NFL ETL.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "import datetime, pandas as pd\n",
    "\n",
    "# Global locations of the raw data, and where to put the results\n",
    "rawDataLocation = './RawData/'\n",
    "resultsDataLocation = './ResultsData/'\n",
    "\n",
    "# True to see print statements throughout\n",
    "verbose = True\n",
    "\n",
    "# Number of games to include; -1 for all; games are randomly selected\n",
    "noOfGames = 45\n",
    "\n",
    "# This unique number is appended to all of the results files\n",
    "# If there is already a result file with this number, that section of the code\n",
    "# is skipped\n",
    "resultFilePrefix = '45Games';\n",
    "#resultFilePrefix = '{date:%Y-%m-%d_%H.%M.%S}'.format(date=datetime.datetime.now())\n",
    "\n",
    "# Files names for each of the steps; saveInterim will save steps 1 and 2\n",
    "saveInterim = False\n",
    "stepOneFileName = resultsDataLocation+resultFilePrefix+'_nfl_allTracking.csv'\n",
    "stepTwoFileName = resultsDataLocation+resultFilePrefix+'_nfl_flat.csv'\n",
    "stepThreeFileName = resultsDataLocation+resultFilePrefix+'_nfl_flatWithTeamUniquePlay.csv'\n",
    "\n",
    "# The dataframe we're going to fill up\n",
    "alltracking = pd.DataFrame()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Reading 45 out of 91 files\n",
      " Read 001 : ./RawData/tracking_gameId_2017100105.csv\n",
      "  - found 264363 records\n",
      " Read 002 : ./RawData/tracking_gameId_2017100800.csv\n",
      "  - found 289368 records\n",
      " Read 003 : ./RawData/tracking_gameId_2017100803.csv\n",
      "  - found 306867 records\n",
      " Read 004 : ./RawData/tracking_gameId_2017100104.csv\n",
      "  - found 297029 records\n",
      " Read 005 : ./RawData/tracking_gameId_2017091003.csv\n",
      "  - found 247642 records\n",
      " Read 006 : ./RawData/tracking_gameId_2017092500.csv\n",
      "  - found 258448 records\n",
      " Read 007 : ./RawData/tracking_gameId_2017092800.csv\n",
      "  - found 250425 records\n",
      " Read 008 : ./RawData/tracking_gameId_2017092400.csv\n",
      "  - found 257406 records\n",
      " Read 009 : ./RawData/tracking_gameId_2017092402.csv\n",
      "  - found 242053 records\n",
      " Read 010 : ./RawData/tracking_gameId_2017101508.csv\n",
      "  - found 283950 records\n",
      " Read 011 : ./RawData/tracking_gameId_2017100809.csv\n",
      "  - found 247270 records\n",
      " Read 012 : ./RawData/tracking_gameId_2017091007.csv\n",
      "  - found 259508 records\n",
      " Read 013 : ./RawData/tracking_gameId_2017101504.csv\n",
      "  - found 331565 records\n",
      " Read 014 : ./RawData/tracking_gameId_2017092408.csv\n",
      "  - found 253045 records\n",
      " Read 015 : ./RawData/tracking_gameId_2017092404.csv\n",
      "  - found 288835 records\n",
      " Read 016 : ./RawData/tracking_gameId_2017100107.csv\n",
      "  - found 264000 records\n",
      " Read 017 : ./RawData/tracking_gameId_2017100108.csv\n",
      "  - found 308873 records\n",
      " Read 018 : ./RawData/tracking_gameId_2017091004.csv\n",
      "  - found 293205 records\n",
      " Read 019 : ./RawData/tracking_gameId_2017100109.csv\n",
      "  - found 326900 records\n",
      " Read 020 : ./RawData/tracking_gameId_2017100801.csv\n",
      "  - found 245381 records\n",
      " Read 021 : ./RawData/tracking_gameId_2017101510.csv\n",
      "  - found 246972 records\n",
      " Read 022 : ./RawData/tracking_gameId_2017091005.csv\n",
      "  - found 258153 records\n",
      " Read 023 : ./RawData/tracking_gameId_2017101503.csv\n",
      "  - found 281015 records\n",
      " Read 024 : ./RawData/tracking_gameId_2017091713.csv\n",
      "  - found 264399 records\n",
      " Read 025 : ./RawData/tracking_gameId_2017100103.csv\n",
      "  - found 270146 records\n",
      " Read 026 : ./RawData/tracking_gameId_2017092405.csv\n",
      "  - found 286064 records\n",
      " Read 027 : ./RawData/tracking_gameId_2017100802.csv\n",
      "  - found 258758 records\n",
      " Read 028 : ./RawData/tracking_gameId_2017101501.csv\n",
      "  - found 351635 records\n",
      " Read 029 : ./RawData/tracking_gameId_2017100113.csv\n",
      "  - found 269206 records\n",
      " Read 030 : ./RawData/tracking_gameId_2017100102.csv\n",
      "  - found 295739 records\n",
      " Read 031 : ./RawData/tracking_gameId_2017100805.csv\n",
      "  - found 307348 records\n",
      " Read 032 : ./RawData/tracking_gameId_2017091009.csv\n",
      "  - found 255890 records\n",
      " Read 033 : ./RawData/tracking_gameId_2017091709.csv\n",
      "  - found 240443 records\n",
      " Read 034 : ./RawData/tracking_gameId_2017091400.csv\n",
      "  - found 280279 records\n",
      " Read 035 : ./RawData/tracking_gameId_2017100811.csv\n",
      "  - found 302867 records\n",
      " Read 036 : ./RawData/tracking_gameId_2017100500.csv\n",
      "  - found 287534 records\n",
      " Read 037 : ./RawData/tracking_gameId_2017091701.csv\n",
      "  - found 239428 records\n",
      " Read 038 : ./RawData/tracking_gameId_2017101500.csv\n",
      "  - found 235613 records\n",
      " Read 039 : ./RawData/tracking_gameId_2017091704.csv\n",
      "  - found 268533 records\n",
      " Read 040 : ./RawData/tracking_gameId_2017092406.csv\n",
      "  - found 251069 records\n",
      " Read 041 : ./RawData/tracking_gameId_2017100111.csv\n",
      "  - found 285175 records\n",
      " Read 042 : ./RawData/tracking_gameId_2017101600.csv\n",
      "  - found 278535 records\n",
      " Read 043 : ./RawData/tracking_gameId_2017100806.csv\n",
      "  - found 261718 records\n",
      " Read 044 : ./RawData/tracking_gameId_2017090700.csv\n",
      "  - found 316026 records\n",
      " Read 045 : ./RawData/tracking_gameId_2017091707.csv\n",
      "  - found 253775 records\n",
      "Removing 45 header rows.\n",
      "Removing 6398 rows with null timestamp.\n",
      "<class 'pandas.core.frame.DataFrame'>\n",
      "Int64Index: 12356010 entries, 1 to 253774\n",
      "Data columns (total 14 columns):\n",
      "time_stamp      object\n",
      "x               float64\n",
      "y               float64\n",
      "s               float64\n",
      "dis             float64\n",
      "dir             float64\n",
      "event           object\n",
      "nflId           int64\n",
      "displayName     object\n",
      "jerseyNumber    int64\n",
      "team            object\n",
      "frame.id        int64\n",
      "gameId          int64\n",
      "playId          int64\n",
      "dtypes: float64(5), int64(5), object(4)\n",
      "memory usage: 1.4+ GB\n",
      "None\n"
     ]
    }
   ],
   "source": [
    "import glob, random, os, numpy as np\n",
    "\n",
    "# Step 1: concat all of the tracking files into a single file\n",
    "# If there is a result file for this step, we can skip it or load it\n",
    "if (os.path.isfile(stepOneFileName)):\n",
    "    if (os.path.isfile(stepTwoFileName)):\n",
    "        if (verbose): print('* Step 1 already exists, skipping : '+stepOneFileName)\n",
    "    else:\n",
    "        if (verbose): print('* Step 1 already exists, loading : '+stepOneFileName)\n",
    "        alltracking = pd.read_csv(stepOneFileName)\n",
    "        if (verbose): print(alltracking.info(verbose=True))\n",
    "else:\n",
    "    # Import all of the tracking csvs and concat them into a single data frame\n",
    "    colnames = ['time_stamp','x','y','s','dis',\n",
    "            'dir','event','nflId','displayName','jerseyNumber',\n",
    "            'team','frame.id','gameId','playId']\n",
    "    allcsvs = rawDataLocation + 'tracking_*.csv'\n",
    "\n",
    "    # Get the glob of all of the files, and the random set of indexes\n",
    "    globAllcsvs = sorted(glob.glob(allcsvs))\n",
    "    if not (-1 < noOfGames < len(globAllcsvs)): noOfGames = len(globAllcsvs)\n",
    "    randomGlobIndexes = random.sample(range(0, len(globAllcsvs)), noOfGames)\n",
    "\n",
    "    # Look through and read the random files, add them to the dataframe\n",
    "    if (verbose): print('Reading '+str(noOfGames)+' out of '+str(len(globAllcsvs))+' files')\n",
    "    for index in range(len(randomGlobIndexes)):\n",
    "        fileName = globAllcsvs[randomGlobIndexes[index]]\n",
    "        if (verbose): print(' Read '+format(index+1, '03')+' : '+fileName)\n",
    "        onefile = pd.read_csv(fileName, names=colnames, low_memory=False,\n",
    "                              parse_dates=['time_stamp'], error_bad_lines=False)\n",
    "        if (verbose): print('  - found '+str(len(onefile.index))+' records')\n",
    "        alltracking = pd.concat([alltracking, onefile], ignore_index=False)\n",
    "\n",
    "    # Get rid of the header rows\n",
    "    if (verbose): print('Removing '\n",
    "                        +str(len(alltracking)-len(alltracking[alltracking['x'] != 'x']))\n",
    "                        +' header rows.')\n",
    "    alltracking = alltracking[alltracking['x'] != 'x']\n",
    "\n",
    "    # Get rid of any rows that don't have anything in the time_stamp column\n",
    "    if (verbose): print('Removing '+str(len(alltracking)-alltracking['time_stamp'].count()\n",
    "                                       +len(alltracking[alltracking['time_stamp'] == 'nan']))\n",
    "                        +' rows with null timestamp.')\n",
    "    alltracking = alltracking.dropna(subset=['time_stamp'])\n",
    "    alltracking = alltracking[alltracking['time_stamp'] != 'nan']\n",
    "\n",
    "    # Make sure all the columns are typed correctly.\n",
    "    alltracking['x'] = alltracking['x'].astype(float);\n",
    "    alltracking['y'] = alltracking['y'].astype(float);\n",
    "    alltracking['s'] = alltracking['s'].astype(float);\n",
    "    alltracking['dis'] = alltracking['dis'].astype(float);\n",
    "    alltracking['dir'] = alltracking['dir'].astype(float);\n",
    "    alltracking['event'] = alltracking['event'].astype(str);\n",
    "    alltracking['nflId'] = alltracking['nflId'].fillna(-1).astype(int);\n",
    "    alltracking['displayName'] = alltracking['displayName'].astype(str);\n",
    "    alltracking['jerseyNumber'] = alltracking['jerseyNumber'].fillna(-1).astype(int);\n",
    "    alltracking['team'] = alltracking['team'].astype(str);\n",
    "    alltracking['frame.id'] = alltracking['frame.id'].astype(int);\n",
    "    alltracking['gameId'] = alltracking['gameId'].astype(int);\n",
    "    alltracking['playId'] = alltracking['playId'].astype(int);\n",
    "    \n",
    "    # Save the interim file\n",
    "    if (saveInterim):\n",
    "        print('Saving step 1 file : '+stepOneFileName)\n",
    "        alltracking.to_csv(stepOneFileName)\n",
    "\n",
    "    if (verbose): print(alltracking.info(verbose=True))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Read plays file : ./RawData/plays.csv\n",
      "  - found 14193 records\n",
      "Merged plays with 12356010 tracking records\n",
      "Read games file : ./RawData/games.csv\n",
      "  - found 91 records\n",
      "Merged games with 12356010 tracking records\n",
      "Read players file : ./RawData/games.csv\n",
      "  - found 1713 records\n",
      "Merged players with 12356010 tracking records\n",
      "<class 'pandas.core.frame.DataFrame'>\n",
      "Int64Index: 12356010 entries, 0 to 12356009\n",
      "Data columns (total 68 columns):\n",
      "time_stamp                object\n",
      "x                         float64\n",
      "y                         float64\n",
      "s                         float64\n",
      "dis                       float64\n",
      "dir                       float64\n",
      "event                     object\n",
      "nflId                     int64\n",
      "displayName               object\n",
      "jerseyNumber              int64\n",
      "team                      object\n",
      "frame.id                  int64\n",
      "gameId                    int64\n",
      "playId                    int64\n",
      "quarter                   int64\n",
      "GameClock                 object\n",
      "down                      int64\n",
      "yardsToGo                 int64\n",
      "possessionTeam            object\n",
      "yardlineSide              object\n",
      "yardlineNumber            float64\n",
      "offenseFormation          object\n",
      "personnel.offense         object\n",
      "defendersInTheBox         float64\n",
      "numberOfPassRushers       float64\n",
      "personnel.defense         object\n",
      "HomeScoreBeforePlay       int64\n",
      "VisitorScoreBeforePlay    int64\n",
      "HomeScoreAfterPlay        int64\n",
      "VisitorScoreAfterPlay     int64\n",
      "isPenalty                 bool\n",
      "isSTPlay                  bool\n",
      "SpecialTeamsPlayType      object\n",
      "KickReturnYardage         float64\n",
      "PassLength                float64\n",
      "PassResult                object\n",
      "YardsAfterCatch           float64\n",
      "PlayResult                int64\n",
      "playDescription           object\n",
      "season                    int64\n",
      "week                      int64\n",
      "gameDate                  object\n",
      "gameTimeEastern           object\n",
      "HomeScore                 int64\n",
      "VisitorScore              int64\n",
      "homeTeamAbbr              object\n",
      "visitorTeamAbbr           object\n",
      "homeDisplayName           object\n",
      "visitorDisplayName        object\n",
      "Stadium                   object\n",
      "Location                  object\n",
      "StadiumType               object\n",
      "Turf                      object\n",
      "GameLength                object\n",
      "GameWeather               object\n",
      "Temperature               float64\n",
      "Humidity                  float64\n",
      "WindSpeed                 object\n",
      "WindDirection             object\n",
      "FirstName                 object\n",
      "LastName                  object\n",
      "PositionAbbr              object\n",
      "EntryYear                 float64\n",
      "DraftRound                float64\n",
      "DraftNumber               float64\n",
      "Height                    object\n",
      "Weight                    float64\n",
      "College                   object\n",
      "dtypes: bool(2), float64(17), int64(17), object(32)\n",
      "memory usage: 6.2+ GB\n",
      "None\n"
     ]
    }
   ],
   "source": [
    "from datetime import time\n",
    "\n",
    "# Step 2: Flatten out by adding the other tables to alltracking\n",
    "# If there is a result file for this step, we can skip it or load it\n",
    "if (os.path.isfile(stepTwoFileName)):\n",
    "    if (os.path.isfile(stepThreeFileName)):\n",
    "        if (verbose): print('* Step 2 already exists, skipping : '+stepTwoFileName)\n",
    "    else:\n",
    "        if (verbose): print('* Step 2 already exists, loading : '+stepTwoFileName)\n",
    "        alltracking = pd.read_csv(stepTwoFileName)\n",
    "        if (verbose): print(alltracking.info(verbose=True))\n",
    "\n",
    "else:\n",
    "    # First add in the plays data, parsing the time to be minutes:seconds\n",
    "    def timeparse (timestamp_str):\n",
    "        timeparsetimes = timestamp_str.split(\":\")\n",
    "        return time(0, int(timeparsetimes[0]), int(timeparsetimes[1]))\n",
    "    playsFile = rawDataLocation + 'plays.csv'\n",
    "    if (verbose): print('Read plays file : '+playsFile)\n",
    "    plays = pd.read_csv(playsFile, parse_dates=['GameClock'], date_parser=timeparse)\n",
    "    if (verbose): print('  - found '+str(len(plays.index))+' records')\n",
    "    alltracking = pd.merge(alltracking, plays, on=['playId','gameId'], how='left')\n",
    "    if (verbose): print('Merged plays with '+str(len(alltracking.index))+' tracking records')\n",
    "\n",
    "    # Next add the games data\n",
    "    gamesFile = rawDataLocation + 'games.csv'\n",
    "    if (verbose): print('Read games file : '+gamesFile)\n",
    "    games = pd.read_csv(gamesFile, parse_dates=['gameDate'])\n",
    "    if (verbose): print('  - found '+str(len(games.index))+' records')\n",
    "    games['gameDate'] = pd.DatetimeIndex(games['gameDate']).date\n",
    "    alltracking = pd.merge(alltracking, games, on=['gameId'], how='left')\n",
    "    if (verbose): print('Merged games with '+str(len(alltracking.index))+' tracking records')\n",
    "\n",
    "    # Lastly, add the player data\n",
    "    playersFile = rawDataLocation + 'players.csv'\n",
    "    if (verbose): print('Read players file : '+gamesFile)\n",
    "    players = pd.read_csv(playersFile)\n",
    "    if (verbose): print('  - found '+str(len(players.index))+' records')\n",
    "    alltracking = pd.merge(alltracking, players, on=['nflId'], how='left')\n",
    "    if (verbose): print('Merged players with '+str(len(alltracking.index))+' tracking records')\n",
    "\n",
    "    # Save the interim file\n",
    "    if (saveInterim):\n",
    "        print('Saving step 2 file : '+stepTwoFileName)\n",
    "        alltracking.to_csv(stepTwoFileName)\n",
    "\n",
    "    if (verbose): print(alltracking.info(verbose=True))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Added team name to 12356010 tracking records\n",
      "Added unique play to 12356010 tracking records\n",
      "Added full game clock to 12356010 tracking records\n",
      "Saving step 3 file : ./ResultsData/45Games_nfl_flatWithTeamUniquePlay.csv\n",
      "<class 'pandas.core.frame.DataFrame'>\n",
      "Int64Index: 12356010 entries, 0 to 12356009\n",
      "Data columns (total 72 columns):\n",
      "time_stamp                object\n",
      "x                         float64\n",
      "y                         float64\n",
      "s                         float64\n",
      "dis                       float64\n",
      "dir                       float64\n",
      "event                     object\n",
      "nflId                     int64\n",
      "displayName               object\n",
      "jerseyNumber              int64\n",
      "team                      object\n",
      "frame.id                  int64\n",
      "gameId                    int64\n",
      "playId                    int64\n",
      "quarter                   int64\n",
      "GameClock                 object\n",
      "down                      int64\n",
      "yardsToGo                 int64\n",
      "possessionTeam            object\n",
      "yardlineSide              object\n",
      "yardlineNumber            float64\n",
      "offenseFormation          object\n",
      "personnel.offense         object\n",
      "defendersInTheBox         float64\n",
      "numberOfPassRushers       float64\n",
      "personnel.defense         object\n",
      "HomeScoreBeforePlay       int64\n",
      "VisitorScoreBeforePlay    int64\n",
      "HomeScoreAfterPlay        int64\n",
      "VisitorScoreAfterPlay     int64\n",
      "isPenalty                 bool\n",
      "isSTPlay                  bool\n",
      "SpecialTeamsPlayType      object\n",
      "KickReturnYardage         float64\n",
      "PassLength                float64\n",
      "PassResult                object\n",
      "YardsAfterCatch           float64\n",
      "PlayResult                int64\n",
      "playDescription           object\n",
      "season                    int64\n",
      "week                      int64\n",
      "gameDate                  object\n",
      "gameTimeEastern           object\n",
      "HomeScore                 int64\n",
      "VisitorScore              int64\n",
      "homeTeamAbbr              object\n",
      "visitorTeamAbbr           object\n",
      "homeDisplayName           object\n",
      "visitorDisplayName        object\n",
      "Stadium                   object\n",
      "Location                  object\n",
      "StadiumType               object\n",
      "Turf                      object\n",
      "GameLength                object\n",
      "GameWeather               object\n",
      "Temperature               float64\n",
      "Humidity                  float64\n",
      "WindSpeed                 object\n",
      "WindDirection             object\n",
      "FirstName                 object\n",
      "LastName                  object\n",
      "PositionAbbr              object\n",
      "EntryYear                 float64\n",
      "DraftRound                float64\n",
      "DraftNumber               float64\n",
      "Height                    object\n",
      "Weight                    float64\n",
      "College                   object\n",
      "teamName                  object\n",
      "uniquePlayId              int64\n",
      "FullGameTime              object\n",
      "FullGameTimeDate          object\n",
      "dtypes: bool(2), float64(17), int64(18), object(35)\n",
      "memory usage: 6.6+ GB\n",
      "None\n"
     ]
    }
   ],
   "source": [
    "from datetime import datetime, timedelta, time\n",
    "\n",
    "# Step 3: Add the game clock, and the unique play index\n",
    "# If there is a result file for this step, we can skip it or load it\n",
    "if (os.path.isfile(stepThreeFileName)):\n",
    "    if (verbose): print('* Step 3 already exists, skipping : '+stepThreeFileName)\n",
    "\n",
    "else:\n",
    "    # Add a team column to every tracking row, based on home or away team\n",
    "    alltracking['teamName'] = np.where(alltracking['team'] == 'away',\n",
    "                                       alltracking['visitorDisplayName'], np.NaN)\n",
    "    alltracking['teamName'] = np.where(alltracking['team'] == 'home',\n",
    "                                       alltracking['homeDisplayName'], alltracking['teamName'])\n",
    "    if (verbose): print('Added team name to '+str(len(alltracking.index))+' tracking records')\n",
    "\n",
    "    # Add a column to give every play a unique id (combination of gameId and playId)\n",
    "    alltracking['uniquePlayId'] = alltracking.apply(lambda row:\n",
    "                                            int(str(row['gameId'])+\n",
    "                                            format(row['playId'], '04')),\n",
    "                                            axis=1)\n",
    "    if (verbose): print('Added unique play to '+str(len(alltracking.index))\n",
    "                        +' tracking records')\n",
    "\n",
    "    # Add a column for full game time that has the game clock for the game, not the qtr\n",
    "    def add_qtr_minutes (row):\n",
    "        addMinutes = 0\n",
    "        if (1 < row['quarter'] < 5 ):\n",
    "            addMinutes = (row['quarter'] - 1) * 15\n",
    "        if (row['quarter'] > 4):\n",
    "            addMinutes = 45 + ((row['quarter'] - 5) * 10)\n",
    "\n",
    "        rowGameClock = datetime.strptime(str(row['GameClock']), '%H:%M:%S').time()\n",
    "        newTotalGameTime = datetime.combine(datetime.today(), time(0,15,0))\n",
    "        newTotalGameTime -= timedelta(minutes=rowGameClock.minute, seconds=rowGameClock.second)\n",
    "        newTotalGameTime += timedelta(minutes=addMinutes)\n",
    "        return newTotalGameTime.time()\n",
    "    alltracking['FullGameTime'] = alltracking.apply(lambda row: add_qtr_minutes(row), axis=1)\n",
    "    if (verbose): print('Added full game clock to '+str(len(alltracking.index))\n",
    "                        +' tracking records')\n",
    "\n",
    "    # Add a new column with the game date and game clock\n",
    "    def add_quarter_minutes_with_date (row):\n",
    "        rowTimeStamp = datetime.strptime(row['time_stamp'], '%Y-%m-%dT%H:%M:%SZ').date()\n",
    "        return str(datetime.combine(rowTimeStamp, row['FullGameTime']))\n",
    "\n",
    "    alltracking['FullGameTimeDate'] = alltracking.apply(lambda row:\n",
    "                                                  add_quarter_minutes_with_date(row), axis=1)\n",
    "\n",
    "    # Save the final file\n",
    "    print('Saving step 3 file : '+stepThreeFileName)\n",
    "    alltracking.to_csv(stepThreeFileName)\n",
    "\n",
    "    if (verbose): print(alltracking.info(verbose=True))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "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": "3.6.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}