Java Atan2() Sort violates general contract?

Question

I'm having trouble with a comparison method being used in Java/Processing.

It says it violates its general contract, which I understand means it doesn't work on the relationships between the items consistently...

But I don't understand why. Since it just returns the angle, theta, based on the ratio of ycor to xcor (like sin to cos), shouldn't it sort them consistently based on their positions from 0 to 360 degrees?

public int compareTo(Orb other) {
double X = Math.atan2(ycor,xcor);
if (Math.atan2(other.ycor,other.xcor) > X) {
  return -1;
}
if (Math.atan2(other.ycor,other.xcor) == X) {
  return 0;
}
return 1;
}

Any help is very appreciated, thanks!

The specific context is that the error occurs when a Collection.Sort() is run.

"Comparison Method Violates Its General Contract!" according to Processing. — Albert Mokrejs, Jun 11 '15 at 22:00
Is it possible that `xcor` and `ycor` are both zero for some of these? I suspect that would result in NaN. Switching to `Double.compare(Math.atan2(ycor, xcor), Math.atan2(other.ycor, other.xcor))` ought to work. — Louis Wasserman, Jun 11 '15 at 22:01
I take it that you solved this already, but if you want a full answer, you'll have better luck if you post an [MCVE](http://stackoverflow.com/help/mcve) that demonstrates the problem. — Kevin Workman, Jun 12 '15 at 00:38

score 0 · Answer 1 · edited Jun 20 '20 at 09:12

0

EDIT: See Louis Wasserman's comment above:

Is it possible that xcor and ycor are both zero for some of these? I suspect that would result in NaN. Switching to Double.compare(Math.atan2(ycor, xcor), Math.atan2(other.ycor, other.xcor)) ought to work.

THE BELOW WAS THE CONTENT OF THIS POST; IT IS PROBABLY WRONG

It's not well-defined whether one angle is "greater" than another - if I'm clockwise of you by 170 degrees, then I'm also counterclockwise of you by 190 degrees.

Even if you choose to use "within 180 degrees of" as greater-than, you still don't have the triangle inequality satisfied.

edited Jun 20 '20 at 09:12

Community

1
1

answered Jun 11 '15 at 22:01

Jakob Weisblat

7,450
9
37
65

That's a good point, thanks! Can you suggest an alternative way to sort in such a way then (Resulting in the clockwise ordering of the points)? – Albert Mokrejs Jun 11 '15 at 22:03
@AlbertMokrejs, can you find a way to express what you *mean* by that? – dfeuer Jun 11 '15 at 22:05
Essentially starting from 0 and going clockwise until 359.99_ which order would one encounter points. – Albert Mokrejs Jun 11 '15 at 22:06
hmm, that seems to be what you're doing already - perhaps my answer is actually wrong? – Jakob Weisblat Jun 11 '15 at 22:11
1

I do not know. The solution suggested by Wasserman works though, so I believe it might be an error involving zeros as suggested. – Albert Mokrejs Jun 11 '15 at 22:27

Java Atan2() Sort violates general contract?

1 Answers1

THE BELOW WAS THE CONTENT OF THIS POST; IT IS PROBABLY WRONG