As an alternative -- you could do an inline type cast....
SELECT SUM(amount)::numeric(10, 2) FROM acc_trans WHERE trans_id=19721 AND
chart_id=10019;
"Karen Hill" <karen_hill22@yahoo.com> wrote in message
news:1156873638.219228.218710@74g2000cwt.googlegroups.com...
>
> "Andrew Baerg" wrote:
>> Hi,
>>
>> I am getting strange results from the sum function as follows:
>>
>> corp=# select amount from acc_trans where trans_id=19721 and
>> chart_id=10019;
>> amount
>> ---------
>> 4.88
>> 117.1
>> -121.98
>> (3 rows)
>>
>> corp=# select sum(amount) from acc_trans where trans_id=19721 and
>> chart_id=10019;
>> sum
>> ----------------------
>> -1.4210854715202e-14
>> (1 row)
>>
>>
>> amount is defined as double precision. I noticed that if I cast amount
>> as numeric, the sum comes out 0 as expected.
>>
>
> Double precision accorrding to the documentation is "8 byte
> variable-precision, inexact". That means when you do the sum, rounding
> occurs. You should use Numeric or Decimal as the datatype. The money
> type is depreciated so don't use it if what you are summing is currency.
>
