From: Robert Klemme <shortcutter@...>
Date: 2011-12-15T17:14:13+09:00
Subject: Re: round float to integer if whole number

On Wed, Dec 14, 2011 at 11:21 PM, Ronnie Aa <liquid98@gmail.com> wrote:
> Gunther Diemant wrote in post #1036785:
>> Possible tests:
>> ff.floor == ff
>> ff % 1 == 0
>>
>> If x is always an integer i would use the mod operator ( x % w == 0
>> ==> x/w integer)
>>
>> 2011/12/14 Ronnie Aa <liquid98@gmail.com>:
>
> thanxs Gunther!

That test does not work reliably:

irb(main):006:0> x = 1.0
=> 1.0
irb(main):007:0> x.class
=> Float
irb(main):008:0> 20.times do |i|
irb(main):009:1*   y = x + (10.0 ** -i)
irb(main):010:1>   printf "%4d %20.16f %s\n", i, y, y == y.floor
irb(main):011:1> end
   0   2.0000000000000000 true
   1   1.1000000000000001 false
   2   1.0100000000000000 false
   3   1.0009999999999999 false
   4   1.0001000000000000 false
   5   1.0000100000000001 false
   6   1.0000009999999999 false
   7   1.0000001000000001 false
   8   1.0000000099999999 false
   9   1.0000000010000001 false
  10   1.0000000001000000 false
  11   1.0000000000100000 false
  12   1.0000000000010001 false
  13   1.0000000000000999 false
  14   1.0000000000000100 false
  15   1.0000000000000011 false
  16   1.0000000000000000 true
  17   1.0000000000000000 true
  18   1.0000000000000000 true
  19   1.0000000000000000 true
=> 20

irb(main):024:0* x = BigDecimal.new "1.0"
=> #<BigDecimal:102e06c0,'0.1E1',9(18)>
irb(main):025:0> 20.times do |i|
irb(main):026:1*   y = x + (BigDecimal.new("10.0") ** -i)
irb(main):027:1>   printf "%4d %30.24f %s\n", i, y, y == y.floor
irb(main):028:1> end
   0     2.000000000000000000000000 true
   1     1.100000000000000088817842 false
   2     1.010000000000000008881784 false
   3     1.000999999999999889865876 false
   4     1.000099999999999988986588 false
   5     1.000010000000000065512040 false
   6     1.000000999999999917733362 false
   7     1.000000100000000058386718 false
   8     1.000000009999999939225290 false
   9     1.000000001000000082740371 false
  10     1.000000000100000008274037 false
  11     1.000000000010000000827404 false
  12     1.000000000001000088900582 false
  13     1.000000000000099920072216 false
  14     1.000000000000009992007222 false
  15     1.000000000000001110223025 false
  16     1.000000000000000000000000 false
  17     1.000000000000000000000000 false
  18     1.000000000000000000000000 false
  19     1.000000000000000000000000 false
=> 20

It is better to define "equality" differently for floating point
numbers.  There are a few common approaches:

1. Define a maximum difference that you accept, e.g.

irb(main):029:0> x = 1.0
=> 1.0
irb(main):030:0> 20.times do |i|
irb(main):031:1*   y = x + (10.0 ** -i)
irb(main):032:1>   printf "%4d %30.24f %s\n", i, y, (y - y.floor).abs < 1e-5
irb(main):033:1> end
   0     2.000000000000000000000000 true
   1     1.100000000000000088817842 false
   2     1.010000000000000008881784 false
   3     1.000999999999999889865876 false
   4     1.000099999999999988986588 false
   5     1.000010000000000065512040 false
   6     1.000000999999999917733362 true
   7     1.000000100000000058386718 true
   8     1.000000009999999939225290 true
   9     1.000000001000000082740371 true
  10     1.000000000100000008274037 true
  11     1.000000000010000000827404 true
  12     1.000000000001000088900582 true
  13     1.000000000000099920072216 true
  14     1.000000000000009992007222 true
  15     1.000000000000001110223025 true
  16     1.000000000000000000000000 true
  17     1.000000000000000000000000 true
  18     1.000000000000000000000000 true
  19     1.000000000000000000000000 true
=> 20

2. Define a maximum delta for the factor to 1

irb(main):034:0> 20.times do |i|
irb(main):035:1*   y = x + (10.0 ** -i)
irb(main):036:1>   printf "%4d %30.24f %s\n", i, y, ((y / y.floor) -
1).abs < 1e-5
irb(main):037:1> end
   0     2.000000000000000000000000 true
   1     1.100000000000000088817842 false
   2     1.010000000000000008881784 false
   3     1.000999999999999889865876 false
   4     1.000099999999999988986588 false
   5     1.000010000000000065512040 false
   6     1.000000999999999917733362 true
   7     1.000000100000000058386718 true
   8     1.000000009999999939225290 true
   9     1.000000001000000082740371 true
  10     1.000000000100000008274037 true
  11     1.000000000010000000827404 true
  12     1.000000000001000088900582 true
  13     1.000000000000099920072216 true
  14     1.000000000000009992007222 true
  15     1.000000000000001110223025 true
  16     1.000000000000000000000000 true
  17     1.000000000000000000000000 true
  18     1.000000000000000000000000 true
  19     1.000000000000000000000000 true
=> 20

Kind regards

robert

-- 
remember.guy do |as, often| as.you_can - without end
http://blog.rubybestpractices.com/