GROUP BY generating addition time ranges

Question

Hello guys,
I have a little problem that I've been working on since this morning and I can't find any solution... So, I'm asking for your help!
I'm using Oracle 10g, and I'm trying to have a full calendar even when records are missing in the base table. For example, let's take this scenario:
CREATE TABLE requests
  code VARCHAR2(100),
  request_time DATE
INSERT INTO requests(code, request_time) VALUES('A', TO_DATE('01/02/2012 13:50', 'DD/MM/YYYY HH24:MI'));
INSERT INTO requests(code, request_time) VALUES('A', TO_DATE('01/02/2012 16:35', 'DD/MM/YYYY HH24:MI'));
INSERT INTO requests(code, request_time) VALUES('A', TO_DATE('01/02/2012 19:11', 'DD/MM/YYYY HH24:MI'));
INSERT INTO requests(code, request_time) VALUES('B', TO_DATE('01/02/2012 19:12', 'DD/MM/YYYY HH24:MI'));
INSERT INTO requests(code, request_time) VALUES('B', TO_DATE('01/02/2012 23:12', 'DD/MM/YYYY HH24:MI'));
INSERT INTO requests(code, request_time) VALUES('C', TO_DATE('03/02/2012 10:05', 'DD/MM/YYYY HH24:MI'));It's a simple table with a code and a request time. Every time a request is processed by a server, we insert a record in that table. What I would like to produce is something like this:
For every day where a request has been processed, I should have a result like this:
A  01/02/2012 00:00 NULL
A  01/02/2012 01:00 NULL
A  01/02/2012 02:00 NULL
A  01/02/2012 03:00 NULL
A  01/02/2012 13:00 1
A  01/02/2012 14:00 NULL
A  01/02/2012 15:00 NULL
A  01/02/2012 16:00 1
A  01/02/2012 17:00 NULL
A  01/02/2012 18:00 NULL
A  01/02/2012 19:00 1
A  01/02/2012 20:00 NULL
A  01/02/2012 21:00 NULL
A  01/02/2012 22:00 NULL
A  01/02/2012 23:00 NULL  -- 24 records per code and per day with requests.
-- same for B
B  01/02/2012 00:00 NULL
B  01/02/2012 01:00 NULL
B  01/02/2012 02:00 NULL
B  01/02/2012 19:00 1
B  01/02/2012 20:00 NULL
B  01/02/2012 23:00 NULL  -- 24 records for the 01/02
-- no request the 02/02 so no records for that day
C  03/02/2012 00:00 NULL
C  03/02/2012 01:00 NULL
C  03/02/2012 10:00 1
C  03/02/2012 22:00 NULL
C  03/02/2012 23:00 NULLI should have 24 records per code (by day) where a request has been logged. So, if we have 2 different codes, I should have 48 records. The problem is that I don't have records all day long. Sometimes, some ranges are empty and I need these empty ranges in the final result.
Anyone can help?? Honestly, I tried but I can't find any solution...
Thanks,

Answer 1

Hi,
user13117585 wrote:
... For example, let's take this scenario:
CREATE TABLE requests
code VARCHAR2(100),
request_time DATE
INSERT INTO requests(code, request_time) VALUES('A', TO_DATE('01/02/2012 13:50', 'DD/MM/YYYY HH24:MI')); ...
Thanks for including the CREATE TABLE and INSERT statements; that's very helpful!
... The problem is that I don't have records all day long...Exactly! You want to display times that aren't in the requests table. You have to SELECT them from some table (or result set, but I'll just say "table"). What table will that be? You can derive such a table from requests. This is called a Counter Table because it "counts" 0, 1, 2, 3, ..., x, up as high as you need. In this case, you don't actually need numbers 0, 1, 2, ..., x; you need DATEs b + 0 hours, b + 1 hour, b + 2 hours, b + 3 hours, ..., b + x hours, where b is a base DATE (midnight of the first date in your table), and x is 24 times the number of days.
Here's one way to do that:
WITH   all_times        AS
     SELECT     start_time + ( (LEVEL - 1)
                    / 24
                    )          AS a_time
     ,     start_time + ( LEVEL
                    / 24
                    )          AS next_time
     FROM     (
              SELECT  TRUNC (MIN (request_time))          AS start_time
              ,         TRUNC (MAX (request_time)) + 1     AS end_time
              FROM    requests
     CONNECT BY     LEVEL     <= 24 * (end_time - start_time)
SELECT       r.code
,       a.a_time
,       NULLIF ( COUNT (r.request_time)
             , 0
           )     AS request_cnt
FROM           all_times  a
LEFT OUTER JOIN      requests   r  PARTITION BY (r.code)
                               ON  r.request_time  >= a.a_time
                               AND r.request_time  <  a.next_time
GROUP BY  r.code
,            a.a_time
ORDER BY  r.code
,            a.a_time
;Edited by: Frank Kulash on Aug 3, 2012 5:41 AM
Added explanation