"""Makaledeki tüm sayısal değerleri çalıştırma çıktılarından yeniden hesaplar.

Önce:  python scripts/run_all.py  ve  python scripts/hata_enjeksiyonu.py
Sonra: python scripts/makale_sayilari.py   ->  outputs/makale_sayilari.json

Kapsam: Tablo 3 (veri kümesi istatistikleri), Tablo 4 (çıkarım), Tablo 5 (EX, Wilson %95 GA,
McNemar), Şekil 3a (altın SQL sonuç dağılımı), Şekil 3b (kategori EX), Tablo 6 (hata enjeksiyonu).
"""
from __future__ import annotations

import collections
import glob
import json
import math
import re
import sys
from pathlib import Path

ROOT = Path(__file__).resolve().parent.parent
sys.path.insert(0, str(ROOT))

from t2s.evaluate import gold_results, load_benchmark, results_match  # noqa: E402
from t2s.schema import connect  # noqa: E402
from t2s.trnum import parse_date  # noqa: E402

OUT = ROOT / "outputs"


def wilson(k: int, n: int, z: float = 1.96):
    p = k / n
    d = 1 + z * z / n
    c = (p + z * z / (2 * n)) / d
    h = z * math.sqrt(p * (1 - p) / n + z * z / (4 * n * n)) / d
    return round(100 * (c - h), 1), round(100 * (c + h), 1)


def mcnemar_exact(b: int, c: int) -> float:
    """İki yönlü kesin McNemar (binom, p=0,5)."""
    n = b + c
    k = min(b, c)
    p = sum(math.comb(n, i) for i in range(0, k + 1)) / 2 ** n
    return min(1.0, 2 * p)


def dataset_stats():
    tabs = [json.loads(Path(f).read_text(encoding="utf-8")) for f in sorted(glob.glob(str(ROOT / "data/ground_truth/*.json")))]
    cells = [c for g in tabs for r in g["govde"] for c in r]
    trnum = sum(1 for c in cells if re.fullmatch(r"-?\(?\d{1,3}(\.\d{3})+(,\d+)?\)?|-?\d+,\d+", c.strip("%")))
    body_merged = 0
    for g in tabs:
        if g["bicim"] == "html":
            html = (ROOT / "data/docs" / g["belge"]).read_text(encoding="utf-8")
            body_merged += len(re.findall(r"<td[^>]*rowspan", html))
    items = load_benchmark()
    return {
        "belge": len({g["belge"] for g in tabs}),
        "tablo": len(tabs),
        "govde_satiri": sum(len(g["govde"]) for g in tabs),
        "govde_hucresi": len(cells),
        "tr_bicimli_sayi": trnum, "tr_bicimli_sayi_yuzde": round(100 * trnum / len(cells), 1),
        "parantezli_negatif": sum(1 for c in cells if c.startswith("(")),
        "yuzde_hucresi": sum(1 for c in cells if "%" in c),
        "tarih_hucresi": sum(1 for c in cells if parse_date(c)),
        "birlesik_baslik": sum(1 for g in tabs for r in g["baslik_satirlari"] for _, cs, rs in r if cs > 1 or rs > 1),
        "birlesik_govde": body_merged,
        "toplam_satirli_tablo": sum(1 for g in tabs if any(re.match(r"(?i)(genel )?toplam", r[0]) for r in g["govde"])),
        "soru": len(items),
        "bolum": dict(collections.Counter(i["bolum"] for i in items)),
        "cevapsiz": sum(1 for i in items if not i["cevaplanabilir"]),
        "zorluk": dict(collections.Counter(i["zorluk"] for i in items)),
        "kanitli": sum(1 for i in items if i["kanit"]),
        "uyarlama": dict(collections.Counter(i["uyarlama"].split(":")[0] for i in items)),
        "coklu_tablo": sum(1 for i in items if len(i["tablolar"]) > 1),
    }


def ablation_detail():
    items = [i for i in load_benchmark() if i["altin_sql"]]
    gold = gold_results(items, OUT / "db/referans.db")
    out = {}
    for m in ("tr", "en", "text"):
        con = connect(OUT / f"db/sistem_{m}.db", readonly=True)
        c = collections.Counter()
        for it in items:
            rows = [tuple(r) for r in con.execute(it["altin_sql"]).fetchall()]
            g = gold[it["id"]]
            if results_match(rows, g):
                c["dogru"] += 1
            elif not rows or all(all(v is None for v in r) for r in rows):
                c["bos_null"] += 1
            elif any(isinstance(v, str) and re.search(r"\d[.,]\d", v) for r in rows for v in r):
                c["metin_sayi"] += 1
            elif any(isinstance(v, (int, float)) for r in rows for v in r):
                c["sayisal_yanlis"] += 1
            else:
                c["diger_yanlis"] += 1
        con.close()
        out[m] = dict(c)
    return out


def main():
    oz = json.loads((OUT / "ozet.json").read_text(encoding="utf-8"))
    rec = {s["sistem"]: [json.loads(l) for l in (OUT / f"sonuclar/sonuclar_{s['sistem']}.jsonl").read_text(encoding="utf-8").splitlines()]
           for s in oz["sistemler"]}
    ex = {}
    for name, rs in rec.items():
        k = sum(r["dogru"] for r in rs)
        t = [r for r in rs if r["bolum"] == "test"]
        kt = sum(r["dogru"] for r in t)
        ex[name] = {"dogru": k, "n": len(rs), "ga95": wilson(k, len(rs)), "test_dogru": kt, "test_n": len(t),
                    "test_ga95": wilson(kt, len(t))}
    base = {r["id"]: r["dogru"] for r in rec["kural_tr"]}
    mc = {}
    for name in rec:
        if name == "kural_tr":
            continue
        other = {r["id"]: r["dogru"] for r in rec[name]}
        b = sum(1 for i in base if base[i] and not other[i])
        c = sum(1 for i in base if not base[i] and other[i])
        mc[name] = {"b": b, "c": c, "p": mcnemar_exact(b, c) if b + c else 1.0}
    res = {"veri_kumesi": dataset_stats(), "ex": ex, "mcnemar": mc, "ablasyon": ablation_detail(),
           "ozet": oz, "hata_enjeksiyonu": json.loads((OUT / "hata_enjeksiyonu.json").read_text(encoding="utf-8"))["ozet"]}
    (OUT / "makale_sayilari.json").write_text(json.dumps(res, ensure_ascii=False, indent=1), encoding="utf-8")
    (OUT / "sayi_ablasyon_detay.json").write_text(json.dumps(res["ablasyon"], ensure_ascii=False), encoding="utf-8")
    print(json.dumps({k: res[k] for k in ("veri_kumesi", "ex", "mcnemar", "ablasyon")}, ensure_ascii=False, indent=1))


if __name__ == "__main__":
    main()