Timeline for How to overcome inaccuracy in Java
Current License: CC BY-SA 3.0
8 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Mar 5, 2014 at 5:41 | comment | added | Hungry Blue Dev | I guess you're right... I must live with it. :-) So, I guess in certain specific cases, we must return expected values, rather than the seemingly-wrong-but-very-close calculated values | |
| Mar 5, 2014 at 5:39 | vote | accept | Hungry Blue Dev | ||
| Mar 5, 2014 at 5:34 | comment | added | yshavit |
No, because there's just no way to represent π/2 exactly with a float or double. If you had a library that took degrees (rather than radians), you could do that -- since it is possible to represent 90.0 exactly. But afaik, StrictMath doesn't do that.
|
|
| Mar 5, 2014 at 5:32 | history | edited | yshavit | CC BY-SA 3.0 |
added 1031 characters in body
|
| Mar 5, 2014 at 5:25 | comment | added | Hungry Blue Dev | So, what if I change my representation system to Radians and not degrees? will it help? | |
| Mar 5, 2014 at 5:24 | comment | added | yshavit |
@ambigram_maker Again, you have to pick some "large-enough" number at which you say "okay, close enough to infinity". Doubles actually can store infinity, but you don't get that result because StrictMath.toRadians(90) returns a number really close to, but not exactly pi/2 (which you can't represent in a finite number of bits).
|
|
| Mar 5, 2014 at 5:20 | comment | added | Hungry Blue Dev |
Well, that's solved the _0.5 stored as ` 0.49999999999999994` problem. What do I do about infinity?
|
|
| Mar 5, 2014 at 5:18 | history | answered | yshavit | CC BY-SA 3.0 |