Specify Grid Model in Scenario List

[rouille]

• Add a new grid model column in the ScenarioList file
• Test changes in the container

[danielolsen]

Related: we currently have some stuff in powersimdata/utility/constants.py that is contant only within the TAMU model (e.g. mapping zone IDs to states), we will want to move this into the usa_tamu folder.

[danielolsen]

One of the things that @rouille and I discussed last week: where should we record the grid model that was used to build a particular scenario? The options we discussed were either as another column in the ScenarioList.csv file (or database equivalent) vs. a field in the case.mat/grid.mat file. I think it comes up when we want to be able to automatically instantiate a 'base' grid to compare against (previously we would call Grid('Eastern'), soon we may need to call Grid('TAMU', 'Eastern') or something similar).

There are probably many other places in the codebase that will be touched by this change.

[BainanXia]

@danielolsen Probably we will need both. Since the model a scenario used is a key fundamental parameter, we definitely need this info in the scenario_info dict and it might also determines whether certain queries are applicable or not in the case that different models consists of different info (some models have more columns used whereas some models have less), which implies the model info is essential for most of the function calls in the current framework.

[danielolsen]

@BainanXia if we have this in the scenario_info dict (i.e. as another column in the ScenarioList), then why would we need it in the case.mat file? Are you thinking that we may need this info in REISE.jl to control some aspects of model building?

[jenhagg]

I like the idea of having it in the scenario list (file/dict/db/etc), the goal being to have a central place with enough information to recreate a scenario (it can include pointers to external data, like we do with profile versions). Technically for full reproducibility we'd need to store a commit hash or something for the exact code that ran it, but that's for a different conversation.

[rouille]

The ScenarioList is the way to go in my opinion. In the Analyze class, the grid model would be enclosed in the scenario info dict, as the engine, and can then be passed to the Grid class to instantiate an object based on the engine/model combo.

[BainanXia]

@danielolsen Exactly, but in other situations, it might be more convenient to look up this info if it exists more than one place.