Getting only the newest records in a Grandchild table

Question

Hi
I have created a new table in the Siebel schema on an Oracle Database 10g Enterprise Edition Release 10.2.0.3.0 - 64bi
This table is called CX_ASSET_XM and is a Grandchild table of S_ASSET
CX_ASSET_XM has a parent table S_ASSET_X
So the structure is like so
S_ASSET  x 1 record
S_ASSET_X x 2 child records
CX_ASSET_XM x 1 grand child records
Therefore the cardinality is one Grandparent can have 2 Grandchild records
This CX_ASSET_XM table will hold information about TCP/IP calls of a Grandchild asset. These calls are not synchronized. therefore one Grandchild asset may make a call today but the other one has not made a call since yesterday.
In my query I find all the grand children records that have been created today (sysdate) and then traverse up the relationship to find the Grandparent, then I traverse back down to find the other twin in the Grandchildren tuple.
The problem with my query is that it returns every single TCP/IP call a pair has made but I only want the most recent callback of each grandchild to analyse
SELECT (cb.SSID), cb.created_by, cb.callback_date, cb.callback_num, asset.desc_text, asset.prom_integ_id, asset.integration_id
--Traversing Up to find the Grandparent record of the
FROM SIEBEL.CX_ASSET_XM cb
    JOIN SIEBEL.S_ASSET_X assetx
      ON cb.SSID = assetx.attrib_37
    JOIN SIEBEL.S_ASSET asset
      ON assetx.par_row_id = asset.row_id
WHERE asset.prom_integ_id IN
--Traversing Down to find the other twin in the pair
(   SELECT asset.prom_integ_id
    FROM SIEBEL.CX_ASSET_XM cb
    JOIN SIEBEL.S_ASSET_X assetx
      ON cb.ssid = assetx.attrib_37
    JOIN SIEBEL.S_ASSET asset
      ON assetx.par_row_id = asset.row_id
ORDER BY cb.callback_date DESC;
e.g. if there are 80 grandchildren which have made two TCP/IP calls that would make a total of 160 records, which is what the query above returns.
However I want 80 records to be returned
e
Grandchildren
Callback Date
1111AAAA (Pair 1)
10-OCT-13
1111AAAB (Pair 1)
09-OCT-13
2222AAAA (Pair 2)
10-OCT-13
2222AAAB (Pair 2)
07-OCT-13
As you can see from the table, atleast one of the pair will have a Callback_Date = Sysdate (Today) which is how I will need to find them in the table before traversing up.
How do I achieve these results. I have tried GROUP BY, DISTINCT, ROWNUM = 1

Answer 1

Hi,
968432 wrote:
Sorry for my ignorance but how do I incorporate the ROW_NUM function into my query, or will it be a new query from scratch.
When you say partition, does that mean the partition would be between my traversal up the tables to the grand parent and then back down again?
You can use ROW_NUMBER with something very much like your existing query; there's no need to start from scratch; you just won't repeat the joins in the WHERE clause.  You'll just be joining the tables once.  If it really helps you to think of traversing either up or down, then go ahead, but you mind find it simpler just to think of joining.  Every row of the join will include some expression(s) that uniquely identifies the grandparent; that's what you'll PARTITION BY.
It would be a lot easier to explain and to understand if we had some sample data and desired results.  Either post CREATE TABLE and INSERT statements for a little sample data, or phrase the question in terms of tables that anybody with Oracle probably has, such as those in the hr schema.  (People in hr.employees are related to deparatments in hr_departments, and departments are related to places in hr.locations, so location is the grandparent of employee.  If you knew how to find the most recently hired employee from each location in the hr schema, then you'd know how to solve your problem in your schema.  For example, if you wanted these results:
CITY                 DEPARTMENT_ID EMPLOYEE_ID LAST_NAME  HIRE_DATE
London                          40         203 Mavris     07-Jun-2002
Munich                          70         204 Baer       07-Jun-2002
Oxford                          80         167 Banda      21-Apr-2008
Seattle                        100         113 Popp       07-Dec-2007
South San Francisco             50         128 Markle     08-Mar-2008
Southlake                       60         104 Ernst      21-May-2007
Toronto                         20         202 Fay        17-Aug-2005
you could get them like this:
WITH  got_r_num  AS
    SELECT  l.city
    ,       d.department_id
    ,       e.employee_id, e.last_name, e.hire_date
    ,       ROW_NUMBER () OVER ( PARTITION BY  l.city
                                 ORDER BY      e.hire_date  DESC
                               )   AS r_num
    FROM    hr.employees    e
    JOIN    hr.departments  d  ON   d.department_id  = e.department_id
    JOIN    hr.locations    l  ON   l.location_id    = d.location_id
SELECT  city
,       department_id
,       employee_id, last_name, hire_date
FROM    got_r_num
WHERE   r_num = 1