Checking Frequency of a record

Question

I have a table that has information about customer payments per day. Now I want to extract the information on which clients made a payment everyday, for the past month. I have an oracle 11g Database.
Edited by: user12184158 on Jul 26, 2012 2:07 AM

Answer 1

user12184158 wrote:
Below is the table structure, sample data and expected output. Note that only custid 300 is in output since he contributed something everyday i.e on the 1 and 2 and 3So you don't want every day in the month, only the days that are actaully in the table, related to some custid.
For example, custid='400' has no data for June 2, but other custids (such as '300') do have data for June 2, therefore you don't want custid='400' in the results.
Custid='300' has no data for June 4, but that's okay, because no other custid has data for June 4, either.
Do I have it right?
CREATE TABLE customerpaytab
SERIALNUMBER NUMBER,
BATCHNUMBER  NUMBER,
CUSTID       VARCHAR2(20 BYTE),
paiddate DATE,
AMOUNT NUMBER
insert into customerpaytab values (55,2,'300','1-Jun-2012',5); ...
Paiddate is a DATE, so don't try to insert a VARCHAR2 value, such as '1-Jun-2012' into it. Use TO_DATE or a DATE literal instead.
sample output
custid,sum(amount)
300,15Here's one way:
WITH     alldates     AS
     SELECT DISTINCT
             TRUNC (paiddate)                    AS adate
     ,          COUNT (DISTINCT TRUNC (paiddate)) OVER ()     AS datecnt
     FROM        customerpaytab
     WHERE     paiddate     >= ADD_MONTHS ( TRUNC (SYSDATE, 'MONTH'), -1)
     AND     paiddate     <               TRUNC (SYSDATE, 'MONTH')
SELECT       p.custid
,       SUM (p.amount)     AS totalamount
FROM       customerpaytab  p
JOIN       alldates       a  ON  a.adate  = TRUNC (p.paiddate)
GROUP BY  p.custid
HAVING       COUNT (DISTINCT TRUNC (p.paiddate))     = MIN (a.datecnt)
;If the time of paiddate is always 00:00:00, then you can say "paiddate" instead of "TRUNC (paiddate)".