[Module] Collision

[Aange10]

Features

This module does collision, so you don't have to! It takes all the stress out of collision, by allowing you to simply input the co-ordinates, and that's it. *Relatively no math involved.
Specifically, checking rectangular collisions becomes easier because this module checks 14 points, increasing accuracy, and keeping you from running needless perimeterfors. *Shutters*

How it Works

Here's a simple breakdown:

CircleToCircle : The module finds the distance between the centers of the two circles and compares it the the sum of the radii.
RectangeToRectangle : The module distinguishes a plane for the Rectangle, and detects 14 key points and looks for plane interception.
RectangleToCircle : The module fabricates a box to represent the Circle, and detects its 14 points along with the Rectangles.

Syntax

The syntax is super simple. [See The Turing Walkthrough for importing modules]

After importing the module you can call it by these parameters:

Turing:

Collision.circleToCircle (Ball_1_x, Ball_1_y, Ball_1_xr, Ball_1_yr, Ball_2_x, Ball_2_y, Ball_2_xr, Ball_2_yr) Collision.rectangleToRectangle (Rect_1_x1, Rect_1_y1, Rect_1_x2, Rect_1_y2, Rect_2_x1, Rect_2_y1, Rect_2_x2, Rect_2_y2) Collision.rectangleToCircle (Rect_x1, Rect_y1, Rect_x2, Rect_y2, Ball_x, Ball_y, Ball_xr, Ball_yr)

The syntax is very self explanatory, and all of the functions output a Boolean that indicates collision.

The parameters are also all integers.

Instructions

1) Import Collision.tu
2) Call the functions with their appropriate syntaxs.
3) Done!

The Module

Just copy and paste this in a Turing File, and name it "Collision.tu"!

[Raknarg]

Pretty useful. However, to be honest, the rectangle to rectangle seems to be like a lot more work than it should've been.

[Aange10]

Why so? A lot of times people settle for checking 4 points (corners), or decide to check the perimeter.

I do see how this could be a bit more, but i built it to reasonably check the entire rectangle, without actually doing so. The 14 strategic points allot a more accuracy than corners, but less time than scanning the perimeter.

However, I also designed it to check for collision at the most common intersections first, and find its result as quickly as possible.

[Raknarg]

Actually, now that I think about it, my way would have taken just as much time haha nevermind

[Aange10]

Out of curiosity, what would your way have been?

[Raknarg]

It would have just been a corner comparison... but I was thinking of my game where I only have to compare 1 corner to 4 corners instead of 4 to 4

[mirhagk]

If x1 and x2 are ordered so x1 is smaller than x2, then your rectangular collision detection can be a lot shorter and faster.

Turing:

%   Rectangle to Rectangle fcn rectangleToRectangle (x1_1, y1_1, x2_1, y2_1, x1_2, y1_2, x2_2, y2_2 : int) : boolean     if (x2_1 > x1_2) then %right side or 1st is greater than left side of 2nd         if (x1_1 < x2_2) then %left side of 1st is less then right side of 2nd             %if both of these are true then we know that the first rectangle has intersecting the 2nd on the x-axis. just do the same with the y             if (y2_1 > y1_2) then                 if (y1_1 < y2_2) then                 end if             end if         end if     end if end rectangleToRectangle

EDIT: Also your circle to circle accepts 2 radius's for each circle, which suggests that the function calculates the intersection of 2 ovals (which it doesn't). You can shorten that to just accept one radius for each, and then only have the 1 if statement (instead of 4)

[Zren]

You assume x1 and y1 are smaller than x2 and y2. You should *not* assume this. Your function should determine the smallest first. Otherwise you end up with a negative width.

Also learn about bitwise operators (and, or, ...). Your if chain could be broken down into:

code:

fcn blarg     result a <= b and b <= c and c <= d end blarg

Turing is also doesn't force you to end a statement at the end of the line. So for really long lines, you can do:

code:

fcn blarg     result a <= b         and b <= c         and c <= d end blarg

[Aange10]

Quote:

Also learn about bitwise operators (and, or, ...). Your if chain could be broken down into:

I do know about these. However, I was thinking that splitting up the if statements would allow the program to determine a false result faster. However my err is in that I forgot that in an or statement the program will stop computing the rest of the statement if one of the ors become true.

My mistake. I'm going to implement both of your suggestions.

[crossley7]

Just wondering why you need to check 14 points for rectangle to rectangle because if they intersect, 1 of the corners will overlap except in the case where 1 is very tall and narrow and the other is shorter and wider and the whole thing overlaps. you just have to check that case so there are really 9 cases. the 8 corners (2 rectangles) inside the other rectangle and the edge case where the centres overlap.

For circle to rectangle, you have to check 9 things again. The radius on corners and then radius overlapping a side and circle completely within rectangle. (Circle larger than rectangle should be covered by all of the first 8)

[mirhagk]

Is there a reason the algorithm I suggested can't be used? Of course you will also need to make sure the points are ordered correctly (but this is easy to do)

[Aange10]

mirhagk @ 18/1/2012, 12:09 am wrote:

Is there a reason the algorithm I suggested can't be used? Of course you will also need to make sure the points are ordered correctly (but this is easy to do)

Nope, I'm working on that right now.

[Aange10]

Okay, I've implemented the new formulas, circle collision also now works with ovals.


Critiques, please?


I'm going to attach it, since it's 190 lines.

[Zren]

Bounding boxes are usually suppose to surround the entire shape. It's usually better to have false positives then not detecting possible collisions (Eg: better to have some places the character can't move to then being half inside a wall).

Also this: http://imgur.com/usygV

There does exist ways to test oval collisions perfectly, but it's rather non-critical. You're better off wondering how to test collisions on regular shapes once those shapes have been rotated.

[Aange10]

Quote:

Bounding boxes are usually suppose to surround the entire shape. It's usually better to have false positives then not detecting possible collisions (Eg: better to have some places the character can't move to then being half inside a wall).

I've made the boxes all bigger.


I'm going to implement that CircleToRectangle formula, and to insure ovals will still work, I'll use my current formula for ovals, and this new one for circles (when comparing to rectangles)