The JVM has no choice: it must faithfully implement the IEEE floating- point spec (http://en.wikipedia.org/wiki/IEEE_754-2008), which specifies limited precision. By asking it to use floats, you are demanding that it accept rounding errors. If you want precision, there are lots of ways to get it; the most common is to cast everything up to integers and perform the computation there, then convert back down later:
user=> (def numbers [12.305 12.3049]) #'user/numbers user=> (/ (apply - (map #(int (* 10000 %)) numbers)) 10000) 1/10000 Or you can tell Java to use BigDecimals, which have arbitrary precision: user=> (- 12.305M 12.3049M) 0.0001M You could use clojure's excellent ratio type instead: user=> (def nums (map #(/ % 10000) [123050 123049])) #'user/nums user=> nums (2461/200 123049/10000) user=> (apply - nums) 1/10000 Computers don't lie; they just do what you tell them to. On Oct 12, 9:17 am, cej38 <junkerme...@gmail.com> wrote: > I keep running into this type of problem: > > user=> (- 12.305 12.3049) > 9.999999999976694E-5 > > The computer (probably the JVM) has just lied to me. Any fourth grade > student will know that this does not equal 0.0001. This would be less > of a problem is the JVM was consistent; if it were consistent then the > following equality would be true: > > user=> (= 0.0001 (- 12.305 12.3049)) > false > > Now it has lied to me again! > > I need a method to reliably compare two numbers. My first choice > would be to have the JVM compute (- 12.305 12.3049) correctly. > Barring that, I need a way to, without fail, compute the truthfulness > of the following: =, <, >, <=, >=. -- You received this message because you are subscribed to the Google Groups "Clojure" group. To post to this group, send email to clojure@googlegroups.com Note that posts from new members are moderated - please be patient with your first post. To unsubscribe from this group, send email to clojure+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/clojure?hl=en
