Use SQL to roll a schedule forward to a random date

Question

Hi,
oracle 10.2.0.4  Linux
We have two largeish  tables ( 10^5 rows in each) that track our workforces scheduled hours.  The first table creates generic schedules that show working days, each schedule must have 1-n rows in it, one for each day in the cycle.  The majority of our workforce are on a rolling 8 day cycle.  In the example below I have two shifts
Sched_4d that is a four day cycle of day-on day-off pairs.
Sched_3d this is a staggered full day, half day, day-off cycle.
From the information below you can see that in 1990 the sched_4d went from an 8 hour day to a 9 hour day.  There is no guarantee that SCHED_4D will not suddenly gain 2 extra days as part of this years union talks.
The second table is a simple assigment table showing when a person goes onto a schedule.
To work out a given day's schedule, you look at the EMP_SHIFT table to work out which schedule is "current", then you look at the date that the person was assigned to the schedule and assume that is SHIFT_ID 1, the next day is SHIFT_ID 2 until you complete the cycle and start again.
CREATE TABLE SCHED_DATA
  SCHED_ID     VARCHAR2(8 CHAR)             NOT NULL,
  ASOFDATE           DATE                          NOT NULL,
  SHIFT_ID        NUMBER(3)           NOT NULL,
  SCHED_HRS       NUMBER(4,2)                   NOT NULL
CREATE UNIQUE INDEX SCHED_DATA on SCHED_DATA (SCHED_ID, ASOFDATE, SHIFT_ID)
CREATE TABLE EMP_SHIFT
EMPID VARCHAR2(15 CHAR) NOT NULL,
ASOFDATE DATE NOT NULL,
SCHED_ID VARCHAR2(8 CHAR) NOT NULL
CREATE UNIQUE INDEX EMP_SHIFT on EMP_SHIFT ( EMPID, ASOFDATE, SCHED_ID)
INSERT INTO SCHED_DATA VALUES ('SCHED_4D', '01-JAN-1980',1,8);
INSERT INTO SCHED_DATA VALUES ('SCHED_4D', '01-JAN-1980',2,0);
INSERT INTO SCHED_DATA VALUES ('SCHED_4D', '01-JAN-1980',3,8);
INSERT INTO SCHED_DATA VALUES ('SCHED_4D', '01-JAN-1980',4,0);
INSERT INTO SCHED_DATA VALUES ('SCHED_4D', '01-JAN-1990',1,9);
INSERT INTO SCHED_DATA VALUES ('SCHED_4D', '01-JAN-1990',2,0);
INSERT INTO SCHED_DATA VALUES ('SCHED_4D', '01-JAN-1990',3,9);
INSERT INTO SCHED_DATA VALUES ('SCHED_4D', '01-JAN-1990',4,0);
INSERT INTO SCHED_DATA VALUES ('SCHED_3D', '01-JAN-1990',1,8);
INSERT INTO SCHED_DATA VALUES ('SCHED_3D', '01-JAN-1990',2,4);
INSERT INTO SCHED_DATA VALUES ('SCHED_3D', '01-JAN-1990',3,0);
INSERT INTO EMP_SHIFT VALUES ('001', '20-DEC-1989', 'SCHED_4D');
INSERT INTO EMP_SHIFT VALUES ('001', '01-JAN-1990', 'SCHED_4D');
INSERT INTO EMP_SHIFT VALUES ('001', '03-JAN-1990', 'SCHED_3D');
Given the above information, I need to write a select that receives 2 dates ( :FROM and :TO) and for all employees in EMP_SHIFT returns a row for each day and the relevant scheduled hours for that day. 
Thus the above data  with a from and to of '21-DEC-1989' : '05-JAN-1990' should return
EMPID, DATE, SCHED_HOURS
001, 21-DEC, 0
001, 22-DEC, 8
001, 23-DEC, 0
001, 24-DEC, 8
001, 25-DEC, 0
001, 26-DEC, 8
001, 27-DEC, 0
001, 28-DEC, 8
001, 29-DEC, 0
001, 30-DEC, 8
001, 31-DEC, 0
001, 01-JAN, 9
001, 02-JAN, 0
001, 03-JAN, 8
001, 04-JAN, 4
001, 05-JAN, 0
The employee started thir assignment to sched_4d on 20-DEC, so that would be SHIFT_ID 1.  Thus 21-DEC is SHIFT ID 2 which is 0 hours.  Cycle until the next event which is 01-JAN, when they are assigned to the new sched_4d which has them working 9 hours on day 1.  On 03-JAN they switch to the new cycle and go on the 8:4:0 rotation.
I can see how I could
SELECT EMPID, DAY_OF_INTEREST, SCHED_ID
FROM EMP_SHIFT A, (SELECT '01-JAN-1979' + rownum "DAY_OF_INTEREST" from dual connect by level <=10000)
WHERE A.ASOFDATE = (SELECT MAX(A1.ASOFDATE) FROM EMP_SHIFT A1 WHERE A1.EMPID = A.EMPID AND A1.ASOFDATE <= DAY_OF_INTEREST)
AND DAY_OF_INTEREST BETWEEN :FROM_DT AND :TO_DT
And I imagine I need to use some kind of MOD( (DAY_OF_INTEREST - EMP_SHIFT.ASOFDATE), (#number of days in shift) ) to work out which shift_id applies for a given Day_of_interest
But I am struggling to do this in a way that would scale to more than one employee,
Do any of the analytic functions achieve this in a neat way ?

Answer 1

Hi,
There are several analytic functions that might help here.  Both tables include starting dates only; we need to know ending dates for employee assignments as well as for schedules.  I used the analytic MIN and LEAD functions to get these in the query below.  Also, the query below needs to know how many days are in each schedule.  I used the analytic COUNT function for that.
WITH    params   AS
   SELECT  TO_DATE ('21-DEC-1989', 'DD_MON-YYYY')  AS start_date
   ,       TO_DATE ('05-JAN-1990', 'DD_MON-YYYY')  AS end_date
   FROM    dual
,       all_dates   AS
    SELECT  start_date + LEVEL - 1    AS a_date
    FROM    params
    CONNECT BY  LEVEL <= (end_date + 1) - start_date
,       sched_data_plus AS
    SELECT  sched_id, asofdate, shift_id, sched_hrs
    ,       NVL ( MIN (asofdate) OVER ( PARTITION BY  sched_id
                                        ORDER BY      asofdate
                                        RANGE BETWEEN 1         FOLLOWING
                                              AND     UNBOUNDED FOLLOWING
                                      ) - 1
                , TO_DATE ('31-DEC-9999', 'DD-MON-YYYY')
                )          AS uptodate
    ,       COUNT (*) OVER ( PARTITION BY  sched_id
                             ,             asofdate
                           )  AS days_in_sched
    FROM    sched_data
,       emp_shift_plus AS
    SELECT  empid, asofdate, sched_id
    ,       NVL ( LEAD (asofdate) OVER ( PARTITION BY  empid
                                         ORDER BY      asofdate
                                       ) - 1
                , TO_DATE ('31-DEC-9999', 'DD-MON-YYYY')
                )    AS uptodate
    FROM    emp_shift
SELECT    e.empid
,         d.a_date
,         s.sched_hrs
FROM      all_dates        d
JOIN      sched_data_plus  s  ON   d.a_date    BETWEEN s.asofdate
                                               AND     s.uptodate
JOIN      emp_shift_plus   e  ON   d.a_date    BETWEEN e.asofdate
                                               AND     e.uptodate
                              AND  e.sched_id  = s.sched_id
                              AND  MOD ( d.a_date - e.asofdate
                                       , s.days_in_sched
                                       ) + 1   = s.shift_id
ORDER BY  e.empid
,         d.a_date
This query starts by getting all the days of interest, that is, all the days between the given start- and end dates.  (When you said "random date", I assume you meant "given date", which may not involve any random element.)
Once we have a list of all the dates of interest, getting the results you want is just a matter of inner-joining to get which schedules were in effect on those dates, and which employees were assigned to those schedules on those dates.
If you're concerned about having more than 1 employee, maybe you should post sample data that involves more than 1 employee.
How do you handle employee terminations?  For example, what if employee 001 had quit on January 4, 1990?  I assume you'd want the output for that employee to stop on January 4, rather than continue to the end of the period of interest.
If you have termination dates in an employee table not shown here, then you can use that termination date instead of December 31, 9999 as the default end date for assignments.  If you have a special schedule for terminations (or leaves of absence, for that matter) you'll probably want to include a WHERE clause in the main query not to display dates after the employee left (or while the employee was on leave).