Bug #9170

Math.sqrt returns different types when mathn is included; breaks various gems - this bug can be reproduced in Ruby 1.8 as well

Added by Jason Hutchens 5 months ago. Updated 3 months ago.

[ruby-core:58653]
Status:Feedback
Priority:Normal
Assignee:-
Category:core
Target version:next minor
ruby -v:ruby 2.0.0p247 (2013-06-27 revision 41674) [x86_64-darwin12.2.1] Backport:1.9.3: UNKNOWN, 2.0.0: UNKNOWN

Description

Math.sqrt() usually returns a Float or throws an exception. Some gems rely on this behaviour (e.g. by calling nan? on the results of a calculation that includes a sqrt).

After requiring mathn, Math.sqrt() returns an Integer in some cases (when the results of the sqrt would be a whole number).

Here is a transcript that demonstrates the problem:

$ rvm use ree
Using /Users/jason/.rvm/gems/ree-1.8.7-2012.02
$ ruby -v
ruby 1.8.7 (2012-02-08 MBARI 8/0x6770 on patchlevel 358) [i686-darwin12.1.0], MBARI 0x6770, Ruby Enterprise Edition 2012.02
$ ruby -e "puts Math.sqrt(4).nan?"
false
$ ruby -e "require 'mathn'; puts Math.sqrt(4).nan?"
-e:1: undefined method nan?' for 2:Fixnum (NoMethodError)
$ rvm use 1.9.3
Using /Users/jason/.rvm/gems/ruby-1.9.3-p448
$ ruby -v
ruby 1.9.3p448 (2013-06-27 revision 41675) [x86_64-darwin12.2.0]
$ ruby -e "puts Math.sqrt(4).nan?"
false
$ ruby -e "require 'mathn'; puts Math.sqrt(4).nan?"
-e:1:in
': undefined method nan?' for 2:Fixnum (NoMethodError)
$ rvm use 2.0
Using /Users/jason/.rvm/gems/ruby-2.0.0-p247
$ ruby -v
ruby 2.0.0p247 (2013-06-27 revision 41674) [x86_64-darwin12.2.1]
$ ruby -e "puts Math.sqrt(4).nan?"
false
$ ruby -e "require 'mathn'; puts Math.sqrt(4).nan?"
-e:1:in
': undefined method nan?' for 2:Fixnum (NoMethodError)
$ rvm use jruby
Using /Users/jason/.rvm/gems/jruby-1.7.4
$ ruby -v
jruby 1.7.4 (1.9.3p392) 2013-05-16 2390d3b on Java HotSpot(TM) 64-Bit Server VM 1.6.0_51-b11-457-11M4509 [darwin-x86_64]
$ ruby -e "puts Math.sqrt(4).nan?"
false
$ ruby -e "require 'mathn'; puts Math.sqrt(4).nan?"
NoMethodError: undefined method
nan?' for 2:Fixnum
(root) at -e:1

Here is a discussion of how the bug manifests itself: https://github.com/olbrich/ruby-units/pull/48

Inclusion of mathn causes issues in RGeo, Alchemist, ActiveShipping and pry-rails

History

#1 Updated by Kenta Murata 4 months ago

  • Category set to core
  • Status changed from Open to Feedback
  • Target version set to next minor

It's not a bug.

Do you want to introduce Numeric#nan? method to resolve this problem?
If so, this issue should be in Feature tracker.

Please tell us how do you think.

#2 Updated by Kenta Murata 4 months ago

I think introducing Numeric#nan? is good.

#3 Updated by Jason Hutchens 4 months ago

mrkn (Kenta Murata) wrote:

It's not a bug.

I'm a middle-man here, but I have gem maintainers claiming that it is a bug, because the documentation for Math.sqrt() indicates that it returns a Float, and that it's therefore OK to assume that it can be treated as a Float by the caller. Ref: http://ruby-doc.org/core-2.1.0/Math.html#method-c-sqrt

#4 Updated by Marc-Andre Lafortune 3 months ago

Isn't this more an issue with the documentation of mathn not specifying all the methods it changes and how?

It might be wise to add a warning that it is suggested not to require it from a library because it might induce unexpected changes in any code using it?

It would be a prime candidate for refinements...

#5 Updated by Zachary Scott 3 months ago

I will check the docs and see if I can fix it.

On Jan 6, 2014, at 12:44 AM, "kranzky (Jason Hutchens)" JasonHutchens@gmail.com wrote:

Issue #9170 has been updated by kranzky (Jason Hutchens).

mrkn (Kenta Murata) wrote:

It's not a bug.

I'm a middle-man here, but I have gem maintainers claiming that it is a bug, because the documentation for Math.sqrt() indicates that it returns a Float, and that it's therefore OK to assume that it can be treated as a Float by the caller. Ref: http://ruby-doc.org/core-2.1.0/Math.html#method-c-sqrt

Bug #9170: Math.sqrt returns different types when mathn is included; breaks various gems - this bug can be reproduced in Ruby 1.8 as well
https://bugs.ruby-lang.org/issues/9170#change-44105

Author: kranzky (Jason Hutchens)
Status: Feedback
Priority: Normal
Assignee:
Category: core
Target version: next minor
ruby -v: ruby 2.0.0p247 (2013-06-27 revision 41674) [x86_64-darwin12.2.1]
Backport: 1.9.3: UNKNOWN, 2.0.0: UNKNOWN

Math.sqrt() usually returns a Float or throws an exception. Some gems rely on this behaviour (e.g. by calling nan? on the results of a calculation that includes a sqrt).

After requiring mathn, Math.sqrt() returns an Integer in some cases (when the results of the sqrt would be a whole number).

Here is a transcript that demonstrates the problem:

$ rvm use ree
Using /Users/jason/.rvm/gems/ree-1.8.7-2012.02
$ ruby -v
ruby 1.8.7 (2012-02-08 MBARI 8/0x6770 on patchlevel 358) [i686-darwin12.1.0], MBARI 0x6770, Ruby Enterprise Edition 2012.02
$ ruby -e "puts Math.sqrt(4).nan?"
false
$ ruby -e "require 'mathn'; puts Math.sqrt(4).nan?"
-e:1: undefined method nan?' for 2:Fixnum (NoMethodError)
$ rvm use 1.9.3
Using /Users/jason/.rvm/gems/ruby-1.9.3-p448
$ ruby -v
ruby 1.9.3p448 (2013-06-27 revision 41675) [x86_64-darwin12.2.0]
$ ruby -e "puts Math.sqrt(4).nan?"
false
$ ruby -e "require 'mathn'; puts Math.sqrt(4).nan?"
-e:1:in
': undefined method nan?' for 2:Fixnum (NoMethodError)
$ rvm use 2.0
Using /Users/jason/.rvm/gems/ruby-2.0.0-p247
$ ruby -v
ruby 2.0.0p247 (2013-06-27 revision 41674) [x86_64-darwin12.2.1]
$ ruby -e "puts Math.sqrt(4).nan?"
false
$ ruby -e "require 'mathn'; puts Math.sqrt(4).nan?"
-e:1:in
': undefined method nan?' for 2:Fixnum (NoMethodError)
$ rvm use jruby
Using /Users/jason/.rvm/gems/jruby-1.7.4
$ ruby -v
jruby 1.7.4 (1.9.3p392) 2013-05-16 2390d3b on Java HotSpot(TM) 64-Bit Server VM 1.6.0_51-b11-457-11M4509 [darwin-x86_64]
$ ruby -e "puts Math.sqrt(4).nan?"
false
$ ruby -e "require 'mathn'; puts Math.sqrt(4).nan?"
NoMethodError: undefined method
nan?' for 2:Fixnum
(root) at -e:1

Here is a discussion of how the bug manifests itself: https://github.com/olbrich/ruby-units/pull/48

Inclusion of mathn causes issues in RGeo, Alchemist, ActiveShipping and pry-rails

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

Also available in: Atom PDF