Bug #5285: Ruby 1.9.2 throws exception on sort of array containing true AND false values

• Status: Rejected
• Assignee: -
• Target version: 1.9.2
• ruby -v: ruby 1.9.2p290 (2011-07-09 revision 32553) [x86_64-linux]

[mcorino (Martin Corino)]

I encountered a very weird bug.

Sorting of arrays containing identical boolean values (ALL true or ALL false) works fine.
As soon as 1 value differs however #sort throws an exception.

ruby-1.9.2-p290 :001 > [true, true].sort
=> [true, true]

ruby-1.9.2-p290 :002 > [false, false].sort
=> [false, false]

ruby-1.9.2-p290 :003 > [false, true].sort
ArgumentError: comparison of FalseClass with true failed
from (irb):3:in sort' from (irb):3 from /home/martin/.rvm/rubies/ruby-1.9.2-p290/bin/irb:16:in '

ruby-1.9.2-p290 :004 > [false, true, false].sort
ArgumentError: comparison of FalseClass with true failed
from (irb):4:in sort' from (irb):4 from /home/martin/.rvm/rubies/ruby-1.9.2-p290/bin/irb:16:in '

ruby-1.9.2-p290 :011 > [true, true, true, true].sort
=> [true, true, true, true]

ruby-1.9.2-p290 :012 > [true, true, true, false].sort
ArgumentError: comparison of TrueClass with false failed
from (irb):12:in sort' from (irb):12 from /home/martin/.rvm/rubies/ruby-1.9.2-p290/bin/irb:16:in '

[naruse (Yui NARUSE)]

• Status changed from Open to Rejected

Sort uses <=>.
true <=> true and false <=> false work, but true <=> false.
Simply because of this.

[pesnk (Pedro Medeiros)]

actually, by doing true <=> false on irb returns nil, while doing true <=> true or false <=> false will return 0. not necessarily an exception. Wouldn't it be right to return -1 or 1 by comparing true with false or false with true, even though true came from TrueClass and false came from FalseClass ?

Booleans are not ordered. It doesn't make sense to say "the logical value true is less than/greater than the logical value false" - that statement doesn't mean anything. I can't help but feel sorting an enumerable of booleans is a nonsensical operation. I'm not even sure what purpose it would serve that wouldn't more reasonably be modeled by another (likely more efficient) operation.

On Sep 6, 2011, at 1:30 PM, Pedro Medeiros [e-mail] wrote:

Issue #5285 has been updated by Pedro Medeiros.

actually, by doing true <