OTOY Forums

This may be premature, and the problem COULD be in my code, but it is not looking like that is the case after some tests.

Attached is a graphic explaining what I am doing. I needed a quick way to make copies of parts in Octane that were pointing in some other desired direction, and that were positioned at some other desired position relative to the original. The "copies" are just the same mesh with transform nodes applied. When a copy is needed, I import the same mesh, scale it to the proper size, rotate it into the desired direction, then translate it into its final position, in that order. I get the values for the rotations and translations from a vb calculator I built.

Please see the attached graphic to see what the problem is. I THINK the transform nodes are not working properly. My simulations for what SHOULD happen come out correct, but Octane's solutions do not come out right. As I post this, I am still looking for some error in my method.

Might be a rounding error?

I thought maybe, too. But I did a test with rounding and the error was not nearly as great as it came out in Octane.

And as an update to my post, I forgot to mention that the problem is in v1.23...I just tested this out in v1.2 and everything comes out OK. I know that a lot of significant changes were made to the node system, so it is looking like a v1.23 bug.

We have changed the order of the operations in the transform node. In 1.23 first the object is scaled, then rotated and then translated. In 1.20 this was the other way around.

--
Roeland

Hm, I had actually had my nodes in both, connected up in the "new" sequence:
These xform nodes were taken from a preexisting v1.2 "template" file I created for speeding up the creation of new scenes. Each node has two of the xforms turned off. So the Scale Node has Rotation and Xlation turned off, the Rotate node has Scale and Xlate turned off, and finally, the Xlate node has Scale and Rotate turned off. I did it this way to make it easier to go in and change values for a particular transformation without accidentally changing one of the other two.

Any clues yet?

Here is a direct comparison between v1.2 and v1.2x:

We switched the order of the transformations between 1.20 and 1.21. So if every of the tree placements node only does a scale, rotation or translation as the name suggests then starting from 1.21 you can put those three into one node. The order in the node graph corresponds to the order used in the new versions.

At least version 1.24 will indeed keep all digits entered although it will display only three digits after the decimal point. Can you verify the left screw moves to a slightly different position if you enter exactly [-0.041, 0, -0.052] as translation?

--
Roeland

If I round down the values, the screw positions do change...They get worse. That was why I thought maybe there was still some internal rounding going on during calculations, but that doesn't make sense since if I keep scale at 1:1, everything is fine.

This is true if I use the old nodes that allowed you to delete a transform (right-click > Delete Node) and the new 1.2x transform node that does not allow you to delete...The effect is the same.

I can't reproduce any issue like this here. Can you send us the scene? Just the two screws is enough.

--
Roeland

Here is the simplified scene and a jpg showing dimensions of the screw placements relative to the base.