#!/usr/bin/env python3
# -*- coding: utf-8 -*-
"""
補完：把白話解說與 OCR 補正併入資料庫，重寫筆記與網頁。
不連網、不重抓、不重新處理 PDF，全部從既有的 _pipeline/資料庫.json 出發。

    cd C:\\Users\\JudyLin\\Documents\\釋字及憲法法庭資料庫
    python _pipeline\\補完.py

產出：
  - 10_釋字/、20_憲判/ 筆記加上「白話解說（AI 生成）」區塊
  - 19 篇原本無文字層的意見書補上 OCR 文字
  - 釋字及憲法法庭資料庫.html         完整版（含意見書全文）
  - 釋字及憲法法庭資料庫_輕量版.html   只到理由書與白話解說，意見書留連結
  - 00_索引/校驗報告.md 更新
"""
import json, sys, time, importlib.util
from pathlib import Path

PIPE = Path(__file__).resolve().parent
ROOT = PIPE.parent

spec = importlib.util.spec_from_file_location("builder", PIPE / "建置資料庫.py")
B = importlib.util.module_from_spec(spec)
spec.loader.exec_module(B)


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

    dbp = PIPE / "資料庫.json"
    if not dbp.exists():
        print("找不到 _pipeline/資料庫.json，請先執行 建置資料庫.py")
        sys.exit(1)
    recs = json.load(open(dbp, encoding="utf-8"))
    print(f"載入 {len(recs)} 筆")

    ai = json.load(open(PIPE / "白話解說.json", encoding="utf-8"))
    ocr = json.load(open(PIPE / "OCR補正.json", encoding="utf-8"))
    opai_p = PIPE / "意見書白話.json"
    opai = json.load(open(opai_p, encoding="utf-8")) if opai_p.exists() else {}
    print(f"白話解說 {len(ai)} 筆；OCR 補正 {len(ocr)} 篇；意見書白話 {len(opai)} 篇")

    n_ai = n_ocr = 0
    for r in recs:
        if r["k"] in ai:
            r["ai"] = ai[r["k"]]
            n_ai += 1
        for i, o in enumerate(r["o"], 1):
            key = f'{r["k"]}#{i}'
            if key in ocr and not o.get("x"):
                o["x"] = ocr[key] + "\n\n（本篇原始 PDF 無文字層，以繁體中文 OCR 擷取，可能有辨識誤差；請以官網 PDF 為準）"
                o.pop("low", None)
                n_ocr += 1
    n_op = 0
    for r in recs:
        for i, o in enumerate(r["o"], 1):
            key = f'{r["k"]}#{i}'
            if key in opai:
                o["ai"] = opai[key]
                n_op += 1
    print(f"併入白話解說 {n_ai} 筆；補上 OCR {n_ocr} 篇；意見書白話 {n_op} 篇")

    roster = B.derive_judges(recs)
    nw = sum(1 for r in recs if r["j"]["w"])
    npro = len({x for r in recs for x in r["j"]["p"]})
    print(f"大法官名冊 {len(roster)} 人；可辨識主筆 {nw} 則；曾提出意見書 {npro} 人")

    print("重寫 Markdown 筆記與法條索引 …")
    nlaw = B.write_notes(recs)
    B.write_index(recs)

    print("重寫網頁（完整版）…")
    out1, s1 = B.write_html(recs)
    print("重寫網頁（輕量版）…")
    out2, s2 = B.write_html(recs, lite=True, filename="釋字及憲法法庭資料庫_輕量版.html")

    still = sum(1 for r in recs for o in r["o"] if not o.get("x"))
    rep = (ROOT / "00_索引" / "校驗報告.md")
    txt = rep.read_text(encoding="utf-8") if rep.exists() else "# 校驗報告\n"
    txt += (f"\n\n## 補完（{time.strftime('%Y-%m-%d %H:%M:%S')}）\n\n"
            f"- 白話解說：{n_ai} / {len(recs)} 則，皆標示為 AI 生成並與原文分區\n"
            f"- 意見書白話解說：{n_op} / {sum(len(r['o']) for r in recs)} 篇\n"
            f"- OCR 補正：{n_ocr} 篇\n"
            f"- 仍無文字之意見書：{still} 篇\n"
            f"- 法條節點：{nlaw}\n"
            f"- 完整版網頁：{s1/1048576:.1f} MB（字元數）\n"
            f"- 輕量版網頁：{s2/1048576:.1f} MB（字元數）\n")
    rep.write_text(txt, encoding="utf-8")

    json.dump(recs, open(dbp, "w", encoding="utf-8"), ensure_ascii=False)
    print("=" * 60)
    print("完成。")
    print("  完整版：", out1)
    print("  輕量版：", out2)
    print("  仍無文字之意見書：", still, "篇")
    print("=" * 60)


if __name__ == "__main__":
    main()