Description
Part 1: Schemas and Architecture
Suppose you are working for the data team at Bird Scooter, a Santa Monica based startup
that aims to revolutionize how people get around on wheels.
How the Bird Scooter service works: a user installs an app on their phone and enters
their credit card information. The app shows a map of deployed scooters nearby. The
user scans a QR code on the scooter using the app, which activates the scooter for use
and begins the clock for per-minute and per-use billing. The user rides the scooter for
a distance, for a certain number of minutes. When the user is done with the scooter,
he/she leaves it somewhere, and marks the trip as complete in the app.
Each scooter has an identifier, and assume that since Bird is a startup, it only has
10,000 scooters deployed. Each scooter also has a flag that specifies a scooter as online
(deployed), offline (not in use for whatever reason), or lost/stolen. Finally, each scooter
is assigned to a home location that rarely changes. For example, some scooters may
be assigned to UCLA, some may be assigned to Santa Monica, and others to Austin.
Occasionally, a Santa Monica scooter may be reassigned to UCLA depending on demand,
but we expect that such a change is very rare.
Assume that Bird currently only 500,000 registered users. A user is someone that has
installed the app. Each user has an identifier, but not all users have a credit card
number on file as many users install the app and then never ride a scooter. There is also
an expiration date [present if a credit card number is present] and an email address.
Assume, for simplicity, that the app communicates over the Internet directly to a database
server. Being a startup, this is probably how it is done, actually.
1
Exercises
(a) Develop a schema for the scooter and user table based on the requirements and use
case described on the previous page. Write the schema as a CREATE TABLE statement.
Specify a primary key, or composite primary key using the correct syntax. Mark
(with a comment) which attributes, if any, are foreign keys (to determine this, you
may have to answer the other parts first). Try to minimize storage space.
(b) Each interaction between a user and a Bird scooter is called a trip. and we will create
a schema for this trip table. Assume each trip has a unique identifier. The app will
use this database table to determine where an available scooter is located. It will also
(somehow) be used to determine the amount to charge the user.
Additionally, data
scientists would like to be able to use this table to determine daily and hourly trends
in when users start and end scooter use and also identify scooter hotspots: areas
where people frequently activate scooters and park them (just assume a location is
a GPS coordinate: latitude and longitude, which can be represented numerically).
Write the CREATE TABLE statement for this table. Specify the primary key. Mark
(with a comment) which attributes, if any, are foreign keys (to determine this, you
may have to answer the other parts first). Try to minimize storage space.
(c) For the trip table, there are two ways that we can write data to the table from
the app. First, we could insert a row when the user activates the scooter, and then
modify the row when the user parks the scooter and ends the session. Second, we
can simply cache the ride data on the phone, and at the end of the session transmit
this data to the database server to be inserted as a row. What are the advantages
and disadvantages of both of these methods? Which would you (an employee at
Bird) prefer and why? Is there an even better way?
(d) Using the tables you just developed in all of Part 1, draw the schema diagram. For
an example of a schema diagram, see Figure 2.8 in the text (page 47), or page 23 in
the lecture slides for Lecture 2.
Some other things to think about:
• Would you include the number of minutes the trip lasted in the trip table? Why
or why not?
• What are the advantages and disadvantages of including the charge to the user in
the trips table?
2
Part 2: SQL and Relational Algebra
If you wish to check your syntax, you can load this dataset into MySQL by following the
directions on CCLE under Homework 1.
Hints: For some of these queries, you will need to use functions on attributes. Check
out the list of date and time functions here, as it should be very useful: https://dev.
mysql.com/doc/refman/5.7/en/date-and-time-functions.html. You do not need
to memorize them! Note that these are not aggregation functions because aggregation
functions may take multiple inputs and produce one output per group. These functions
take one input and return one output without using any grouping variables. This does
not mean that your queries will not use the aggregation functions we discussed in class
though.
Bay Area Rapid Transit (BART) is a subway system that stops at various places on the
Bay Area Peninsula, City of San Francisco, underwater to Oakland and the East Bay.
When a user inserts a ticket, or scans a pass card on a turnstile, BART records that a
user entered the subway system. On exit, the passenger does the same thing to exit the
station (inserts a ticket or scans a fare card) and BART records that the user’s journey is
complete. Throughput can be defined as the number of passengers that entered at origin
A, and exited at destination B.
In this exercise, we will do various queries on this data to answer several interesting
questions.
This dataset consists of BART ridership data from 2017. The schema for the two tables
in this database are provided below in case you do not wish to use the data.
— This table is for your own information to see which station is which.
CREATE TABLE station(
Abbreviation CHAR(4),
Description VARCHAR(1000),
Location VARCHAR(23),
Name VARCHAR(50)
);
CREATE TABLE rides2017(
Origin CHAR(4),
Destination CHAR(4),
Throughput INT,
DateTime DATETIME
);
3
Exercises
(a) Write a query to compute the total throughput (passengers, or number of trips) by
time of day for the year of 2017. The result should contain only the hour of day,
named hour, and the number of trips named trips.
(b) Write a query that lists the one pair of station codes that had the largest throughput
on the weekdays in 2017.
(c) Write a query that returns the 5 destinations that saw the highest average throughput on Mondays between 7am and 10am, and rank them from highest to lowest.
Return the destinations and their averages.
(d) It is 2018. Suppose you are working with BART to expand stations and services
based on 2017 ridership data. You want to know which stations are overcrowded.
Suppose we consider an originating station as overcrowded if the maximum hourly
throughput across all one-hour periods in the year is greater than 100 times the
average hourly throughput across all one-hour periods in the year. Write a query
that retrieves all originating stations that were overcrowded in 2017. (Passengers
spend more time in originating stations because they must wait for the train there.)
(e) Suppose we take the result from part (a) and call it hourly ridership. Given the
following query, write the equivalent expression in the relational algebra.
SELECT
hour,
trips / 100
FROM hourly_ridership
WHERE (hour >= 7 AND hour < 10) OR (hour >= 17 OR hour < 19);
(f) Suppose we want to study how the weather affects how busy particular stations
are. In the Occupancy relation, we have the name of a station called Station, the
DateTime, and the number of people passing through the station called Riders, as
attributes.
In the Weather relation, we have the Station, DateTime, Temperature,
Condition (for simplicity assume a string, like “cloudy”) attributes. We are only
interested in comparing occupancy during “sunny” hours and “rainy” hours, and we
only care about Station, DateTime, Riders and Condition. Write an expression
in the relational algebra that represents this context.
4
