Database Design Course notes

Concepts

  • Entity - Table name (e.g. user, article)
  • Attribute / Field / Column
  • Relation: relation between entities
    • 1:1 - One to one
    • 1:N - One to many
    • M:N - Many to many
  • Tuple / Row / Record / Entry
  • Table / File: Collection of tuples or rows and attribute names
  • Database design: Process of designing data tables to guarantee data integrity
  • Schema: structure of the database
  • Keys: Used to guarantee uniqueness. Usually IDs.
  • SQL: Structured Query Language
    • DDL: Data definition language. Part of SQL to define schema.
    • DML: Data management language. Part of SQL to query data.
  • First formal form (1NF): https://en.wikipedia.org/wiki/First_normal_form
    • Atomic columns
      • Value should be 1 thing. E.g. Full name should be splitted into First name and Last name
    • Singular columns
      • E.g. "favourite movies" would not work because you could store multiple items in the same field.
    • No table column can have tables as values. Most RDBMS enforce 1NF. No SQL systems allow an attribute to have relations and subvalues.
  • Parent / Child:
    • In a relationship, there is always a parent and a child
    • Parent: Has Primary Key (PK)
    • Child: Has Foreign Key (FK)
  • Intermediary table / Juntion table:
    • For implementing M:N relationships
  • Binary relatioships:
    • Usually, relationships are between two entities.
    • It could be between more than two entities
  • Index:
    • A database index is a data structure that improves the speed of data retrieval operations on a database table at the cost of additional writes and storage space to maintain the index data structure. https://en.wikipedia.org/wiki/Database_index
  • Look up table
    • Instead of repeating a membership like gold, silver, etc. Have a second table and reference it. Kind of Enum tables.
    • user:
      • | membership_id |
    • membership:
      • | id | membership |

Keys

  • Superkey: Any number of columns that create a unique key combining the values
    • There are many combinations of columns that create a unique key, so there are many superkeys.
  • Candidate key: Superkey with the least number of columns
    • Could have many candidate keys: email, username, ID.
  • Primary key:
    • Unique
    • Never changing
    • Never null
  • Alternate key
    • Unique, never changing, never null.
    • Other candidate keys that are not the primary
  • Surrogate key: Key that is created to guarantee uniqueness, oposed to natural keys. E.g. user_id.
    • Could be autoincrement or UUID
    • Surrogate are not usually given to the user. Because if you share the ID, it starts to gain meaning and become a natural key.
  • Natural key: Key that occurs naturally in data. E.g. username could be a natural key.
  • Surrogate key vs Natural key:
    • Usually easier to use surrogate keys than natural keys
      • Don't need to worry about data changing, because they will be unique
    • Be consistent and always use surrogate or natural keys
  • Foreign key:
    • Reference to a primary key from a different table
    • NOT NULL: to require and ensure there is a relationship
    • Constraints
    • Foreign key constraints ^986268
      • ON UPDATE (if the foreign key in the parent changes)
        • CASCADE: Update the foreign key to reflect the change
        • RESTRICT (NO ACTION): Restrict the change in the parent table
        • SET NULL: Set NULL in the foreign key column
      • ON DELETE (if row with primary key referenced in foreign key is deleted)
        • CASCADE: Delete the row with the foreign key
        • RESTRICT (NO ACTION): Restrict the change in the parent table
        • SET NULL: Set NULL in the foreign key column
      • RESTRICT vs NO ACTION: In Postgres, the difference between RESTRICT and NO ACTION only arises when you define a constraint as DEFERRABLE with an INITIALLY DEFERRED or INITIALLY IMMEDIATE mode. https://www.postgresql.org/docs/current/sql-set-constraints.html
  • Simple key: One column
  • Composite key: Multiple columns
  • Compound key: Same as composite keys. Normally used interchangably, but better to stick to Composite key. Compound keys usually have all of the attributes being keys to other entities (e.g. many-to-many relationships). https://dba.stackexchange.com/questions/3134/in-sql-is-it-composite-or-compound-keys

Data integrity

  • Entity integrity: Ensure uniqueness in every entity. Usually solved with an ID column.
  • Relational integrity / Referencial integrity: Usually managed by RDBMS to ensure relations exist and are safe.
    • Foreign keys should have contraints [[#^986268]]
  • Domain integrity: Data is valid and follows the expected type. Partly solved with types. Also adding validations would help.

Naming convention

  • All lowercase
    • E.g. user, instead of User
  • Singular entity name
    • E.g. user, article
  • Underscores to separate entity words
    • E.g. card_payment instead of CardPayment
  • Name ID columns with the table name: E.g. "user" table has "user_id". This way, all the tables referencing user_id would have the same name for it, including the "user" table.

Database Relationships

  • One to one
  • One to many
  • Many to many

One to one relationship (1:1)

Could be:

  • Attribute
    • If only need one data point to define the entity.
    • E.g. username: text
  • Foreign key to another table
    • If you have more than one attribute for the same entity
      • E.g. Card number, card name, card issue date, etc.
    • Add foreign key on parent table
      • | user | card_id |
      • | card |

One to many relationship (1:N)

  • Add foreign key to child table
    • | user |
    • | card | user_id |

Many to many relationship (M:N)

  • Create a Intermediary / Junction table:
    • | user |
    • | user_id | card_id |
    • | card |

Entity relationship modeling

Standard for drawing schemas

Cardinality
Modality

Normalization

  • 1NF: First normal form
    • Making everything atomic
    • A relation is in first normal form if and only if no attribute domain have relations as elements.
    • Basically, if a field has multiple values, that should be split into two tables.
      • | user_id |
      • | email_id | email | user_id (fk) | (this allows for multiple emails)
  • 2NF: Second normal form
    • Removing partial dependencies
    • A relation is in 2NF if it is in 1NF.
    • There are no partial dependencies for the relationship.
    • E.g.
      • user: | user_id |
      • book_author: | book_id | user_id | book_summary (wrong because it depends on the book_id) |
      • book: | book_id | title |
  • 3NF: Third normal form
    • Removing transitive dependencies
    • E.g. poor design:
      • | review_id | stars | stars_meaning |
      • Star meaning changes with stars, so depends on something that is not a key
    • E.g. better design:
      • | review_id | star_id |
      • | star_id | stars | stars_meaning |

Indexes

  • Clustered
    • Reorganizes the data
    • Can only have one
  • Non clustered
    • Points to the data
    • E.g. end of the book index
    • Can have multiple
  • Composite
    • Index on multiple columns

Pros:

  • Faster queries

Cons:

  • Storage and memory
  • Need to update the index when changes on the data happen. Writes are slower.

Joins

  • https://en.wikipedia.org/wiki/Join_(SQL)
  • INNER JOIN (default JOIN)
    • Default join between two tables
    • Only returns the rows that match the condition in both tables
  • Outer joins
    • Include all columns that don't match one in the joined table, including null values.
    • When the joined column is a foreign key and NOT NULL, outer joins would return the same than inner joins.
    • Types:
      • Left outer join (default): Include all non-matching rows in the left table
      • Right outer join: Include all non-matching rows in the right table
      • Full outer join: Include non-matching rows from all tables
  • Self join:
    • Join a table with itself
    • For example:
      • | user_id | name | referred_by |
      • Self join on the referred_by to find the user and the referral
SELECT u1.name, u2.name AS "referred_by"
FROM user AS u1
INNER JOIN user AS u2
ON u1.referred_by = u2.user_id