Split dates into different shifts.

Question

Hello all,
I'm trying to setup a query to get my date ranges into 8 hour shifts which start at 23:00, 07:00, and 15:00.
create table t (start_date, end_date)
insert into t values (to_date('11-21-12 21:02:15', 'MM-dd-yy HH24:MI:ss'), to_date('11-21-12 21:02:51', 'MM-dd-yy HH24:MI:ss'));
insert into t values (to_date('11-21-12 21:09:06', 'MM-dd-yy HH24:MI:ss'), to_date('11-24-12 23:30:18', 'MM-dd-yy HH24:MI:ss'));
insert into t values (to_date('11-24-12 23:35:28', 'MM-dd-yy HH24:MI:ss'), to_date('11-24-12 23:49:34', 'MM-dd-yy HH24:MI:ss'));Desired output:
START_DATE        | END_DATE
11-24-12 23:35:28 | 11-24-12 23:49:34
11-24-12 23:00:00 | 11-24-12 23:30:18
11-24-12 15:00:00 | 11-24-12 23:00:00
11-24-12 07:00:00 | 11-24-12 15:00:00
11-23-12 23:00:00 | 11-24-12 07:00:00
11-23-12 15:00:00 | 11-23-12 23:00:00
11-23-12 07:00:00 | 11-23-12 15:00:00
11-22-12 23:00:00 | 11-23-12 07:00:00
11-22-12 15:00:00 | 11-22-12 23:00:00
11-22-12 07:00:00 | 11-22-12 15:00:00
11-21-12 23:00:00 | 11-22-12 07:00:00
11-21-12 21:09:06 | 11-21-12 23:00:00
11-21-12 21:02:15 | 11-21-12 21:02:51Version:
Oracle Database 11g Enterprise Edition Release 11.2.0.2.0 - 64bit Production
PL/SQL Release 11.2.0.2.0 - Production
Is this possible?
Thanks in advance!

Answer 1

Hi,
Tarianz wrote:
... How would I go about adding another column into this?
I've been messing around with the query trying to add it myself but have been unsuccessful.
Similar to:
*ID is the order the dates were inserted into the table.
ID| START_DATE        | END_DATE
3| 11-24-12 23:35:28 | 11-24-12 23:49:34
2| 11-24-12 23:00:00 | 11-24-12 23:30:18
2| 11-24-12 15:00:00 | 11-24-12 23:00:00 ...
That sounds like a job for the analytic ROW_NUMBER function.
ROW_NUMBER () OVER (ORDER BY x, y, z)     AS t_id   -- id is an Oracle keywordwhere x, y and z are the columns or expressions that indicate the order you want. You can have any number of such expressions (1 or more), and you can use <b>DESC</b>ending order if you want to.
If the combination (x, y, z) is not unique, that is, if 2 or more rows were added at the same time, then ROW_NUMBER will give them distinct numbers, but there's no telling which number will be lower.
You need to do this when first using table t, before the rows are split into shifts. In the query I suggested, you could add the expression above in sub-query got_shift_num.
So if the rows were INSERTed into the table in the same order as start_date, then you could do this:
WITH     got_base_date     AS
     SELECT     TRUNC (SYSDATE) + (7 / 24)     AS base_date
     FROM     dual
,     got_shift_num     AS
     SELECT     t.start_date, t.end_date
     ,     b.base_date
     ,     FLOOR ( (t.start_date - b.base_date)
                * 3
                )          AS start_shift_num
     ,     FLOOR ( (t.end_date - b.base_date)
                * 3
                )          AS end_shift_num
     ,     ROW_NUMBER () OVER (ORDER BY t.start_date)     -- ***** NEW *****
                       AS t_id                         -- ***** NEW *****
     FROM    got_base_date  b
     CROSS JOIN            t
,     cntr     AS
     SELECT  LEVEL - 1     AS n
     FROM     (
              SELECT  MAX ( end_shift_num
                            - start_shift_num
                    )     AS max_shift_cnt
              FROM    got_shift_num
     CONNECT BY     LEVEL <= 1 + max_shift_cnt
SELECT       s.t_id                              -- ***** NEW *****
,       GREATEST ( s.start_date
               , s.base_date + ( ( s.start_shift_num
                              + c.n
                       / 3
             )     AS shift_start_date
,       LEAST    ( s.end_date
               , s.base_date + ( ( s.start_shift_num
                              + c.n
                         + 1
                       / 3
             )     AS end_start_date
,       start_date, end_date          -- For debugging
FROM       got_shift_num  s
JOIN       cntr           c  ON c.n  <= s.end_shift_num - s.start_shift_num
ORDER BY  shift_start_date  DESC
;This is exactly what I posted yesterday, only with 3 lines, marked "***** NEW *****", added.
You could modify Solomon's suggestion, using ROW_NUMBER in sub-query a.