R: Rounding problems - Mailing list pgsql-general

From Paolo Saudin
Subject R: Rounding problems
Date
Msg-id 004401c9cc04$945a9670$bd0fc350$@it
Whole thread Raw
In response to Re: Rounding problems  (Tom Lane <tgl@sss.pgh.pa.us>)
List pgsql-general
>>"Paolo Saudin" <paolo@ecometer.it> writes:
>> I have a problem with a query wich simple aggregate values. In the sample
>> below I have two values, 1.3 and 1.4. Rounding their average with one
>> decimals, should give 1.4.
>
>You seem way overoptimistic about float4 values being exact.  They are
>not.  The actual computation being done here is more like
>
>regression=# select (1.3::real + 1.4::real) / 2 ;
>     ?column?
>------------------
> 1.34999990463257
> (1 row)
>
>If you want an exact sum with no roundoff error you should be storing
>all your values as numeric (and taking the consequent speed and space
>hit :-().
>
>            regards, tom lane
>


I converted all the fields in numeric type instead of real and now both
queries return the same result !
Now I need to test about performances ...

Thank you very much !!
Paolo Saudin


pgsql-general by date:

Previous
From: "Paolo Saudin"
Date:
Subject: R: Rounding problems
Next
From: Wojtek
Date:
Subject: keeping track of function execution