Fix assignment of compartment grid points

[mogres]

Problem

This PR includes various updates to grid creation and visualization and fixes errors in main while trying to pack nested compartments.

When assigning whether grid points are inside or outside a compartment, only grid points with a compartment id of 0 were checked and updated. This caused an issue if the compartments were nested, and the outer compartment was processed before the inner (as it should).

Solution

points_in_mesh now updates the identity of grid points where compartment id <= the id of the current compartment being assigned. This assumes that larger compartments are created before smaller ones.

The buildgrid_trimesh function and distance calculations in gradient.py was updated to streamline and speed up grid creation. Visualization of grid points was cleaned up to remove unused values, improve readability and visibility.

Change summary:

Gradient calculations:

• removed normalization of surface distances while setting gradient attributes, this is now done while calculating the weights
• ignore nan values while scaling distances
• avoid calculating surface distances if not needed for a gradient. This calculation is computationally expensive and only needs to be performed if the gradient mode is surface

Grid creation:

• reduced chunk size for determining grid points inside-outside compartments
• allow reassigning grid points for downstream compartments: this assumes that outer compartments are created before the inner nested compartments.

Grid visualization:

• sort grid point values and remove nans before visualizing
• set the size of grid points relative to the grid spacing

Steps to Verify:

pack -r examples/recipes/v2/peroxisomes_surface_gradient.json -c examples/packing-configs/peroxisome_packing_config.json

Screenshots (optional):

Time for packing peroxisomes_surface_gradient:

fix/multiple_compartments:

Keyfiles (delete if not relevant):

1. Compartment.py
2. Gradient.py

[codecov-commenter]

Codecov Report

Patch and project coverage have no change.

Comparison is base (2c9db6d) 98.52% compared to head (8d76176) 98.52%.

❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #192   +/-   ##
=======================================
  Coverage   98.52%   98.52%           
=======================================
  Files          16       16           
  Lines         476      476           
=======================================
  Hits          469      469           
  Misses          7        7           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[github-actions]

Packing analysis report

Analysis for packing results located at cellpack/tests/outputs/test_spheres/spheresSST

Ingredient name	Encapsulating radius	Average number packed
ext_A	25	236.0

Packing image

Distance analysis

Expected minimum distance: 50.00
Actual minimum distance: 50.01

Ingredient key	Pairwise distance distribution
ext_A

[meganrm]

Looks good to me

[rugeli]

Thanks for fixing it! I can finally run and pack the peroxisomes_surface_gradient recipe. However, it took me 1212.10s. Should we look into it further?

[mogres]

Thanks for fixing it! I can finally run and pack the peroxisomes_surface_gradient recipe. However, it took me 1212.10s. Should we look into it further?

Did you run this on your local machine? If so, the timing sounds about right! I was running this on the dgx workstation which would be faster than a laptop.