EE477

Relational Model

Steven Euijong Whang 1

Recap
● A database is a collection of files with related data
● A DBMS is a piece of software that manages databases
○ Allows users to create new databases and specify their schemas (logical structure of data)
○ Give users ability to query and modify the data using a query language
○ Support the storage of very large amounts of data
○ Enable recovery of the database in the face of failures, errors, or intentional misuse
○ Control access to data from many users at once

2
Data model
● A notation for describing data or information
● Consists of:
○ Structure of the data
○ Operations on the data
○ Constraints on the data

3
Structure of the data
● Referred to as a “conceptual model” of the data
● Higher level than “physical models”, i.e., data structures like arrays and lists
● Example: a relation consists of a schema, attributes, and tuples

title year length genre

Oldboy 2003 120 mystery

Ponyo 2008 103 anime

Frozen 2013 102 anime

4
Operations on the data
● Usually a limited set of operations that can be performed
○ Queries (operations that retrieve information)
SELECT *
○ Modifications (operations that change the database)
FROM Movies
● This is a strength, not a weakness WHERE studioName = ‘Disney’
○ Programmers can describe operations at a very high level AND year = 2013;
○ The DBMS implements them efficiently
○ Not easy to do when coding in C

5
Constraints on the data
● Usually have limitations on the data
● Examples
○ Day of a week is an integer between 1 and 7
○ Age is larger than 0
○ Student IDs are unique

6
Data models
● Relational Most DBMS’s
● Key/Value
● Graph No SQL
● Document (Semi-structured)
● Column-family
● Array/Matrix Machine Learning

● Hierarchical
Obsolete
● Network

7
The relational model
● Structure
○ Based on tables (relations)
○ Looks like an array of structs in C, but this is just one possible implementation
○ In database systems, tables are not stored as main-memory structures and must take into account the
need to access relations on disk

title year length genre

Oldboy 2003 120 mystery

Ponyo 2008 103 anime

Frozen 2013 102 anime

8
The relational model
● Operations
○ Relational algebra
○ E.g., all the rows where genre is “anime”
● Constraints
○ E.g., Genre must be one of a fixed list of values,
no two movies can have the same title

title year length genre

Oldboy 2003 120 mystery

Ponyo 2008 103 anime

Frozen 2013 102 anime

9
The semi-structured model
● Structure <Movies>
○ Resembles trees or graphs, rather than tables or arrays <Movie title=”Oldboy”>
○ <Year>2003</Year>
Represent data by hierarchically nested tagged elements
<Length>120</Length>
● Operations <Genre>mystery</Genre>
</Movie>
○ Involve following path from element to subelements
<Movie title=”Ponyo”>
● Constraints <Year>2008</Year>
…
○ Involve types of values associated with tags </Movies>
○ E.g., <Length> tag values are integers,
each <Movie> element must have a <Length>

10
The key-value model
● Structure
key value
○ (key, value) pairs
○ Key is a string or integer 1000 (oldboy, 2003)
○ Value can be any blob of data
1001 (ponyo, 2008)
● Operations
○ get (key), put(key, value) 1002 (frozen, 2013)
○ Operations on values not supported
● Constraints
○ E.g., key is unique, value is not NULL

11
Comparison of modeling approaches
● Relational model
○ Simple and limited, but reasonably versatile
○ Limited, but useful operations
○ Efficient access to large data
○ A few lines of SQL can do the work of 1000’s of lines of C code
○ Preferred in DBMS’s
● Semi-structured model
○ More flexible, but slower to query
● Key-value model
○ Even more flexible, but cannot query

12
Basics of the relational model
● Relation: two-dimensional table containing data
● Schema: relation name and set of attributes
○ Movies(title, year, length, genre)
● Database schema: set of schemas for the relations of a database columns /
attributes /
● A tuple has one component for each attribute
fields
○ (Oldboy, 2003, 120, mystery)

title year length genre

rows /
tuples / Oldboy 2003 120 mystery
records Ponyo 2008 103 anime

Frozen 2013 102 anime

13
Domains
● Each attribute has a domain
○ A particular elementary type, e.g., integer and string
○ Cannot be a type that can be broken down into components, e.g., set, list, array

Movies(title:string, year:integer, length:integer, genre:string)

14
Equivalent representations of a relation
● A relation is a set of tuples (not a list)
● A schema is a set of attributes (not a list)
● Hence, the order of tuples or attributes of a relation is immaterial

title year length genre year genre title length

Oldboy 2003 120 mystery 2013 anime Frozen 102

Ponyo 2008 103 anime 2003 mystery Oldboy 120

Frozen 2013 102 anime 2008 anime Ponyo 103

15
Pop quiz
How many ways are there to represent this relation?

title year length genre

Oldboy 2003 120 mystery

Ponyo 2008 103 anime

Frozen 2013 102 anime

16
Relation instances
● A relation is not static
○ Tuples may be inserted, deleted, or updated
○ The schema can also change, but this can be expensive
● Instance: the set of tuples of a relation
● A conventional database system only maintains the current instance

17
Keys of relations
● A set of attributes that uniquely identifies a record
● Many real-world databases use artificial keys like ID’s
○ Employee IDs, student IDs, license numbers, …

Key

Movies movieId title year length genre

111 Oldboy 2003 120 mystery

123 Ponyo 2008 103 anime

456 Frozen 2013 102 anime

18
Keys of relations
● There can be multiple keys, but only one can be a primary key

Primary key Another key

Movies movieId title year length genre

111 Oldboy 2003 120 mystery

123 Ponyo 2008 103 anime

456 Frozen 2013 102 anime

19
Keys of relations
● Foreign key: attribute(s) whose value is a key of a record in another relation

StarsIn actor movie foreign key to Movies.movieId

John 123

Will 456

Movies movieId title year length genre

111 Oldboy 2003 120 mystery

123 Ponyo 2008 103 anime

456 Frozen 2013 102 anime

20
Defining a relation schema in SQL
● SQL (Structured Query Language)
○ Principal language to describe and manipulate relational databases
○ Declarative
○ Supports
■ Data Definition Language (DDL): for declaring database schemas
■ Data Manipulation Language (DML): for querying and modifying the database
● Similar analogies in C or Java
○ I.e., declare data vs executable code

21
Relations in SQL
● Tables
○ Can be modified and queried
● Views
○ Relations defined by computation
○ Not stored, but constructed (whole/part) when needed
● Temporary tables
○ Constructed by SQL processor for queries and data modifications
○ Thrown away after use and not stored

22
Data types
● CHAR(n), VARCHAR(n)
○ If CHAR(n) string has fewer than n characters, padded with spaces
○ I.e., if n=4, ‘foo’ is assumed to be ‘foo ’
● BIT(n), BIT VARYING(n)
● BOOLEAN (can have TRUE, FALSE, and UNKNOWN)
● INT, SHORTINT
● FLOAT, REAL, DOUBLE PRECISION, DECIMAL(n,d)
● DATE, TIME

23
Simple table declarations
● To create a table, use CREATE TABLE

CREATE TABLE Movies ( CREATE TABLE MovieStar (

title CHAR(100), name CHAR(30),
year INT, address VARCHAR(30),
length INT, gender CHAR(1),
genre CHAR(10), birthdate DATE
studioName CHAR(30), );
producer INT
);

24
Modifying relation schemas
● To modify a table, use ALTER TABLE and DROP TABLE

DROP TABLE R;
Existing tuples
ALTER TABLE MovieStar ADD phone CHAR(16); will have NULL
values for attribute
ALTER TABLE MovieStar DROP birthdate; phone

25
Default values
● If you prefer not to have NULLs, use default values

gender CHAR(1) DEFAULT ‘?’,

Birthdate DATE DEFAULT DATE ‘0000-00-00’

ALTER TABLE MovieStar ADD phone CHAR(16) DEFAULT ‘unlisted’;

26
Declaring keys
● Declare one attribute to be a key
● Add separate declaration which attributes form a key
○ Need to use this method for multiple-attribute keys

CREATE TABLE MovieStar ( CREATE TABLE MovieStar (

name CHAR(30) PRIMARY KEY, name CHAR(30),
address VARCHAR(30), address VARCHAR(30),
gender CHAR(1), gender CHAR(1),
birthdate DATE birthdate DATE,
); PRIMARY KEY (name, address)
);

27
Declaring keys
● Can also use UNIQUE instead of PRIMARY KEY to indicate keyness
● Some differences
○ Primary key attributes cannot have NULL values
○ There can only be one primary key,
but there can be multiple unique keys CREATE TABLE MovieStar (
name CHAR(30),
address VARCHAR(30),
gender CHAR(1),
birthdate DATE,
UNIQUE (name, address)
);

28
Declaring foreign-key constraints
● Declare attribute(s) of a relation to be a foreign key
○ The referenced attribute(s) of second relation must be UNIQUE or PRIMARY KEY
○ Values of the foreign key must appear in referenced attributes of some tuple

CREATE TABLE Studio (

name CHAR(30) PRIMARY KEY,
address VARCHAR(255),
presidentCert INT REFERENCES MovieExec(cert)
);

CREATE TABLE Studio (

name CHAR(30) PRIMARY KEY,
address VARCHAR(255),
presidentCert INT,
FOREIGN KEY (presidentCert) REFERENCES MovieExec(cert)
); 29
Declaring indexes
● An index is a data structure for efficiently finding tuples with a certain field value
● The most important one is a B-tree, which we will cover later

SELECT *
CREATE INDEX YearIndex ON Movies(year);
FROM Movies
DROP INDEX YearIndex;
WHERE year = 2008

30
Inserting tuples
● A new tuple can be inserted into the relation R using an insertion statement.
○ For any missing attributes of R, the tuple has default values
○ If we provide values for all attributes, the list of attributes can be omitted

INSERT INTO Movies(title, year, length, genre, studio)

VALUES (‘Ponyo’, 2008, 103, ‘anime’, ‘Ghibli’);

producer will have a NULL default value

31
Deleting tuples
● Use a delete statement to delete every tuple satisfying a condition
○ The tuple must be described by a WHERE clause
○ Be careful: omitting the WHERE clause removes all tuples from table

DELETE FROM Movies

WHERE year >= 2008
AND length > 100
AND genre = ‘anime’;

32
Updating tuples
● Change the components of existing tuples in the database
○ Multiple assignments are separated by commas

UPDATE Movies
SET length = 110, Producer = 123
WHERE title = ‘Ponyo’
AND year = 2008;

33
Tables discussion
● Attributes and tuples are not ordered
● Tables are “flat”
● Physical data independence
○ Tables can be implemented / stored on disk differently without affecting the application program

34
Table implementation
● This table can be stored “row major”, i.e., an array of objects
title year length genre

Oldboy 2003 120 mystery

Ponyo 2008 103 anime

Frozen 2013 102 anime

Oldboy Ponyo Frozen

2003 2008 2013
120 103 102
mystery anime anime

35
Table implementation
● or “column major”, i.e., one array per attribute
title year length genre

Oldboy 2003 120 mystery

Ponyo 2008 103 anime

Frozen 2013 102 anime

Oldboy Ponyo Frozen

2003 2008 2013

120 103 102

mystery anime anime 36

Summary
● Data model
○ Structure, operations, constraints
○ Relational, semi-structured, key-value
● Relational model
○ Tables, schema, SQL (DDL)

● Next: SQL (DML)

37