Draft: Blender conversion and vectorisation

Draft. My experimental branch.

Blender conversion

Examples live in examples/vectorised. Run them with

blender --python examples/vectorised/new_blender_api.py

Interesting examples are

• examples/vectorised/new_blender_api.py - for a prototype API
• examples/vectorised/scatter.py - for tubes and spheres
• examples/vectorised/scatter_bicolour.py - for bicolouring tubes and spheres, but scattering and displaying is decoupled
• examples/vectorised/scatter_link_bicolour.py - for bicolouring tubes and spheres and scattering and displaying is coupled, just like in develop
• examples/vectorised/discrete_curve.py, examples/vectorised/discrete_net.py, examples/vectorised/half_edge.py - for converting library objects

This branch is also a testbed for decoupling visualisation from Blender. For example, first run

poetry install --sync --with dev --with docs --with tests --with coverage --with plot

Notice the --with plot. This installs plotly. Most examples can also be visualised in plotly

python3 examples/vectorised/sphere.py

Of course, everything is vectorised. Points are stored in columns rather than rows. Both have advantages and disadvantages, e.g.

• with column-data, A @ points, where A is a matrix works as expected, row-data would require points @ A.T
• with row-data, points + translation, where translation.shape == (3,) works as expected, but column-data requires points + translation.reshape((3, 1)) or similar
• seems like scipy, scikit etc. usually store data in the rows, haven't checked thoroughly yet