Mesh/Line/Points: Fix copy().

[Mugen87]

Fixed #16224.

This PR implements the approach I have outlined here: #16224 (comment)

The idea is to remove all clone() methods in mesh, points and line classes and always use the version of Object3D. The copy() methods are now responsible for copying all properties.

[Mugen87]

The difference in the new implementation is the invocation of updateMorphTargets(). Previously, the method processed the geometry of the source mesh. Now, it is processing the default geometry and thus producing no result. However, morphTargetInfluences and morphTargetDictionary are copied in copy() so you still get the same object status.

In any event, there might be use cases where code could break. E.g.

var mesh1 = new THREE.Mesh();
mesh1.geometry = geometry; // assign geometry with morph attributes
var mesh2 = mesh1.clone();

Previously, mesh2 would have valid morphTargetInfluences and morphTargetDictionary properties. Now, these properties are missing since copy() replicates the status of mesh1(which has no morphTargetInfluences and morphTargetDictionary). However, I think the new behavior is the result you would normally expect.

[WestLangley]

This is getting better. Thank you for working on it.

Even so, I have long-advocated removing clone() for a host of reasons -- except for the math classes, maybe. ***

---

*** But I do not like it there either because it violates our so-called policy of requiring users to allocate memory at the application level. Plus, inexperienced users are known to call clone() unnecessarily in tight loops.

[Mugen87]

How about discussing a potential removal of Object3D.clone() in a new issue? AFAICS, the knowledge about this topic is spread over multiple issues and PRs. Maybe it's easier to draw attention to the downsides by having a single issue.

[WestLangley]

I think this PR an improvement, so let's merge it.

You can wait until r.118 if you want to be safe.

[Mugen87]

You can wait until r.118 if you want to be safe.

I would prefer that so we have more time for testing 😇 . Although I think this change should be safe.

[Mugen87]

@mrdoob Since we are at the beginning of a new dev cycle, can we merge this PR? 😇

[WestLangley]

I think this is worth a try. But we will see...

I think @mrdoob prefers to wait several days after a release to ensure there are no hot-fixes required.

[mrdoob]

Yeah, I think we're good. Only GLTFLoader had a minor issue but I updated npm only with the fixed GLTFLoader.

[mrdoob]

Thanks!