# -*- coding: utf-8 -*-
"""voice 引用一致性校验（26 号推广工程配套）。

对每款深度重构游戏，扫描 game.js（含 index.html / sessions）里引用的全部
'voice/<id>' 字面量，对照 tools/gen_voice_dual.py 的台词表（GAMES + NPCS）：

  1. 引用了但台词表没有 → 多半是接线笔误（未生成就先报出来）
  2. NPC 表 / 平铺目录里存在但双性别表没有 → 合法（角色音平铺），单独计数
  3. 台词表里有但代码从不引用 → 死句（列出来人工裁定：可能被 session 或
     链式构造间接引用，只警告不判错）

用法：.venv-tts/bin/python tools/validate_voice_refs.py [game ...]
不带参数 = 校验全部 15 款。
"""
import importlib.util
import json
import os
import re
import sys

ROOT = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
sys.path.insert(0, os.path.join(ROOT, "tools"))

spec = importlib.util.spec_from_file_location(
    "gen_voice_dual", os.path.join(ROOT, "tools", "gen_voice_dual.py"))
gvd = importlib.util.module_from_spec(spec)
spec.loader.exec_module(gvd)

VOICE_RE = re.compile(r"['\"]voice/([A-Za-z0-9_]+)['\"]")

# 每个 game 要扫描的源文件（相对 web/<game>/）
SRC_FILES = ["game.js", "index.html"]
# 会话剧本也引用 id（transcribe 用），只用于「死句」判定的豁免
SESSION_GLOB = os.path.join(ROOT, "web", "harness", "sessions", "{game}.mjs")


def scan_game(game):
    refs = set()
    gdir = os.path.join(ROOT, "web", game)
    for name in SRC_FILES:
        p = os.path.join(gdir, name)
        if not os.path.exists(p):
            continue
        with open(p, encoding="utf-8") as f:
            refs |= set(VOICE_RE.findall(f.read()))
    # 会话剧本（豁免死句用）
    ses = set()
    sp = SESSION_GLOB.format(game=game)
    if os.path.exists(sp):
        with open(sp, encoding="utf-8") as f:
            ses |= set(VOICE_RE.findall(f.read()))
    return refs, ses


def flat_ids(game):
    """已有资产 id：顶层平铺（角色/NPC 单音色）+ male/female 目录文件名
    （含早期轮次落盘但不在台词表的「在库沿用」资产，如 rhythm 的 gd_*）。"""
    vdir = os.path.join(ROOT, "web", game, "audio", "voice")
    if not os.path.isdir(vdir):
        return set()
    ids = {f.rsplit(".", 1)[0] for f in os.listdir(vdir) if "." in f}
    for sub in ("male", "female"):
        d = os.path.join(vdir, sub)
        if os.path.isdir(d):
            ids |= {f.rsplit(".", 1)[0] for f in os.listdir(d) if "." in f}
    return ids


def main():
    games = sys.argv[1:] or list(gvd.GAMES.keys())
    bad = 0
    for game in games:
        # 双性别台词表 + NPC 平铺表 + 平铺资产文件 + 键盘变体（同 id）
        table = {fid for fid, _i, _t in gvd.GAMES.get(game, [])}
        npc = {fid for fid, _s, _in, _t in gvd.NPCS.get(game, [])}
        flat = flat_ids(game)
        key = {fid for fid, _t in gvd.KEY_LINES.get(game, [])}
        refs, ses = scan_game(game)
        # 引用解析：id 直配；动态前缀引用（'voice/' + x 拼接）不在本工具能力内，
        # 游戏若声明 DYNAMIC_VOICE_PREFIX 则跳过该项
        # 尾下划线 = 动态前缀拼接（'voice/btl_loot_' + n），不在直配能力内
        unresolved = sorted(
            r for r in refs
            if not r.endswith("_") and r not in table and r not in npc and r not in flat)
        # 死句：表里有、代码与会话都不引用
        dead = sorted(
            fid for fid in table
            if fid not in refs and fid not in ses and fid not in key)
        status = "OK " if not unresolved else "BAD"
        if unresolved:
            bad += 1
        print(f"[{status}] {game}: 引用 {len(refs)}，表内 {len(table)}，NPC {len(npc)}，未解析 {len(unresolved)}，疑似死句 {len(dead)}")
        for u in unresolved:
            print(f"    未解析: voice/{u}")
        for d in dead[:10]:
            print(f"    死句?: {d}")
        if len(dead) > 10:
            print(f"    …另有 {len(dead) - 10} 条")
    sys.exit(1 if bad else 0)


if __name__ == "__main__":
    main()