Re: math.floor and rounding

[James Graves <ansible@...>]

On Mon, Sep 20, 2010 at 11:53:47PM +0800, KHMan wrote:
> On 9/20/2010 11:32 PM, Jose wrote:
> > I want to round numbers with two decimal digits.
> >
> > I use  math.floor( n * 100 )  but sometimes I have situations like:
> >
> > n = 10.20
> > math.floor( n * 100 ) -->  1019
> >
> > My current solution is to always add 0.001 to n, but is this the right
> > approach ?
> 
> If it's *precisely* 2 decimal places, consider instead to do your 
> calculations without any fractions, i.e. use 1020 etc.

If you want to _round_ off a floating point number, instead of
flooring it, here is the rounding function I have in my standard
library:

    function round(num) return math.floor(num+.5) end

It works better when you've got more digits of precision than you
wanted:

    n = 10.2049
    round(n * 100)   -->  1020

    n = 10.2051
    round(n * 100)   -->  1021

Are you handling negative numbers?  math.floor may not be what you want
in that case:

    n = -30.000001
    math.floor(n * 100)   --> -3001
    round(n * 100)        --> -3000

I encourage you to read up on round, floor and ceiling:

	http://en.wikipedia.org/wiki/Floating_point#Representable_numbers.2C_conversion_and_rounding

It is my believe that what people usually want when converting a
floating point number to an integer is the nearest integer.  That
depends on the domain, obviously.  So if your doing accounting, the
rules a different than if you're just drawing some graphics on the
screen.

James Graves