Delete merge_metadata.py

merge_metadata.py

@@ -1,113 +0,0 @@
-#!/usr/bin/env python3
-"""
-将 gmo-svs/annotation 和 soulx-singer-eval/annotation 下的 JSONL 文件
-整理到一个 metadata.csv 文件中。
-"""
-
-import json
-import csv
-from pathlib import Path
-
-
-def load_jsonl(filepath: Path) -> list[dict]:
-    """加载 JSONL 文件，返回记录列表"""
-    records = []
-    with open(filepath, "r", encoding="utf-8") as f:
-        for line in f:
-            line = line.strip()
-            if line:
-                records.append(json.loads(line))
-    return records
-
-
-def process_record(record: dict, source: str) -> dict:
-    """
-    处理单条记录：
-    - 添加 source 字段区分 gmo-svs 和 soulx-singer-eval
-    - 将 wav_fn 重命名为 wav_path（"wa v" 为拼写错误，同等处理）
-    """
-    processed = {}
-    wav_path = None
-    for key, value in record.items():
-        if key == "wav_fn":
-            wav_path = value  # 优先使用 wav_fn
-        elif key == "wa v":  # 原数据中的拼写错误
-            if wav_path is None:
-                wav_path = value
-            # 不加入 processed，避免输出错误列名
-        else:
-            processed[key] = value
-    processed["wav_path"] = wav_path
-    processed["source"] = source
-    return processed
-
-
-def serialize_value(value):
-    """将值序列化为 CSV 可存储的字符串"""
-    if isinstance(value, (list, tuple)):
-        return json.dumps(value, ensure_ascii=False)
-    return value
-
-
-def flatten_record_for_csv(record: dict) -> dict:
-    """
-    展平记录用于 CSV 写入。
-    确保 source 在合适位置，列表类型转为 JSON 字符串。
-    """
-    # 定义列顺序：source 放前面，wav_path 在 wav_fn 位置
-    result = {}
-    for key, value in record.items():
-        result[key] = serialize_value(value)
-    return result
-
-
-def main():
-    base_dir = Path(__file__).parent
-
-    # 定义数据源：(annotation目录路径, 来源标识)
-    sources = [
-        (base_dir / "gmo-svs" / "annotation", "gmo-svs"),
-        (base_dir / "soulx-singer-eval" / "annotation", "soulx-singer-eval"),
-    ]
-
-    all_records = []
-
-    for ann_dir, source_name in sources:
-        if not ann_dir.exists():
-            print(f"警告: 目录不存在，跳过: {ann_dir}")
-            continue
-
-        for jsonl_file in sorted(ann_dir.glob("*.jsonl")):
-            records = load_jsonl(jsonl_file)
-            for rec in records:
-                processed = process_record(rec, source_name)
-                all_records.append(processed)
-            print(f"已加载 {jsonl_file.name}: {len(records)} 条")
-
-    if not all_records:
-        print("未找到任何记录")
-        return
-
-    # 收集所有列名，确保 source 和 wav_path 在合适位置
-    all_keys = set()
-    for rec in all_records:
-        all_keys.update(rec.keys())
-
-    # 列顺序：source 优先，wav_path 替代 wav_fn
-    preferred_order = ["source", "item_name", "dataset", "song_name", "language", "singer", "wav_path"]
-    remaining = sorted(all_keys - set(preferred_order))
-    fieldnames = [k for k in preferred_order if k in all_keys] + remaining
-
-    output_path = base_dir / "metadata.csv"
-    with open(output_path, "w", encoding="utf-8", newline="") as f:
-        writer = csv.DictWriter(f, fieldnames=fieldnames, extrasaction="ignore")
-        writer.writeheader()
-        for rec in all_records:
-            row = flatten_record_for_csv(rec)
-            writer.writerow(row)
-
-    print(f"\n完成! 共 {len(all_records)} 条记录已写入 {output_path}")
-
-
-if __name__ == "__main__":
-    main()