{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "e782d207-09de-4ca6-a40a-7e2d65d26998",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Found 122635 videos. Starting scan with 64 threads...\n",
      "\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Scanning Videos:   4%|▍         | 4672/122635 [01:51<37:18, 52.69file/s]  "
     ]
    }
   ],
   "source": [
    "import os\n",
    "import av\n",
    "from pathlib import Path\n",
    "from concurrent.futures import ThreadPoolExecutor, as_completed\n",
    "from tqdm import tqdm\n",
    "\n",
    "# --- PYAV LOG SUPPRESSION ---\n",
    "# Suppresses verbose warnings from PyAV so it doesn't break the progress bar\n",
    "av.logging.set_level(av.logging.FATAL)\n",
    "\n",
    "# --- CONFIGURATION ---\n",
    "TARGET_FOLDER = r\"/workspace/musubi-tuner/dataset/ltxxx\" \n",
    "MAX_THREADS = 64\n",
    "# ---------------------\n",
    "\n",
    "def delete_corrupted(file_path, reason):\n",
    "    \"\"\"Helper function to handle the deletion and message formatting.\"\"\"\n",
    "    result_msg = f\"[CORRUPT] {reason}\\n\"\n",
    "    result_msg += f\"    -> File: {file_path.parent.name}/{file_path.name}\\n\"\n",
    "    result_msg += f\"    -> Deleting {file_path.name}...\"\n",
    "    try:\n",
    "        file_path.unlink()  # Deletes the video\n",
    "        return (\"CORRUPTED\", result_msg + \" SUCCESS.\")\n",
    "    except Exception as del_e:\n",
    "        return (\"CORRUPTED\", result_msg + f\" FAILED: {del_e}\")\n",
    "\n",
    "def process_video(file_path):\n",
    "    \"\"\"\n",
    "    Worker function executed by the threads. \n",
    "    Uses PyAV to open the container, precisely mirroring the musubi_tuner dataloader.\n",
    "    \"\"\"\n",
    "    try:\n",
    "        # Opening the container triggers the metadata avdict_to_dict parse.\n",
    "        # This is the exact line that failed in your stack trace.\n",
    "        with av.open(str(file_path)) as container:\n",
    "            \n",
    "            # Optionally, we can also force it to read the first frame \n",
    "            # to catch standard physical corruptions while we are here.\n",
    "            stream = container.streams.video[0]\n",
    "            for frame in container.decode(stream):\n",
    "                break # Just decode one frame and stop\n",
    "\n",
    "        return (\"OK\", \"\")\n",
    "        \n",
    "    except UnicodeDecodeError as e:\n",
    "        # Catches the exact error breaking your dataloader\n",
    "        return delete_corrupted(file_path, f\"UnicodeDecodeError (Metadata corrupted): {e}\")\n",
    "        \n",
    "    except av.error.InvalidDataError as e:\n",
    "        # Catches natively caught PyAV stream corruptions\n",
    "        return delete_corrupted(file_path, f\"PyAV InvalidDataError (Stream corrupted): {e}\")\n",
    "        \n",
    "    except Exception as e:\n",
    "        # Catches any other reason PyAV might refuse to read the file\n",
    "        return delete_corrupted(file_path, f\"PyAV Error: {e}\")\n",
    "\n",
    "def check_and_clean_videos(base_path):\n",
    "    base_dir = Path(base_path)\n",
    "    \n",
    "    if not base_dir.is_dir():\n",
    "        print(f\"Error: The path '{base_dir}' is not a valid directory.\")\n",
    "        return\n",
    "\n",
    "    video_extensions = {'.mp4', '.mkv', '.avi', '.mov', '.wmv', '.flv', '.webm', '.m4v'}\n",
    "    files_to_check =[]\n",
    "\n",
    "    # 1. Gather all matching video files first\n",
    "    for folder in base_dir.iterdir():\n",
    "        if folder.is_dir() and folder.name.lower().startswith(\"videos\") and folder.name.lower() != \"videos\":\n",
    "            \n",
    "            # Using rglob(\"*\") to search recursively\n",
    "            for file_path in folder.rglob(\"*\"):\n",
    "                if file_path.is_file() and file_path.suffix.lower() in video_extensions:\n",
    "                    files_to_check.append(file_path)\n",
    "\n",
    "    if not files_to_check:\n",
    "        print(\"No videos found matching the criteria.\")\n",
    "        return\n",
    "\n",
    "    print(f\"Found {len(files_to_check)} videos. Starting scan with {MAX_THREADS} threads...\\n\")\n",
    "\n",
    "    # 2. Process the gathered files\n",
    "    with ThreadPoolExecutor(max_workers=MAX_THREADS) as executor:\n",
    "        futures =[executor.submit(process_video, path) for path in files_to_check]\n",
    "        \n",
    "        for future in tqdm(as_completed(futures), total=len(files_to_check), desc=\"Scanning Videos\", unit=\"file\"):\n",
    "            status, message = future.result()\n",
    "            \n",
    "            # Print if corrupted\n",
    "            if status != \"OK\":\n",
    "                tqdm.write(message)\n",
    "\n",
    "if __name__ == \"__main__\":\n",
    "    check_and_clean_videos(TARGET_FOLDER)\n",
    "    print(\"\\nScan complete.\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "71101124-eb53-4a33-a4af-e3467d222365",
   "metadata": {},
   "outputs": [],
   "source": [
    " "
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python3 (ipykernel)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.12.13"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}