offset on a very simple mesh crashes my Ubuntu VM about meshlib HOT 8 CLOSED

Hello Fabio!

Can you please attach full script that you run, and what error is printed when it crashes.
Anyway there are some issues with the code you send:

1. params.voxelSize should be greather then zero (as far as voxel is kind of 3d pixel, zero size requers infinite resolution of voxels grid, so it have to be greather)
2. shell mode requeres positive offset value that is at least half of voxel size

from meshlib.

Thanks, Grantim!

The script is very long, what I put here is sufficient to reproduce the problem. I don't have an error, I see the RAM be greater till the complete saturation of the RAM, blocking the VM

I've put a positive offset of 2 and voxelSize to 0.16 and it works.

Do you suggest 4 as Voxel size?
is possible that 4 means a coarse result?

Thanks a lot!
Fabio

from meshlib.

Glad that worked!

Voxel size should depend of object size

some exapmles of shell with different voxel size:
Voxel size 0.05, offset 0.15

Voxel size 0.001, offset 0.03

less voxel size means better quality but takes more resources, if you need offset 2 i would suggest 0.1 voxel size. (Offset 2 is 10 times bigger than the mesh, so most likely it would look like sphere)

from meshlib.

thanks a lot for your time and your consideration.
Yes, in the end, it shall be similar to a sphere.
This is a small part of a bigger mesh made of parts with different offsets.
If I want at the end a thickness of 2 I have to put an offset of 1, isn't it?

As you are so kind, I have also a pair of questions about boolean operations.
They take a lot of resources, there is a way to limit it?
I have seen that all the resources are not freed at the end of the elaboration, there is something to release?

Thanks
Fabio

from meshlib.

In this case offset value is close to radius, so if you mean diameter by thickness you need offset value 1.

About boolean, we have to different implementations, one voxel boolean, that need amount of resouces based on voxel size, and also we have mesh-based boolean (it can produce an error if input meshses has self-intersections) that take less resources.

from meshlib import mrmeshpy as mm
help(mm.boolean) # mesh-based
help(mm.voxelBooleanSubtract)
help(mm.voxelBooleanUnite)
help(mm.voxelBooleanIntersect)

both methods do not require manual releasing, so I think it may be a bug, could you please provide small example so we can investigate it?

from meshlib.

Hi,
thanks again, I've tried boolean voxel functions seem more stable in terms of memory (but slower).
Regarding the memory leek, When I've some minutes I try to provide you with an example, perhaps in a new issue.

from meshlib.

Thanks!

from meshlib.

Closing for now, if you will have any other issues please feel free to contact us!

from meshlib.