Tools#

The policyengine_core.tools module contains miscellaneous utility functions, including for running YAML tests.

run_tests#

policyengine_core.tools.test_runner.run_tests(tax_benefit_system, paths, options=None)#

Runs all the YAML tests contained in a file or a directory.

If path is a directory, subdirectories will be recursively explored.

Parameters:

  • tax_benefit_system (TaxBenefitSystem) – the tax-benefit system to use to run the tests

  • paths ((str/list)) – A path, or a list of paths, towards the files or directories containing the tests to run. If a path is a directory, subdirectories will be recursively explored.

  • options (dict) – See more details below.

Raises:

AssertionError – if a test does not pass

Returns:

the number of sucessful tests excecuted

Testing options:

dump_simulation#

policyengine_core.tools.simulation_dumper.dump_simulation(simulation, directory)#

Write simulation data to directory, so that it can be restored later.

Writes the simulation’s branch values, including inherited values, and lists the periods it calculated (see restore_simulation).

restore_simulation#

policyengine_core.tools.simulation_dumper.restore_simulation(directory, tax_benefit_system, **kwargs)#

Restore simulation from directory

Each value is restored as the dumped simulation stored it: one it calculated (listed in the variable’s derived_periods.txt) as a calculated value, and every other value as an input, recorded in _user_input_keys as set_input records one, so apply_reform keeps the inputs and drops the calculated values, as it would have in the dumped simulation.

A dump lists which values were calculated, not where an input came from: a value the dumped simulation stored with put_in_cache without derived=True (an input to auto-carry-over and uprating that set_input did not record, so that apply_reform dropped it there) is restored as a recorded input too.

A dump written before policyengine-core 3.32.16 lists no calculated value, so every value in it is restored as an input. A dump without a format marker that lists none may be one of those, so restoring it warns: apply_reform then keeps any calculated value in it as dumped instead of recalculating it.