How the mesh uploader decides which object from dae file becomes the root?

[lucynes1]

Hello! I would like to ask something.. how does mesh uploader decides which object from dae file becomes the root? After making few tests I couldn't find any pattern
It doesn't choose first or last object from dae, it doesn't choose by the name neither. It can't be random though because every time it chooses the same object for the same dae.

Thank you in advance for help! ❤️

[arton Rotaru]

6 hours ago, lucynes1 said:

It can't be random though because every time it chooses the same object for the same dae.

Just uploaded the same dae 2 times. Result:

[lucynes1]

Oh! Maybe I was just very lucky then to get same results. You used firestorm to upload yes?

[arton Rotaru]

4 minutes ago, lucynes1 said:

Oh! Maybe I was just very lucky then to get same results. You used firestorm to upload yes?

No, I use the LL viewer.
Perhaps rather than lucky, I could imagine that you may have set MeshImportUseSLM to True in Debug settings, and that's why it's always the same root after the first upload.

Edited December 26, 2022 by arton Rotaru

[animats]

I looked at that uploader code in the viewer once. The link order is determined when the data items for the objects are sorted by memory address. Really. That's why it seems random.

When I was uploading my escalators in various sizes, I hit this. There's a range of escalator lengths. The resizing was done in Blender, with a Python script to resize the center section, adjust the UVs for the moving handrail, and make the appropriate number of steps for the length. Then I uploaded all the different sizes. About half were linked wrong. I then had to go unlink and relink the ones that uploaded wrong. Which is why I dug into the viewer code to see what was happening inside.

It's a COLLADA thing. There's no explicit object order in COLLADA export, so the uploader has to do something arbitrary. I once suggested sorting by name, but people who upload everything as "Object" would still get randomness.

LL is working on gLTF uploads, in addition to COLLADA. gLTF format, which is what everybody else uses now, has hierarchies and order. So you get consistency. This is supposed to be the next step after gLTF materials, which is in test now on the beta grid. Discussed at Creator User Group.

Once that's in, you should be able to make it in Blender, or anything else that speaks gLTF, and what you see in the creation program is what you see in world. This does mean uploading the textures with the model, rather than uploading in grey and fussing with the textures in world. You'll still be able to change them in world, or, if you really want to, upload with no textures, but the initial upload can be a complete set. With all the new texture layers, you'll probably want to debug the look in Blender and then upload the whole thing.

(This assumes LL finishes the job, which, in the past, has not always happened.)

 

 

[OptimoMaximo]

I remember having done some experimenting in such direction, long time ago, and i seem to recall that putting all objects as children of the item you want to be root directly in your program before exporting to collards actually maintained the root consistently once uploaded.

[Aquila Kytori]

17 hours ago, animats said:

It's a COLLADA thing. There's no explicit object order in COLLADA export,

Isn't that what the Sort by Object name option in Blenders Collada exporter is for ?

6 Objects to be exported:

 

 

17 hours ago, animats said:

so the uploader has to do something arbitrary.

Or have it fixed so that it can read and use the order of objects it finds in the .dae file. ?

 

17 hours ago, animats said:

I once suggested sorting by name, but people who who upload everything as "Object" would still get randomness.

The people who "upload everything as "Object" " obviously don't care what order the objects are dealt with.

 

 

Edited December 27, 2022 by Aquila Kytori