_normalise_period_key drops the stop date of start.stop reform keys #724

_normalise_period_key (projects/policyengine-simulation-executor/src/policyengine_simulation_executor/simulation_runtime.py) reduces a start.stop period key to its start date:

def _normalise_period_key(period_key: Any) -> str:
    """Convert legacy ``start.stop`` period keys to v4 effective dates."""
    text = str(period_key)
    parts = text.split(".")
    if len(parts) > 1 and len(parts[0]) == 10:
        return parts[0]
    return text

_normalise_policy applies it to both the baseline and the reform policy before they reach policyengine.core.Simulation(policy=...). In policyengine.py main:

• compile_reform_to_policy turns each {path: {date: value}} entry into a ParameterValue with end_date=None (tax_benefit_models/common/reform.py);
• reform_dict_from_parameter_values (utils/parametric_reforms.py) runs such an open-ended value until the day before the next value of the same parameter starts, or to 2100-12-31.

So the stop date of a bounded app value never reaches the simulation. For example, {"2026-01-01.2026-12-31": 0.05} stays in force after 2026, until the next start date the policy gives that parameter, or to 2100. In a budget window, later years then use the reform value instead of the baseline.

This comes from reading sim-api main and policyengine.py main on 2026-10-06; I have not run it. The release bundle sim-api pins may resolve an older policyengine.py. Found while reviewing PolicyEngine/policyengine-us#9628.