Data Lake

Udacity Data Engineer Nanodegree - Project 4

Sparkify has grown their user base and song database even more and want to move their data warehouse to a data lake. Their data resides in S3, in a directory of JSON logs on user activity on the app, as well as a directory with JSON metadata on the songs in their app.

This project entails building an ETL pipeline that extracts their data from S3, processes them using Spark, and loads the data back into S3 as a set of dimensional tables. This will allow their analytics team to continue finding insights in what songs their users are listening to.

Song Dataset

The first dataset is a subset of real data from the Million Song Dataset. Each file is in JSON format and contains metadata about a song and the artist of that song. The files are partitioned by the first three letters of each song's track ID. For example, here are filepaths to two files in this dataset:

• song_data/A/B/C/TRABCEI128F424C983.json
• song_data/A/A/B/TRAABJL12903CDCF1A.json

And below is an example of what a single song file, TRAABJL12903CDCF1A.json, looks like.

{"num_songs": 1, "artist_id": "ARJIE2Y1187B994AB7", "artist_latitude": null, "artist_longitude": null, "artist_location": "", "artist_name": "Line Renaud", "song_id": "SOUPIRU12A6D4FA1E1", "title": "Der Kleine Dompfaff", "duration": 152.92036, "year": 0}

Log Dataset

The second dataset consists of log files in JSON format generated by this event simulator based on the songs in the dataset above. These simulate app activity logs from an imaginary music streaming app based on configuration settings.

The log files in the dataset you'll be working with are partitioned by year and month. For example, here are filepaths to two files in this dataset.

• log_data/2018/11/2018-11-12-events.json
• log_data/2018/11/2018-11-13-events.json

Below is an example of what the data in a log file, 2018-11-12-events.json, looks like.

Schema for Song Play Analysis

Using the song and log datasets, you'll need to create a star schema optimized for queries on song play analysis. This includes the following tables.

Fact Table

table name: songplays

description: records in log data associated with song plays i.e. records with page NextSong

column	type	description
songplay_id	monotonically_increasing_id	unique identifier for songplay
start_time	timestamp	TODO
user_id	integer	TODO
level	varchar	TODO
song_id	varchar	TODO
artist_id	varchar	TODO
session_id	integer	TODO
location	varchar	TODO
user_agent	varchar	TODO

Dimension Tables

users - users in the app

column	type	description
user_id	integer	TODO
first_name	varchar	TODO
last_name	varchar	TODO
gender	varchar	TODO
level	varchar	TODO

songs - songs in music database

column	type	description
song_id	varchar	TODO
title	varchar	TODO
artist_id	varchar	TODO
year	varchar	TODO
duration	varchar	TODO

artists - artists in music database

column	type	description
artist_id	varchar	TODO
name	varchar	TODO
location	varchar	TODO
latitude	varchar	TODO
longitude	varchar	TODO

time - timestamps of records in songplays broken down into specific units

column	type	description
start_time	integer	TODO
hour	integer	TODO
day	integer	TODO
week	integer	TODO
month	integer	TODO
year	integer	TODO
weekday	integer	TODO

Rationale for schema design

users: there should be a unique row per user, with no duplicates for user_id. It's possible that a user shows up in both the free and paid tiers, but these are not considered duplicates.

songs: duplicates on song_id are dropped here as well as they don't convey any additional useful information. We don't need multiple rows for the same song.

artists: same thing as songs, drop any duplicate artist_id

time: not sure what the purpose of this table has been throughout these projects, but this is just a list of date parts for each songplay from the log events. Duplicates are not dropped here, as multiple users may listen at the same time. That said, there's nothing intrinsic to this table alone that gives any meaning to any of the timestamps, whether or not they are duplicates.

songplays: These are log events where the page column of the event is equal to NextSong. These pages indicate a discrete songplay. Duplicates are dropped for the column start_time are dropped if they exist.

ETL explanation

The job reads in a config file to get the AWS credentials and target S3 bucket. From there, it creates a SparkSession, reads song data and log events from a destination bucket. It does a number of transformations to get the appropriate dataframe structures, and writes to the target bucket.