{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {
    "editable": true
   },
   "outputs": [],
   "source": [
    "import configparser\n",
    "from datetime import datetime\n",
    "from pyspark.sql.functions import dayofweek\n",
    "import os\n",
    "from pyspark.sql import SparkSession\n",
    "from pyspark.sql.functions import udf, col\n",
    "from pyspark.sql.functions import year, month, dayofmonth, hour, weekofyear, date_format"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "editable": true
   },
   "outputs": [],
   "source": [
    "def create_spark_session():\n",
    "    spark = SparkSession \\\n",
    "        .builder \\\n",
    "        .config(\"spark.jars.packages\", \"org.apache.hadoop:hadoop-aws:2.7.0\") \\\n",
    "        .getOrCreate()\n",
    "    return spark"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "editable": true
   },
   "outputs": [],
   "source": [
    "def process_song_data(spark, input_data, output_data):\n",
    "    \"\"\"\n",
    "    This function reads the song data from S3, processes that data using Spark, and writes them back to S3\n",
    "\n",
    "    Parameters\n",
    "    ----------\n",
    "        spark: This is the spark session\n",
    "        input_date: This is the path to song_date into S3 bucket\n",
    "        output_data: This is the path where parquet files will be written\n",
    "    \n",
    "    \"\"\"\n",
    "    \n",
    "    # get filepath to song data file\n",
    "    # e.g. from S3 file location: song_data/A/B/C/TRABCEI128F424C983.json\n",
    "    song_data = input_data + 'song_data/*/*/*/*.json'\n",
    "    \n",
    "    # read song data file\n",
    "    df = spark.read.json(song_data)\n",
    "\n",
    "    # extract columns to create songs table\n",
    "    songs_table = df.select('song_id', 'title', 'artist_id', 'year', 'duration').dropDuplicates()\n",
    "    songs_table.createOrReplaceTempView('songs')\n",
    "    \n",
    "    # write songs table to parquet files partitioned by year and artist\n",
    "    songs_table.write.partitionBy('year', 'artist_id').parquet(os.path.join(output_data, 'songs/songs.parquet'), 'overwrite')\n",
    "\n",
    "    # extract columns to create artists table\n",
    "    artists_table = df.select('artist_id', 'artist_name', 'artist_location', 'artist_latitude', 'artist_longitude') \\\n",
    "                      .withColumnRenamed('artist_name', 'name') \\\n",
    "                      .withColumnRenamed('artist_location', 'location') \\\n",
    "                      .withColumnRenamed('artist_latitude', 'latitude') \\\n",
    "                      .withColumnRenamed('artist_longitude', 'longitude').dropDuplicates()\n",
    "    \n",
    "    artists_table.createOrReplaceTempView('artists')\n",
    "    \n",
    "    \n",
    "    # write artists table to parquet files\n",
    "    artists_table.write.parquet(os.path.join(output_data, 'artists/artists.parquet'), 'overwrite')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "metadata": {
    "editable": true
   },
   "outputs": [],
   "source": [
    "def process_log_data(spark, input_data, output_data):\n",
    "    \"\"\"\n",
    "    This function reads the log data from S3, processes that data using Spark, and writes them back to S3\n",
    "\n",
    "    Parameters\n",
    "    ----------\n",
    "        spark: This is the spark session\n",
    "        input_date: This is the path to log_date into S3 bucket\n",
    "        output_data: This is the path where parquet files will be written\n",
    "    \n",
    "    \"\"\"\n",
    "    \n",
    "    \n",
    "    # get filepath to log data file\n",
    "    # e.g. from S3 file location: log_data/2018/11/2018-11-12-events.json\n",
    "    log_data = input_data + 'log_data/*.json'\n",
    "\n",
    "    # read log data file\n",
    "    df = spark.read.json(log_data)\n",
    "    \n",
    "    # filter by actions for song plays\n",
    "    # df_actions = df.filter(df.page == \"NextSong\").select('ts', 'userId', 'level', 'song', 'artist', 'sessionId', 'location', 'userAgent')\n",
    "    df_actions = df.where(df.page == 'NextSong')\n",
    "    df_actions.select('ts', 'userId', 'level', 'song', 'artist', 'sessionId', 'location', 'userAgent')\n",
    "    \n",
    "    # extract columns for users table    \n",
    "    users_table = df.select('userId', 'firstName', 'lastName', 'gender', 'level').dropDuplicates()\n",
    "    users_table.createOrReplaceTempView('users')\n",
    "    \n",
    "    # write users table to parquet files\n",
    "    users_table.write.parquet(os.path.join(output_data, 'users/users.parquet'), 'overwrite')\n",
    "\n",
    "    # create timestamp column from original timestamp column\n",
    "    get_timestamp = udf(lambda x: str(int(int(x)/1000)))\n",
    "    df = df.withColumn('timestamp', get_timestamp(df_actions.ts))\n",
    "    \n",
    "    # create datetime column from original timestamp column\n",
    "    get_datetime = udf(lambda x: str(datetime.fromtimestamp(int(x) / 1000)))\n",
    "    df = df.withColumn('datetime', get_datetime(df_actions.ts))\n",
    "    \n",
    "    # extract columns to create time table\n",
    "    time_table = df.select('datetime') \\\n",
    "                           .withColumn('start_time', df.datetime) \\\n",
    "                           .withColumn('hour', hour('datetime')) \\\n",
    "                           .withColumn('day', dayofmonth('datetime')) \\\n",
    "                           .withColumn('week', weekofyear('datetime')) \\\n",
    "                           .withColumn('month', month('datetime')) \\\n",
    "                           .withColumn('year', year('datetime')) \\\n",
    "                           .withColumn('weekday', dayofweek('datetime')) \\\n",
    "                           .dropDuplicates()\n",
    "    \n",
    "    # write time table to parquet files partitioned by year and month\n",
    "    time_table.write.partitionBy('year', 'month').parquet(os.path.join(output_data,'time/time.parquet'), 'overwrite')\n",
    "\n",
    "    # read in song data to use for songplays table\n",
    "    song_df = spark.read.json(input_data + 'song_data/*/*/*/*.json')\n",
    "\n",
    "    # extract columns from joined song and log datasets to create songplays table\n",
    "    df = df.alias('log_df')\n",
    "    song_df = song_df.alias('song_df')\n",
    "    log_and_songs = df.join(song_df, col('log_df.artist') == col('song_df.artist_name'), 'inner')\n",
    "    \n",
    "    \n",
    "    songplays_table = log_and_songs.select(\n",
    "        col('log_df.datetime').alias('start_time'),\n",
    "        col('log_df.userId').alias('user_id'),\n",
    "        col('log_df.level').alias('level'),\n",
    "        col('song_df.song_id').alias('song_id'),\n",
    "        col('song_df.artist_id').alias('artist_id'),\n",
    "        col('log_df.sessionId').alias('session_id'),\n",
    "        col('log_df.location').alias('location'), \n",
    "        col('log_df.userAgent').alias('user_agent'),\n",
    "        year('log_df.datetime').alias('year'),\n",
    "        month('log_df.datetime').alias('month'))\n",
    "\n",
    "    songplays_table.createOrReplaceTempView('songplays')\n",
    "    \n",
    "    # write songplays table to parquet files partitioned by year and month\n",
    "    time_table = time_table.alias('timetable')\n",
    "\n",
    "    songplays_table.write.partitionBy('year', 'month').parquet(os.path.join(output_data,'songplays/songplays.parquet'),'overwrite')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "editable": true
   },
   "outputs": [],
   "source": [
    "    '''Start spark session and define S3 locations where the files will be readed and written'''\n",
    "    spark = create_spark_session()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "editable": true
   },
   "outputs": [],
   "source": [
    "    '''Paths for local testing'''\n",
    "    input_data = \"data/\"\n",
    "    output_data = \"data/outputs/\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "metadata": {
    "editable": true
   },
   "outputs": [],
   "source": [
    "    '''Run above functions according to the previous parameters '''\n",
    "    process_song_data(spark, input_data, output_data)    \n",
    "    process_log_data(spark, input_data, output_data)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {
    "editable": true
   },
   "outputs": [],
   "source": [
    "from pyspark.sql.functions import dayofweek\n",
    "df = spark.read.json(\"data/log_data/*.json\")\n",
    "df_actions = df.where(df.page == 'NextSong')\n",
    "df_actions.select('ts', 'userId', 'level', 'song', 'artist', 'sessionId', 'location', 'userAgent')\n",
    "\n",
    "# create timestamp column from original timestamp column\n",
    "get_timestamp = udf(lambda x: str(int(int(x)/1000)))\n",
    "df = df.withColumn('timestamp', get_timestamp(df_actions.ts))\n",
    "\n",
    "# create datetime column from original timestamp column\n",
    "get_datetime = udf(lambda x: str(datetime.fromtimestamp(int(x) / 1000)))\n",
    "df = df.withColumn('datetime', get_datetime(df_actions.ts))\n",
    "\n",
    "# extract columns to create time table\n",
    "time_table = df.select('datetime') \\\n",
    "                           .withColumn('start_time', df.datetime) \\\n",
    "                           .withColumn('hour', hour('datetime')) \\\n",
    "                           .withColumn('day', dayofmonth('datetime')) \\\n",
    "                           .withColumn('week', weekofyear('datetime')) \\\n",
    "                           .withColumn('month', month('datetime')) \\\n",
    "                           .withColumn('year', year('datetime')) \\\n",
    "                           .withColumn('weekday', dayofweek('datetime')) \\\n",
    "                           .dropDuplicates()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {
    "editable": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+--------------------+--------------------+----+---+----+-----+----+-------+\n",
      "|            datetime|          start_time|hour|day|week|month|year|weekday|\n",
      "+--------------------+--------------------+----+---+----+-----+----+-------+\n",
      "|2018-11-01 21:05:...|2018-11-01 21:05:...|  21|  1|  44|   11|2018|      5|\n",
      "|2018-11-01 21:42:...|2018-11-01 21:42:...|  21|  1|  44|   11|2018|      5|\n",
      "|2018-11-01 21:17:...|2018-11-01 21:17:...|  21|  1|  44|   11|2018|      5|\n",
      "+--------------------+--------------------+----+---+----+-----+----+-------+\n",
      "only showing top 3 rows\n",
      "\n"
     ]
    }
   ],
   "source": [
    "time_table.show(3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "editable": true
   },
   "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": "3.6.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}