[BEHAVIOR] Chipmunk Physics

R0J0hound , awesome..

found a bug when you set object to immovable yes after adding a joint, it will cause an error.. I did this during on post collision- not sure if it happens outside of that.

I'm also getting some errors sometimes- not sure why yet.. saying that can't set body angle to nan.. I think it happens when it tries to set angle towards position and the position is ontop of itself.. or something.. not sure though.. this happens during on post collision.

R0J0hound

Does this mean we can create bridge building type games where the stress on joints can be calculated ?

R0J0hound, Am playing with the plugin and wanted to create Newtons cradle, something like this in the video - ( after 25 seconds).

Attached is my .capx file. I am facing 3 issues -

1. Issue#1 - When I have only one ball dragged to the left and released, it works fine. But when I drag the ball to almost horizontal(90 degrees) and release, then I notice the right most ball goes beyond 90 degrees ( against conservation energy law)...Any idea what could be the issue?

2. Issue#2 - When DragTwo group is enabled (let chipmunk Immovable to Yes be still disabled), then it drags two balls to the left on touch. When touch is released, somehow the collisions are not working as they way they should in Newtons cradle...The third ball from the left doesnt go as much high as the right most 4th one...

3. Issue#3 - When I try to set ChipmunkImmovable to Yes on touch and then to "No" on touch released, it gives me a javascript error in the plugin behavior. Not sure if this is a bug.

Please let me know if this can be done as demonstrated in the video above? Having done many physics simulations, I have always found achieving NewtonsCradle to be pretty difficult using Box2D physics - revolute joints somehow...Chipmunk helps me to get this a bit further but still I seem to be struck...

kmsravindra

This version works more like it should, although there are some odd bouncing effects.

spongehammer, Thanks...But all the 3 issues that I have mentioned above still exist...None of them go away

Newton's cradle in 3 events The key is to use the circle collision shape, max elasticity & zero friction.

Still not true physics though as i am pretty sure real steel balls dont have any elasticity.

Edit: Corrected by R0J0hound below . I learn something new everyday.

mattb, sorry.. it doesn't work at all the way a newtons cradle should...There are all types of things happening with collisions in your example...The first thing is that I don't think using drag-drop works well with chipmunk physics to drag chipmunk objects over other...

It has always been stated that the physics behavior shouldn't be mixed with the other behaviors, as unforeseen things can happen.

Ok here's another version without the drag-drop, & just an impulse to start things off. There's still some movement of the inner balls (which gets worse if you enable the infinity sleep time action) but it's pretty close to the expected result i think, & probably as good as the physics will allow.

Prominent thanks for the report. I'll be looking into it.

spongehammer

You can already get the impulse applied to a joint so yes you can make something like that.

kmsravindra

#1 Physics libraries are only a approximation, so energy isn't always conserved. You can increase the "global settings"->iterations to make it more accurate. You can also make the fixed timestep smaller to make collisions more accurate as well.

#2 Toggling immovable isn't the best way to drag objects around. The switch discards any contact points and isn't very physically correct. A better way is to attach a joint to an object that is moved directly. See bottom of post.

#3 If it causes a js error then it's a bug. I'll look into it.

spongehammer

It's correct, elasticity in physics is how a collision reacts, not how rubbery it is. An elasticity of 1 should mean that no energy is lost in a collision, which is pretty much how ball bearings act when they collide.

Attached is a physics way to drag objects around. When you click on an object you create a joint to a no collision, immovable sprite that is moved around by the mouse. Then when the mouse is released the joint is removed.

Also the joints in the capx are set to break if you're too rough.

if it helps, this is the error I get about the angle thing..

Error: Assertion failed: Internal Error: Attempting to set body's angle to NaN

at assert (http://localhost:50000/rojoChipmunkBeta ... or.js:51:9)

at Body.setAngleInternal (http://localhost:50000/rojoChipmunkBeta ... .js:1725:2)

at Body.position_func (http://localhost:50000/rojoChipmunkBeta ... .js:1768:7)

at Space.step (http://localhost:50000/rojoChipmunkBeta ... js:4998:14)

at behinstProto.tick (http://localhost:50000/rojoChipmunkBeta ... js:6983:17)

at Runtime.logic (http://localhost:50000/preview.js:2346:29)

at Runtime.tick (http://localhost:50000/preview.js:2066:8)

at tickFunc (http://localhost:50000/preview.js:629:38)

R0J0hound - a couple of things:

1. Are the single contact points of the green boxes below an expected behavior?

[attachment=1:3pql80c7][/attachment:3pql80c7]

[attachment=0:3pql80c7][/attachment:3pql80c7]

2. Could you give a quick explanation or example of how to use the hexadecimal numbers for collision layers?

Thanks

Also been getting this error.. not sure what is causing it though.. edit:seems to occur when a line collides in two places at the same time...

Error: Assertion failed: Cannot remove a shape that was not added to the space. (Removed twice maybe?)

at assert (http://localhost:50000/rojoChipmunkBeta ... or.js:51:9)

at Space.removeShape (http://localhost:50000/rojoChipmunkBeta ... .js:3889:3)

at behinstProto.updateShape (http://localhost:50000/rojoChipmunkBeta ... js:6642:16)

at behinstProto.changeShapeCheck (http://localhost:50000/rojoChipmunkBeta ... js:6956:10)

at behinstProto.getBody2 (http://localhost:50000/rojoChipmunkBeta ... .js:7598:8)

at Acts.addPivotJoint (http://localhost:50000/rojoChipmunkBeta ... js:7643:20)

at Action.run_object (http://localhost:50000/eveng.js:1843:10)

at EventBlock.run_actions_and_subevents (http://localhost:50000/eveng.js:929:38)

at EventBlock.run (http://localhost:50000/eveng.js:886:9)

at EventBlock.run_subevents (http://localhost:50000/eveng.js:994:11)

at EventBlock.run_actions_and_subevents (http://localhost:50000/eveng.js:933:8)

at EventBlock.run (http://localhost:50000/eveng.js:886:9)

EDIT: seems to occur when a line collides in two places at the same time...

R0J0hound , also, I'm noticing that if a line is created overlapping another object so that it extends across two sides, it'll position the line twice the depth rather than just once- this causes the line to be pushed out way too far. I'm guessing that it somehow positions it for each collision point, which is why it pushes it out too far.

edit: actually.. to be more accurate, I tried to push it out via collision depth * normal.. but it seems to be pushing it too far out when it extends across to sides. If it extend across one side it pushes out fine.

edit2: actually maybe it's fine. Since I move it to the collision point, and then push it, seems like I don't need to push it out. For some reason though, sometimes the collision point is inside objects, and I'm not sure how to get the collision point on the edge of the object when this happens..