如何编写这个SQL来链接DB2中的数据？

Question

For example, there are two columns - ID1 and ID2 which are linked to each other. It is possible that the IDs are linked with other IDs in the table.

ID1         ID2
------------------
001         002
001         003
004         005
002         006
005         007

In the above table, 001, 002, 003, 006 are linked and 004, 005, 007 are linked.

Is it possible to query this information in SQL for DB2?

The format is similar as the following:

Group       ID
--------------
1           001
1           002
1           003
1           006
2           004
2           005
2           007

On the other hand, if one more record (008, 007) is added to the table

ID1         ID2
------------------
001         002
001         003
004         005
002         006
005         007
008         007  (Newly added)

The expected result will be:

Group       ID
--------------
1           001
1           002
1           003
1           006
2           004
2           005
2           007
2           008

because 004, 005, 007, 008 are linked.

The DB2 Version is 9.7.

Answer 1

Sure you can! It requires a recursive query:

WITH Recur (grp, root, leaf) as (SELECT ROW_NUMBER() OVER(ORDER BY root.id1), 
                                        CAST(NULL as CHAR(3)), 
                                        root.id1
                                 FROM Linked as root
                                 EXCEPTION JOIN Linked as leaf
                                 ON leaf.id2 = root.id1
                                 GROUP BY root.id1
                                 UNION ALL 
                                 SELECT grp, leaf, id2
                                 FROM Recur 
                                 JOIN Linked
                                 ON id1 = leaf)

SELECT grp, leaf 
FROM Recur
ORDER BY grp, leaf

(Tested on my local iSeries, and have a working SQL Fiddle example , which has to use the LEFT JOIN -style exception to work in SQL Server)

Yields the expected output:

grp   leaf
=============
1     001 
1     002 
1     003 
1     006 
2     004 
2     005 
2     007