improve workflow for stepwise evaluation

[DamienGilliard]

This PR wants to merge small improvements on the workflow into the implement_pca branch

[DamienGilliard]

from meeting, two changes:

• Instead of setting the beam ID at the CAD_segmentation step, add a component that allows to only focus on a sub-selection of the assembly.
• with the history: output as datatree instead of dict. We can't do anything with a dict in gh

[DamienGilliard]

UPDATE: the history is now a gh_tree and a new component has been created to truncate the assembly:

[DamienGilliard]

• For FRTruncateAssembly, have i_truncate_index be the index of the last beam we keep (i_truncate_index = 0 --> 1 beam in assembly)

[eleniv3d]

@DamienGilliard Do you have maybe an example file for this?

[eleniv3d]

Do you understand this? :)

[DamienGilliard]

If you recompile the C++ it should resolve this issue. I added the get_principal_axes in the python binding and the GetPrincipalAxis in the C++ DFPointCloud class

[eleniv3d]

If you recompile the C++ it should resolve this issue. I added the get_principal_axes in the python binding and the GetPrincipalAxis in the C++ DFPointCloud class

i thought i did but will try again :)

[DamienGilliard]

I have added a fallback solution: when the K-means clustering on the normals does not give proper vectors, we fall back on the oriented bounding box. On the data that gave errors, we don't have them anymore :)

[DamienGilliard]

ah yes indeed, well spotted ! 1fad289

[eleniv3d]

made some small changes which are commented in every commit. otherwise, let's merge it!

[DamienGilliard]

Thanks a lot 🙏