Recursive SQL query to find all matching identifiers - sql

Question

I have a table with following structure
CREATE TABLE Source
(
[ID1] INT,
[ID2] INT
);
INSERT INTO Source ([ID1], [ID2])
VALUES (1, 2), (2, 3), (4, 5),
(2, 5), (6, 7)
Example of Source and Result tables:
Source table basically stores which id is matching which another id. From the diagram it can be seen that 1, 2, 3, 4, 5 are identical. And 6, 7 are identical. I need a SQL query to get a Result table with all matches between ids.
I found this item on the site - Recursive query in SQL Server
similar to my task, but with a different result.
I tried to edit the code for my task, but it does not work. "The statement terminated. The maximum recursion 100 has been exhausted before statement completion."
;WITH CTE
AS
(
SELECT DISTINCT
M1.ID1,
M1.ID1 as ID2
FROM Source M1
LEFT JOIN Source M2
ON M1.ID1 = M2.ID2
WHERE M2.ID2 IS NULL
UNION ALL
SELECT
C.ID2,
M.ID1
FROM CTE C
JOIN Source M
ON C.ID1 = M.ID1
)
SELECT * FROM CTE ORDER BY ID1
Thanks a lot for the help!

Answer 1

This is a challenging question. You are trying to walk through a graph in two directions. There are two key ideas:
Add "reverse" edges, so the graph behaves like a digraph but with edges in both directions.
Keep a list of edges that have been visited. In SQL Server, strings are one method.
So:
with s as (
select id1, id2 from source
union -- on purpose
select id2, id1 from source
),
cte as (
select s.id1, s.id2, ',' + cast(s.id1 as varchar(max)) + ',' + cast(s.id2 as varchar(max)) + ',' as ids
from s
union all
select cte.id1, s.id2, ids + cast(s.id2 as varchar(max)) + ','
from cte join
s
on cte.id2 = s.id1
where cte.ids not like '%,' + cast(s.id2 as varchar(max)) + ',%'
)
select *
from cte
order by 1, 2;
Here is a db<>fiddle.

Answer 2

Since all node connections are bidirectional - add reversed relations to the original list
Find all possible paths from each node; almost usual recursion, the only difference is - we need to keep root id1
Avoid cycles - we need to be aware of it because we don't have directions
source:
;with src as(
select id1, id2 from source
union
-- reversed connections
select id2, id1 from source
), rec as (
select id1, id2, CAST(CONCAT('/', src.id1, '/', src.id2, '/') as varchar(8000)) path
from src
union all
-- keep the root id1 from the start of each path
select rec.id1, src.id2, CAST(CONCAT(rec.path, src.id2, '/') as varchar(8000))
from rec
-- usual recursion
inner join src on src.id1 = rec.id2
-- avoid cycles
where rec.path not like CONCAT('%/', src.id2, '/%')
)
select id1, id2, path
from rec
order by 1, 2
output
| id1 | id2 | path |
|-----|-----|-----------|
| 1 | 2 | /1/2/ |
| 1 | 3 | /1/2/3/ |
| 1 | 4 | /1/2/5/4/ |
| 1 | 5 | /1/2/5/ |
| 2 | 1 | /2/1/ |
| 2 | 3 | /2/3/ |
| 2 | 4 | /2/5/4/ |
| 2 | 5 | /2/5/ |
| 3 | 1 | /3/2/1/ |
| 3 | 2 | /3/2/ |
| 3 | 4 | /3/2/5/4/ |
| 3 | 5 | /3/2/5/ |
| 4 | 1 | /4/5/2/1/ |
| 4 | 2 | /4/5/2/ |
| 4 | 3 | /4/5/2/3/ |
| 4 | 5 | /4/5/ |
| 5 | 1 | /5/2/1/ |
| 5 | 2 | /5/2/ |
| 5 | 3 | /5/2/3/ |
| 5 | 4 | /5/4/ |
| 6 | 7 | /6/7/ |
| 7 | 6 | /7/6/ |
http://sqlfiddle.com/#!18/76114/13
source table will contain about 100,000 records
There is nothing that can help you with this. The task is unpleasant - finding all possible connections. Almost CROSS JOIN. With even more connections in the end.

Answer 3

Looks like I came up with a similar answer as the other posters. My approach was to insert the existing value pairs, and then insert the reverse of each pair.
Once you expand the list of value pairs, you can transverse the table to find all the pairs.
CREATE TABLE #Source
([ID1] int, [ID2] int);
INSERT INTO #Source
(
[ID1]
,[ID2]
)
VALUES
(1, 2)
,(2, 3)
,(4, 5)
,(2, 5)
,(6, 7)
INSERT INTO #Source
(
[ID1]
,[ID2]
)
SELECT
[ID2]
,[ID1]
FROM #Source
;WITH expanded AS
(
SELECT DISTINCT
ID1 = s1.ID1
,ID2 = s1.ID2
FROM #Source s1
LEFT JOIN #Source s2 ON s1.ID2 = s2.ID1
UNION
SELECT DISTINCT
ID1 = s1.ID1
,ID2 = s2.ID2
FROM #Source s1
LEFT JOIN #Source s2 ON s1.ID2 = s2.ID1
WHERE s1.ID1 <> s2.ID2
)
,recur AS
(
SELECT DISTINCT
e1.ID1
,e1.ID2
FROM expanded e1
LEFT JOIN expanded e2 ON e1.ID2 = e2.ID1
WHERE e1.ID1 <> e1.ID2
UNION ALL
SELECT DISTINCT
e1.ID1
,e2.ID2
FROM expanded e1
INNER JOIN expanded e2 ON e1.ID2 = e2.ID1
WHERE e1.ID1 <> e2.ID2
)
SELECT DISTINCT
ID1, ID2
FROM recur
ORDER BY ID1, ID2
DROP TABLE #Source

Answer 4

This is a way to get that output by brute force, but may not be the best solution with a different/larger data set:
select sub1.rnk as ID1
,sub2.rnk as ID2
from
(
select a.*
,rank() over (partition by 1 order by id1, id2) as RNK
from source a
) sub1
cross join
(
select a.*
,rank() over (partition by 1 order by id1, id2) as RNK
from source a
) sub2
where sub1.rnk <> sub2.rnk
union all
select id1 as ID1
,id2 as ID2
from source
where id1 = 6
union all
select id2 as ID1
,id1 as ID2
from source
where id1 = 6;