Give md_ccs_service_unit an explicit UNIT_1 select default

[MaxGhenis]

Summary

Follow-up to PolicyEngine/policyengine-us#9400. That PR's sweep of Enum-valued variables found md_ccs_service_unit as the one remaining select whose conditions are not exhaustive. Its unit_hours bracket returns 0 units for negative or NaN childcare_hours_per_day (verified: calc([-1, 0, 2, 3, 3.0001, 6, 10, nan]) gives [0, 1, 1, 1, 2, 3, 3, 0]), and no condition matched 0, so numpy's select filled those rows with the integer 0.

What happens today

• The raw formula output for a household with an adult (0 hours) and a child with -1 hours is the object array [UNIT_1, 0]. Encoding that directly raises AttributeError: 'int' object has no attribute 'index'; with the unmatched row first it raises ValueError: Invalid value(s) ['0', 'MDCCSServiceUnit.UNIT_1'] for enum MDCCSServiceUnit.
• A full Simulation does not crash today. The variable has defined_for = StateCode.MD, and policyengine-core's defined_for path (simulation.py: np.where(mask, array, default_value) followed by item.index if isinstance(item, Enum) else item) passes the bare 0 through as Enum index 0, which happens to be UNIT_1. Setting defined_for = None on the loaded variable reproduces the crash. So the current answer is right by coincidence of member order; this PR makes it right by rule.

Fix

Drop the redundant units == 1 branch and give the select default=MDCCSServiceUnit.UNIT_1, mirroring #9400. UNIT_1 is the semantically right default: COMAR 13A.14.06.02B(59)(a) defines one unit as 3 hours or less per day, so anything at or below that ceiling, including out-of-range hours, is one unit. It also matches how zero hours is already treated. No output changes for any hours at or above 0 (units 1, 2, 3 map exactly as before); negative hours keep producing UNIT_1. The variable's reference now cites the definitions section (13A.14.06.02) alongside the payment section (.11).

Consumers are consistent with the zero-hours treatment: md_ccs_weekly_copay charges $1/week for UNIT_1 (eligible children only; eligibility does not depend on hours) and md_ccs_payment_rate scales the UNIT_3 base rate by 1/3.

Considered and not done: clamping with max_(hours, 0). It is redundant given the default for negative hours, and numpy.maximum propagates NaN, so it would not cover the NaN case anyway. NaN cannot be supplied as a direct input (core rejects it in set_input); the default guards the computed path regardless.

Tests

• md_ccs_service_unit.yaml: negative hours alone; an adult listed first with a negative-hours child second (the ordering that raises AttributeError on the raw path); a negative-hours child first with a full-time sibling (the ValueError ordering).
• edge_cases.yaml Case 38: negative hours give the $1/week copay, the same as Case 36 (zero hours).
• md_ccs_payment_rate.yaml: negative hours in a Region X licensed center give $127/week ($381 UNIT_3 base × 1/3).
• New policyengine_us/tests/test_md_ccs_service_unit_enum_default.py (collected by make test-other-python): encodes the raw formula output directly, which is the path a variable without defined_for takes, for both row orderings and for NaN. Two of its three tests fail on unfixed main (04a961a) with the object array [UNIT_1, 0]; all three pass with the fix.
• The YAML cases also pass on unfixed main because of the defined_for passthrough, so they pin the intended behaviour rather than catch the crash; the pytest is the regression guard.
• All 86 MD CCS YAML tests pass:

.venv/bin/policyengine-core test policyengine_us/tests/policy/baseline/gov/states/md/msde/ccs -c policyengine_us

• uvx ruff@0.9.0 format and ruff check are clean on the changed Python files. Local environment: Python 3.14.4, policyengine-core 3.30.2.

Related

• #9400 fixes the same bug class in ccdf_age_group and left this variable for a follow-up.

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 100.00%. Comparing base (e21799b) to head (a03e82e).
⚠️ Report is 18 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff            @@
##              main     #9402   +/-   ##
=========================================
  Coverage   100.00%   100.00%           
=========================================
  Files            3         1    -2     
  Lines           33        19   -14     
=========================================
- Hits            33        19   -14     

Flag	Coverage Δ
unittests	100.00% <100.00%> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.