Query for Finding the Date History

Question

Hi Geeks,
My Table looks like this
Employee Assignment Position     Job     Grade     Start_date     End_date
101 132 2055 Clerk C01 10-Mar-00 11-Feb-01
101 132 2055 Clerk C02 12-Feb-01 15-Dec-01
101 132 2056 Clerk C02 16-Dec-01 22-Jul-03
101 132 2055 Clerk C03 23-Jul-03 31-Dec-4000 --(unassigned)
Now I need output like
Employee Assignment Position Start_date End_date
101 132 2055 10-Mar-2000 15-Dec-2001
101 132 2056 16-Dec-2001 22-Jul-2003
101 132 2055 23-Jul-2003 31-Dec-4000 --(unassigned)
The following important things are to be considered
1. Rows to be grouped by position
2. Each row much have the start date and end date of the employee position. If the employee is the same postion and different grade or job that doesnt matter. But if the employee jumps from his position (2055) to 2056 and after sometime he comes to the same position (i.e. 2055) the date history should be separate.
I am unable to derive the logic also. I think we need to create a function for this.
Kindly help me out.
Thanks in Advance
Punithavel
Message was edited by:
Punithavel
Message was edited by:
Punithavel

Answer 1

Too difficult to read. I have no database access to test (usually I use a step by step approach, proceeding with the next step when the current one is OK).
I can describe a general idea (may post something afterwards with no promise it will work)
step 1: mark the position changes
step 2: retain only mark 1 and mark max(mark) of each group
step 3: make a single row from the two rows of each group
Regards
Etbin
with
marking as
(select employee,assignment,position,start_date,end_date,
        row_number() over (partition by employee,assignment,position order by start_date) the_mark
   from the_table
collecting as
(select employee,assignment,position,start_date,end_date,the_mark
   from marking m
  where the_mark = 1
     or the_mark = (select max(the_mark)
                      from marking
                     where employee = m.employee
                       and assignment = m.assignment
                       and position = m.position
                     group by employee,assignment,position
                     having count(*) > 1
combining as
(select employee,assignment,position,
        case when the_mark = 1 then start_date end start_date,
        case when the_mark = 1
             then case when lead(position,1) over (partition by employee,assignment,position order by start_date) = position
                       then lead(end_date,1) over (partition by employee,assignment,position order by start_date)
                       else end_date
                  end
        end end_date
   from collecting
select employee,assignment,position,start_date,end_date
  from combining
where start_date is not null
order by employee,assignment,position,start_date*** not tested ***
Message was edited by: Etbin
user596003
on a lazy afternoon I took a look at your friend's solution too and I like it.
It's more concise than mine (defining starts and ends).
Not being able to test it the only problem in his solution might be the absence of order by clauses to assign correct rownums
WITH lag_and_lead AS
(SELECT employee,assignment,position,start_date,end_date,
        LAG (position) OVER (PARTITION BY employee,assignment ORDER BY start_date) AS lag_position,
        LEAD (position) OVER (PARTITION BY employee,assignment ORDER BY start_date) AS lead_position
   FROM your_table
  ORDER BY employee,assignment,start_date
starts AS
(SELECT employee,assignment,position,start_date,ROWNUM rn
   FROM lag_and_lead
  WHERE position <> lag_position
     OR lag_position IS NULL
  order by employee,assignment,start_date
ends AS
(SELECT employee,assignment,position,end_date,ROWNUM rn
   FROM lag_and_lead
  WHERE position <> lead_position
     OR lead_position IS NULL
  order by employee,assignment,start_date
SELECT starts.employee,starts.assignment,starts.position,starts.start_date,ends.end_date
  FROM starts,ends
WHERE starts.employee = ends.employee
   AND starts.assignment = ends.assignment
   AND starts.position = ends.position
   AND starts.rn = ends.rn
EMPLOYEE ASSIGNMENT POSITION JOB   GRADE START_DATE  END_DATE
     101        132     2055 Clerk C01   10-Mar-2000 11-Feb-2001
     101        132     2055 Clerk C02   12-Feb-2001 15-Dec-2001
     101        132     2056 Clerk C02   16-Dec-2001 22-Jul-2003
     101        132     2055 Clerk C03   23-Jul-2003 31-Dec-4000
lag_and_lead
EMPLOYEE ASSIGNMENT POSITION START_DATE  END_DATE    lag_position lead_position
     101        132     2055 10-Mar-2000 11-Feb-2001         null          2055
     101        132     2055 12-Feb-2001 15-Dec-2001         2055          2056
     101        132     2056 16-Dec-2001 22-Jul-2003         2055          2055
     101        132     2055 23-Jul-2003 31-Dec-4000         2056          null
starts
EMPLOYEE ASSIGNMENT POSITION START_DATE  rn
     101        132     2055 10-Mar-2000  1
     101        132     2056 16-Dec-2001  2
     101        132     2055 23-Jul-2003  3
ends
EMPLOYEE ASSIGNMENT POSITION END_DATE    rn
     101        132     2055 15-Dec-2001  1
     101        132     2056 22-Jul-2003  2
     101        132     2055 31-Dec-4000  3
EMPLOYEE ASSIGNMENT POSITION START_DATE  END_DATE
     101        132     2055 10-Mar-2000 15-Dec-2001
     101        132     2056 16-Dec-2001 22-Jul-2003
     101        132     2055 23-Jul-2003 31-Dec-4000Regards
Etbin
Message was edited by: Etbin
user596003