Lo que todo programador debería saber sobre aritmética de punto flotante
allow for user-supplied epsilon
+15
-6
+14
-4
content/errors/NearlyEqualsTest.java
+14
-4
content/errors/NearlyEqualsTest.java
···
4
4
import org.junit.Test;
5
5
6
6
/**
7
-
* Demonstrates a good method for comparing floating-point values
8
-
* using an epsilon. Run via JUnit 4
7
+
* Test suite to demonstrate a good method for comparing
8
+
* floating-point values using an epsilon. Run via JUnit 4.
9
+
*
10
+
* Note: this function attempts a "one size fits all" solution. There
11
+
* may be some edge cases for which it still produces unexpected results,
12
+
* and some of the tests it was developed to pass probably specify behaviour
13
+
* that is not appropriate for some applications. Before using it, make
14
+
* sure it's appropriate for your application!
9
15
*
10
16
* From http://floating-point-gui.de
11
17
*
···
13
19
*/
14
20
public class NearlyEqualsTest
15
21
{
16
-
public static boolean nearlyEqual(float a, float b)
22
+
public static boolean nearlyEqual(float a, float b, float epsilon)
17
23
{
18
-
final float epsilon = 0.000001f;
19
24
final float absA = Math.abs(a);
20
25
final float absB = Math.abs(b);
21
26
final float diff = Math.abs(a-b);
···
26
31
} else { // use relative error
27
32
return diff / (absA+absB) < epsilon;
28
33
}
34
+
}
35
+
36
+
public static boolean nearlyEqual(float a, float b)
37
+
{
38
+
return nearlyEqual(a, b, 0.000001f);
29
39
}
30
40
31
41
/** Regular large numbers - generally not problematic */
+1
-2
content/errors/comparison.html
+1
-2
content/errors/comparison.html
···
38
38
39
39
Also, the result is not commutative (`nearlyEquals(a,b)` is not always the same as `nearlyEquals(b,a)`). To fix these problems, the code has to get a lot more complex, so we really need to put it into a function of its own:
40
40
41
-
public static boolean nearlyEqual(float a, float b)
41
+
public static boolean nearlyEqual(float a, float b, float epsilon)
42
42
{
43
-
float epsilon = 0.000001f;
44
43
float absA = Math.abs(a);
45
44
float absB = Math.abs(b);
46
45
float diff = Math.abs(a-b);