Complex Query on a huge table

Question

Hi,
I have a table with around 100 million records. This table has 3 columns.
Customer_id,
start_date
end_date
Customer ID may be repeated with different ranges of start_date and end_date. Below is an example
Customer ID Start_Date End_Date
1 01-JAN-2013 31-MAR-2013
1 01-APR-2013 31-MAY-2013
1 01-JUL-2013 31-DEC-2013
2 01-MAY-2013 NULL
2 01-JAN-2013 NULL
I need to create another table with the below output
Customer ID Start_Date End_Date
1 01-JAN-2013 31-MAY-2013 ---This got merged as the end date (31-MAR-2013) and start date (01-APR-2013) can be merged to 1 record.
1 01-JUL-2013 31-DEC-2013 --- This did not get merged as end date (31-MAY-2013) and start date (01-JUL-2013) have a gap
2 01-JAN-2013 NULL --- Here null to be treated as an active record, so we pick the MIN of the dates.
Is thr a way to do this using a query. I have a pl-sql script where I have added this logic.
Sampel data and table structure provided. Thanks
create table xx_temp (customer_id number,
start_date date,
end_date);
insert into xx_temp values (1,to_date('01012013','ddmmyyyy'), to_date('31032013','ddmmyyyy'));
insert into xx_temp values (1,to_date('01042013','ddmmyyyy'), to_date('31052013','ddmmyyyy'));
insert into xx_temp values (1,to_date('01072013','ddmmyyyy'), to_date('31122013','ddmmyyyy'));
insert into xx_temp values (2,to_date('01012013','ddmmyyyy'), null);
insert into xx_temp values (2,to_date('01052013','ddmmyyyy'), null);

Answer 1

Hi.
{message:id=10947774} shows how to do that.
Use
NVL ( end_date
     , DATE '9999-12-31'
     )to equate NULL end_dates with an impossibly late date.
Edited by: Frank Kulash on Apr 7, 2013 7:01 AM
I see you added CREATE TABLE and INSERT statements, so I can test it. Thanks.
Here's what you requested:
CREATE TABLE     another_table
AS
WITH     got_new_grp     AS
     SELECT     customer_id, start_date, end_date
     ,     CASE
              WHEN  start_date > 1 + MAX (end_date) OVER
                                         ( PARTITION BY  customer_id
                                  ORDER BY      start_date
                                  RANGE BETWEEN UNBOUNDED PRECEDING
                                          AND     1      PRECEDING
              THEN  1
              ELSE  0
          END          AS new_grp
     FROM     xx_temp
--     WHERE     ...     -- if you need any filtering, put it here
,     got_grp          AS
     SELECT  customer_id, start_date, end_date
     ,     SUM (new_grp) OVER ( PARTITION BY  customer_id
                                   ORDER BY          start_date
                       )         AS grp
     FROM    got_new_grp
SELECT       customer_id
,       MIN (start_date)  AS start_date
,       MAX (end_date)    AS end_date
FROM       got_grp
GROUP BY  customer_id
,            grp
;There's no need to equate NULLs with an impossibly late date in this case. Ignore what I said about NVL.
Since you consider an end_date of March 31, 2013 to overlap with a start_date of 1 April, 2013, I changed
start_date >= MAX (end_date) ....from the earlier thread to
start_date > 1 + MAX (end_date) ....above.