Group By with nested calculations

Question

Hi Gurus,
I have an Acct table with the follwig fields
Bal_date
Bal_amt
Acc_No
I want to calculate the average balance based on this formula:
Average=Amount * no_of_days/no_of_days_in_month
Where amount= Bal_amt (i.e. daily balance)
no_of_days = number of days the account has been holding that amount
Calculation for average negatives is done and reported separately
For instance if account 123456 has the following balance history:
01-JAN-09 to 08-JAN-09 = 1000
09-JAN-09 to 21-JAN-09 = -2000
22-JAN-09 to 31-JAN-09 = 3000
For Positive_Average = (1000*8/31)+(3000*10/31)
=258.06 + 967.74 = 1225.81
Calculation for negative:
     (-2000*13/31) = -838.71
I have used a case to separate positive balances from negative balances
I want the output like this
ACC_NO          AVG_AMT          POS_NEG_MARKER
123456          1225.81          1
123456          838.71          0
The negative balance is reported in its absolute form so I have a marker that checks if bal was -ve then 0 else 1
My Approach:
select Acc_No,
sum((AVG(Bal_amt)*(((max(Bal_date)-min(bal_date))/to_char(last_day(ab_date),'DD'))))),
(case when Bal_amt < 0 then '0' else '1' end) as POS_NEG_MARKER
from ACCT
Where Bal_date between '01-JAN-09' and '31-JAN-09'
Group by Acc_No, (case when Bal_amt < 0 then '0' else '1' )
This definitely thows the ORA-00937: not a single-group group function and if I play with the select elements I get ORA-00979: not a GROUP BY expression
I am sure there is a better way to handle this requirement. Please assist with ideas and corrections where necessary.
Thanx in advance

Answer 1

Your welcome. How do you want the query to behave when you have a zero balance? An extra record?
What the sign () does is return a 1, 0, -1 depending on whether the value is positive, zero or negative.
SQL> select val, sign (val)
  2    from (select 100 val from dual
  3          union all
  4          select 0 from dual
  5          union all
  6          select -123 from dual
  7  )
  8  /
       VAL  SIGN(VAL)
       100          1
         0          0
      -123         -1you could do something like this:
SQL> select acc_no
  2       , round (abs (sum (case when ind = 1
  3              then pos_amt
  4              when ind = -1
  5              then neg_amt
  6              else 0
  7              end * no_of_days/days_in_month
  8             )), 2) avg_amt
  9       , greatest (ind, 0) pos_neg_marker
10    from (
11  select decode (sign (bal_amt), 1, bal_amt) pos_amt
12       , decode (sign (bal_amt), -1, bal_amt) neg_amt
13       , decode (sign (bal_amt), 0, bal_amt) zero_amt
14       , sign (bal_amt) ind
15       , acc_no
16       , lead (bal_date, 1, last_day (bal_date))
17                  over (partition by acc_no
18                            order by bal_date
19                       ) - bal_date no_of_days
20       , last_day (bal_date) - trunc (bal_date, 'MM') + 1 days_in_month
21    from test
22  )
23   group by acc_no, ind
24  /
    ACC_NO    AVG_AMT POS_NEG_MARKER
         1          0              0
         1    1129.03              1
         1      387.1              0