|
Nov 2013
9:53am, 14 Nov 2013
13,657 posts
|
fetcheveryone
Yay, I killed the thread with maths 
|
|
Nov 2013
9:59am, 14 Nov 2013
19,182 posts
|
eL Bee!
Some of us are missing from the [All Time] Leaderboard
|
|
Nov 2013
9:59am, 14 Nov 2013
1,567 posts
|
Drell
That's really interesting. You just have to make sure that arccos isn't fed a number larger than 1. But also interesting that the rounding error could take it over 1 in the first place.
|
|
Nov 2013
10:08am, 14 Nov 2013
13,658 posts
|
fetcheveryone
I tracked it down to the sin() function (and I expect the cos() function is equally troubled).
Here's how I calculate the distance between two points (hopefully the site won't truncate it). Both machines were producing identical numbers until the row with the sin() functions. I've added a line below that to pull the value back down to 1 if it's crept over. If by chance, the troubled line undercalculates the value, it will have the effect of slightly increasing the returned distance - but shouldn't fall over.
function calculateDistance($lat1,$lng1,$lat2,$lng2) {
if ($lat1==$lat2 && $lng1==$lng2) {
return 0;
}
$radlat1 = pi() * $lat1/180;
$radlat2 = pi() * $lat2/180;
$radlon1 = pi() * $lng1/180;
$radlon2 = pi() * $lng2/180;
$theta = $lng1-$lng2;
$radtheta = pi() * $theta/180;
$dist = sin($radlat1) * sin($radlat2) + cos($radlat1) * cos($radlat2) * cos($radtheta);
$dist = acos($dist);
$dist = $dist * 180/pi();
$dist = $dist * 60 * 1.1515;
return $dist;
}
|
|
Nov 2013
10:17am, 14 Nov 2013
13,659 posts
|
fetcheveryone
Right then, fix in place, flowers.log closely monitoring the situation, and eL Bee! installed in the all time leaderboard at number two. Although... given the vastness of the flowers bug, I may well give the existing trails six weeks to burn out, and then drop back to monthly and yearly leaderboards for 2014.
|
|
Nov 2013
10:19am, 14 Nov 2013
1,568 posts
|
Drell
Well any error from truncating down to 1 is going to be pretty small compared to the variation in the earth's radius
|
|
Nov 2013
10:44am, 14 Nov 2013
16,196 posts
|
Night-owl
Well Fetch. I have no idea what you are talking about. But thankfully you do, and if it makes the game fairer even in 6 weeks time (ie. Christmas) Then that is good.
|
|
Nov 2013
10:52am, 14 Nov 2013
229 posts
|
Ceratonia
That value (1.000000000000000222044604925031) is the smallest possible floating point number above 1 that can be represented with the amount of binary precision you have. So it's definitely a rounding error, exactly the kind of thing that can happen when you move code between different floating point libraries.
|
|
Nov 2013
11:36am, 14 Nov 2013
1,032 posts
|
Molesy - 41 sleeps until Christmas
So what were the "certain circumstances" that triggered the bug?
|
|
Nov 2013
11:39am, 14 Nov 2013
2,310 posts
|
CStar (600)
Ah, it all makes sense now. (No it doesn't. I haven't a clue what you mean, but thank you for sorting out the problem!)
|
