Set JupyterViz as stable

[rht]

This PR stabilizes the new Jupyter Visualisation. A new tutorial is available here: https://mesa.readthedocs.io/en/latest/tutorials/visualization_tutorial.html

If you already use the new visualisation, and want to keep using it, you need to change the import from from mesa.experimental.jupyter_viz import JupyterViz to from mesa.visualization.jupyter_viz import JupyterViz

If you want to want to keep using the old visualisation, you can do so by using the Mesa 2.3.x releases. Mesa 3.0 won't be supporting it.

[github-actions]

Performance benchmarks:

Model	Size	Init time [95% CI]	Run time [95% CI]
Schelling	small	🔵 +0.3% [+0.1%, +0.5%]	🔵 +0.3% [+0.1%, +0.5%]
Schelling	large	🔵 +0.1% [-0.3%, +0.5%]	🔵 +0.7% [+0.3%, +1.1%]
WolfSheep	small	🔵 +0.0% [-0.5%, +0.5%]	🔵 +0.9% [+0.7%, +1.0%]
WolfSheep	large	🔵 +0.0% [-0.4%, +0.5%]	🔵 +0.3% [-0.4%, +0.9%]
BoidFlockers	small	🔵 +0.9% [+0.3%, +1.6%]	🔵 +0.2% [-0.5%, +0.8%]
BoidFlockers	large	🔵 +1.1% [+0.7%, +1.6%]	🔵 +0.2% [-0.9%, +1.4%]

[EwoutH]

Does this break existing models?

[rht]

This is an API breaking change because of the first commit.

[EwoutH]

Good to know! Should we do it in two stages:

• For 2.3, just add a warning we're going to (re)move it
• For 3.0, actually (re)move it

I might also be a proponent of just removing it in 3.0. You can always keep using the older versions if you want or need to. That keeps the codebase clean and lean.

[EwoutH]

I think we can't include this in 2.3 since it's breaking. So we would add a warning for 2.3.0 that we will remove it in 3.0. @rht could you open a separate PR that adds a deprecation warning to the old visualisation?

Then we can merge this PR as soon as 2.3 is branched - which will be this week.

[tpike3]

To be clear on the process:
1 - Separate PR that has a deprecation warning.
2 - Merge this in 3.0

I would agree that we want to remove the old visualization code from the codebase entirely.

But will add, don't we want push people to mesa-viz-tornado? Interestingly other people updated it in February - update

As an aside always thinking about the ABM and Mesa ecosystem this add some intriguing questions on package management.

[EwoutH]

2.3 is released, so we can start with 3.0 development work, including moving this PR forward!

Deprecation warning for 2.3.x would still be nice though.

[rht]

Should I remove the Tornado viz altogether? This will free up the default dependencies and documentation.

[github-actions]

Performance benchmarks:

Model	Size	Init time [95% CI]	Run time [95% CI]
Schelling	small	🔴 +16.5% [+16.1%, +16.9%]	🔴 +25.3% [+25.1%, +25.5%]
Schelling	large	🔴 +16.2% [+15.8%, +16.6%]	🔴 +20.7% [+20.0%, +21.2%]
WolfSheep	small	🔴 +23.6% [+22.2%, +25.1%]	🔴 +205.3% [+200.5%, +210.6%]
WolfSheep	large	🔴 +22.4% [+21.7%, +23.1%]	🔴 +159.7% [+156.6%, +162.9%]
BoidFlockers	small	🔴 +14.2% [+13.5%, +15.0%]	🔴 +10.3% [+9.5%, +11.1%]
BoidFlockers	large	🔴 +13.1% [+12.3%, +13.8%]	🔴 +13.4% [+12.8%, +14.0%]

[EwoutH]

Should I remove the Tornado viz altogether?

I'm good with that. Maybe we need to write a (short) migration guide for 3.0, and include a deprecation warning linking to it when trying to import the old tornado viz.

[rht]

I already wrote on Matrix several days ago

can't put warning on 2.3.0 because the old viz is a direct module import of mesa-viz-tornado classes
you can add to the release note instead

[EwoutH]

Right, then we do that.

[EwoutH]

I would like a second review on this, but from my perspective, this is good to go.

[rht]

I'm merging now because if merge conflicts happen again, this is hard to redo.

[EwoutH]

Agreed! Let’s continue to test and develop this, we have quite some time before 3.0 stable.