search

openfisca / country-template

Start modelling the tax and benefit system of your country in a few minutes.
https://legislation.demo.openfisca.org
GNU Affero General Public License v3.0
33 stars 15 forks source link

Added example script for running the situation examples #149

Open verbman opened 6 months ago

verbman commented 6 months ago

The OpenFisca docs link to a France specific example here that is out of date and should use SimulationBuilder. This PR adapts the situation examples example from the tutorial and applies it here. This would provide us the opportunity to provide simple code tests to the documentation that isn't France specific. For instance here we could instead use this repository as the example and provide the script python -m openfisca_country_template.situation_examples.run to prove installation.

MattiSG commented 6 months ago

Thank you for this suggestion.

The doc page you linked to is called “How to test your changes on “ready to use” situations (for OpenFisca-France)”. It is clearly country-specific and should be erased as such. The idea it points at, however, is relevant and interesting: the idea of ensuring each country model shares and maintains these “ready to use situations”, or “cas types” in French (apparently translated as “typical cases”) for reuse. It is already in large part implemented by the existence of these “situation examples”. I would suggest renaming this folder to typical_cases to surface their usage, and adapt the documentation to point that every country package is encouraged to provide their own typical cases, and that their default location is in this folder.

I am not to see the point of providing trivial Python code to then load these typical cases. If you believe it can be worthwhile, I would suggest making this file an actual part of the test suite to systematically validate the validity of the typical cases 🙂 In this way, every country model would be pushed to maintain them. Ideally, we would find a way to check this validity without requesting calculations that depend on the model variables (i.e. just make sure that every input variable referenced in the typical case is a valid one, without requesting any variable to be computed, so that the tests don't need to be updated when the output variables would change).