Why are collision groups removed from "REV" version? about oimo.js HOT 7 CLOSED

is the dev version
speed improved, more broadphase, more logical
rev is the basic test dev the best optimised version and yes many differency.
cylinder suport on rev is not finish and have many bug.

think like this
rev is the old grand'pa dev is the young child

from oimo.js.

hmm okay, i figured so much after reading through the code. I'm trying to figure out how to port over cylinders. Most of the code seems to be copyable from rev to dev, but the whole collision handling system has changed as you mention. so far i'm trying to understand how the collision handling has improved / changed from using "result.addContactInfo" in REV to "manifold.addPoint" in DEV

ContactManifold seems to have completely superseded CollisionResult then?

can you explain some of the main changes between the two that i'd have to be aware of when porting cylinders over? i made the necessary changes to the proxy generator and updated it to use setup mass, but my objects just disappear when they collide, so i'm missing something, it might be an error in some of the other stuff i changed though. I'll have to do more investigation but any help would be appreciated. I'm using this to figure out just how contacts / shapes are handled so i can figure out how to improve the system

from oimo.js.

the main process of dev is in collision/narrowphase
for each new shape you have to add collision for other shape
and next add final declaration to world.js line 59 this.detectors...

from oimo.js.

I've done the last two points you made, i'll investigate collision/narrowphase.

I think i've gotten sphere-cylinder to work right now somewhat, but no luck with cylinder+cylinder as of yet.

Also for some reason in your cylinder code from REV, the moment of inertia around the y axis is calculated as [(1/2) x radius x radius], but i'm pretty sure it should be [mass x radius x radius] for a cylinder. Any reason why it was done that way or just a typo?

from oimo.js.

yea rev code of cylinder is not good

from oimo.js.

I've modified the DEV version to use a modified version of the rev cylinder code which uses manifolds / etc. I'm in the process of making it use OIMO_ARRAY_TYPE instead of normalwidth etc. Any extra changes I should make for it to be suitably merged into DEV? also how should i go about including these changes. Right now i've got a directly modified "oimo.dev" build, but i don't know if i have to make the pull requests on individual components.

Also, theres a variable name "dimentions" used in shapes, but the word should probably be "dimensions" instead, as this is the correct spelling.

Going to try and implement some triangle mesh / polygon soup support after this. I'm thinking we'll need octree for a secondary "broadphase"/"midphase", or at least a spatial hash (but the triangles will be rigid/static so i'd imagine an octree would work better in this case) but i'm not sure if this is the best way to tackle it. Have any insight that might help? this is a nice discussion on the topic : http://www.bulletphysics.org/Bullet/phpBB3/viewtopic.php?f=4&t=592

as far as i can tell you're using SAT for most collisions right? i haven't looked deep into your code as i'm not planning to reverse engineer anything if i can just ask you :)

I'd also like to investigate the problem of objects "tunneling" into each other, it seems they aren't pushed out of a penetration adequately.

from oimo.js.

wow cool, yes you can pull requests or if many change create new version DEV2
OIMO_ARRAY_TYPE is better if you have fixed length
yes dimentions is in french :)
triangle mesh / polygon should be the best update but is not easy maybe look at cannon.js is a pure javascript engine to.
SAT is far the best method in any case

from oimo.js.