Update to latest version of three-to-cannon. Fixes #160.

[MignonBelongie]

Fixes #160. To see the corrected behavior, set debug: true in examples/cannon.html, and see that the wireframe for the cone now has the correct orientation.

See https://github.com/donmccurdy/three-to-cannon/pull/27 for the fix.

[Galadirith]

Thanks so much for preparing this and your much larger associated contributions in donmccurdy/three-to-cannon#27 😊 I think package-lock.json should also be updated to reflect the changes to package.json. This can be done with (reference)

npm install --save --force three-to-cannon@3.0.2

three.js has also become a peer dependency for three-to-cannon since v2.1.0 which means that with the installed version (rather than linked package) we'll get test and build errors in APS. I think it should be sufficient to also just add three as a peer dependency just like in https://github.com/donmccurdy/three-to-cannon/commit/d85ced9cc9ab677ff4bc50901aa7ad4c8805da32. And then install it locally to get package-lock.json to update

npm install three@0.115.0

We will also need to update our distribution command to exclude three from the bundle, again just as was done in three-to-cannon treating three as an external library that will be available globally for users. browserify has a similar --external option, and I think something like the following should be a sufficient replacement

browserify index.js -o dist/aframe-physics-system.js -x three 
budo index.js:bundle.js -d examples --port 8000 -x three

However I don't think browserify has the same --globals option. Looking at some of the distributions for three-to-cannon I'm not actually sure what --globals three=THREE is doing in the microbundle execution in the dist command. I assumed it would just declare three=THREE in the output bundle, but I don't see that. So I'm not sure right now what is the right thing to do here, do you have any thoughts on this?

Running the tests do work with these changes. I think technically the tests do include three in the test bundles, which means it's not exactly the same as the distributed bundles, but I think the same is also true in three-to-cannon where three is not excluded. So I would be happy to leave the test commands as they are.

[Galadirith]

Really sorry, I made a mistake in the budo dev command, the correct one should be

budo index.js:bundle.js -d examples --port 8000 -- -x three

However running this actually fails with an error

Uncaught Error: Cannot find module 'three'
    at o (_prelude.js:1)
    at _prelude.js:1
    at Object.6.cannon-es (three-to-cannon.js:1)
    at o (_prelude.js:1)
    at _prelude.js:1
    at Object.10.../../../lib/CANNON-shape2mesh (body.js:2)
    at o (_prelude.js:1)
    at _prelude.js:1
    at Object.1../src/components/ammo-constraint (index.js:5)
    at o (_prelude.js:1)

when trying to view http://localhost:8000/spring.html. So there is something wrong going on here, and it probably has something to do with the --globals option. I'm happy to look into this further if that would be helpful? When Don's status is no longer on vacation maybe we could also ask for some advice on this, and what --globals is doing in the microbundle command.

[MignonBelongie]

I've been trying to figure this out. The problem, as I understand it, is not that aframe-physics-system.js doesn't work, but just that it contains all of three.js, and thus is huge. Is that right?

aframe.js defines THREE, and thus it is already defined when aframe-physics-system.js is loaded. Previously, that was good enough for three-to-cannon, which also used THREE. Now, instead, three-to-cannon imports from 'three', which is why I had to add 'three' as a dependency in package.json. To make things more complicated, aframe uses 'super-three', not 'three'. I don't think we can expect three-to-cannon to change to import from 'super-three' instead of 'three', and I don't even know if that would help, so, I think the question is how do we make browserify only include the parts of three that we need (instead of all 1.2 MB (600K for three.min.js)). Am I understanding the problem correctly?

[Galadirith]

I've been trying to figure this out. The problem, as I understand it, is not that aframe-physics-system.js doesn't work, but just that it contains all of three.js, and thus is huge. Is that right?

Yeh that's a much better way to describe it 😊

For the issue with super-three vs three, I don't think it matters because in the end we want to just say THREE will be exposed globally, and both super-three and three do that. And super-three only really makes additions to three https://github.com/mrdoob/three.js/compare/dev...supermedium:dev, and so the THREE global that super-three exposes would have everything three-to-ammo needs.

I think the question is how do we make browserify only include the parts of three that we need (instead of all 1.2 MB (600K for three.min.js)). Am I understanding the problem correctly?

Yeh I think that is the question, and really we should be able to exclude all of three. There should be some mechanism for browserify to effectively replace the require('three') statement in three-to-cannon with something that references the global THREE. We could manually write something to do this, but I'm sure we're not the first project to have this problem, and it must already have a solution.

[MignonBelongie]

There should be some mechanism for browserify to effectively replace the require('three') statement in three-to-cannon with something that references the global THREE. We could manually write something to do this, but I'm sure we're not the first project to have this problem, and it must already have a solution.

Now I understand why you were talking about --globals three=THREE. I didn't know you could do what you're describing. I'll try to figure it out.

[MignonBelongie]

@Galadirith Thanks for the -g suggestion in https://github.com/donmccurdy/three-to-cannon/issues/28#issuecomment-706487966 Does this look OK to you now?

[Galadirith]

That looks great, thanks so much for making those changes @MignonBelongie 😊 I'd be happy to suggest for you to merge this now.