Bug #3563: problems with exponentiation, cbrt

• Status: Closed
• Assignee: mame (Yusuke Endoh)
• Target version: -
• ruby -v: ruby 1.9.2dev (2010-07-02 revision 28524) [i386-darwin9.8.0]

[kolbrich (Kevin Olbrich)]

=begin

nothing required

(-27)**(Rational(1,3)) #=> (1.5000000000000004+2.598076211353316i)

I would expect that this would return -3

Complex(1.5000000000000004,2.598076211353316)**3 #=> (-27.000000000000007+1.2434497875801753e-14i)

I can understand how this might happen given round off errors, but the following identity should be true

Math.cbrt(-27) == (-27)**(Rational(1,3)) #=> false

Also note that if you do this...

require 'complex'
Math.cbrt(-27) #=> (1.5000000000000004+2.598076211353316i)
=end

[mame (Yusuke Endoh)]

• Status changed from Open to Assigned
• Assignee set to mame (Yusuke Endoh)

=begin
Hi,

2010/7/13 Kevin Olbrich [e-mail]:

(-27)**(Rational(1,3)) #=> (1.5000000000000004+2.598076211353316i)

I would expect that this would return -3

I think this is not a bug. 1.9.1 and 1.9.2 should be changed.
But I understand your expectation. To change this behavior in the
future, I propose to add a note to rdoc of Rational#**:

diff --git a/rational.c b/rational.c
index f5a6d26..734131b 100644
--- a/rational.c
+++ b/rational.c
@@ -907,6 +907,9 @@ nurat_fdiv(VALUE self, VALUE other)
*

• Performs exponentiation.

• • When the solution is underspecified, it is not specified which
• • solution this method returns.
• • For example:
  • Rational(2) ** Rational(3) #=> (8/1)

Also note that if you do this...

require 'complex'
Math.cbrt(-27) #=> (1.5000000000000004+2.598076211353316i)

complex.rb is deprecated, but mathn.rb causes the same.
I think this is a bug of cmath:

diff --git a/lib/cmath.rb b/lib/cmath.rb
index b23dac2..976f269 100644
--- a/lib/cmath.rb
+++ b/lib/cmath.rb
@@ -85,7 +85,7 @@ module CMath
end

def cbrt(z)

• if z.real? and z >= 0

• if z.real?
  cbrt!(z)
  else
  Complex(z) ** (1.0/3)

Unless there is objection, I'll commit these patches to both trunk
and ruby_1_9_2.

--
Yusuke Endoh [e-mail]
=end

[kolbrich (Kevin Olbrich)]

=begin
I'm fine with adding a comment, however the one you added does not really clarify the matter (at least not to me).
=end

[mame (Yusuke Endoh)]

=begin
Umm. Could you make a draft?

2010/7/14 Kevin Olbrich [e-mail]:

I'm fine with adding a comment, however the one you added does not really clarify the matter (at least not to me).

--
Yusuke Endoh [e-mail]

=end

[mame (Yusuke Endoh)]

=begin
I'll commit only the patch for cmath.rb and close this ticket.

If you (or anyone) show me a draft in a few days, I'll import it.
But if you don't, we will release 1.9.2 without any note, which
may make it difficult to change the behavior in 1.9.3 or later.

2010/7/14 Yusuke ENDOH [e-mail]:

Umm.  Could you make a draft?

2010/7/14 Kevin Olbrich [e-mail]:

I'm fine with adding a comment, however the one you added does not really clarify the matter (at least not to me).

--
Yusuke Endoh [e-mail]

--
Yusuke Endoh [e-mail]

=end

[mame (Yusuke Endoh)]

• Status changed from Assigned to Closed
• % Done changed from 0 to 100

=begin
This issue was solved with changeset r28698.
Kevin, thank you for reporting this issue.
Your contribution to Ruby is greatly appreciated.
May Ruby be with you.

=end