[usdview] Layered alembic transform payloads not always working

[AlexSchwank]

Description of Issue

I'm running into an issue when I use multiple alembic payloads for a prim for transformations.
I have two files: teapot.abc containing a teapot mesh and teapotTransform.abc containing a transform for the teapot. Layering them using payloads is causing issues in usdview (see Case 2).

Case 1: no layers

def Xform "myTeapot_original" {
    over "teapot_teapot" (
        payload = [
            @teapot.abc@</teapot_teapot>
        ]
    )
    {}
}

A non-transformed teapot. No surprise here.

Case 2: abc transform (error case)

def Xform "myTeapot_abcTransform"
{
    over "teapot_teapot" (
        payload = [
            @teapotTransform.abc@</teapot_teapot>,
            @teapot.abc@</teapot_teapot>
        ]
    )
    {}
}

This is the case where it doesn't work. The teapot is not transformed. However, usdview is showing the transform attributes for teapot_teapot.

Case 3: usd transform

def Xform "myTeapot_usdTransform"
{
    over "teapot_teapot" (
        payload = [
            @teapotTransform.usda@</teapot_teapot>,
            @teapot.abc@</teapot_teapot>
        ]
    )
    {}
}

This works as expected. The teapot is transformed. teapotTransform.usda was generated by usdcat teapotTransform.abc -o teapotTransform.usda. usdview is showing the same transform attributes for teapot_teapot as in Case 2.

Case 4: flattened stage

def Scope "Assets"
{
    def Xform "myTeapot_usdTransform"
    {
        def Xform "teapot_teapot"
        {
            float3 xformOp:rotateZXY = (1, 2, 3)
            float3 xformOp:scale = (1.1, 1.2, 1.3)
            double3 xformOp:translate = (2, 3, 4)
            uniform token[] xformOpOrder = ["xformOp:translate", "xformOp:rotateZXY", "xformOp:scale"]

            def Mesh "teapot_teapotShape"
            {
                #...
            }
        }
    }

    def Xform "myTeapot_abcTransform"
    {
        def Xform "teapot_teapot"
        {
            float3 xformOp:rotateZXY = (1, 2, 3)
            float3 xformOp:scale = (1.1, 1.2, 1.3)
            double3 xformOp:translate = (2, 3, 4)
            uniform token[] xformOpOrder = ["xformOp:translate", "xformOp:rotateZXY", "xformOp:scale"]

            def Mesh "teapot_teapotShape"
            {
                #...
            }
        }
    }

    def Xform "myTeapot_original"
    {
        def Xform "teapot_teapot"
        {
            def Mesh "teapot_teapotShape"
            {
                #...
            }
        }
    }
}

This works as well. Both myTeapot_usdTransform and myTeapot_abcTransform are transformed correctly.

teapot_files.zip

Package Versions

Tested in 19.07 and 19.11

[spiffmon]

Hi @AlexSchwank !
The behavior you are describing sounds incredibly weird. Unfortunately, the transform values in your example are really subtle and I am having a hard time seeing any difference in the three pictures you provided... If you can reword it to throw in a 90 degree rotation or a larger scale, I can at least try to verify the behavior here. Honestly, if we see the same behavior I think it’s unlikely we’d be able to investigate anytime soon... this is a rather unexpected use for alembic files in USD (though I can’t imagine how it would screw up if usdcat gives the right answer). Can I point you at. PR #1099 , which we are hoping to accept soon, which would allow you to use ALembic’s own layering feature to compose your alembic files in Alembic-space, which might give you bettter ransform comPosing, since we flatten out all the Alembic xform ops?

[AlexSchwank]

Hi @spiffmon ,

The teapot in the top right corner is the transformed one. The teapot at the origin has not transform.
I'll check out the PR - that might solve some problems on our end.

[jilliene]

Filed as internal issue #USD-5879

[spitzak]

Isn't usd ignoring the root transform of abc files? Or am I thinking of something in-house here

…

On Mon, Feb 24, 2020 at 11:34 AM jilliene ***@***.***> wrote: Filed as internal issue #USD-5879 — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#1118>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAJMCEIWCETGLUEKITGSHZLREQONRANCNFSM4KXXSGWQ> .