DragControls: Fix drag object in local coordinate

[lafflan]

when drag an object who's local position and rotation not equal world position and rotation will cause a wrong effect

[lafflan]

https://jsfiddle.net/sweerwen/3foLr7sn/199/ see this demo

https://jsfiddle.net/sweerwen/3foLr7sn/201/ fixed

[Mugen87]

For testing: https://raw.githack.com/sweerwen/three.js/patch-1/examples/webgl_interactive_draggablecubes.html

[WestLangley]

when drag an object who's local position and rotation not equal world position and rotation will cause a wrong effect

I expect the problem is simply caused by the rotated parent object. Can you confirm?

---

Also, I suggest you avoid getWorldPosition() as it is not efficient. World matrices should have been updated already by the renderer in this case.

[lafflan]

I expect the problem is simply caused by the rotated parent object. Can you confirm?

maybe...but i'm not very clear.I found this problem incidentally when I was using it.

Also, I suggest you avoid getWorldPosition() as it is not efficient. World matrices should have been updated already by the renderer in this case.

but It seems that there is no other way to get the world position.

[WestLangley]

but It seems that there is no other way to get the world position.

Check the source code for .getWorldPosition().

vector.setFromMatrixPosition( object.matrixWorld );

[arodic]

I expect the problem is simply caused by the rotated parent object. Can you confirm?

Exactly. Therefore, parent's transform should be taken into account when performing world space transformations.

I would expect similar issue to arise with scaled parent (a common case when importing objects from different world units)

[WestLangley]

In your use case,

_selected.parent.worldToLocal()

inverts the parent's world matrix repeatedly. This method should not be called in tight loops.

Can you compute the necessary inverse and reuse it?

[mrdoob]

@WestLangley Just double-checking... Are you suggesting changes to this PR?

[WestLangley]

Are you suggesting changes to this PR?

I am suggesting the OP avoid calling worldToLocal() in tight loops. I doubt the parent's inverse will be changing while dragging.

[lafflan]

@WestLangley thanks， could you please help to check

[WestLangley]

@sweerwen Yes, that is what I had in mind. I have not tested your code, however.

These convenience methods are just that: convenient. However, they are not necessarily efficient.

Thanks for making the change.

[Mugen87]

Looks good 👍

[mrdoob]

Thanks!