> -----Original Message-----
> From: Paul Tillotson [mailto:pntil(at)shentel(dot)net]
> Sent: Monday, April 26, 2004 4:41 PM
> To: pgsql-general(at)postgresql(dot)org
> Subject: Re: [GENERAL] Arbitrary precision modulo operation
> I see there are a few misconceptions about numeric and
> modulus on here:
> (1) A modulus operation on a numeric type should NOT have
> rounding errors. The whole point of numeric is that it is an
> arbitrary precision BASE 10 representation of your number.
This is true
> The modulus returns the (whole
> number) remainder as a result of a division.
This is true if the numeric values are integers.
When the values are not integral, some non-integral results can be
2.50 % 2.50 is 0
But 13.89 modulo 3.50 is 3.39
If you work it out on paper, you will see that 3.39 is the correct
> (2) the modulus operator/function is, AFAIK, supposed to
> return the modulus with the SAME SIGN as the divisor, so I
> think this is a bug. That's what every other modulus operator
> that I have ever seen does. Would you mind doing
I would agree that a positive modulus is preferable. However, the
negative result is also mathematically correct.
> foodb=> SELECT version();
> (3) MySQL just rounds large numbers to the highest value that
> the type will support, and apparently, no arbitrary precision
> types are listed on this page:
> You can't expect to get a modulus from an out-of-range number.
Should it not (therefore) throw an error of some sort?
pgsql-general by date
|Next:||From: Alvaro Herrera||Date: 2004-04-27 01:48:44|
|Subject: Re: 7.3.4 on Linux: UPDATE .. foo=foo+1 degrades massivly|
|Previous:||From: Paul Tillotson||Date: 2004-04-26 23:41:16|
|Subject: Re: Arbitrary precision modulo operation|