> | <guid8> | Hola! | ZenMate
These are relations between apps. I want to show that Avast and Hitman has a relation but in this view i do not need to show in what "direction" they have an relation. It's a given in this view that the relation goes both ways.
EDIT: Seems like my example was to simple. The solution doesn't work with more data.
DECLARE #a TABLE (ID INT, Name1 VARCHAR(50), Name2 VARCHAR(50))
INSERT INTO #a VALUES ( 1, 'XMind', 'MindNode' )
INSERT INTO #a VALUES ( 2, 'MindNode', 'XMind' )
INSERT INTO #a VALUES ( 3, 'avast', 'Hitman Pro' )
INSERT INTO #a VALUES ( 4, 'Hitman Pro', 'avast' )
INSERT INTO #a VALUES ( 5, 'PPLive Video Accelerator', 'Hola! Better Internet' )
INSERT INTO #a VALUES ( 6, 'ZenMate', 'Hola! Better Internet' )
INSERT INTO #a VALUES ( 7, 'Hola! Better Internet', 'PPLive Video Accelerator' )
INSERT INTO #a VALUES ( 8, 'Hola! Better Internet', 'ZenMate' )
SELECT a1.ID AS ID1 ,
a2.ID AS ID2 ,
a1.Name1 ,
a2.Name1 AS Name2
FROM #a a1
JOIN #a a2 ON a1.Name1 = a2.Name2
AND a1.ID < a2.ID -- avoid duplicates
This works however so i guess it's the Guid that is messing with me.
EDIT AGAIN:
I haven't looked at this for a while and i thought it worked but i just realized it does not. I've struggled all morning with this but i must admit that SQL is not really my strong suite. The thing is this.
DECLARE #a TABLE (ID int, Name1 VARCHAR(50), Name2 VARCHAR(50))
INSERT INTO #a VALUES ( 1, 'XMind', 'MindNode' )
INSERT INTO #a VALUES ( 2, 'MindNode', 'XMind' )
INSERT INTO #a VALUES ( 3, 'avast', 'Hitman Pro' )
INSERT INTO #a VALUES ( 4, 'PPLive Video Accelerator', 'Hola! Better Internet' )
INSERT INTO #a VALUES ( 5, 'ZenMate', 'Hola! Better Internet' )
INSERT INTO #a VALUES ( 6, 'Hitman Pro', 'avast' )
INSERT INTO #a VALUES ( 7, 'Hola! Better Internet', 'PPLive Video Accelerator' )
INSERT INTO #a VALUES ( 8, 'Hola! Better Internet', 'ZenMate' )
INSERT INTO #a VALUES ( 9, 'XX', 'A' )
INSERT INTO #a VALUES ( 10, 'XX', 'BB' )
INSERT INTO #a VALUES ( 11, 'BB', 'XX' )
INSERT INTO #a VALUES ( 12, 'A', 'XX' )
INSERT INTO #a VALUES ( 13, 'XX', 'CC' )
INSERT INTO #a VALUES ( 14, 'CC', 'XX' )
;With CTE as
(
SELECT a1.ID AS ID1 ,
a2.ID AS ID2 ,
a1.Name1 ,
a2.Name1 AS Name2,
CheckSum(Case when a1.Name1>a2.Name1 then a2.Name1+a1.Name1 else a1.Name1+a2.Name1 end) ck, -- just for display
Row_Number() over (Partition by CheckSum(Case when a1.Name1>a2.Name1 then a2.Name1+a1.Name1 else a1.Name1+a2.Name1 end)
order by CheckSum(Case when a1.Name1>a2.Name1 then a2.Name1+a1.Name1 else a1.Name1+a2.Name1 end)) as rn
FROM #a a1
JOIN #a a2 ON a1.Name1 = a2.Name2
)
Select ID1, ID2,Name1, Name2
from CTE C1
where rn=1
When i use this code it sure works fine with the names but it doesn't match the ID's correctly.
The result is
ID1 | ID2 | Name1 | Name2
12 | 9 | A | X (Correct)
7 | 5 | Hola! | ZenMate (Not Correct)
[..]
I've pulled my hair all morning but i can't figure this out. I still use Guid's as ID's and just use Int's here to make it a bit more readable.

Answer 1

DECLARE #a TABLE (ID INT, Name1 VARCHAR(50), Name2 VARCHAR(50))
INSERT INTO #a VALUES ( 1, 'XMind', 'MindNode' )
INSERT INTO #a VALUES ( 2, 'MindNode', 'XMind' )
INSERT INTO #a VALUES ( 3, 'avast', 'Hitman Pro' )
INSERT INTO #a VALUES ( 4, 'Hitman Pro', 'avast' )
SELECT a1.ID AS ID1 ,
a2.ID AS ID2 ,
a1.Name1 ,
a2.Name1 AS Name2
FROM #a a1
JOIN #a a2 ON a1.Name1 = a2.Name2
AND a1.ID < a2.ID -- avoid duplicates
Referring to the amendment and extension of your question, a more complicated solution is required.
We form a CHECKSUM on a1.Name1,a2.Name (to get an identical we exchanged on size).
Using this we generate with ROW_NUMBER (Transact-SQL) a number and use only rows from the result with number 1.
DECLARE #a TABLE (ID uniqueIdentifier, Name1 VARCHAR(50), Name2 VARCHAR(50))
INSERT INTO #a VALUES ( NewID(), 'XMind', 'MindNode' )
INSERT INTO #a VALUES ( NewID(), 'MindNode', 'XMind' )
INSERT INTO #a VALUES ( NewID(), 'avast', 'Hitman Pro' )
INSERT INTO #a VALUES ( NewID(), 'Hitman Pro', 'avast' )
INSERT INTO #a VALUES ( NewID(), 'PPLive Video Accelerator', 'Hola! Better Internet' )
INSERT INTO #a VALUES ( NewID(), 'ZenMate', 'Hola! Better Internet' )
INSERT INTO #a VALUES ( NewID(), 'Hola! Better Internet', 'PPLive Video Accelerator' )
INSERT INTO #a VALUES ( NewID(), 'Hola! Better Internet', 'ZenMate' )
INSERT INTO #a VALUES ( NewID(), 'XX', 'A' )
INSERT INTO #a VALUES ( NewID(), 'A', 'XX' )
INSERT INTO #a VALUES ( NewID(), 'XX', 'BB' )
INSERT INTO #a VALUES ( NewID(), 'BB', 'XX' )
INSERT INTO #a VALUES ( NewID(), 'XX', 'CC' )
INSERT INTO #a VALUES ( NewID(), 'CC', 'XX' )
;With CTE as
(
SELECT a1.ID AS ID1 ,
a2.ID AS ID2 ,
a1.Name1 ,
a2.Name1 AS Name2,
CheckSum(Case when a1.Name1>a2.Name1 then a2.Name1+a1.Name1 else a1.Name1+a2.Name1 end) ck, -- just for display
Row_Number() over (Partition by CheckSum(Case when a1.Name1>a2.Name1 then a2.Name1+a1.Name1 else a1.Name1+a2.Name1 end)
order by CheckSum(Case when a1.Name1>a2.Name1 then a2.Name1+a1.Name1 else a1.Name1+a2.Name1 end)) as rn
FROM #a a1
JOIN #a a2 ON a1.Name1 = a2.Name2
)
Select *
from CTE C1
where rn=1
Edit:
If you only want to get those where both fields are fitting the needed query would simply be:
SELECT a1.ID AS ID1 , a2.ID AS ID2 , a1.Name1 , a2.Name1 AS Name2
FROM #a a1
JOIN #a a2 ON a1.Name1 = a2.Name2 and a1.Name2 = a2.Name1 AND a1.ID < a2.ID

Answer 2

If the output should contain only two-way relations ('XX' + 'A') AND ('A' + 'XX'), try this:
;
WITH m (ID1, ID2, Name1, Name2) AS (
SELECT ID1, ID2, Name1, Name2
FROM (
SELECT a1.ID AS ID1
,a2.ID AS ID2
,a1.Name1 AS Name1
,a2.Name1 AS Name2
,ROW_NUMBER() OVER (PARTITION BY a1.Name1, a2.Name1 ORDER BY (SELECT 1)) AS n
FROM #a AS a1
JOIN #a AS a2
ON a1.Name1 = a2.Name2
AND a1.Name2 = a2.Name1
) AS T
WHERE n = 1
)
SELECT DISTINCT *
FROM (
SELECT ID1, ID2, Name1, Name2
FROM m
WHERE ID1 <= ID2
UNION ALL
SELECT ID2, ID1, Name2, Name1
FROM m
WHERE ID1 > ID2
) AS dm
It produces the output as follows:
+------+-----+--------------------------+-----------------------+
| ID1 | ID2 | Name1 | Name2 |
+------+-----+--------------------------+-----------------------+
| 1 | 2 | XMind | MindNode |
| 3 | 6 | avast | Hitman Pro |
| 4 | 7 | PPLive Video Accelerator | Hola! Better Internet |
| 5 | 8 | ZenMate | Hola! Better Internet |
| 9 | 12 | XX | A |
| 10 | 11 | XX | BB |
| 13 | 14 | XX | CC |
+------+-----+--------------------------+-----------------------+

Answer 3

Just rank your rows with ROW_NUMBER function and use this rank in join instead of original ID column:
DECLARE #a TABLE (ID UNIQUEIDENTIFIER, Name1 VARCHAR(50), Name2 VARCHAR(50))
INSERT INTO #a VALUES ( NEWID(), 'XMind', 'MindNode' )
INSERT INTO #a VALUES ( NEWID(), 'MindNode', 'XMind' )
INSERT INTO #a VALUES ( NEWID(), 'avast', 'Hitman Pro' )
INSERT INTO #a VALUES ( NEWID(), 'Hitman Pro', 'avast' )
INSERT INTO #a VALUES ( NEWID(), 'PPLive Video Accelerator', 'Hola! Better Internet' )
INSERT INTO #a VALUES ( NEWID(), 'ZenMate', 'Hola! Better Internet' )
INSERT INTO #a VALUES ( NEWID(), 'Hola! Better Internet', 'PPLive Video Accelerator' )
INSERT INTO #a VALUES ( NEWID(), 'Hola! Better Internet', 'ZenMate' )
;WITH cte AS(SELECT *, ROW_NUMBER() OVER (ORDER BY (SELECT 1)) rn FROM #a)
SELECT a1.ID AS ID1 ,
a2.ID AS ID2 ,
a1.Name1 ,
a2.Name1 AS Name2
FROM cte a1
JOIN cte a2 ON a1.Name1 = a2.Name2 AND
a2.Name1 = a1.Name2 AND
a1.rn < a2.rn
Output:
ID1 ID2 Name1 Name2
Guid Guid XMind MindNode
Guid Guid avast Hitman Pro
Guid Guid PPLive Video Accelerator Hola! Better Internet
Guid Guid ZenMate Hola! Better Internet

Answer 4

I suggest you to use this simple way:
SELECT
t2.ID, t3.ID ID2,
t1.Name1,t1.Name2
FROM (
SELECT DISTINCT
CASE WHEN Name1 <= Name2 THEN Name1 ELSE Name2 END AS Name1,
CASE WHEN Name1 <= Name2 THEN Name2 ELSE Name1 END AS Name2
FROM
#a) t1
JOIN
#a t2 ON t1.Name1+t1.Name2 = t2.Name1+t2.Name2
JOIN
#a t3 ON t1.Name1+t1.Name2 = t3.Name2+t3.Name1
For this:
ID | ID2 | Name1 | Name2
----+-----+-----------------------+---------------------------
12 | 9 | A | XX
3 | 4 | avast | Hitman Pro
11 | 10 | BB | XX
14 | 13 | CC | XX
7 | 5 | Hola! Better Internet | PPLive Video Accelerator
8 | 6 | Hola! Better Internet | ZenMate
2 | 1 | MindNode | XMind

Answer 5

You can solve this using a CROSS APPLY
SELECT a2.ID ID_1,a1.ID ID_2, a2.Name1 , a2.Name2
FROM #a a1
CROSS APPLY
(
SELECT ID, Name2, Name1
FROM #a aa
WHERE aa.Name1 = a1.Name2 AND a1.Name1 = aa.Name2 AND a1.ID > aa.ID
) a2

Answer 6

You can try also:
select min(ID) ID1,
max(ID) ID2,
Name1,
Name2
from ( -- Here I get all the IDs and each couple sorted
-- Change > to < if you don't like the order
select ID,
case
when Name1 > Name2 then Name1
else Name2
end Name1,
case
when Name1 > Name2 then Name2
else Name1
end Name2
from table1
) as t
group by Name1,
Name2
You can even tansform this in a simgle query, without the inner one, but I think in this way it's more readable and you can understand better my approach.