-- 039_bugetar_uat_pattern_match.sql
-- High-precision CUI match for bugetar.entitate UAT entries by stripping
-- ONRC parenthetical suffix "X (PRIMARIA Y)" and comparing normalized names.
--
-- Replaces the trgm-based 038 attempt which was too slow (90+ min, low yield)
-- and hit false positives on the COMUNA-PRIMARIA naming gap.
--
-- Insight: ONRC stores comune/orașe with parenthetical suffix:
--   "Comuna Surduc (Primaria Comunei Surduc)"
--   "COMUNA CIZER (PRIMARIA)"
--   "Comuna Mesesenii de Jos (Primaria Mesesenii de Jos Jud. Salaj)"
-- Bugetar entries are clean: "COMUNA SURDUC". Stripping ' (...)' from ONRC
-- name and comparing normalized → exact match → high-confidence resolve.
--
-- Idempotent: UPDATEs only WHERE cui IS NULL. Threshold-agnostic.
-- Test runtime: ~1.7s per judet, 42 judete → ~70s total.

\timing on
SET pg_trgm.similarity_threshold = 0.78;  -- not used here but reset for safety

WITH judet_map AS (
  SELECT * FROM (VALUES
    ('AB','ALBA'),('AG','ARGES'),('AR','ARAD'),('B','MUNICIPIUL BUCURESTI'),
    ('BC','BACAU'),('BH','BIHOR'),('BN','BISTRITA-NASAUD'),('BR','BRAILA'),
    ('BT','BOTOSANI'),('BV','BRASOV'),('BZ','BUZAU'),('CJ','CLUJ'),
    ('CL','CALARASI'),('CS','CARAS-SEVERIN'),('CT','CONSTANTA'),('CV','COVASNA'),
    ('DB','DAMBOVITA'),('DJ','DOLJ'),('GJ','GORJ'),('GL','GALATI'),
    ('GR','GIURGIU'),('HD','HUNEDOARA'),('HR','HARGHITA'),('IF','ILFOV'),
    ('IL','IALOMITA'),('IS','IASI'),('MH','MEHEDINTI'),('MM','MARAMURES'),
    ('MS','MURES'),('NT','NEAMT'),('OT','OLT'),('PH','PRAHOVA'),
    ('SB','SIBIU'),('SJ','SALAJ'),('SM','SATU MARE'),('SV','SUCEAVA'),
    ('TL','TULCEA'),('TM','TIMIS'),('TR','TELEORMAN'),('VL','VALCEA'),
    ('VN','VRANCEA'),('VS','VASLUI')
  ) AS m(code, name_ascii)
),
candidates AS (
  SELECT DISTINCT ON (b.id) b.id, e.cui, e.name AS firm_name
  FROM bugetar.entitate b
  JOIN judet_map jm ON jm.code = b.judet
  JOIN firms.entities e
    ON firms.normalize_company_name(regexp_replace(e.name, '\s*\(.*$', '')) = firms.normalize_company_name(b.entity_name)
   AND UPPER(translate(COALESCE(e.adr_judet,''), 'ŞȘŢȚăâîĂÂÎ', 'SSTTAAIAAI')) = jm.name_ascii
  WHERE b.cui IS NULL
  ORDER BY b.id, e.cui  -- deterministic when multiple firms share normalized stripped name
)
UPDATE bugetar.entitate b
SET cui = c.cui,
    cui_match_score = 0.95::numeric(5,2),  -- high-confidence stripped-exact match
    cui_match_method = 'uat_pattern',
    updated_at = now()
FROM candidates c
WHERE b.id = c.id AND b.cui IS NULL;

-- Final stats
SELECT cui_match_method, count(*),
       round(avg(cui_match_score)::numeric, 2) AS avg_score
FROM bugetar.entitate
WHERE cui IS NOT NULL
GROUP BY 1
ORDER BY 2 DESC;

SELECT 'unmatched' AS t, count(*) FROM bugetar.entitate WHERE cui IS NULL;