Need help with special logic...

Question

Hi,
i have a query which selects all status changes for tasks. For every NEW status a record is inserted where the new status TYPE and ENTRY_DATE are stored.
A task can have several statuses based on this information (unlimited number of status changes is allowed):
1.) Open (the last inserted status record has type: 'Open')
2.) Closed (the last inserted status record has type 'Closed')
I need to get the following result:
TASK_ID     START_TIME        STOP_TIME          
333450      09-01-09 10:00     09-01-09 11:00    
333450      12-01-09 10:00     12-01-09 15:00    
333450      12-01-09 16:00     12-01-09 16:00    
333450      12-01-09 16:00     CURRENT_DATE    
Scenarios:
- Multiple 'Close' records after each other: The START_TIME in the 3d tuple should be copied from the STOP_TIME in the real data.
- No 'Close' record: The STOP_TIME in the 4th tuple should be the CURRENT_DATE.
- Multiple 'Open' records after each other: The STOP_TIME of the previous 'Open' record in the realdata should be the ENTRY_DATE of the 'current' record in the realdata. (THIS ONE ISN'T PRESENT IN THE DATA EXAMPLE!!!)
From this data:
with REALDATA as (
select 104993 PLAN_ID, to_date('09-01-09 10:00','dd-MM-yy HH24:MI') ENTRY_DATE, 'Open' STATUS_DESCRIPTION, 'Open' STATUS_TYPE, 333450 TASK_ID, 'Assembly' TASK_DESCRIPTION from dual
union
select 104993, to_date('09-01-09 11:00','dd-MM-yy HH24:MI'), 'Completed', 'Closed', 333450, 'Assembly' from dual
union
select 104993, to_date('12-01-09 10:00','dd-MM-yy HH24:MI'), 'Open', 'Open', 333450, 'Assembly' from dual
union
select 104993, to_date('12-01-09 15:00','dd-MM-yy HH24:MI'), 'Completed', 'Closed', 333450, 'Assembly' from dual
union
select 104993, to_date('12-01-09 16:00','dd-MM-yy HH24:MI'), 'Completed', 'Closed', 333450, 'Assembly' from dual
union
select 104993, to_date('13-01-09 15:00','dd-MM-yy HH24:MI'), 'Open', 'Open', 333450, 'Assembly' from dual
select TASK_ID, TO_CHAR(ENTRY_DATE,'dd-MM-yy HH24:MI'), STATUS_DESCRIPTION, STATUS_TYPE, TASK_DESCRIPTION from realdata order by ENTRY_DATE
Extra abstraction:_
So the data should be sorted by ENTRY_DATE:
then it should be 'grouped' like this (very abstract: tuples split by comma, between (brackets) is which ENTRY_DATE the data should come from)):
- normal pair
OPEN, CLOSED
result: START_TIME (OPEN) | STOP_TIME (CLOSED)
- multiple open records
OPEN, OPEN, OPEN, CLOSED
result: START_TIME (OPEN) | STOP_TIME (next OPEN), START_TIME (OPEN) | STOP_TIME (next OPEN), START_TIME (OPEN) | STOP_TIME (CLOSED)
- multiple closed records
OPEN, CLOSED, CLOSED
result: START_TIME (OPEN) | STOP_TIME (CLOSED), START_TIME (CLOSED) | STOP_TIME (CLOSED)
- no open record
CLOSED[,CLOSED]
result: START_TIME (CLOSED) | STOP_TIME (CLOSED)[,START_TIME (CLOSED) | STOP_TIME (CLOSED)]
- no closed record
OPEN
result: START_TIME (OPEN) | STOP_TIME (CURRENT_DATE)
- no closed record with multiple open records
OPEN, OPEN
result: START_TIME (OPEN) | STOP_TIME (next OPEN), START_TIME (OPEN) | STOP_TIME (CURRENT_DATE)Edited by: user574699 on Jan 14, 2009 5:09 AM
Edited by: user574699 on Jan 14, 2009 5:30 AM - Extra abstraction at the bottom!

Answer 1

Based on your original data, I'd go with something like:
with REALDATA as (select 104993 PLAN_ID, to_date('09-01-09 10:00','dd-MM-yy HH24:MI') ENTRY_DATE, 'Open' STATUS_DESCRIPTION, 'Open' STATUS_TYPE, 333450 TASK_ID, 'Assembly' TASK_DESCRIPTION from dual
                  union all
                  select 104993, to_date('09-01-09 11:00','dd-MM-yy HH24:MI'), 'Completed', 'Closed', 333450, 'Assembly' from dual
                  union all
                  select 104993, to_date('12-01-09 10:00','dd-MM-yy HH24:MI'), 'Open', 'Open', 333450, 'Assembly' from dual
                  union all
                  select 104993, to_date('12-01-09 15:00','dd-MM-yy HH24:MI'), 'Completed', 'Closed', 333450, 'Assembly' from dual
                  union all
                  select 104993, to_date('12-01-09 16:00','dd-MM-yy HH24:MI'), 'Completed', 'Closed', 333450, 'Assembly' from dual
                  union all
                  select 104993, to_date('13-01-09 15:00','dd-MM-yy HH24:MI'), 'Open', 'Open', 333450, 'Assembly' from dual),
   trans_data as (select task_id,
                         task_description,
                         plan_id,
                         entry_date,
                         nvl(lead(entry_date) over (partition by task_id, task_description, plan_id
                                                    order by entry_date), sysdate) stop_date,
                         status_type status,
                         lag(status_type) over (partition by task_id, task_description, plan_id
                                                order by entry_date) prev_status,
                         lead(status_type) over (partition by task_id, task_description, plan_id
                                                 order by entry_date) next_status
                  from   realdata)
select task_id,
       task_description,
       plan_id,
       entry_date,
       case when status = 'Closed' and next_status != 'Closed' then
                 entry_date
            else stop_date
       end
from   trans_data
where  case when status = 'Closed' and prev_status = 'Open' then
                 1
            else 0
       end != 1But it looks like your actual data is more complicated than the original set of data (extra statuses for one). Also, you haven't said what should happen (I don't think) in the case of "Open, Close, Open Close" - I've taken it that the 2nd Open should have a start time of the 2nd Open, rather than the stop time of the 1st close.