Added property layer viz to sugarscape

[sanika-n]

This PR addresses Issue #2580, I have replaced the normal matplotlib visualization in sugarscape to the property layer visualization with matplotlib as backend.

Before

After

[github-actions]

Performance benchmarks:

Model	Size	Init time [95% CI]	Run time [95% CI]
BoltzmannWealth	small	🔴 +6.5% [+4.8%, +8.2%]	🔵 +0.3% [+0.2%, +0.4%]
BoltzmannWealth	large	🔵 +0.5% [-0.3%, +1.6%]	🔵 +0.5% [-2.2%, +2.5%]
Schelling	small	🔵 +0.7% [+0.4%, +1.0%]	🔵 -1.1% [-1.2%, -1.0%]
Schelling	large	🔵 -0.3% [-0.6%, +0.0%]	🔵 +2.8% [+1.6%, +4.0%]
WolfSheep	small	🔵 +2.0% [+1.7%, +2.2%]	🔵 -0.5% [-0.7%, -0.3%]
WolfSheep	large	🔵 +1.6% [+0.9%, +2.3%]	🔴 +9.9% [+8.3%, +11.3%]
BoidFlockers	small	🔵 +1.3% [+0.2%, +2.4%]	🔵 -0.5% [-0.7%, -0.3%]
BoidFlockers	large	🔵 +0.2% [-0.5%, +1.0%]	🔵 -0.8% [-1.1%, -0.4%]

[tpike3]

@sanika-n The formatting here seems off --- are you using the ruff linter

[sanika-n]

Oh, I actually wrote spice in the same way as sugar but when I ran ruff it was changing the formatting of spice and it didn't allow me to keep it like sugar, so now I have made sugar formatted like spice...

[tpike3]

Thanks @sanika-n -- this code shows the power of property layer.

Just one question as the formatting does not seem to be uniform.

[quaquel]

I had a look at sugarscape a few weeks back as well. It made me realize that the property layer portrayal needs some more work at the API level (see also #2640). However, in the meantime, this looks fine to me. However, could you experiment a bit with colormaps that are more distinct that the current red and blue? Likewise, what happens if you lower the alpha to say 0.5?

[sanika-n]

So here are some plots with different values of alpha, the reason why there is not much opacity even alpha=1 is that I set vmax to 10, and the sugar and spice levels in this example only go up to 4.

Given that the first property layer is sugar and the second is spice, in areas where both spice and sugar equal 10 we will only see a bright red because both colors are completely opaque (alpha=1 and value=vmax) and therefore spice will cover up sugar. I set vmax to 10 to avoid this. However, this case is rare, and I noticed it doesn't come up much in this example, and setting vmax to 4 made the plot more vibrant.

Coming to colormaps, I tried a few configurations, but for most of them, I wasn't too happy with how it turned out. The colors were mixing quite a lot with each other.

Example 1:

propertylayer_portrayal = {
    "sugar": {"colormap": "winter", "alpha": 1, "colorbar": True, "vmin": 0, "vmax": 4},
    "spice": {"colormap": "afmhot", "alpha": 0.5, "colorbar": True, "vmin": 0, "vmax": 10},
}

Example 2

propertylayer_portrayal = {
    "sugar": {"colormap": "winter", "alpha": 1, "colorbar": True, "vmin": 0, "vmax": 4},
    "spice": {"colormap": "Reds", "alpha": 0.4, "colorbar": True, "vmin": 0, "vmax": 10},
}

Example 3:

propertylayer_portrayal = {
    "sugar": {"colormap": "viridis", "alpha": 0.5, "colorbar": True, "vmin": 0, "vmax": 4},
    "spice": {"colormap": "plasma", "alpha": 0.5, "colorbar": True, "vmin": 0, "vmax": 4},
}

[quaquel]

Thanks for this. I'll think about this a bit more (also in light of the various other visualization and API issues that are popping up). I am inclined to accept this PR while we hash out a better overall API level solution.

[tpike3]

Thanks for this. I'll think about this a bit more (also in light of the various other visualization and API issues that are popping up). I am inclined to accept this PR while we hash out a better overall API level solution.

I concur @quaquel