Plotting of comp morphology

[jnsbck]

I got side-tracked while working on a different PR and implemented plotting of the actual compartment structure:

The compartments are rendered in 3D and then the correct projection of the 3d mesh is plotted on whatever plane.

fig, ax = plt.subplots(2, 3, figsize=(10, 5))

for i, orientation in enumerate([np.array([1,1,2]), np.array([0,1,0.1]), np.array([0,0.1,1])]):
    plot_cylinder_projection(orientation, length=5, radius=1, center=np.array([0,0,0]), dims=[0,1], ax=ax[0,i])
    plot_cylinder_projection(orientation, length=5, radius=1, center=np.array([0,0,0]), dims=[1,2], ax=ax[1,i])
    ax[0,i].set_title(f"Orientation: {orientation}")
plt.show()

Hence the compartment radii and lengths are correctly reflected in the plot.

However, plotting an imported morphology looked weird for some reason.

While looking into this, I noticed two things:
Since the cylindrical compartments are straight, they run from one compartment edge to the other on a straight line. However, they model a potentially very zigzaggy branch/path structure. This means:
a) Assuming the comp center is at loc=0.5 along the path (currently how we do it), can put the comp center off axis.
b) Setting the comp length to the path length can result in comps which appear longer than the distance between their two endpoints.

This can be illustrated by adding a bit of noisy to the xyz coordinates of the example above.

Changing the compartment centers to be the center of the cylinder, already produces much better looking plots.

While I dont think changing b) makes sense (apart from for visualization maybe) I think changing a) might be a good idea.

Lemme know your thoughts.

PR is missing documentation, some cleanup and handling of single point cases.

[michaeldeistler]

Hi Jonas, thanks, this can be super useful, especially for the radiuses.

I don't understand this sentence: Assuming the comp center is at loc=0.5 along the path (currently how we do it), can put the comp center off axis.

Afaik we currently have no notion of "comp center", so I don't understand what you mean.

[jnsbck]

We do in '_update_nodes_with_xyz'.
When the locations of the compartments are computed they are assumed to lie on the traced morphology (xyzr). For a branch that gets split into 4 comps this means the comp Centers are spaced at equidistant points along xyzr.

[michaeldeistler]

Okay! Then yes, I think I would be pro changing the compartment centers to be the center of the cylinder.

[jnsbck]

Ready for review. Now cell.vis(type="volume") plots the compartments as seen below.

Not sure if "volume" is the best kwarg tho.

Q: Should I also add col kwarg for color like vis?

[michaeldeistler]

Thanks, awesome! Please add a test (just ensuring that there is no error) for a compartment, branch, cell, network, swc file, and net.cell(1).vis(...)

[jnsbck]

Thanks for the feedback!

Should I also add the col kwarg for consistency? #432 (comment)
@michaeldeistler ?

[michaeldeistler]

Yes I think it would be great to have a col argument!

[jnsbck]

@michaeldeistler added tests and addressed all comments. Feel free to merge if you approve.

[jnsbck]

swc files to reproduce the effect of zigzaging morphologies on the comp lengths. file extension changed to txt since swc not supported

noisy_test.txt
test.txt