Grouping to be done by comparing 2 rows in a table

Question

Hi
I'm using oracle 10g.
I have a table with 3 columns
id--------origin--------destination
1---------N1-----------N3
2---------N2-----------N4
3---------N2-----------N4
4---------N1-----------N3
5---------N2-----------N6
6---------N1-----------N3
7---------N2-----------N6
if there are common id's contains the same origin and same destination then we need to group it like
id--------origin--------destination
1---------N1-----------N3-----G1
2---------N2-----------N4-----G2
3---------N2-----------N4-----G2
4---------N1-----------N3-----G1
5---------N2-----------N6-----G3
6---------N1-----------N3-----G1
7---------N2-----------N6-----G3
please help me
thanks
sandeep

Answer 1

Hi,
Whenever you have a problem, please post CREATE TABLE and INSERT statements for the sample data. You'll get better answers sooner.
What role (if any) does main_group play in this problem? Should all the rows with the same new group also have the same main_group? If so:
WITH     got_group_flag     AS
     SELECT     main_group, id, start_date, end_date
     ,     CASE
              WHEN  start_date <= MAX (end_date) OVER ( PARTITION BY  main_group
                                                               ORDER BY      start_date
                                         ,                  id
                                         ROWS BETWEEN UNBOUNDED PRECEDING
                                                 AND        1         PRECEDING
              THEN  0
              ELSE  1
          END     AS group_flag
     FROM    table_x
--     WHERE     ...          -- If you need any filtering, do it here
SELECT       main_group, id, start_date, end_date
,       'G' || SUM (group_flag) OVER ( ORDER BY      main_group
                                            ,             start_date
                          ,                 id
                           ) AS new_group
FROM       got_group_flag
ORDER BY  main_group
,            start_date
;As you can see, GROUP_FLAG = 1 if a new_group starts with the current row, when sorted by start_date and id. If start_date is not unique, then we need something like id in the ordering, to make sure that the order in SUM (in the main query) will be the same as the order used in MAX (in the sub-query).
Once we know where the new_groups start, finding to which new_group each row belongs is just a matter of adding up all the starting points so far.
If main_group is not a factor in determining the new_group, then omit references to main_group in both analytic clauses.
This assumes that id is unique, and that start_date <= end_date on each row.
This does NOT assume anything else about the dates, e.g., that start_date or end_date is unique, or that, if row A has an earlier start_date than row B, then A will have an earlier end_date, too.
Edited by: Frank Kulash on Feb 15, 2011 2:23 PM
Added explanation.