9 ms·

Why 0.1 + 0.2 == 0.3 is false?

15 points by watermel0n 12y ago

pmoriarty 12y ago
In guile[1]: % guile guile> (= (+ 0.1 0.2) 0.3) #f [1] - https://www.gnu.org/software/guile/ https://www.gnu.org/software/guile/
- jgrant27 12y ago
  Guile(and most Lisps such as Common Lisp) support rationals specifically to avoid this problem, why not just use them ? scheme@(guile-user)> (= (+ 1/10 2/10) 3/10) $1 = #t
  - draegtun 12y ago
    perl6 also use rationals (by default)... $ perl6 > 0.1 + 0.2 == 0.3 True > (0.1 + 0.2).perl 3/10 For perl5 you must use the bigrat pragma: $ re.pl > 0.1 + 0.2 != 0.3 1 > use bigrat; > 0.1 + 0.2 == 0.3 1 ref: http://perldoc.perl.org/bigrat.html http://perldoc.perl.org/bigrat.html
  - pmoriarty 12y ago
    "why not just use them ?" Sometimes decimal notation is the most natural, and sometimes you're copying numbers in decimal notation out of a book, wikipedia, or some other publication.
JackC 12y ago
This headline is backwards -- it's the opposite of the actual stackoverflow question. The question is why, in the D language, `0.1 + 0.2 == 0.3` is true, when it should be false given the underlying floating point implementation. (Spoiler alert: the answer is that the addition is done at compile time rather than run time.)
sheetjs 12y ago
On a side note, Excel is infuriatingly inconsistent: some operations work in a way that .1 + .2 == .3 (for example, try `=0.1+0.2=0.3`) but others fall apart (displayed as a fraction with one digit, 0.3 -> 2/7 but 0.1+0.2 -> 1/3)
- taf2 12y ago
  floating point is hard. backwards compatibility is even harder
  - wlievens 12y ago
    For a modern spreadsheet it would not be unreasonable for a user to expect symbolic arithmetic rather than floating point error soup.
dottrap 12y ago
Douglas Crockford (JavaScript: The Good Parts) has talked about this problem with binary floating point to do decimal (human) math, on many occasions. Here is one such instance. http://www.yuiblog.com/blog/2009/03/10/when-you-cant-count-on-your-numbers/ http://www.yuiblog.com/blog/2009/03/10/when-you-cant-count-o... He has been on a crusade to introduce a decimal based floating point standard because he believes it is finally time we revisit this decision to use binary floating point. Edit: spelling

15 points by watermel0n 12y ago

In guile[1]: % guile guile> (= (+ 0.1 0.2) 0.3) #f [1] - https://www.gnu.org/software/guile/ https://www.gnu.org/software/guile/

jgrant27 12y ago
Guile(and most Lisps such as Common Lisp) support rationals specifically to avoid this problem, why not just use them ? scheme@(guile-user)> (= (+ 1/10 2/10) 3/10) $1 = #t
- draegtun 12y ago
  perl6 also use rationals (by default)... $ perl6 > 0.1 + 0.2 == 0.3 True > (0.1 + 0.2).perl 3/10 For perl5 you must use the bigrat pragma: $ re.pl > 0.1 + 0.2 != 0.3 1 > use bigrat; > 0.1 + 0.2 == 0.3 1 ref: http://perldoc.perl.org/bigrat.html http://perldoc.perl.org/bigrat.html
- pmoriarty 12y ago
  "why not just use them ?" Sometimes decimal notation is the most natural, and sometimes you're copying numbers in decimal notation out of a book, wikipedia, or some other publication.

JackC 12y ago

This headline is backwards -- it's the opposite of the actual stackoverflow question. The question is why, in the D language, `0.1 + 0.2 == 0.3` is true, when it should be false given the underlying floating point implementation. (Spoiler alert: the answer is that the addition is done at compile time rather than run time.)

sheetjs 12y ago

On a side note, Excel is infuriatingly inconsistent: some operations work in a way that .1 + .2 == .3 (for example, try `=0.1+0.2=0.3`) but others fall apart (displayed as a fraction with one digit, 0.3 -> 2/7 but 0.1+0.2 -> 1/3)

taf2 12y ago
floating point is hard. backwards compatibility is even harder
- wlievens 12y ago
  For a modern spreadsheet it would not be unreasonable for a user to expect symbolic arithmetic rather than floating point error soup.

dottrap 12y ago

Douglas Crockford (JavaScript: The Good Parts) has talked about this problem with binary floating point to do decimal (human) math, on many occasions. Here is one such instance. http://www.yuiblog.com/blog/2009/03/10/when-you-cant-count-on-your-numbers/ http://www.yuiblog.com/blog/2009/03/10/when-you-cant-count-o... He has been on a crusade to introduce a decimal based floating point standard because he believes it is finally time we revisit this decision to use binary floating point. Edit: spelling