Re: math.floor ... bad trunc ?

[voidptr <voidptr69@...>]

On 2015-08-08 1:09 AM, Coda Highland wrote:
> On Fri, Aug 7, 2015 at 9:53 PM, voidptr <voidptr69@gmail.com> wrote:
> > from  Zerobrane IDE with lua 5.3 and 5.2
> >
> > a = 1.45555
> > b = a * 10^4
> > -- b = 14555.5
> > b = b + 0.5
> > -- b = 14556
> > c = math.floor(b)
> > -- c = 14555
> > c = b // 1
> > -- c = 14555
> >
> > it there a way to force a good  float to int conversion ???
> >
> > // voidptr
> This is a standard floating-point issue.
>
> You asked for truncation, you got truncation. b's value was never
> 14.556; it was 14.5559something -- because a's value is not 1.45555;
> that can't be represented precisely in base 2. Its value is
> 1.455549something.

darn so silly I am   !!
I agree with binary representation this NOT exactly what I meant ...
b = b + 0.5
-- b = 14556.0
should give me something like this
14555.999999999998994582028899458236992359161376953125   +/- 0.01 (kindof)


I miss to include the .0 decimal
in fact the interpreter show me 14556.0
so I thought if the interpreter show me 14556.0 that  .0 means in was 
securely rounded to that decimal.
and number have been correctly reinterpreted ...

so never believe your eyes !


> If you want rounding, then round. If you want lopsided rounding, then
> add an epsilon and floor. (For example, you might add 0.01 before
> flooring if you want to say "14555.99 is okay to call 14556, but
> 14555.98 should be truncated to 14555.")
>
> /s/ Adam