Question

I have a hierarchy table with the following data :

SOURCE  TARGET  Level   ID
0       1       1       1
0       2       1       2
2       3       2       3
2       4       2       4
2       5       2       5
1       3       2       6
1       4       2       7
1       5       2       8
5       3       3       9
5       3       3       10
4       3       3       11
4       3       3       12
3       6       3       13
3       6       3       14
3       6       4       15
3       6       4       16
3       6       4       17
3       6       4       18

The SOURCE and TARGET rows are the original data and are used to connect between parents and children. for example, the third row (SOURCE 2, TARGET 3 on LEVEL 2) connects to the second row (SOURCE 0, TARGET 2 on LEVEL 1) since the Source of the first equals the target of the second.
The ID column is added at the end using a ROW_NUMBER function and is used to give each row a unique ID.
It may be easier to understand if SOURCE is replaced with PARENT and TARGET with CHILD.

I join the table to itself in order to find the “parent”.
I want each “instance” of a “source” on each level to connect to one of its parents. It’s not important which ones connect but all need to be connected and to different parents.

The final results should look something like this:

SOURCE  TARGET  Level   ID   P_ID
0       1       1       1    NULL
0       2       1       2    NULL
2       3       2       3    2
2       4       2       4    2
2       5       2       5    2
1       3       2       6    1
1       4       2       7    1
1       5       2       8    1
5       3       3       9    5
5       3       3       10   8
4       3       3       11   4
4       3       3       12   7
3       6       3       13   3
3       6       3       14   6
3       6       4       15   9
3       6       4       16   10
3       6       4       17   11
3       6       4       18   12

Any suggestions on how to write a good ms-sql query for this?

Answer 1

Link to sample data and SQL Fiddle

The query to use is below.

;with cte as (
  select *,rn=row_number() over (partition by level, target
                                 order by id),
           lc=count(1) over (partition by level, target)
  from tbl
)

select a.*, b.id as parent_id
from cte a
left join cte b on b.level=a.level-1
               and b.target=a.source
               and b.rn=(a.rn-1)%b.lc+1
order by id

1. Items are sequenced at each level/target combination
2. Children are linked to parents using by sequence, however if there are more children than parents, the MOD (%) operator takes care of going back to the first parent and continues distribution