The world without float literal
?
0 likes?1,017 views
The document discusses the difficulties of using floating point numbers (Floats) in Ruby and proposes using Rational numbers instead. Some key points made include: - Floats are stored internally as binary which can lead to representation errors, while we think of them as decimal numbers - Floats actually represent a range of values rather than a single number - Using Rational would allow keeping numbers as exact fractions, avoiding errors from Float representations - While Rational may seem slower, optimizations are possible for common denominators and FastMath libraries could be used for Floats if needed
The world without float literal
- 1. The world without the literal notation for ?oating-point numbers , Kenta Murata GeneticLab Co., Ltd. Tuesday, July 19, 2011 1
- 2. CRuby Committer (BigDecimal) Ruby Sapporo Kosen Conf http://www.flickr.com/photos/koichiroo/5244581973/ Tuesday, July 19, 2011 2
- 3. Tuesday, 1 2010 3 July 19, 2011 3
- 4. RubyKaigi n !! Tuesday, 1 2010 3 July 19, 2011 4
- 5. Ruby 04 Sapporo RubyKaigi 04 Summer 2012 Tuesday, July 19, 2011 5
- 6. I have some question. Tuesday, July 19, 2011 6
- 7. Do you use Float values? Tuesday, July 19, 2011 7
- 8. Do you understand Float well? Tuesday, July 19, 2011 8
- 9. Float is too dif?cult. Tuesday, July 19, 2011 9
- 10. Reading dif?culty Tuesday, July 19, 2011 10
- 11. ruby-1.9.2-p290 :001 > 76287755398823936.0 => 76287755398823940.0 ruby-1.9.2-p290 :002 > Tuesday, July 19, 2011 11
- 12. Comparing dif?culty Tuesday, July 19, 2011 12
- 13. ruby-1.9.2-p290 :001 > 12.4 * 1.05 == 13.02 => false ruby-1.9.2-p290 :002 > 12.4 * 1.05 => 13.020000000000001 ruby-1.9.2-p290 :003 > Tuesday, July 19, 2011 13
- 14. Representation difference Tuesday, July 19, 2011 14
- 15. Human: Decimal Internal: Binary Tuesday, July 19, 2011 15
- 16. Concept difference Tuesday, July 19, 2011 16
- 17. We think Float is a number. Tuesday, July 19, 2011 17
- 18. It is real that Float is a range. Tuesday, July 19, 2011 18
- 19. They¨re troublesome Tuesday, July 19, 2011 19
- 20. There is a solution Tuesday, July 19, 2011 20
- 21. Why not use Rational. Tuesday, July 19, 2011 21
- 22. It is dif?cult, too. Tuesday, July 19, 2011 22
- 23. Ruby doesn¨t have any literal notations for Rational Tuesday, July 19, 2011 23
- 24. Does a fractional notation solve the problems? Tuesday, July 19, 2011 24
- 25. No! Tuesday, July 19, 2011 25
- 26. We naturally use decimal point notation. Tuesday, July 19, 2011 26
- 27. We need a decimal- point notation for representing a Rational. Tuesday, July 19, 2011 27
- 28. ruby--p290 :001 > 13.02.class => Rational ruby--p290 :001 > 12.4 * 1.05 == 13.02 => true ruby-1.9.2-p290 :002 > 12.4 * 1.05 => 13.02 ruby-1.9.2-p290 :003 > 13.02 / 3 => (217/50) Tuesday, July 19, 2011 28
- 29. Do you worry about Rational is slow? Tuesday, July 19, 2011 29
- 30. We can optimize for denominators of 10n Tuesday, July 19, 2011 30
- 31. Float isn¨t as so fast as you think Tuesday, July 19, 2011 31
- 32. If you need fast Float calculation Tuesday, July 19, 2011 32
- 33. Write it in C as extension library Tuesday, July 19, 2011 33
- 34. If you need to calculate with arrays of Float Tuesday, July 19, 2011 34
- 35. Use NArray Tuesday, July 19, 2011 35
- 36. Rational may be what you want Tuesday, July 19, 2011 36
- 37. This talk¨s background, demonstrations, and more detail descriptions Tuesday, July 19, 2011 37
- 38. Tuesday, July 19, 2011 38