From: David Garamond <lists@...6.isreserved.com>
Date: 2004-06-28T03:31:49+09:00
Subject: Re: rounding to n decimal digits

Michael C. Libby wrote:
>>>  class Float
>>>    def round_n(n=0)
>>>      raise FloatDomainError unless self.finite?
>>>      raise ArgumentError unless n.class == Fixnum && n >= 0
>>>      sprintf("%.#{n}f", self).to_f
>>>    end
>>>
>>>    def ceil_n(n=0)
>>>      raise FloatDomainError unless self.finite?
>>>      raise ArgumentError unless n.class == Fixnum && n >= 0
>>>      (self*10**n).ceil.to_f / 10**n
>>>    end
>>>
>>>    def floor_n(n=0)
>>>      raise FloatDomainError unless self.finite?
>>>      raise ArgumentError unless n.class == Fixnum && n >= 0
>>>      (self*10**n).floor.to_f / 10**n
>>>    end
>>>  end
>>>
> 
> That's the same technique he's using for ceil and floor-- seems like the
> best approach available (the only thing left to do now is test whether
> multiply and divide is faster than sprintf). Except that instead of
> defining round_n, ceil_n, and floor_n, I would probably just redefine
> the existing Float#round, Float#ceil and Float#floor methods. Because
> he's got n=0 in the method signature existing code that depends on the
> default behavior of Float should be fine.

Btw, if you want to override #round, #ceil, #floor, remember to return a 
Fixnum if n=0 to avoid breaking other code.

-- 
dave