Do numbers with a final digit of five and
rounding precision greater than zero ever
round up?
Bob Dankert wrote:
According to that, it rounds to the nearest even number. Shouldn't 9.95
go to 10 then, and 9.85 go to 9.8?
After additional testing with SQLite 3.2.2, I have the following
results:
Round(9.95,1) -> 9.9 *Rounded Down*
Round(9.85,1) -> 9.8 *Rounded Down*
Round(9.5,0) -> 10 *Rounded Up*
Round(9.995,2) -> 9.99 *Rounded Down*
I really see no pattern or sense to the results.
Bob
Envision Information Technologies
Associate
[EMAIL PROTECTED]
v. 608.256.5680
f. 608.256.3780
-----Original Message-----
From: Kervin L. Pierre [mailto:[EMAIL PROTECTED]
Sent: Tuesday, August 30, 2005 12:18 PM
To: sqlite-users@sqlite.org
Subject: Re: [sqlite] Unlucky number for the ROUND function
Interesting...
Bankers' Rounding
http://blogs.msdn.com/ericlippert/archive/2003/09/26/53107.aspx
Brass Tilde wrote:
From: "Bob Dankert"
Using an older version which I compiled, I get 9.9, though it
seems it should round up to 10.0?
This may be dependent upon the math library linked into the
application
by the compiler used to build SQLite. Some libraries appear to now be
using so-called "banking rounding" (though there are other names),
where
a "5" digit is rounded up or down depending upon the digit immediately
to its left. If that digit is odd, it rounds one way, if even, the
other. It looks like in this case, 9.95 would round to 9.9, while
9.85
would likely round to 9.9 as well. Try rounding 9.85 and see what you
get.
Brad
.
