Float round bug in ruby?

Question

ruby-1.8.7 > 1.55.round(1)
 => 1.6 
ruby-1.8.7 > 1.555.round(2)
 => 1.56 
ruby-1.8.7 > 1.155.round(2)
 => 1.16 
ruby-1.8.7 > 10.156.round(2)
 => 10.16
ruby-1.8.7 > 10.155.round(2)
 => 10.15 
ruby-1.8.7 > 10.165.round(2)
 => 10.16

What gives? Am I missing something?

EDIT

ruby-1.9.2 > 10.155.round(2)
 => 10.15 
ruby-1.9.2 > 10.165.round(2)
 => 10.16

Probably related: http://stackoverflow.com/questions/1950541/consistent-rounding-of-floating-points-in-ruby — Thilo, Jan 05 '12 at 12:18
Thanks, this at least provides a workaround for me. And maybe it really is related to floating point imprecision... — tbk, Jan 05 '12 at 12:29
Also related: http://stackoverflow.com/questions/4055618/ruby-floating-point-errors — Rick Regan, Jan 05 '12 at 13:27
[Don't](http://bugs.ruby-lang.org/projects/ruby/wiki/HowToReject) file it as a bug! — Andrew Grimm, Jan 05 '12 at 22:17

score 4 · Accepted Answer · answered Jan 05 '12 at 12:31

Floating point values are not precise. Your 10.165 is represented on paper/the screen as 10.165, but in memory it's represented as something extremely close to 10.165... whether it rounds up or down is a matter of which direction the error happens to fall in.

If you need accurate handling of fractional numbers, you can either represent them as BigDecimal (with a decimal scale/precision notation), or as Rational (with a fractional numerator/denominator notation).

score 2 · Answer 2 · answered Jan 09 '12 at 13:06

2

The Python docs have a good explanation of the fundamental limitations of binary floating-point numbers: http://docs.python.org/tutorial/floatingpoint.html

answered Jan 09 '12 at 13:06

Ned Batchelder

364,293
75
561
662

Float round bug in ruby?

2 Answers2