From: Benoit Daloze <eregontp@...>
Date: 2010-07-08T19:40:45+09:00
Subject: [ruby-core:31127] Re: [Backport #3550] Floating Point Representation 	Prevents Raising to Fractional Power

On 8 July 2010 10:32, Milind Pandit <redmine@ruby-lang.org> wrote:
> Backport #3550: Floating Point Representation Prevents Raising to Fractional Power
> http://redmine.ruby-lang.org/issues/show/3550
>
> Author: Milind Pandit
> Status: Open, Priority: Normal
>
> Not sure why this literal float can be raised successfully to a fractional power, but when assigned to a variable, it returns NaN.
>
> $ ruby --version
> ruby 1.8.7 (2009-06-12 patchlevel 174) [universal-darwin10.0]
> $ irb
> -2.21792114695341 ** 0.1
> => -1.08291560040828
>>> -2.21792114695341 ** 1.1
> => -2.40182141051126
>>> a = -2.21792114695341
> => -2.21792114695341
>>> a ** 0.1
> => NaN # Expected: -1.08291560040828
>>> a ** 1.1
> => NaN # Expected: -2.40182141051126
>>> -2.21792114695341.to_f ** 1.1
> => NaN # Expected: -1.08291560040828
>>> -2.21792114695341.to_f ** 0.1
> => NaN # Expected: -2.40182141051126
>>> a - -2.21792114695341
> => 0.0
>>> (a - -2.21792114695341).zero?
> => true
>

Because the answer is complex, and when you use both literal, the
precedence of #** is higher than #-@
> (-2.21792114695341) ** 1.1
=> NaN
But
> -2.21792114695341 ** 1.1
=> -2.40182141051126

In 1.9, Complex are in the core and so:
> (-2.21792114695341) ** 1.1
=> (-2.2842679034439537-0.742203633301588i)

I believe it is possible in 1.8, but "require 'complex'" did not
resolve this for me.