Project

General

Profile

Actions

Bug #10979

closed

Hash fails to distinguish 0.0 & -0.0

Added by avondrak (Alex Vondrak) about 9 years ago. Updated almost 9 years ago.

Status:
Closed
Assignee:
-
Target version:
-
ruby -v:
ruby 2.2.0p0 (2014-12-25 revision 49005) [x86_64-darwin13]
[ruby-core:<unknown>]

Description

According to the docs ( http://ruby-doc.org/core-2.2.1/Hash.html ):

Two objects refer to the same hash key when their hash value is identical and the two objects are eql? to each other.

However, that does not appear to be the case for 0.0 and -0.0:

$ irb
2.2.0 :001 > (0.0).hash
 => -204267982040607835 
2.2.0 :002 > (-0.0).hash
 => -204267982040607835 
2.2.0 :003 > (0.0).eql?(-0.0)
 => true 
2.2.0 :004 > (-0.0).eql?(0.0)
 => true 
2.2.0 :005 > { 0.0 => :positive }.has_key?(-0.0)
 => false 
2.2.0 :006 > { -0.0 => :negative }.has_key?(0.0)
 => false 
2.2.0 :007 > { 0.0 => :positive }[-0.0]
 => nil 
2.2.0 :008 > { -0.0 => :negative }[0.0]
 => nil 
2.2.0 :009 > { 0.0 => :positive }.keys.map(&:hash)
 => [-204267982040607835] 
2.2.0 :010 > { -0.0 => :negative }.keys.map(&:hash)
 => [-204267982040607835]

I realize that floating point numbers are weird, and that 0.0 & -0.0 aren't literally the same machine representation. However, they are #eql? to each other have the same hash codes. I figure either the docs are misleading or there's a bug in the Hash primitives.

This also has carry-over effects to Array differences, for example:

$ irb
2.2.0 :001 > [0.0] - [-0.0]
 => [0.0] 
2.2.0 :002 > [-0.0] - [0.0]
 => [-0.0]
2.2.0 :003 > [0.0, -0.0] - [-0.0]
 => [0.0]

This was the initial behavior I observed in Ruby 2.1.3 that led me to file this bug. I tried Ruby 2.2.0 and the same behavior still seems to persist.

Actions

Also available in: Atom PDF

Like0
Like0Like0Like0Like0