{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "0336fd0c",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "✅ База загружена. Треков: 1744\n",
      "🔍 Собираем акустические признаки...\n",
      "\n",
      "🚀 ГОТОВО! Обогащенная база сохранена: ../../dataset/DEAM/music_db_enriched.csv\n",
      "Собрано фичей для 1744 из 1744 треков.\n",
      "   song_id  valence  arousal    energy      flux    centroid       pitch  \\\n",
      "0        2      3.1      3.0  0.097268  0.846947  483.421751   93.884056   \n",
      "1        3      3.5      3.3  0.126809  0.959460  173.219616   62.682589   \n",
      "2        4      5.7      5.5  0.156699  1.333944  466.434797   92.850316   \n",
      "3        5      4.4      5.3  0.126455  1.009927  546.152506  158.673853   \n",
      "4        7      5.8      6.4  0.268180  1.589191  175.369162   83.823484   \n",
      "\n",
      "         hnr       zcr   entropy  sharpness  \n",
      "0   3.615380  0.034270  3.299075   0.426490  \n",
      "1  -2.600122  0.017893  2.294971   0.165583  \n",
      "2  -0.579130  0.042936  3.258138   0.395410  \n",
      "3   1.751148  0.043781  3.514585   0.494367  \n",
      "4  12.006770  0.014783  2.177862   0.170058  \n"
     ]
    }
   ],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "from pathlib import Path\n",
    "from tqdm import tqdm # для красивого прогресс-бара, если не установлен - убери\n",
    "\n",
    "# 1. Пути к файлам\n",
    "base_dir = Path(\"../../dataset/DEAM\") # Поправь, если запускаешь из другого места\n",
    "music_db_path = base_dir / \"music_db.csv\"\n",
    "features_dir = base_dir / \"features\" / \"features\"\n",
    "output_path = base_dir / \"music_db_enriched.csv\"\n",
    "\n",
    "# 2. Наш \"Золотой список\" (8 признаков)\n",
    "target_columns = {\n",
    "    'pcm_RMSenergy_sma_amean': 'energy',\n",
    "    'pcm_fftMag_spectralFlux_sma_amean': 'flux',\n",
    "    'pcm_fftMag_spectralCentroid_sma_amean': 'centroid',\n",
    "    'F0final_sma_amean': 'pitch',\n",
    "    'logHNR_sma_amean': 'hnr',\n",
    "    'pcm_zcr_sma_amean': 'zcr',\n",
    "    'pcm_fftMag_spectralEntropy_sma_amean': 'entropy',\n",
    "    'pcm_fftMag_psySharpness_sma_amean': 'sharpness'\n",
    "}\n",
    "\n",
    "# 3. Загружаем текущую базу с V/A\n",
    "if not music_db_path.exists():\n",
    "    print(f\"❌ ОШИБКА: Не найден файл {music_db_path}\")\n",
    "else:\n",
    "    df_main = pd.read_csv(music_db_path)\n",
    "    print(f\"✅ База загружена. Треков: {len(df_main)}\")\n",
    "\n",
    "    # Подготавливаем новые колонки\n",
    "    for col_name in target_columns.values():\n",
    "        df_main[col_name] = np.nan\n",
    "\n",
    "    # 4. Проходимся по всем трекам и ищем их акустические CSV\n",
    "    print(\"🔍 Собираем акустические признаки...\")\n",
    "    found_count = 0\n",
    "    \n",
    "    for index, row in df_main.iterrows():\n",
    "        song_id = int(row['song_id'])\n",
    "        feature_file = features_dir / f\"{song_id}.csv\"\n",
    "        \n",
    "        if feature_file.exists():\n",
    "            try:\n",
    "                # Читаем CSV с признаками (разделитель там обычно точка с запятой)\n",
    "                df_feat = pd.read_csv(feature_file, sep=';')\n",
    "                \n",
    "                # Усредняем значения по всем фреймам (одна песня разбита на сотни строк-фреймов)\n",
    "                mean_features = df_feat[list(target_columns.keys())].mean()\n",
    "                \n",
    "                # Записываем в главную базу\n",
    "                for orig_col, new_col in target_columns.items():\n",
    "                    df_main.at[index, new_col] = mean_features[orig_col]\n",
    "                    \n",
    "                found_count += 1\n",
    "            except Exception as e:\n",
    "                print(f\"Ошибка чтения {feature_file}: {e}\")\n",
    "                \n",
    "    # 5. Сохраняем результат\n",
    "    # Удаляем треки, для которых не нашлось фичей (если такие есть)\n",
    "    df_main = df_main.dropna(subset=list(target_columns.values()))\n",
    "    \n",
    "    df_main.to_csv(output_path, index=False)\n",
    "    print(f\"\\n🚀 ГОТОВО! Обогащенная база сохранена: {output_path}\")\n",
    "    print(f\"Собрано фичей для {found_count} из {len(df_main)} треков.\")\n",
    "    print(df_main.head())"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python (thesis)",
   "language": "python",
   "name": "thesis"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.11.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}