Query with 2 level of breaks overlapings, one with a distinct constraint

Question

My boss asked me, for an Apex graph, to give for one day
every 10 min the number of UNIQUE people (id) connected in the last 30 minutes.
It is because this Unique ID count that I cannot solve this, but I am convinced it can be done using analytical or model clause.
The folllwiing test case perfectly emulate the set of data as they comes from Websphere logs:
--  We want :
--    One row every 10 minutes that gives the count of UNIQUE  ID
--        connected in the last 30 min before this point
-- in this test case we assume that at 08h00 one user connects
--     every second during one hour.
with v as (
  select trunc(dbms_random.value(1,2000)+1) id,
       trunc(sysdate) + 80/24 + level/86400 sec from           <- this is  a typo, I wanted 8/24
  dual connect by level < 3600
select id, to_char(sec,'YYYY-MM-DD HH24:MI:SS') from v
      ID   Time
       568 2011-07-23 08:48:19
      1057 2011-07-23 08:48:20
      1864 2011-07-23 08:48:21
      1500 2011-07-23 08:48:22
       917 2011-07-23 08:48:23
      1844 2011-07-23 08:48:24
       779 2011-07-23 08:48:25
       448 2011-07-23 08:48:26
       382 2011-07-23 08:48:27
       256 2011-07-23 08:48:28
      1821 2011-07-23 08:48:29
       872 2011-07-23 08:48:30
       658 2011-07-23 08:48:31
.Edited by: bpolarski on 20 juil. 2011 12:10 added typo comment

Answer 1

Hi Solomon
I don't think the numbers are correct. There are 1200 rows before 8:20, but your query is showing 1569 distinct sessions.
with v as (
           select  --+materialize
                    trunc(dbms_random.value(1,2000)+1) id,
                   trunc(sysdate) + 80/24 + level/86400 sec
             from  dual connect by level < 3600
select
    count(*),
    count(distinct id)
FROM
    v where sec <= TO_DATE('23/07/2011 08:20:00','dd/mm/yyyy hh24:mi:ss')
SQL> with v as (
  2             select  --+materialize
  3                      trunc(dbms_random.value(1,2000)+1) id,
  4                     trunc(sysdate) + 80/24 + level/86400 sec
  5               from  dual connect by level < 3600
  6            )
  7  select
  8      count(*),
  9      count(distinct id)
10  FROM
11      v where sec <= TO_DATE('23/07/2011 08:20:00','dd/mm/yyyy hh24:mi:ss')
12  /
  COUNT(*) COUNT(DISTINCTID)
      1200               910I know this might not be the neatest way to get the result, but I think this is the correct one. I tried with analytics but I kept hitting problems with the ranges...
with v as (
           select  trunc(dbms_random.value(1,2000)+1) id,
                   trunc(sysdate) + 80/24 + level/86400 sec
             from  dual connect by level < 3600
SELECT
    sec_10_min,
    (   SELECT
            count(distinct l.id)
        FROM
            v l
        WHERE
            l.sec >= sec_10_min - NUMTODSINTERVAL(30,'MINUTE')
        AND
            l.sec <= sec_10_min
     ) ct
FROM
    (   SELECT
            sec_10_min
        FROM
                select  trunc(sec,'HH') + trunc(to_char(sec,'MI') / 10) / 144 sec_10_min
                from  v
                WHERE
                    rownum > 0
        group by
            sec_10_min
  order by sec_10_min
SQL> with v as (
  2             select  trunc(dbms_random.value(1,2000)+1) id,
  3                     trunc(sysdate) + 80/24 + level/86400 sec
  4               from  dual connect by level < 3600
  5            )
  6  SELECT
  7      sec_10_min,
  8      (   SELECT
  9              count(distinct l.id)
10          FROM
11              v l
12          WHERE
13              l.sec >= sec_10_min - NUMTODSINTERVAL(30,'MINUTE')
14          AND
15              l.sec <= sec_10_min
16       ) ct
17  FROM
18      (   SELECT
19              sec_10_min
20          FROM
21              (
22                  select  trunc(sec,'HH') + trunc(to_char(sec,'MI') / 10) / 144 sec_10_min
23                  from  v
24                  WHERE
25                      rownum > 0
26              )
27          group by
28              sec_10_min
29      )
30    order by sec_10_min
31  /
SEC_10_MIN                  CT
23/07/2011 08:00:00          0
23/07/2011 08:10:00        511
23/07/2011 08:20:00        897
23/07/2011 08:30:00       1176
23/07/2011 08:40:00       1179
23/07/2011 08:50:00       1188HTH
David
Edited by: Bravid on Jul 20, 2011 2:43 PM
Edited by: Bravid on Jul 20, 2011 2:48 PM
Updated NLS to show times in the output