#!/usr/bin/env python3
"""
使用 AnimeWwise 引擎无头提取昔涟语音。
不需要 GUI，直接调用 extract.py 的核心逻辑。
"""

import os
import sys

# 添加 AnimeWwise 到路径
AWW_DIR = r"D:\Project\Code\Uni\Cyrene-Voice-Model\tools\AnimeWwise"
sys.path.insert(0, AWW_DIR)
os.chdir(AWW_DIR)

from extract import WwiseExtract

# 输入目录
HSR_AUDIO = r"D:\MeowG\Honkai：Star_Rail\StarRail_Data\Persistent\Audio\AudioPackage\Windows\Chinese(PRC)"
OUTPUT_DIR = r"D:\Project\Code\Uni\Cyrene-Voice-Model\data\cyrene_extracted"
MAP_FILE = "hkrpg.map"

print("=" * 60)
print("昔涟语音提取 (AnimeWwise Headless)")
print("=" * 60)

# 1. 初始化提取器
extractor = WwiseExtract()

# 2. 加载映射
print("\n[1/4] 加载角色映射...")
extractor.load_map(MAP_FILE)

# 3. 加载音频文件
print("\n[2/4] 加载 .pck 文件...")
pck_files = sorted([
    os.path.join(HSR_AUDIO, f)
    for f in os.listdir(HSR_AUDIO)
    if f.endswith(".pck") and ("VoBanks" in f or "External_del_3." in f)
])

print(f"  找到 {len(pck_files)} 个目标 .pck 文件")

# 进度回调
def progress(data):
    if data[0] == "total":
        pct = int(data[1])
        if pct % 20 == 0:
            print(f"  加载进度: {pct:.0f}%")

file_structure = extractor.load_folder(
    _map=MAP_FILE,
    files=pck_files,
    diff_path="",
    base_path=HSR_AUDIO,
    progress=progress,
)

# 4. 搜索昔涟
print("\n[3/4] 搜索昔涟语音...")

def search_cyrene(structure, path=""):
    """递归搜索文件结构中包含昔涟相关名称的文件"""
    results = []
    cyrene_terms = ['cyrene', 'mimi', 'xilian', 'mem', 'de_moi_ge']

    # 搜索文件夹
    folders = structure.get("folders", {})
    for folder_name, folder_content in folders.items():
        sub_path = f"{path}/{folder_name}" if path else folder_name
        results.extend(search_cyrene(folder_content, sub_path))

    # 搜索文件
    for file_entry in structure.get("files", []):
        file_name = file_entry[0].lower()
        file_path = f"{path}/{file_entry[0]}"
        for term in cyrene_terms:
            if term in file_name or term in file_path.lower():
                results.append((path, file_entry[0], file_entry[1]))
                break

    return results

matches = search_cyrene(file_structure)
print(f"  找到 {len(matches)} 个昔涟相关文件")

if len(matches) == 0:
    print("\n  未匹配到昔涟，显示所有顶层目录结构:")
    folders = file_structure.get("folders", {})
    for name in sorted(folders.keys()):
        count = len(folders[name].get("files", []))
        # 递归统计
        def count_all(s):
            total = len(s.get("files", []))
            for v in s.get("folders", {}).values():
                total += count_all(v)
            return total
        total = count_all(folders[name])
        print(f"    {name}/ ({total} files)")

    # 也搜一下其他可能的名称
    print("\n  搜索所有包含'voice'的路径...")
    voice_matches = search_cyrene(file_structure)
    # 换个方式搜
    def search_all(structure, prefix=""):
        for folder_name, folder_content in structure.get("folders", {}).items():
            sub = f"{prefix}/{folder_name}" if prefix else folder_name
            sub_lower = sub.lower()
            if any(t in sub_lower for t in ['voice', 'char', 'npc', 'player', 'avatar']):
                print(f"    {sub}/")
            search_all(folder_content, sub)

    search_all(file_structure)
else:
    for path, name, meta in matches[:30]:
        print(f"    {path}/{name}")

# 5. 提取
if matches:
    print(f"\n[4/4] 提取 {len(matches)} 个文件到 {OUTPUT_DIR}...")
    output_files = []
    for path, name, meta in matches:
        output_files.append({
            "path": path.split("/") if path else [],
            "name": name,
            "source": meta["source"],
            "offset": meta["offset"],
            "size": meta["size"],
            "original_name": meta["original_name"],
        })

    extractor.extract_files(
        _input=HSR_AUDIO,
        files=output_files,
        output=OUTPUT_DIR,
        _format="wav",
        progress=progress,
    )
    print(f"\n完成！文件保存在 {OUTPUT_DIR}")
else:
    print("\n未找到昔涟语音。可能需要用 AnimeWwise GUI 手动浏览文件结构。")

extractor.reset()