Add support for drawing discrete grids

[quaquel]

Add support for drawing experimental Grid Spaces (i.e. HexGrid, OrthogonalMooreGrid, OrthogonalVonNeumannGrid).

Closes #2377. This code needs further refinement (e.g., duplication of agent potrayal code from voroinoi grid, offsets for hexgrids, and adding experiment NetworkGrids).

[github-actions]

Performance benchmarks:

Model	Size	Init time [95% CI]	Run time [95% CI]
BoltzmannWealth	small	🔵 -0.8% [-2.2%, +0.7%]	🔵 +0.5% [+0.3%, +0.7%]
BoltzmannWealth	large	🔵 -1.4% [-1.8%, -1.0%]	🔵 -2.5% [-3.0%, -1.9%]
Schelling	small	🔵 -1.6% [-1.9%, -1.4%]	🔵 -1.4% [-1.6%, -1.1%]
Schelling	large	🔵 -0.9% [-1.9%, -0.0%]	🟢 -5.2% [-6.3%, -3.9%]
WolfSheep	small	🔵 -1.0% [-1.3%, -0.8%]	🔵 -0.9% [-1.1%, -0.7%]
WolfSheep	large	🔵 +0.8% [-0.0%, +1.8%]	🔵 +0.3% [-1.7%, +2.7%]
BoidFlockers	small	🔵 -1.0% [-1.7%, -0.2%]	🔵 -0.9% [-1.8%, +0.1%]
BoidFlockers	large	🔵 -1.1% [-1.8%, -0.4%]	🔵 -1.2% [-1.8%, -0.7%]

[EwoutH]

Uninstall your current mesa, pip install . -e and then solara run app.py is what I found out working.

[quaquel]

Uninstall your current mesa, pip install . -e and then solara run app.py is what I found out working.

Thanks! This gave me the idea for a slightly different approach: insert the correct version of mesa into sys.path at the top. Based on this, I can confirm that this code works as intended.

Let me know what else would be needed to complete this PR since its my first contribution to the visualization side of things.

[EwoutH]

insert the correct version of mesa into sys.path at the top. Based on this, I can confirm that this code works as intended.

That's the other obvious way to go, both have their advantages based on what your working on.

@Corvince I'm also quite new to this stack, could you please review this one?

[Corvince]

I think this PR is very fine and a good first step. I pre-approve this, just please also add OrthogonalVonNeumann

[Corvince]

This should also work for OrthogonalVonNeumannGrid

[quaquel]

yes, I'll fix that! Just need to figure out how to capture this in the match case statement.

I guess it even will work for hexgrids (although we might want to visualize these a bit differently with offsets for every odd row.

[wang-boyu]

Hi @quaquel, I understand that this is already merged. May I ask whether it is intentional to remove the call to _draw_grid() here? Thanks!

[quaquel]

I guess I made a mistake. I'll put in a PR to fix this asap.

Thanks for the post-merge review!

[quaquel]

and fixed via #2398

[wang-boyu]

Great! Thanks!

[EwoutH]

One improvement for this function would be to use the _get_agent_data when drawing the new grids. I got a weird error, and using _get_agent_data would have prevented and handled that. I wrote it up here:

• SolaraViz: Use _get_agent_data consistently in visualisation of all spaces #2407

[quaquel]

One improvement for this function would be to use the _get_agent_data when drawing the new grids. I got a weird error, and using _get_agent_data would have prevented and handled that

I looked at that and I could not use it because that method is hard coded to only work with _Grid: for agents, pos in space.coord_iter().