#!/usr/bin/env python3
# -*- coding: utf-8 -*-
"""
聲請人歸屬標註　第 2 步：彙整與 QC
==================================
把 50_聲請人標註/全量/*.json 併成一張表，逐筆驗證證據句可逐字還原，
輸出可併檔的 CSV 與樣本流程檔。

    python _pipeline\\聲請人_彙整.py

輸出於 50_聲請人標註/：
    聲請人歸屬.csv     一筆確定終局裁判一列，可用「釋憲案＋案號」與 40_案件歷程 併檔
    聲請人歸屬.json    同上，機器可讀
    sample_flow.json   樣本流程（各步驟之 n 與排除理由）
    QC報告.md          驗證結果
"""

import csv, json, glob, os, re, sys, collections, unicodedata
from pathlib import Path

ROOT = Path(__file__).resolve().parent.parent
OUT = ROOT / "50_聲請人標註"
S_OK = {"S0", "S1", "S2", "S3", "S4"}
D_OK = {"D0a", "D0b", "D1", "D2", "D3", "D4", "D5"}
SN = {"S0": "不明", "S1": "自然人", "S2": "法人或團體", "S3": "法院或法官", "S4": "機關"}
DN = {"D1": "直接指名", "D2": "兩段式對應", "D3": "單一聲請人當然歸屬",
      "D4": "群組已定個別未定", "D5": "原文明示非本件確定終局裁判",
      "D0a": "待查—原文未提及", "D0b": "待查—提及但不能確定"}


def norm(s):
    return re.sub(r"[\s　]", "", unicodedata.normalize("NFKC", s or ""))


def main():
    try:
        sys.stdout.reconfigure(encoding="utf-8")
    except Exception:
        pass

    W = {w["k"]: w for w in json.load(open(OUT / "工作檔.json", encoding="utf-8"))}
    files = sorted(glob.glob(str(OUT / "全量" / "*.json")))
    done = {os.path.basename(f)[:-5] for f in files}
    missing = sorted(set(W) - done)

    rows, bad = [], []
    for f in files:
        k = os.path.basename(f)[:-5]
        rs = json.load(open(f, encoding="utf-8"))
        body = (OUT / "卡片" / f"{k}.txt").read_text(encoding="utf-8").split("=" * 78, 1)[-1]
        cn = norm(body)
        want = [c["案號"] for c in W[k]["cases"]]
        if len(rs) != len(want):
            bad.append(f"{k}：筆數 {len(rs)} 與清單 {len(want)} 不符")
        for i, r in enumerate(rs):
            r["釋憲案"] = k
            r["釋憲字號"] = W[k]["z"]
            rows.append(r)
            if i < len(want) and r.get("案號") != want[i]:
                bad.append(f"{k} idx{r.get('idx')}：案號與清單不符")
            if r.get("S") not in S_OK:
                bad.append(f"{k} idx{r.get('idx')}：非法 S 碼 {r.get('S')}")
            if r.get("D") not in D_OK:
                bad.append(f"{k} idx{r.get('idx')}：非法 D 碼 {r.get('D')}")
            if not r.get("證據句1"):
                bad.append(f"{k} idx{r.get('idx')}：無證據句")
            if r.get("D") != "D0a":
                for fld in ("證據句1", "證據句2"):
                    ev = r.get(fld) or ""
                    if not ev:
                        continue
                    parts = [p for p in re.split(r"…+|\.{3,}|⋯+", ev) if len(norm(p)) >= 8]
                    if [p for p in parts if norm(p) not in cn]:
                        bad.append(f"{k} idx{r.get('idx')}：{fld} 無法在原文逐字還原")

    with open(OUT / "聲請人歸屬.csv", "w", encoding="utf-8-sig", newline="") as fh:
        w = csv.writer(fh)
        w.writerow(["釋憲案", "釋憲字號", "確定終局裁判", "S碼", "聲請人身分別",
                    "D碼", "歸屬確定度", "聲請人序號", "聲請人稱謂原文", "多人合稱",
                    "D信心", "S信心", "證據句1", "證據句2", "註記"])
        for r in rows:
            w.writerow([r["釋憲案"], r["釋憲字號"], r.get("案號", ""),
                        r.get("S", ""), SN.get(r.get("S"), ""),
                        r.get("D", ""), DN.get(r.get("D"), ""),
                        r.get("聲請人序號", ""), r.get("聲請人稱謂原文", ""),
                        r.get("多人合稱", ""), r.get("信心", ""), r.get("S信心", ""),
                        r.get("證據句1", ""), r.get("證據句2", ""), r.get("註記", "")])
    json.dump(rows, open(OUT / "聲請人歸屬.json", "w", encoding="utf-8"), ensure_ascii=False)

    dcnt = collections.Counter(r["D"] for r in rows)
    scnt = collections.Counter(r["S"] for r in rows)
    ided = sum(v for k, v in dcnt.items() if k in ("D1", "D2", "D3"))
    flow = {
        "釋憲案總數": 870,
        "官網可取得確定終局裁判字號者_件": len(W),
        "標註母體_裁判筆數": sum(len(w["cases"]) for w in W.values()),
        "已標註_筆數": len(rows),
        "未標註之釋憲案": missing,
        "排除_未取得裁判字號_件": 769,
        "D碼分布": dict(dcnt.most_common()),
        "S碼分布": dict(scnt.most_common()),
        "可個別歸屬_D1D2D3": ided,
        "群組層級_D4": dcnt.get("D4", 0),
        "非本件確定終局裁判_D5": dcnt.get("D5", 0),
        "待查_D0a_D0b": dcnt.get("D0a", 0) + dcnt.get("D0b", 0),
        "D信心低": sum(1 for r in rows if r.get("信心") == "低"),
        "S信心低": sum(1 for r in rows if r.get("S信心") == "低"),
    }
    json.dump(flow, open(OUT / "sample_flow.json", "w", encoding="utf-8"),
              ensure_ascii=False, indent=1)

    rep = ["# 聲請人歸屬標註　QC 報告", "",
           f"- 標註件數：{len(done)} / {len(W)}",
           f"- 標註筆數：{len(rows)} / {flow['標註母體_裁判筆數']}",
           f"- 未標註之釋憲案：{missing or '無'}",
           f"- **QC 問題：{len(bad)} 項**", ""]
    if bad:
        rep += [f"- {b}" for b in bad[:60]]
    rep += ["", "## D 碼分布（歸屬確定度）", "", "| 碼 | 名稱 | 筆數 | 占比 |", "|---|---|---|---|"]
    for d, n in sorted(dcnt.items()):
        rep.append(f"| {d} | {DN[d]} | {n} | {n/len(rows):.1%} |")
    rep += ["", "## S 碼分布（聲請人身分別）", "", "| 碼 | 名稱 | 筆數 | 占比 |", "|---|---|---|---|"]
    for s, n in sorted(scnt.items()):
        rep.append(f"| {s} | {SN[s]} | {n} | {n/len(rows):.1%} |")
    rep += ["", "## 交叉表 S × D", "",
            "| S＼D | " + " | ".join(sorted(dcnt)) + " |",
            "|---" * (len(dcnt) + 1) + "|"]
    cross = collections.Counter((r["S"], r["D"]) for r in rows)
    for s in sorted(scnt):
        rep.append(f"| {s} {SN[s]} | " +
                   " | ".join(str(cross.get((s, d), 0)) for d in sorted(dcnt)) + " |")
    rep += ["", "## 驗證方法", "",
            "1. 筆數與 `確定終局裁判.csv` 清單逐案比對，案號逐字比對。",
            "2. S／D 碼值域檢查。",
            "3. 每則證據句以「……」切段後，逐段確認為官網原文（解釋文／主文＋理由書／理由）之連續子字串。",
            "   D0a 之證據句為固定語句，免驗。",
            "4. 標註卡片不含任何既有碼，標註者無從確認前次結果。"]
    (OUT / "QC報告.md").write_text("\n".join(rep) + "\n", encoding="utf-8")

    print(f"標註 {len(done)}/{len(W)} 件、{len(rows)}/{flow['標註母體_裁判筆數']} 筆")
    print(f"QC 問題：{len(bad)} 項")
    print("D 碼：", dict(dcnt.most_common()))
    print("S 碼：", dict(scnt.most_common()))
    print("輸出：", OUT / "聲請人歸屬.csv")


if __name__ == "__main__":
    main()