From: Reuven Lerner <redmine@...>
Date: 2010-05-17T20:25:42+09:00
Subject: [ruby-core:30275] [Backport #3304] Bug in BigDecimal#to_f

Backport #3304: Bug in BigDecimal#to_f
http://redmine.ruby-lang.org/issues/show/3304

Author: Reuven Lerner
Status: Open, Priority: Normal

I'm using ruby 1.8.7 (2009-06-08 patchlevel 173) [universal-darwin10.0], from OS X Snow Leopard.    I am using BigDecimal in a project that requires a large number
of significant figures after the decimal point.  I know that there are issues using floats for exact math, but I think that I have discovered something surprising 
even when you take that into consideration:

In IRB:

>> require 'bigdecimal'
=> true
>> number = BigDecimal.new('2.50000')
=> #<BigDecimal:102ba0128,'0.25E1',8(12)>
>> number.to_f
=> 2.5
>> number += BigDecimal.new('1000000')
=> #<BigDecimal:102b8d0f0,'0.10025E7',12(20)>
>> number.to_f
=> 1002500.0
>> number.to_s("F").to_f
=> 1000002.5

Am I missing something?  I'd like to think that this is my mistake, but these results definitely violated the Principle of Least Surprise!

Reuven


----------------------------------------
http://redmine.ruby-lang.org