On Mon, Apr 15, 2019 at 2:19 AM Miki Tebeka <miki.teb...@gmail.com> wrote:
> Can anyone explain the below?
> (When printing out b with %.20f it prints 1.20999999999999996447)
>
> package main
>
> import "fmt"
>
> func main() {
> a, b := 1.1*1.1, 1.21
> fmt.Println(a == b) // true
> fmt.Println(1.1*1.1 == 1.21) // false
> }
>
>
As David Riley points out in his reply you should never compare floats
(other than a few special cases such as zero) for equality. Consider the
program below. If a C `double` had infinite precision or used a
representation where 1e-10 and all multiples could be represented exactly
then the result of the `for` loop would be exactly 1.21. It isn't because
the `double` type has neither attribute. The issues are fundamentally the
same for Go notwithstanding the rules for untyped constants. The output of
this program on my system is:
1e-10 d9d7bdbb3ddb7cdf
1.21 f5c28f5c3ff35c28
1.21000008685 0d12ec183ff35c29
#include <stdio.h>
int main() {
double f = 1.21;
double g = 1.0 / 1e10;
double h = 0.0;
for (long long i = 121 * 1e8; i; i--) h += g;
fprintf(stdout, "%.12g %08x%08x\n", g, ((int *)&g)[0], ((int *)&g)[1]);
fprintf(stdout, "%.12g %08x%08x\n", f, ((int *)&f)[0], ((int *)&f)[1]);
fprintf(stdout, "%.12g %08x%08x\n", h, ((int *)&h)[0], ((int *)&h)[1]);
}
--
Kurtis Rader
Caretaker of the exceptional canines Junior and Hank
--
You received this message because you are subscribed to the Google Groups
"golang-nuts" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to golang-nuts+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.
