Numbering lines based on two columns

Question

Hi Everyone,
I´d like to know if there is a way to achieve the numbering presented on the second table below with only native Oracle functions, like ROW_COUNT() over partition, etc.
I´m using Oracle 10g.
The logic used is:
Starting from 1, increment one every time the ORIGIN is the same as the FIRST ORIGIN of the group of rows (ID).
ID    ORIGIN    DESTINATION    ORDER
1     A         B              1
1     B         A              2
1     A         B              3
1     B         C              4
1     C         A              5
ID     ORIGIN    DESTINATION    ORDER    NUMBERING
1      A         B              1        1
1      B         A              2        1
1      A         B              3        2
1      B         C              4        2
1      C         A              5        2In order to compare the ORIGIN of each row with the FIRST ORIGIN of the group I used the LAG function to create a column that will have the FIRST ORIGIN of the group value.
However I was not able to number the lines as shown above (column NUMBERING).
Any help will be much appreciated.
Test query:
WITH T AS
  SELECT 1 ID, 'A' ORIGIN, 'B' DESTINATION, 1 ORDERING FROM DUAL UNION ALL
  SELECT 1 ID, 'B' ORIGIN, 'A' DESTINATION, 2 ORDERING FROM DUAL UNION ALL
  SELECT 1 ID, 'A' ORIGIN, 'B' DESTINATION, 3 ORDERING FROM DUAL UNION ALL
  SELECT 1 ID, 'B' ORIGIN, 'C' DESTINATION, 4 ORDERING FROM DUAL UNION ALL
  SELECT 1 ID, 'C' ORIGIN, 'A' DESTINATION, 5 ORDERING FROM DUAL
  SELECT T.ID
       , T.ORIGIN
       , T.DESTINATION
       , T.ORDERING
       , LAG (T.ORIGIN, T.ORDERING -1, 0) OVER (PARTITION BY T.ID
                                                    ORDER BY T.ID
                                                           , T.ORDERING) FIRST_ORIGIN_OF_GROUP
    FROM T
ORDER BY T.ID
       , T.ORDERING

Answer 1

Hi,
Here's one way:
WITH     got_first_origin     AS
     SELECT     id, origin, destination, ordering
     ,     FIRST_VALUE (origin) OVER ( PARTITION BY  id
                                      ORDER BY         ordering
                                 ) AS first_origin
     FROM    t
SELECT     id, origin, destination, ordering
,     COUNT ( CASE
                    WHEN  origin = first_origin
              THEN  1
                END
           )     OVER ( PARTITION BY  id
                       ORDER BY      ordering
               ) AS numbering
FROM     got_first_origin
;This assumes that the combination of id and ordering is unique. Within an id, ordering does not have to be consecutive integers, or anything like that.
Analytic functions can't be nested (the argument to the anlytic COUNT function can't call the analytic FIRST_VALUE function); that's why the sub-query is necessary.
You could do something with LAG, like you tried, rather than FIRST_VALUE, but you'd still need a sub-query, for the same reason.