Model/domain
Khalim Conn-Kowlessar c79f574e99 Slice S0380.90: 6 strict-raise dispatches + UnmappedSapCode promoted to shared module
Bundled slice closing the next 6 silent-fallback dispatch sites flagged
by the post-S0380.89 audit per [[reference-unmapped-sap-code]]:

  1. PV pitch (RdSAP 10 §11.1 — codes 1..5 → 0/30/45/60/90°)
  2. PV overshading (SAP 10.2 Table M1 — codes 1..4 → 1.0/0.8/0.5/0.35)
  3. Meter type (RdSAP cert enum 1..5 → Tariff enum)
  4. Tariff → (high, low) rate (RdSAP 10 Table 32 — 4 of 5 Tariffs)
  5. Heat-network DLF by age band (SAP 10.2 Table 12c — A..M)
  6. Secondary heating fraction by main_heating_category (SAP Table 11)

Each dispatch follows the established strict / total split:
  - Absent lodging (None / 0 / "") → cascade's modal-default value
  - Lodging present but unmapped → `UnmappedSapCode(field, value)`

`UnmappedSapCode` promoted from `cert_to_inputs.py` to new module
`domain/sap10_calculator/exceptions.py` so `tables/table_12a.py` can
raise it too (the meter-type dispatch lives there). `cert_to_inputs`
re-exports it for backward compat with existing test imports.

Corpus audit at HEAD 6d02d205 (full JSON sweep):

  PV pitch codes:           {2, 3}        — covered
  PV overshading codes:     {1, 2}        — covered
  meter_type codes:         {1, 2, 3}     — covered (incl. digit-string '2')
  main_heating_category:    {2, 4, 6, 7, 10} — covered

All corpus codes already in dispatch dicts — no production regression
expected.

**One silent runtime fix surfaced by the strict-raise rollout**: the
GOV.UK API lodges `meter_type` as a digit-string (e.g. `'2'`) on many
certs, but the original `_METER_STR_TO_INT` dict only had word aliases
("single", "dual", "unknown"). Pre-S0380.90 the digit-string fell
through to the silent `return Tariff.STANDARD` default. Adding a
`key.isdigit() → int(key)` short-circuit routes these through the int
enum correctly. Confirmed 125 golden cert fixtures previously running
on this silent default — all now passing with explicit STANDARD via
the int dispatch path (not via the silent fallback).

Tests (6 new, AAA-structure):

  - `test_pv_pitch_deg_full_table_coverage_per_rdsap_10_section_11_1`
  - `test_pv_overshading_factor_full_table_m1_coverage`
  - `test_meter_type_dispatch_full_table_12a_coverage` (incl. digit-string)
  - `test_tariff_high_low_rates_full_dispatch_coverage`
  - `test_heat_network_dlf_full_table_12c_age_band_coverage`
  - `test_secondary_heating_fraction_for_category_full_table_11_coverage`

Each test pins: spec-correct codes → expected dispatch result; absent
lodging → modal default; lodging present but unmapped → `UnmappedSapCode`
with field + value attached.

Test baseline: 574 pass (was 568 + 6 new) + 9 expected
`test_sap_result_pin[000565-*]` fails unchanged. Cohort + golden +
cert 9501 unaffected. Pyright net-zero per touched file.

Open silent-fallback inventory now empty per
[[reference-unmapped-sap-code]] — the cascade dispatch boundary is
now fully strict-raise-gated for code translations. Cascade VALUE
defaults (u_wall, u_floor, etc.) remain total per RdSAP §6.2.3.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
2026-06-01 16:28:47 +00:00
..
addresses standardist Address 2026-05-22 10:13:32 +00:00
data_transformation moved classifier data transformation to an easy one 2026-06-01 14:53:34 +00:00
epc pr review, move domain and orhcestration 2026-06-01 14:00:31 +00:00
sap10_calculator Slice S0380.90: 6 strict-raise dispatches + UnmappedSapCode promoted to shared module 2026-06-01 16:28:47 +00:00
sap10_ml Slice S0380.86: §5.6 thin-wall stone + §5.8 dry-line closes BP[0] alt1 cascade gap 2026-06-01 16:28:47 +00:00
tasks added postcode splitter rewrite to ddd 2026-05-19 16:35:09 +00:00
postcode.py get rid of comments 2026-05-20 13:21:11 +00:00