DEV SESSION 11 NOV --- 2000 UTC+8/1400 CET/1200 GMT/ 0600 EST

[tpike3]

Where: via Matrix

Proposed Agenda:

-Small Development Grant and Visualization Improvements
-Mesa-Geo
-PRs and Issues

@rht , @jackiekazil , @Corvince , @wang-boyu , @ankitk50, @rlskoeser

[tpike3]

1st apologies for messing up the time as the US east coast switch to daylight savings time. (I will get it right in December)

Below are the notes:

• We are going to do a Mesa 2.1.5 release this week once we merge docs: Convert HISTORY.rst to HISTORY.md #1866 and possibly convert the README to markdown as we continue to convert all .rst files to md. The primary reason for this is @rlskoeser identified that under the new construct the plots are flickering as the render each time. and @rht edit with jupyterviz agent space and measures flash on mesa 2.1.4 #1867 fixed this.
• AAG mesa and mesa-geo submission is complete for April conference.
• The Mesa community is growing we added several more people to the distro list after CSSSA and the last release.
• As we look at the front end @ankitk50 is working on building tests for the new front end and doing some additional cleanup. As we look at additional features we are thinking the ability to zoom in (imagine a plot with lots and lots of agents with a similar feature to a typical GIS zoom ability) and the ability to scroll over to get information about specific agents. The question for the team (specifically @rht, @Corvince and @wang-boyu ) Should we try and do this organic with matplotlib, or use another library like mpl-interactions, plotly, or bokeh? Please let us know your thought in this discussion.

[wang-boyu]

1st apologies for messing up the time as the US east coast switch to daylight savings time. (I will get it right in December)

Oh no the fault is mine - I should've kept track of that. Sorry.

[wang-boyu]

Regarding interactive plotting, Solara already has support for Plotly...? If users can extract agents as dataframe, they should be able to do something like this: https://github.com/wang-boyu/mesa-examples/blob/gis%2Fsolara-frontend/examples/schelling/app.ipynb

There're some small issues with the code above, but it's also inline with what we discussed here projectmesa/mesa-geo#157 (comment)

[rlskoeser]

I've been using altair for my mesa project because I'm more familiar with it and it solved a problem with plotting my agent space for something I couldn't easily do in matplotlib (I think it was setting colors based on a data attribute, but I may be misremembering). I don't have any particular objection to plotly. Would be good to keep it possible for people working with mesa to still use altair if they want to, since solara supports both.

[tpike3]

Based on on this it seem like altair is the way to go. Any disagreements?

[rht]

According to vega/vega#1738, Vega has 3D visualization, but this has yet to be wrapped in Altair.
My take would be to go with Solara's route, with several configurable plotting backends, but defaults to Altair. When users want to display network space or 3D space, we should raise an error message that says that they need to select the Plotly backend, and to have Plotly installed. This way, pip install mesa, by default, doesn't take much space, just like in pip install solara. This should enable running the visualization code on Binder.

[rht]

This should enable running the visualization code on Binder.

This has yet to be tested. I wasn't able to recall why the current JupyterViz can't run on Binder. It might have something to do with resource limitation.

[jackiekazil]

@tpike3 agree to altair.