examples: Clean-up Epstein Civil Violence and PD grid

[EwoutH]

Clean up the epstein civil voilence models.

One issue is that a deepcopy currently isn't possible, and Solara currently uses that.

model1 = EpsteinCivilViolence()
import copy
# make deepcopy
model2 = copy.deepcopy(model1)

@quaquel would you like to pick this further up?

[github-actions]

Performance benchmarks:

Model	Size	Init time [95% CI]	Run time [95% CI]
BoltzmannWealth	small	🟢 -16.1% [-17.5%, -14.6%]	🔵 +0.2% [+0.0%, +0.3%]
BoltzmannWealth	large	🔵 +2.1% [+1.6%, +2.6%]	🔵 +1.4% [+0.7%, +2.1%]
Schelling	small	🔵 -1.8% [-2.4%, -1.2%]	🔵 -0.6% [-0.8%, -0.4%]
Schelling	large	🔵 -0.9% [-2.4%, +0.5%]	🔵 +0.0% [-1.9%, +2.5%]
WolfSheep	small	🔵 -0.7% [-1.0%, -0.3%]	🔵 +1.1% [+0.8%, +1.3%]
WolfSheep	large	🔵 +1.4% [-0.2%, +2.9%]	🔴 +7.1% [+4.8%, +9.6%]
BoidFlockers	small	🔵 -3.2% [-3.5%, -2.9%]	🔵 +0.6% [-0.2%, +1.3%]
BoidFlockers	large	🔵 -0.5% [-1.4%, +0.2%]	🔵 +1.2% [+0.6%, +1.8%]

[quaquel]

This is a properly obscure problem. It arises from trying to deepcopy the OrthogonalMooreSpace. Even worse, whether or not the error triggers depends on the size of the grid (which suggests to me that perhaps recurssion error is triggered to early?). See also here

To test this yourself, try to deepcopy a moore / von neumann grid with varying sizes and see when it errrors out.

My hunch is that the problem starts from Cell.connections. If I comment out self.connections[key] = other in Cell.connect the problem disappears, supporting my hunch. One solution is to implement __setstate__ and __getstate__ explicitly for both Cell and Grid. Basically, don't return Cell.connections, and in Grid.__setstate__ after having initialized everything, rebuild the connections explicitly.

This is a major bug because it means that large experimental grids are currently not pickleable....

@Corvince any thoughts?

[quaquel]

I have committed a possible fix in line with my previous comment. In short, In Cell.__getstate__, I set "connections" to an empty dict. In DiscreteSpace.__setstate__, I rebuild the connections via DiscreteSpace._connect_cells(). This latter method already existed in Grid and Voronoi, but not yet in network. I have thus moved the method stub to DiscreteSpace and implemented it for Network.

It needs further testing and additional unit tests for all DiscreteSpaces and large sizes, but it seems to work.

Also, this is a fix for Grids, not sure if the problem is present in all Discrete spaces....

[EwoutH]

Thanks! I will continue updating the examples, if this works for all of them than we can include this in a separate PR.

[quaquel]

Thanks! I will continue updating the examples, if this works for all of them than we can include this in a separate PR.

Ok, I'll move the fix into a separate PR and add additional unit tests. Let me know how you get on with the other examples.

[EwoutH]

Awesome!

Updated pd_grid (Demographic Prisoner's Dilemma on a Grid). One issue is that we can't visualize any of the cell spaces currently. Without that the visualisations are quite useless. Tracking issue here:

• Visualize Cell Space #2377

[quaquel]

I guess it would make sense to merge this even if #2377 is not addressed yet. It means the docs can progress and would give us something to use when developing #2377.

[github-actions]

Performance benchmarks:

Model	Size	Init time [95% CI]	Run time [95% CI]
BoltzmannWealth	small	🟢 -8.1% [-9.4%, -6.6%]	🔵 -0.9% [-1.1%, -0.6%]
BoltzmannWealth	large	🔵 +0.2% [-0.1%, +0.4%]	🔵 -3.6% [-4.8%, -2.4%]
Schelling	small	🔵 -0.8% [-1.1%, -0.4%]	🔵 +0.1% [-0.1%, +0.3%]
Schelling	large	🔵 +0.4% [-0.4%, +1.3%]	🔵 +1.1% [-1.5%, +4.1%]
WolfSheep	small	🔵 +0.7% [+0.5%, +0.9%]	🔵 -0.2% [-0.4%, -0.0%]
WolfSheep	large	🔵 +0.1% [-0.9%, +0.9%]	🔵 -1.6% [-3.3%, +0.2%]
BoidFlockers	small	🔵 -2.1% [-2.8%, -1.4%]	🔵 -0.2% [-0.8%, +0.5%]
BoidFlockers	large	🔵 -1.1% [-1.6%, -0.6%]	🔵 -0.8% [-1.2%, -0.2%]

[quaquel]

What's there is, is good. Would it make sense to remove the notebooks?

[EwoutH]

This merge conflict, fuck.... couldn't we have merged this before the examples move?

[quaquel]

Yep the merge conflict is massive (as with the docs stuff). Might be better to start a new branch and copy these changes into it (As you are doing with docs atm).

[quaquel]

Do you mind if I pick this up as a new PR given the rebase/merge mess? I have time this afternoon (Wednesday 23 October) for simple cleanup work (but not for figuring out complicated stuff like the solara run import problem).

[quaquel]

Superseded by #2408