Bug #16843: A bug with floating point multiplication

• Status: Rejected
• Assignee: -
• Target version: -
• ruby -v: 2.6.6, 2.7.1
• Backport: 2.5: UNKNOWN, 2.6: UNKNOWN, 2.7: UNKNOWN

[FedorKK (Fedor Koshel)]

I've reproduced it with both currently stable versions: 2.6.6 and 2.7.1.

irb(main):001:0> 0.29 * 100
=> 28.999999999999996
irb(main):002:0> 0.29 * 100.0
=> 28.999999999999996
irb(main):003:0> 0.28 * 100
=> 28.000000000000004
irb(main):004:0> 0.28 * 100.0
=> 28.000000000000004
irb(main):005:0> 0.27 * 100.0
=> 27.0
irb(main):006:0> 0.27 * 100
=> 27.0

$ ruby -v                                                                                                                                                                                                 
ruby 2.7.1p83 (2020-03-31 revision a0c7c23c9c) [x86_64-linux]

irb(main):001:0> 0.29 * 100
=> 28.999999999999996
irb(main):002:0> 0.29 * 100.0
=> 28.999999999999996
irb(main):003:0> 0.28 * 100.0
=> 28.000000000000004
irb(main):004:0> 0.28 * 100
=> 28.000000000000004
irb(main):005:0> 0.27 * 100
=> 27.0
irb(main):006:0> 0.27 * 100.0
=> 27.0


$ ruby -v                                                                                                                                                                                                 
ruby 2.6.6p146 (2020-03-31 revision 67876) [x86_64-linux]

[sawa (Tsuyoshi Sawada)]

What is the bug?

[FedorKK (Fedor Koshel)]

The behaviour is super unpredictable. I mean, the problem with floats is known, but it usually happens with precision values, or huge numbers. In this case, if you try to get an integer from 0.29*100 you'll get 28. Which is strange.

If you think, this is just another float discuss, you can close the ticket. But for me the work with simple numbers should be more predictable.

[ko1 (Koichi Sasada)]

FYI:

$ python
Python 2.7.17 (default, Apr 15 2020, 17:20:14)
[GCC 7.5.0] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> 0.29 * 100
28.999999999999996
>>> 0.29 * 100.0
28.999999999999996
>>> 0.28 * 100
28.000000000000004
>>> 0.28 * 100.0
28.000000000000004
>>> 0.27 * 100.0
27.0
>>> 0.27 * 100
27.0

[FedorKK (Fedor Koshel)]

Yes, but for example in Go:

func main() {
	fmt.Printf("%f", 0.29 * 100)
	fmt.Println()
	fmt.Printf("%f", 0.28 * 100)
	fmt.Println()
	fmt.Printf("%f", 0.27 * 100)
}

29.000000
28.000000
27.000000

[sawa (Tsuyoshi Sawada)]

as well in Ruby:

"%f" % (0.29 * 100) # => "29.000000"
"%f" % (0.28 * 100) # => "28.000000"
"%f" % (0.27 * 100) # => "27.000000"

[FedorKK (Fedor Koshel)]

Ok, let's close it. It's really doesn't make any sense to improve the standard representation. But still:

func main() {
        fmt.Println(0.29 * 100 == 29.0)
}
true

irb(main):001:0> 0.29 * 100 == 29.0
=> false

[hsbt (Hiroshi SHIBATA)]

• Status changed from Open to Rejected

[duerst (Martin Dürst)]

Just as an add-on:
Go has similar effects to those reported here for Ruby, see e.g. https://stackoverflow.com/questions/33640943/golang-float-arithmetic.
@FedorKK, if you can find out why e.g. fmt.Println(0.29 * 100 == 29.0) prints true in Go, please tell us.