"""Missão 3 / Bloco B — de onde vem o CST/alíquota de PIS/COFINS emitido, somente leitura.
Testa correlação e determinismo por NCM, CODGRUPO (via C000025), CFOP, CSOSN/CST_ICMS,
GRUPO_ICMS. Não decide regra fiscal - só mede taxa de reprodução.
"""
import json
import os
import sys
from pathlib import Path
from collections import Counter, defaultdict

ROOT = Path(__file__).resolve().parents[1]
sys.path.insert(0, str(ROOT)); sys.path.insert(0, str(ROOT / "scripts")); sys.path.insert(0, str(ROOT / "agent" / "src"))
import fdb  # noqa: E402
import run_fiscal_decision_v1 as base  # noqa: E402

DBS = {
    "GTI": r"D:\PROJETOS\GTISISTEMAS\Banco\GTI.FDB",
    "CONSTRUCAO": str(ROOT / "Banco" / "CONSTRUCAO.FDB"),
    "MMMODAS": str(ROOT / "Banco" / "MMMODAS.FDB"),
    "RINCAO": str(ROOT / "Banco" / "RINCAO.FDB"),
    "UNIAO": str(ROOT / "Banco" / "UNIAO.FDB"),
}


def connect(path):
    return fdb.connect(database=path, user=os.environ["FB_USER"], password=os.environ["FB_PASSWORD"],
                        fb_library_name=os.environ.get("FIREBIRD_CLIENT_LIBRARY"), isolation_level=fdb.ISOLATION_LEVEL_READ_COMMITED_RO)


def analyze(name, path):
    conn = connect(path); cur = conn.cursor()
    cur.execute("SELECT COUNT(*), COUNT(NULLIF(TRIM(CST_PIS_COFINS),'')) FROM NFCE_ITENS")
    tot, filled = cur.fetchone()
    out = {"name": name, "items_total": tot, "items_cst_filled": filled}
    if filled == 0:
        out["conclusion"] = "NAO_APLICAVEL_SEM_DADOS_PARA_CORRELACIONAR"
        conn.close()
        return out
    # candidate determinants, only over the filled subset
    for dim, expr in (("NCM", "i.NCM"), ("CODGRUPO", "p.CODGRUPO"), ("CFOP", "i.CFOP"), ("CSOSN", "i.CSOSN"), ("GRUPO_ICMS", "i.GRUPO_ICMS")):
        try:
            cur.execute(f"""
                SELECT {expr}, TRIM(i.CST_PIS_COFINS), COUNT(*)
                FROM NFCE_ITENS i LEFT JOIN C000025 p ON p.CODIGO = i.CODPRODUTO
                WHERE COALESCE(TRIM(i.CST_PIS_COFINS),'') != ''
                GROUP BY 1, 2
            """)
            rows = cur.fetchall()
            by_key = defaultdict(Counter)
            for k, cst, n in rows:
                by_key[k][cst] += n
            deterministic = sum(sum(c.values()) for c in by_key.values() if len(c) == 1)
            total_filled_dim = sum(sum(c.values()) for c in by_key.values())
            out[f"determinism_{dim}"] = {"distinct_keys": len(by_key), "items_covered": total_filled_dim,
                                         "items_with_single_cst_for_key": deterministic,
                                         "reproduction_rate_pct": round(100.0 * deterministic / total_filled_dim, 2) if total_filled_dim else None}
        except Exception as e:  # noqa: BLE001
            out[f"determinism_{dim}_error"] = str(e)[:150]
    # aliquota_pis fill vs cst fill: same items or different?
    cur.execute("SELECT COUNT(*) FROM NFCE_ITENS WHERE ALIQUOTA_PIS IS NOT NULL AND COALESCE(TRIM(CST_PIS_COFINS),'')=''")
    out["aliquota_filled_but_cst_blank"] = cur.fetchone()[0]
    cur.execute("SELECT COUNT(*) FROM NFCE_ITENS WHERE ALIQUOTA_PIS IS NULL AND COALESCE(TRIM(CST_PIS_COFINS),'')!=''")
    out["cst_filled_but_aliquota_null"] = cur.fetchone()[0]
    conn.close()
    return out


def main():
    base.load_dotenv(ROOT / ".env")
    results = {}
    for name, path in DBS.items():
        print(f"== {name}", flush=True)
        results[name] = analyze(name, path)
        print(json.dumps(results[name], ensure_ascii=False, indent=1, default=str), flush=True)
    out_path = ROOT / "docs" / "runs" / "BLOCOB_PIS_COFINS_SOURCE_V1.json"
    out_path.write_text(json.dumps(results, ensure_ascii=False, indent=1, default=str), encoding="utf-8")
    print(f"-> {out_path}")


if __name__ == "__main__":
    main()