mirror of
https://github.com/Hestia-Homes/Model.git
synced 2026-07-12 13:29:04 +00:00
dee72bbdfa
_has_suspended_timber_floor_per_spec only read the Main bp's per-part floor_construction_type lodgement; the gov-API mapper frequently leaves that None even when the global epc.floors[].description carries an explicit "Suspended, ..." observation. _main_floor_u_value already fell back to this joined description for the U-value calc — the infiltration rule did not, so a genuinely suspended-timber floor silently entered (12)=0 instead of 0.2 unsealed. Extracted the shared _effective_floor_description helper and added a tri-state resolver (explicit lodgement -> description, excluding "not timber" -> Table 19 footnote-1 age-A/B default) so both paths agree. Cert 100061275133 (Elmhurst-validated build, PR #1439 handoff): engine 77.24 -> 76.33, now an exact match to lodged (76). Corpus gauge 77.7% -> 77.8% within-0.5, MAE 0.637 -> 0.636; floors ratcheted. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> |
||
|---|---|---|
| .. | ||
| addresses | ||
| billing | ||
| data_transformation | ||
| elmhurst | ||
| epc | ||
| epc_prediction | ||
| fuel_rates | ||
| geospatial | ||
| magicplan | ||
| modelling | ||
| property | ||
| property_baseline | ||
| sap10_calculator | ||
| sap10_ml | ||
| tasks | ||
| building_geometry.py | ||
| postcode.py | ||