# -*- coding: utf-8 -*-
"""
从 数据.jsonl 提取【最终状态是错误/无字幕】的视频 COS 路径，写到 txt(可当任务列表重跑测试)。
按每个视频的【最后一条记录】判定最终状态(忽略中间重试)。

用法:
  python temp/extract_errors.py 2号机器数据.jsonl error_videos.txt
  python temp/extract_errors.py 2号机器数据.jsonl out.txt error,no_subtitle   # 也要无字幕的
"""
import json
import sys
from collections import Counter, defaultdict

JSONL = sys.argv[1] if len(sys.argv) > 1 else "数据.jsonl"
OUT = sys.argv[2] if len(sys.argv) > 2 else "error_videos.txt"
STATUSES = sys.argv[3].split(",") if len(sys.argv) > 3 else ["error"]


def main():
    try:
        f = open(JSONL, encoding="utf-8")
    except FileNotFoundError:
        print(f"找不到文件: {JSONL}")
        return

    latest = {}          # video_cos_path -> (status, error_reason)
    with f:
        for line in f:
            line = line.strip()
            if not line:
                continue
            try:
                r = json.loads(line)
            except json.JSONDecodeError:
                continue
            vid = r.get("video_cos_path")
            if not vid:
                continue
            latest[vid] = (r.get("status"), r.get("error_reason", ""))

    # 最终状态计数
    status_count = Counter(s for s, _ in latest.values())
    # 选中的(最终状态在 STATUSES 里)
    selected = [(v, er) for v, (s, er) in latest.items() if s in STATUSES]

    # 写 txt(每行一个 COS 路径)
    with open(OUT, "w", encoding="utf-8") as w:
        for v, _ in selected:
            w.write(v + "\n")

    print(f"文件: {JSONL}")
    print(f"去重后视频数: {len(latest)}")
    print(f"最终状态分布: {dict(status_count)}")
    print(f"提取状态 {STATUSES}: {len(selected)} 个 → 写入 {OUT}")

    # 错误原因 Top（帮你看主要错在哪）
    reasons = Counter()
    for _, er in selected:
        if er:
            key = er.split(":")[0][:60] if ":" in er else er[:60]
            reasons[key] += 1
    if reasons:
        print("\n错误原因 Top:")
        for k, c in reasons.most_common(10):
            print(f"  {c:>5}  {k}")


if __name__ == "__main__":
    main()