Java (Math.max) var0 != var0

Asked Jan 27 '18 at 22:03

Active Jan 28 '18 at 00:37

Viewed 83 times

Can anyone please explain to me what is the point of var0 != var0 in Math.max(a, b) method?

public static float max(float var0, float var1) {
  if (var0 != var0) {
    return var0;
  } else if (var0 == 0.0F && var1 == 0.0F && (long)Float.floatToRawIntBits(var0) == negativeZeroFloatBits) {
    return var1;
  } else {
    return var0 >= var1 ? var0 : var1;
  }
}

edited Jan 28 '18 at 00:37

Korashen

2,144
2
18
28

asked Jan 27 '18 at 22:03

Kamil Bęben

1,064
8
12

1

NaN perhaps? Not sure. – Hovercraft Full Of Eels Jan 27 '18 at 22:04
Also what's the idea behind converting `Float.floatToRawIntBits()` to long? You don't gain anything by doing that. – markspace Jan 27 '18 at 22:05
3

if (a != a) return a; // a is NaN – dreamcrash Jan 27 '18 at 22:05
Does that matter though? The last line should return var0 anyway if var0 is NAN. – markspace Jan 27 '18 at 22:06
https://stackoverflow.com/questions/10034149/why-is-nan-not-equal-to-nan – dreamcrash Jan 27 '18 at 22:07
Yes, NaN is it! – Hovercraft Full Of Eels Jan 27 '18 at 22:07
I think it's a bad test though, they should be testing var1 for NAN, because that won't be returned by the last line if either input is NAN. – markspace Jan 27 '18 at 22:08
@markspace The last condition returns `var1` if it's `NaN`, not `var0` since `var0 >= Float.NaN == false`. – Kayaman Jan 27 '18 at 22:12
Oh, ternary is true first, then false. Sorry got those mixed up. @Kayaman Now I'm just confused about the middle part. Testing for +0 vs -0? – markspace Jan 27 '18 at 22:15
1

@markspace yes, negative 0 is smaller than positive 0 in this case. Not the most straightforward method to read, considering all the intricacies involved. – Kayaman Jan 27 '18 at 22:17
As for why the check is done as `var0 != var0` instead of `Float.isNaN(var0)`, that's an optimization. – Kayaman Jan 28 '18 at 10:16

Java (Math.max) var0 != var0

0 Answers0