ENH: Top-level module structure of `_gettsim`

[lars-reimann]

Is your feature request related to a problem?

The top-level module structure of _gettsim mixes several concerns:

• Built-in policy functions (aka internal functions) and their parameters:

  • social_insurance_contributions
  • taxes
  • transfers
  • demographic_vars.py
  • functions.py
  • parameters

• Synthetic data:

  • synthetic_data
  • synthetic.py

• "Core" functionality:

  • aggregation.py
  • aggregation_jax.py
  • aggregation_numpy.py
  • config.py
  • functions_loader.py
  • gettsim_typing.py
  • interface.py
  • piecewise_functions.py
  • policy_environment.py
  • shared.py
  • vectorization.py
  • visualization.py

Problem: Because of this it is not immediately obvious, which files must be touched to add a new policy / synthetic data / core functionality. A PR will likely focus on one of the areas.

Other minor issue: config.py needs to describe the PATHS_TO_INTERNAL_FUNCTIONS, which are used by the function loader. It's not possible to simply point to one module that contains all built-in policy functions.

---

Describe the solution you would like to see

Introduce three modules directly beneath _gettsim:

• built_in_policies
• synthetic_data
• core

Potentially add further modules below core (e.g. aggregation).

This helps developers find the code they need to adjust for a new feature.

---

Interface breaking implications

The public gettsim interface can stay unchanged. Any client code that directly depends on the internal _gettsim interface breaks, however.

---

Describe alternatives you've considered

• The names can be replaced by something closer to your domain model.
• built_in_policies and synthetic_data are related and could also be grouped together.

[hmgaudecker]

Makes a lot of sense. Probably useful try to think about what is a general framework and what is specific to the German taxes/transfers system. Not that we have any ambition to expand to other countries, but it may make sense to have a very simple system with just a couple of functions for didactical purposes or in research projects where taxes/transfers play a minor role.