Record counts by hour,day - report

Question

Record count, grouped by day, hour
Intended output:
DATE | 00:00 | 01:00 | 02:00 | .....
07-21-2011 | 100 | 200 | 300
07-22-2011 | 500 | 0 | 200
The below returns a version, but it doesn't account for when there are no records for a given hour.
I've created a successful script, but it involves too many subqueries, and is an abmonination
Any help would be awesome!
(any table with a date field will work)
with
hrly_ttls as
select to_char(sdate,'YYYY-MM-DD') as dte, to_char(sdate,'HH24') as hr,
count(*) as cnt from df_audit_batches
group by to_char(sdate,'YYYY-MM-DD'),to_char(sdate,'HH24')
select dte,
max(ltrim(sys_connect_by_path(cnt,' '),' ')) keep
(dense_rank last order by level) as Ttls
from (
select hrly_ttls.*, row_number() over(partition by dte order by hr) as rn
from hrly_ttls
start with rn = 1
connect by prior dte = dte
and prior rn = rn - 1
group by dte;
==> returns
DTE TTLS
2011-07-01     2 1 1 1
2011-07-02     1 2 1 1

Answer 1

Hi,
Welcome to the forum!
Whenever you post a question, please include CREATE TABLE and INSERT statements for some sample data. That way, people who want to help you can re-create the problem and test their ideas. Also post the results you want from that data. Simplify if possible. For example, if you need a query for all 24 hours in a day, you can post a question that only has data and results for maybe 3 hours. Adapting the result should be easy.
Always say which version of oracle you're using.
Since this is your first post, I'll post some sample data for you:
CREATE TABLE     df_audit_branches
(     sdate     DATE
INSERT INTO df_audit_branches (sdate) VALUES (TO_DATE ('21-Jul-2011 00:00', 'DD-Mon-YYYY HH24:MI'));
INSERT INTO df_audit_branches (sdate) VALUES (TO_DATE ('21-Jul-2011 00:01', 'DD-Mon-YYYY HH24:MI'));
INSERT INTO df_audit_branches (sdate) VALUES (TO_DATE ('21-Jul-2011 02:02', 'DD-Mon-YYYY HH24:MI'));
INSERT INTO df_audit_branches (sdate) VALUES (TO_DATE ('22-Jul-2011 01:40', 'DD-Mon-YYYY HH24:MI'));And the results you want from that data might look like this:
DAY         TXT
21-Jul-2011    2   0   1
22-Jul-2011    0   1   0In the real problem, of course, you would need 24 "columns" in txt, not just 3.
To include all hours, even if there is no data for a particular hour, do an outer join to a "table" (I used a sub-query, all_hours, not a real table) that has one row per hour.
Here's one way to do that:
WITH     got_cnt          AS
     SELECT       TRUNC (sdate)               AS day
     ,       TO_CHAR (sdate, 'HH24')     as hour
     ,       COUNT (*)               AS cnt
     FROM       df_audit_branches
     GROUP BY  TRUNC (sdate)
     ,       TO_CHAR (sdate, 'HH24')
,     all_hours     AS
     SELECT     LEVEL               AS n
     ,     TO_CHAR ( LEVEL - 1
               , 'FM00'
               )          AS hour
     FROM     dual
     CONNECT BY     LEVEL <= 3     -- or 24, for the real query
SELECT       c.day
,       REPLACE ( SYS_CONNECT_BY_PATH ( TO_CHAR ( NVL (c.cnt, 0)
                                , '999'
            )          AS txt
FROM            all_hours     h
LEFT OUTER JOIN     got_cnt          c       PARTITION BY (day)
                         ON     c.hour     = h.hour
WHERE     CONNECT_BY_ISLEAF     = 1
START WITH     h.n     = 1
CONNECT BY     h.n     = PRIOR h.n + 1
     AND     c.day     = PRIOR     c.day
ORDER BY  c.day
;This will work in Oracle 10.1 (and up).