Bug in Float? (was: NaN Error)

Tim_Pease · 17 April 2007 18:41

I ran across some peculiar behavior yesterday when working with large
floating point values. Here is a small script that demonstrates the
behavior. My question is, is this the expected behavior or is this a
bug in Float?

cat float_bug.rb

puts "ruby #{RUBY_VERSION} " +
"(#{RUBY_RELEASE_DATE} patchlevel #{RUBY_PATCHLEVEL}) " +
"[#{RUBY_PLATFORM}]"

puts "Float::MAX = #{Float::MAX}"
puts "1.0e+300 = #{1.0e+300}"
puts "1.0e+301 = #{1.0e+301}"
puts "1.0e+308 = #{1.0e+308}"
puts "1.0e+700 = #{1.0e+700}"

# EOF

When I run this script on my Windows machine, it produces the output I
would expect -- i.e. numbers above Float::MAX are "Infinity" and
numbers below Float::MAX are valid floating point numbers.

ruby -w float_bug.rb

ruby 1.8.5 (2006-12-25 patchlevel 12) [i386-cygwin]
Float::MAX = 1.79769313486232e+308
1.0e+300 = 1.0e+300
1.0e+301 = 1.0e+301
1.0e+308 = 1.0e+308
1.0e+700 = Infinity

Now, when I run the exact same script on a Solaris machine, some odd
things start to happen

ruby -w float_bug.rb

ruby 1.8.5 (2006-12-25 patchlevel 12) [sparc-solaris2.8]
Float::MAX = 1.79769313486232e+308
1.0e+300 = 1.0e+300
1.0e+301 = NaN
1.0e+308 = NaN
1.0e+700 = 8.55180271847103e+116

Whey are valid floats being treated as NaN? Better yet, why does
1.0e+700 get evaluated to 8.5518027e+116?

Once more on the Windows 1-click installed version of Ruby.

ruby -w float_bug.rb

ruby 1.8.6 (2007-03-13 patchlevel 0) [i386-mswin32]
Float::MAX = 1.79769313486232e+308
1.0e+300 = 1.0e+300
1.0e+301 = NaN
1.0e+308 = NaN
1.0e+700 = 1.0e+189

Any thoughts on this one? I encourage you to copy the float_bug.rb
script and try it out for yourself.

Blessings,
TwP

Dan_Stevens_IAmAI · 17 April 2007 19:23

$ ruby float_bug.rb
ruby 1.8.5 (2006-08-25) [i486-linux]
Float::MAX = 1.79769313486232e+308
1.0e+300 = 1.0e+300
1.0e+301 = 1.0e+301
1.0e+308 = 1.0e+308
1.0e+700 = 1.02189775487683e+36

···

On 17/04/07, Tim Pease <tim.pease@gmail.com> wrote:

I ran across some peculiar behavior yesterday when working with large
floating point values. Here is a small script that demonstrates the
behavior. My question is, is this the expected behavior or is this a
bug in Float?

> cat float_bug.rb
puts "ruby #{RUBY_VERSION} " +
"(#{RUBY_RELEASE_DATE} patchlevel #{RUBY_PATCHLEVEL}) " +
"[#{RUBY_PLATFORM}]"

puts "Float::MAX = #{Float::MAX}"
puts "1.0e+300 = #{1.0e+300}"
puts "1.0e+301 = #{1.0e+301}"
puts "1.0e+308 = #{1.0e+308}"
puts "1.0e+700 = #{1.0e+700}"

# EOF

When I run this script on my Windows machine, it produces the output I
would expect -- i.e. numbers above Float::MAX are "Infinity" and
numbers below Float::MAX are valid floating point numbers.

> ruby -w float_bug.rb
ruby 1.8.5 (2006-12-25 patchlevel 12) [i386-cygwin]
Float::MAX = 1.79769313486232e+308
1.0e+300 = 1.0e+300
1.0e+301 = 1.0e+301
1.0e+308 = 1.0e+308
1.0e+700 = Infinity

Now, when I run the exact same script on a Solaris machine, some odd
things start to happen

> ruby -w float_bug.rb
ruby 1.8.5 (2006-12-25 patchlevel 12) [sparc-solaris2.8]
Float::MAX = 1.79769313486232e+308
1.0e+300 = 1.0e+300
1.0e+301 = NaN
1.0e+308 = NaN
1.0e+700 = 8.55180271847103e+116

Whey are valid floats being treated as NaN? Better yet, why does
1.0e+700 get evaluated to 8.5518027e+116?

Once more on the Windows 1-click installed version of Ruby.

> ruby -w float_bug.rb
ruby 1.8.6 (2007-03-13 patchlevel 0) [i386-mswin32]
Float::MAX = 1.79769313486232e+308
1.0e+300 = 1.0e+300
1.0e+301 = NaN
1.0e+308 = NaN
1.0e+700 = 1.0e+189

Any thoughts on this one? I encourage you to copy the float_bug.rb
script and try it out for yourself.

Blessings,
TwP

Gavin_Kistner2 · 17 April 2007 19:45

Further, (Windows, 1-click), same version:

C:\>ruby -v
ruby 1.8.6 (2007-03-13 patchlevel 0) [i386-mswin32]

C:\>irb -v
irb 0.9.5(05/04/13)

C:\>irb
irb(main):001:0> 1.0e+800
=> 1.0e+289
irb(main):002:0> 1.0e+900
=> NaN
irb(main):003:0> 1.0e+1000
=> NaN
irb(main):004:0> 1.0e+1100
(irb):4: warning: Float 1.0e+1100 out of range
=> Infinity

···

On Apr 17, 11:41 am, "Tim Pease" <tim.pe...@gmail.com> wrote:

Once more on the Windows 1-click installed version of Ruby.

> ruby -w float_bug.rb

ruby 1.8.6 (2007-03-13 patchlevel 0) [i386-mswin32]
Float::MAX = 1.79769313486232e+308
1.0e+300 = 1.0e+300
1.0e+301 = NaN
1.0e+308 = NaN
1.0e+700 = 1.0e+189

Chris_Shea1 · 17 April 2007 21:20

On OS X 10.4.9

mvb:~ cms$ ruby float_bug.rb
ruby 1.8.6 (2007-03-13 patchlevel 0)
Float::MAX = 1.79769313486232e+308
1.0e+300 = 1.0e+300
1.0e+301 = NaN
1.0e+308 = NaN
1.0e+700 = 0.0

And my Xubuntu box (pre-RUBY_PATCHLEVEL apparently):

ruby 1.8.4 (2005-12-24)[i486-linux]
Float::MAX = 1.79769313486232e+308
1.0e+300 = 1.0e+300
1.0e+301 = 1.0e+301
1.0e+308 = 1.0e+308
1.0e+700 = 0.0

···

On Apr 17, 12:41 pm, "Tim Pease" <tim.pe...@gmail.com> wrote:

I ran across some peculiar behavior yesterday when working with large
floating point values. Here is a small script that demonstrates the
behavior. My question is, is this the expected behavior or is this a
bug in Float?

> cat float_bug.rb

puts "ruby #{RUBY_VERSION} " +
"(#{RUBY_RELEASE_DATE} patchlevel #{RUBY_PATCHLEVEL}) " +
"[#{RUBY_PLATFORM}]"

puts "Float::MAX = #{Float::MAX}"
puts "1.0e+300 = #{1.0e+300}"
puts "1.0e+301 = #{1.0e+301}"
puts "1.0e+308 = #{1.0e+308}"
puts "1.0e+700 = #{1.0e+700}"

# EOF

When I run this script on my Windows machine, it produces the output I
would expect -- i.e. numbers above Float::MAX are "Infinity" and
numbers below Float::MAX are valid floating point numbers.

> ruby -w float_bug.rb

ruby 1.8.5 (2006-12-25 patchlevel 12) [i386-cygwin]
Float::MAX = 1.79769313486232e+308
1.0e+300 = 1.0e+300
1.0e+301 = 1.0e+301
1.0e+308 = 1.0e+308
1.0e+700 = Infinity

Now, when I run the exact same script on a Solaris machine, some odd
things start to happen

> ruby -w float_bug.rb

ruby 1.8.5 (2006-12-25 patchlevel 12) [sparc-solaris2.8]
Float::MAX = 1.79769313486232e+308
1.0e+300 = 1.0e+300
1.0e+301 = NaN
1.0e+308 = NaN
1.0e+700 = 8.55180271847103e+116

Whey are valid floats being treated as NaN? Better yet, why does
1.0e+700 get evaluated to 8.5518027e+116?

Once more on the Windows 1-click installed version of Ruby.

> ruby -w float_bug.rb

ruby 1.8.6 (2007-03-13 patchlevel 0) [i386-mswin32]
Float::MAX = 1.79769313486232e+308
1.0e+300 = 1.0e+300
1.0e+301 = NaN
1.0e+308 = NaN
1.0e+700 = 1.0e+189

Any thoughts on this one? I encourage you to copy the float_bug.rb
script and try it out for yourself.

Blessings,
TwP

Nobuyoshi_Nakada1 · 18 April 2007 01:42

Hi,

At Wed, 18 Apr 2007 03:41:22 +0900,
Tim Pease wrote in [ruby-talk:248272]:

I ran across some peculiar behavior yesterday when working with large
floating point values. Here is a small script that demonstrates the
behavior. My question is, is this the expected behavior or is this a
bug in Float?

Thank you, fixed now.

···

--
Nobu Nakada

Tim_Pease · 18 April 2007 14:57

Thank you very much, Nobu.

TwP

···

On 4/17/07, Nobuyoshi Nakada <nobu@ruby-lang.org> wrote:

Hi,

At Wed, 18 Apr 2007 03:41:22 +0900,
Tim Pease wrote in [ruby-talk:248272]:
> I ran across some peculiar behavior yesterday when working with large
> floating point values. Here is a small script that demonstrates the
> behavior. My question is, is this the expected behavior or is this a
> bug in Float?

Thank you, fixed now.

Topic		Replies	Views
NaN Error ruby-talk	2	71	17 April 2007
Is this a float-bug or naive comprehension? ruby-talk	2	125	18 July 2011
Floating point problem... floating indeed :( ruby-talk	3	118	1 December 2006
Possible bug in rb_float_new()? ruby-talk	2	143	26 July 2009
"nan".to_f? ruby-talk	28	341	16 October 2004

Bug in Float? (was: NaN Error)

Related topics